diff options
Diffstat (limited to 'synapse')
252 files changed, 9919 insertions, 7220 deletions
diff --git a/synapse/__init__.py b/synapse/__init__.py index 1c27d68009..0dd538d804 100644 --- a/synapse/__init__.py +++ b/synapse/__init__.py @@ -36,7 +36,7 @@ try: except ImportError: pass -__version__ = "1.5.1" +__version__ = "1.8.0" if bool(os.environ.get("SYNAPSE_TEST_PATCH_LOG_CONTEXTS", False)): # We import here so that we don't have to install a bunch of deps when diff --git a/synapse/api/auth.py b/synapse/api/auth.py index a1fc542c76..86f7f5b144 100644 --- a/synapse/api/auth.py +++ b/synapse/api/auth.py @@ -25,13 +25,7 @@ from twisted.internet import defer import synapse.logging.opentracing as opentracing import synapse.types from synapse import event_auth -from synapse.api.constants import ( - EventTypes, - JoinRules, - LimitBlockingTypes, - Membership, - UserTypes, -) +from synapse.api.constants import EventTypes, LimitBlockingTypes, Membership, UserTypes from synapse.api.errors import ( AuthError, Codes, @@ -40,8 +34,13 @@ from synapse.api.errors import ( ResourceLimitError, ) from synapse.config.server import is_threepid_reserved +<<<<<<< HEAD from synapse.types import UserID from synapse.util.caches import register_cache +======= +from synapse.types import StateMap, UserID +from synapse.util.caches import CACHE_SIZE_FACTOR, register_cache +>>>>>>> origin/develop from synapse.util.caches.lrucache import LruCache from synapse.util.metrics import Measure @@ -84,7 +83,7 @@ class Auth(object): @defer.inlineCallbacks def check_from_context(self, room_version, event, context, do_sig_check=True): - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() auth_events_ids = yield self.compute_auth_events( event, prev_state_ids, for_verification=True ) @@ -513,71 +512,40 @@ class Auth(object): """ return self.store.is_server_admin(user) - @defer.inlineCallbacks - def compute_auth_events(self, event, current_state_ids, for_verification=False): - if event.type == EventTypes.Create: - return [] - - auth_ids = [] - - key = (EventTypes.PowerLevels, "") - power_level_event_id = current_state_ids.get(key) - - if power_level_event_id: - auth_ids.append(power_level_event_id) - - key = (EventTypes.JoinRules, "") - join_rule_event_id = current_state_ids.get(key) + def compute_auth_events( + self, event, current_state_ids: StateMap[str], for_verification: bool = False, + ): + """Given an event and current state return the list of event IDs used + to auth an event. - key = (EventTypes.Member, event.sender) - member_event_id = current_state_ids.get(key) + If `for_verification` is False then only return auth events that + should be added to the event's `auth_events`. - key = (EventTypes.Create, "") - create_event_id = current_state_ids.get(key) - if create_event_id: - auth_ids.append(create_event_id) + Returns: + defer.Deferred(list[str]): List of event IDs. + """ - if join_rule_event_id: - join_rule_event = yield self.store.get_event(join_rule_event_id) - join_rule = join_rule_event.content.get("join_rule") - is_public = join_rule == JoinRules.PUBLIC if join_rule else False - else: - is_public = False + if event.type == EventTypes.Create: + return defer.succeed([]) + + # Currently we ignore the `for_verification` flag even though there are + # some situations where we can drop particular auth events when adding + # to the event's `auth_events` (e.g. joins pointing to previous joins + # when room is publically joinable). Dropping event IDs has the + # advantage that the auth chain for the room grows slower, but we use + # the auth chain in state resolution v2 to order events, which means + # care must be taken if dropping events to ensure that it doesn't + # introduce undesirable "state reset" behaviour. + # + # All of which sounds a bit tricky so we don't bother for now. - if event.type == EventTypes.Member: - e_type = event.content["membership"] - if e_type in [Membership.JOIN, Membership.INVITE]: - if join_rule_event_id: - auth_ids.append(join_rule_event_id) + auth_ids = [] + for etype, state_key in event_auth.auth_types_for_event(event): + auth_ev_id = current_state_ids.get((etype, state_key)) + if auth_ev_id: + auth_ids.append(auth_ev_id) - if e_type == Membership.JOIN: - if member_event_id and not is_public: - auth_ids.append(member_event_id) - else: - if member_event_id: - auth_ids.append(member_event_id) - - if for_verification: - key = (EventTypes.Member, event.state_key) - existing_event_id = current_state_ids.get(key) - if existing_event_id: - auth_ids.append(existing_event_id) - - if e_type == Membership.INVITE: - if "third_party_invite" in event.content: - key = ( - EventTypes.ThirdPartyInvite, - event.content["third_party_invite"]["signed"]["token"], - ) - third_party_invite_id = current_state_ids.get(key) - if third_party_invite_id: - auth_ids.append(third_party_invite_id) - elif member_event_id: - member_event = yield self.store.get_event(member_event_id) - if member_event.content["membership"] == Membership.JOIN: - auth_ids.append(member_event.event_id) - - return auth_ids + return defer.succeed(auth_ids) @defer.inlineCallbacks def check_can_change_room_list(self, room_id, user): diff --git a/synapse/api/constants.py b/synapse/api/constants.py index 49c4b85054..0ade47e624 100644 --- a/synapse/api/constants.py +++ b/synapse/api/constants.py @@ -1,7 +1,8 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 OpenMarket Ltd # Copyright 2017 Vector Creations Ltd -# Copyright 2018 New Vector Ltd +# Copyright 2018-2019 New Vector Ltd +# Copyright 2019 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -94,6 +95,8 @@ class EventTypes(object): ServerACL = "m.room.server_acl" Pinned = "m.room.pinned_events" + Retention = "m.room.retention" + class RejectedReason(object): AUTH_ERROR = "auth_error" @@ -145,3 +148,7 @@ class EventContentFields(object): # Labels for the event, cf https://github.com/matrix-org/matrix-doc/pull/2326 LABELS = "org.matrix.labels" + + # Timestamp to delete the event after + # cf https://github.com/matrix-org/matrix-doc/pull/2228 + SELF_DESTRUCT_AFTER = "org.matrix.self_destruct_after" diff --git a/synapse/api/errors.py b/synapse/api/errors.py index 5853a54c95..1c9456e583 100644 --- a/synapse/api/errors.py +++ b/synapse/api/errors.py @@ -17,13 +17,15 @@ """Contains exceptions and error codes.""" import logging -from typing import Dict +from typing import Dict, List from six import iteritems from six.moves import http_client from canonicaljson import json +from twisted.web import http + logger = logging.getLogger(__name__) @@ -80,6 +82,29 @@ class CodeMessageException(RuntimeError): self.msg = msg +class RedirectException(CodeMessageException): + """A pseudo-error indicating that we want to redirect the client to a different + location + + Attributes: + cookies: a list of set-cookies values to add to the response. For example: + b"sessionId=a3fWa; Expires=Wed, 21 Oct 2015 07:28:00 GMT" + """ + + def __init__(self, location: bytes, http_code: int = http.FOUND): + """ + + Args: + location: the URI to redirect to + http_code: the HTTP response code + """ + msg = "Redirect to %s" % (location.decode("utf-8"),) + super().__init__(code=http_code, msg=msg) + self.location = location + + self.cookies = [] # type: List[bytes] + + class SynapseError(CodeMessageException): """A base exception type for matrix errors which have an errcode and error message (as well as an HTTP status code). @@ -158,12 +183,6 @@ class UserDeactivatedError(SynapseError): ) -class RegistrationError(SynapseError): - """An error raised when a registration event fails.""" - - pass - - class FederationDeniedError(SynapseError): """An error raised when the server tries to federate with a server which is not on its federation whitelist. diff --git a/synapse/api/filtering.py b/synapse/api/filtering.py index bec13f08d8..6eab1f13f0 100644 --- a/synapse/api/filtering.py +++ b/synapse/api/filtering.py @@ -1,5 +1,8 @@ # -*- coding: utf-8 -*- # Copyright 2015, 2016 OpenMarket Ltd +# Copyright 2017 Vector Creations Ltd +# Copyright 2018-2019 New Vector Ltd +# Copyright 2019 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. diff --git a/synapse/api/urls.py b/synapse/api/urls.py index ff1f39e86c..f34434bd67 100644 --- a/synapse/api/urls.py +++ b/synapse/api/urls.py @@ -29,7 +29,6 @@ FEDERATION_V2_PREFIX = FEDERATION_PREFIX + "/v2" FEDERATION_UNSTABLE_PREFIX = FEDERATION_PREFIX + "/unstable" STATIC_PREFIX = "/_matrix/static" WEB_CLIENT_PREFIX = "/_matrix/client" -CONTENT_REPO_PREFIX = "/_matrix/content" SERVER_KEY_V2_PREFIX = "/_matrix/key/v2" MEDIA_PREFIX = "/_matrix/media/r0" LEGACY_MEDIA_PREFIX = "/_matrix/media/v1" diff --git a/synapse/app/_base.py b/synapse/app/_base.py index 2ac7d5c064..0e8b467a3e 100644 --- a/synapse/app/_base.py +++ b/synapse/app/_base.py @@ -237,6 +237,12 @@ def start(hs, listeners=None): """ Start a Synapse server or worker. + Should be called once the reactor is running and (if we're using ACME) the + TLS certificates are in place. + + Will start the main HTTP listeners and do some other startup tasks, and then + notify systemd. + Args: hs (synapse.server.HomeServer) listeners (list[dict]): Listener configuration ('listeners' in homeserver.yaml) @@ -269,7 +275,7 @@ def start(hs, listeners=None): # It is now safe to start your Synapse. hs.start_listening(listeners) - hs.get_datastore().start_profiling() + hs.get_datastore().db.start_profiling() setup_sentry(hs) setup_sdnotify(hs) @@ -311,9 +317,7 @@ def setup_sdnotify(hs): # Tell systemd our state, if we're using it. This will silently fail if # we're not using systemd. - hs.get_reactor().addSystemEventTrigger( - "after", "startup", sdnotify, b"READY=1\nMAINPID=%i" % (os.getpid(),) - ) + sdnotify(b"READY=1\nMAINPID=%i" % (os.getpid(),)) hs.get_reactor().addSystemEventTrigger( "before", "shutdown", sdnotify, b"STOPPING=1" diff --git a/synapse/app/admin_cmd.py b/synapse/app/admin_cmd.py index 04751a6a5e..1c7c6ec0c8 100644 --- a/synapse/app/admin_cmd.py +++ b/synapse/app/admin_cmd.py @@ -45,7 +45,6 @@ from synapse.replication.slave.storage.registration import SlavedRegistrationSto from synapse.replication.slave.storage.room import RoomStore from synapse.replication.tcp.client import ReplicationClientHandler from synapse.server import HomeServer -from synapse.storage.engines import create_engine from synapse.util.logcontext import LoggingContext from synapse.util.versionstring import get_version_string @@ -85,8 +84,7 @@ class AdminCmdServer(HomeServer): class AdminCmdReplicationHandler(ReplicationClientHandler): - @defer.inlineCallbacks - def on_rdata(self, stream_name, token, rows): + async def on_rdata(self, stream_name, token, rows): pass def get_streams_to_replicate(self): @@ -105,8 +103,10 @@ def export_data_command(hs, args): user_id = args.user_id directory = args.output_directory - res = yield hs.get_handlers().admin_handler.export_user_data( - user_id, FileExfiltrationWriter(user_id, directory=directory) + res = yield defer.ensureDeferred( + hs.get_handlers().admin_handler.export_user_data( + user_id, FileExfiltrationWriter(user_id, directory=directory) + ) ) print(res) @@ -229,14 +229,10 @@ def start(config_options): synapse.events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - ss = AdminCmdServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ss, config, use_worker_options=True) diff --git a/synapse/app/appservice.py b/synapse/app/appservice.py index 02b900f382..2217d4a4fb 100644 --- a/synapse/app/appservice.py +++ b/synapse/app/appservice.py @@ -34,7 +34,6 @@ from synapse.replication.slave.storage.events import SlavedEventStore from synapse.replication.slave.storage.registration import SlavedRegistrationStore from synapse.replication.tcp.client import ReplicationClientHandler from synapse.server import HomeServer -from synapse.storage.engines import create_engine from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.versionstring import get_version_string @@ -116,9 +115,8 @@ class ASReplicationHandler(ReplicationClientHandler): super(ASReplicationHandler, self).__init__(hs.get_datastore()) self.appservice_handler = hs.get_application_service_handler() - @defer.inlineCallbacks - def on_rdata(self, stream_name, token, rows): - yield super(ASReplicationHandler, self).on_rdata(stream_name, token, rows) + async def on_rdata(self, stream_name, token, rows): + await super(ASReplicationHandler, self).on_rdata(stream_name, token, rows) if stream_name == "events": max_stream_id = self.store.get_room_max_stream_ordering() @@ -143,8 +141,6 @@ def start(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - if config.notify_appservices: sys.stderr.write( "\nThe appservices must be disabled in the main synapse process" @@ -159,10 +155,8 @@ def start(config_options): ps = AppserviceServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ps, config, use_worker_options=True) diff --git a/synapse/app/client_reader.py b/synapse/app/client_reader.py index dadb487d5f..3edfe19567 100644 --- a/synapse/app/client_reader.py +++ b/synapse/app/client_reader.py @@ -62,7 +62,6 @@ from synapse.rest.client.v2_alpha.keys import KeyChangesServlet, KeyQueryServlet from synapse.rest.client.v2_alpha.register import RegisterRestServlet from synapse.rest.client.versions import VersionsRestServlet from synapse.server import HomeServer -from synapse.storage.engines import create_engine from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.versionstring import get_version_string @@ -181,14 +180,10 @@ def start(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - ss = ClientReaderServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ss, config, use_worker_options=True) diff --git a/synapse/app/event_creator.py b/synapse/app/event_creator.py index d110599a35..d0ddbe38fc 100644 --- a/synapse/app/event_creator.py +++ b/synapse/app/event_creator.py @@ -57,7 +57,6 @@ from synapse.rest.client.v1.room import ( ) from synapse.server import HomeServer from synapse.storage.data_stores.main.user_directory import UserDirectoryStore -from synapse.storage.engines import create_engine from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.versionstring import get_version_string @@ -180,14 +179,10 @@ def start(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - ss = EventCreatorServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ss, config, use_worker_options=True) diff --git a/synapse/app/federation_reader.py b/synapse/app/federation_reader.py index 418c086254..311523e0ed 100644 --- a/synapse/app/federation_reader.py +++ b/synapse/app/federation_reader.py @@ -46,7 +46,6 @@ from synapse.replication.slave.storage.transactions import SlavedTransactionStor from synapse.replication.tcp.client import ReplicationClientHandler from synapse.rest.key.v2 import KeyApiV2Resource from synapse.server import HomeServer -from synapse.storage.engines import create_engine from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.versionstring import get_version_string @@ -162,14 +161,10 @@ def start(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - ss = FederationReaderServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ss, config, use_worker_options=True) diff --git a/synapse/app/federation_sender.py b/synapse/app/federation_sender.py index 139221ad34..a57cf991ac 100644 --- a/synapse/app/federation_sender.py +++ b/synapse/app/federation_sender.py @@ -40,7 +40,7 @@ from synapse.replication.slave.storage.transactions import SlavedTransactionStor from synapse.replication.tcp.client import ReplicationClientHandler from synapse.replication.tcp.streams._base import ReceiptsStream from synapse.server import HomeServer -from synapse.storage.engines import create_engine +from synapse.storage.database import Database from synapse.types import ReadReceipt from synapse.util.async_helpers import Linearizer from synapse.util.httpresourcetree import create_resource_tree @@ -59,8 +59,8 @@ class FederationSenderSlaveStore( SlavedDeviceStore, SlavedPresenceStore, ): - def __init__(self, db_conn, hs): - super(FederationSenderSlaveStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(FederationSenderSlaveStore, self).__init__(database, db_conn, hs) # We pull out the current federation stream position now so that we # always have a known value for the federation position in memory so @@ -69,7 +69,7 @@ class FederationSenderSlaveStore( self.federation_out_pos_startup = self._get_federation_out_pos(db_conn) def _get_federation_out_pos(self, db_conn): - sql = "SELECT stream_id FROM federation_stream_position" " WHERE type = ?" + sql = "SELECT stream_id FROM federation_stream_position WHERE type = ?" sql = self.database_engine.convert_param_style(sql) txn = db_conn.cursor() @@ -145,9 +145,8 @@ class FederationSenderReplicationHandler(ReplicationClientHandler): super(FederationSenderReplicationHandler, self).__init__(hs.get_datastore()) self.send_handler = FederationSenderHandler(hs, self) - @defer.inlineCallbacks - def on_rdata(self, stream_name, token, rows): - yield super(FederationSenderReplicationHandler, self).on_rdata( + async def on_rdata(self, stream_name, token, rows): + await super(FederationSenderReplicationHandler, self).on_rdata( stream_name, token, rows ) self.send_handler.process_replication_rows(stream_name, token, rows) @@ -173,8 +172,6 @@ def start(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - if config.send_federation: sys.stderr.write( "\nThe send_federation must be disabled in the main synapse process" @@ -189,10 +186,8 @@ def start(config_options): ss = FederationSenderServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ss, config, use_worker_options=True) diff --git a/synapse/app/frontend_proxy.py b/synapse/app/frontend_proxy.py index e647459d0e..30e435eead 100644 --- a/synapse/app/frontend_proxy.py +++ b/synapse/app/frontend_proxy.py @@ -39,7 +39,6 @@ from synapse.replication.slave.storage.registration import SlavedRegistrationSto from synapse.replication.tcp.client import ReplicationClientHandler from synapse.rest.client.v2_alpha._base import client_patterns from synapse.server import HomeServer -from synapse.storage.engines import create_engine from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.versionstring import get_version_string @@ -234,14 +233,10 @@ def start(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - ss = FrontendProxyServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ss, config, use_worker_options=True) diff --git a/synapse/app/homeserver.py b/synapse/app/homeserver.py index 98e528187a..d86d924bea 100644 --- a/synapse/app/homeserver.py +++ b/synapse/app/homeserver.py @@ -31,7 +31,7 @@ from prometheus_client import Gauge from twisted.application import service from twisted.internet import defer, reactor from twisted.python.failure import Failure -from twisted.web.resource import EncodingResourceWrapper, NoResource +from twisted.web.resource import EncodingResourceWrapper, IResource, NoResource from twisted.web.server import GzipEncoderFactory from twisted.web.static import File @@ -39,7 +39,6 @@ import synapse import synapse.config.logger from synapse import events from synapse.api.urls import ( - CONTENT_REPO_PREFIX, FEDERATION_PREFIX, LEGACY_MEDIA_PREFIX, MEDIA_PREFIX, @@ -65,12 +64,18 @@ from synapse.replication.tcp.resource import ReplicationStreamProtocolFactory from synapse.rest import ClientRestResource from synapse.rest.admin import AdminRestResource from synapse.rest.key.v2 import KeyApiV2Resource -from synapse.rest.media.v0.content_repository import ContentRepoResource from synapse.rest.well_known import WellKnownResource from synapse.server import HomeServer +<<<<<<< HEAD from synapse.storage import DataStore, are_all_users_on_domain from synapse.storage.engines import IncorrectDatabaseSetup, create_engine from synapse.storage.prepare_database import UpgradeDatabaseException, prepare_database +======= +from synapse.storage import DataStore +from synapse.storage.engines import IncorrectDatabaseSetup +from synapse.storage.prepare_database import UpgradeDatabaseException +from synapse.util.caches import CACHE_SIZE_FACTOR +>>>>>>> origin/develop from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.module_loader import load_module @@ -110,7 +115,16 @@ class SynapseHomeServer(HomeServer): for path, resmodule in additional_resources.items(): handler_cls, config = load_module(resmodule) handler = handler_cls(config, module_api) - resources[path] = AdditionalResource(self, handler.handle_request) + if IResource.providedBy(handler): + resource = handler + elif hasattr(handler, "handle_request"): + resource = AdditionalResource(self, handler.handle_request) + else: + raise ConfigError( + "additional_resource %s does not implement a known interface" + % (resmodule["module"],) + ) + resources[path] = resource # try to find something useful to redirect '/' to if WEB_CLIENT_PREFIX in resources: @@ -222,13 +236,7 @@ class SynapseHomeServer(HomeServer): if self.get_config().enable_media_repo: media_repo = self.get_media_repository_resource() resources.update( - { - MEDIA_PREFIX: media_repo, - LEGACY_MEDIA_PREFIX: media_repo, - CONTENT_REPO_PREFIX: ContentRepoResource( - self, self.config.uploads_path - ), - } + {MEDIA_PREFIX: media_repo, LEGACY_MEDIA_PREFIX: media_repo} ) elif name == "media": raise ConfigError( @@ -293,22 +301,6 @@ class SynapseHomeServer(HomeServer): else: logger.warning("Unrecognized listener type: %s", listener["type"]) - def run_startup_checks(self, db_conn, database_engine): - all_users_native = are_all_users_on_domain( - db_conn.cursor(), database_engine, self.hostname - ) - if not all_users_native: - quit_with_error( - "Found users in database not native to %s!\n" - "You cannot changed a synapse server_name after it's been configured" - % (self.hostname,) - ) - - try: - database_engine.check_database(db_conn.cursor()) - except IncorrectDatabaseSetup as e: - quit_with_error(str(e)) - # Gauges to expose monthly active user control metrics current_mau_gauge = Gauge("synapse_admin_mau:current", "Current MAU") @@ -333,7 +325,7 @@ def setup(config_options): "Synapse Homeserver", config_options ) except ConfigError as e: - sys.stderr.write("\n" + str(e) + "\n") + sys.stderr.write("\nERROR: %s\n" % (e,)) sys.exit(1) if not config: @@ -343,40 +335,23 @@ def setup(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - config.database_config["args"]["cp_openfun"] = database_engine.on_new_connection - hs = SynapseHomeServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) synapse.config.logger.setup_logging(hs, config, use_worker_options=False) - logger.info("Preparing database: %s...", config.database_config["name"]) + logger.info("Setting up server") try: - with hs.get_db_conn(run_new_connection=False) as db_conn: - prepare_database(db_conn, database_engine, config=config) - database_engine.on_new_connection(db_conn) - - hs.run_startup_checks(db_conn, database_engine) - - db_conn.commit() - except UpgradeDatabaseException: - sys.stderr.write( - "\nFailed to upgrade database.\n" - "Have you checked for version specific instructions in" - " UPGRADES.rst?\n" - ) - sys.exit(1) - - logger.info("Database prepared in %s.", config.database_config["name"]) + hs.setup() + except IncorrectDatabaseSetup as e: + quit_with_error(str(e)) + except UpgradeDatabaseException as e: + quit_with_error("Failed to upgrade database: %s" % (e,)) - hs.setup() hs.setup_master() @defer.inlineCallbacks @@ -435,7 +410,7 @@ def setup(config_options): _base.start(hs, config.listeners) hs.get_pusherpool().start() - hs.get_datastore().start_doing_background_updates() + hs.get_datastore().db.updates.start_doing_background_updates() except Exception: # Print the exception and bail out. print("Error during startup:", file=sys.stderr) @@ -541,8 +516,10 @@ def phone_stats_home(hs, stats, stats_process=_stats_process): # Database version # - stats["database_engine"] = hs.get_datastore().database_engine_name - stats["database_server_version"] = hs.get_datastore().get_server_version() + # This only reports info about the *main* database. + stats["database_engine"] = hs.get_datastore().db.engine.module.__name__ + stats["database_server_version"] = hs.get_datastore().db.engine.server_version + logger.info("Reporting stats to %s: %s" % (hs.config.report_stats_endpoint, stats)) try: yield hs.get_proxied_http_client().put_json( @@ -584,7 +561,7 @@ def run(hs): def performance_stats_init(): _stats_process.clear() _stats_process.append( - (int(hs.get_clock().time(), resource.getrusage(resource.RUSAGE_SELF))) + (int(hs.get_clock().time()), resource.getrusage(resource.RUSAGE_SELF)) ) def start_phone_stats_home(): diff --git a/synapse/app/media_repository.py b/synapse/app/media_repository.py index 2c6dd3ef02..5b5832214a 100644 --- a/synapse/app/media_repository.py +++ b/synapse/app/media_repository.py @@ -21,7 +21,7 @@ from twisted.web.resource import NoResource import synapse from synapse import events -from synapse.api.urls import CONTENT_REPO_PREFIX, LEGACY_MEDIA_PREFIX, MEDIA_PREFIX +from synapse.api.urls import LEGACY_MEDIA_PREFIX, MEDIA_PREFIX from synapse.app import _base from synapse.config._base import ConfigError from synapse.config.homeserver import HomeServerConfig @@ -34,13 +34,12 @@ from synapse.replication.slave.storage._base import BaseSlavedStore from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore from synapse.replication.slave.storage.client_ips import SlavedClientIpStore from synapse.replication.slave.storage.registration import SlavedRegistrationStore +from synapse.replication.slave.storage.room import RoomStore from synapse.replication.slave.storage.transactions import SlavedTransactionStore from synapse.replication.tcp.client import ReplicationClientHandler from synapse.rest.admin import register_servlets_for_media_repo -from synapse.rest.media.v0.content_repository import ContentRepoResource from synapse.server import HomeServer from synapse.storage.data_stores.main.media_repository import MediaRepositoryStore -from synapse.storage.engines import create_engine from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.versionstring import get_version_string @@ -49,6 +48,7 @@ logger = logging.getLogger("synapse.app.media_repository") class MediaRepositorySlavedStore( + RoomStore, SlavedApplicationServiceStore, SlavedRegistrationStore, SlavedClientIpStore, @@ -83,9 +83,6 @@ class MediaRepositoryServer(HomeServer): { MEDIA_PREFIX: media_repo, LEGACY_MEDIA_PREFIX: media_repo, - CONTENT_REPO_PREFIX: ContentRepoResource( - self, self.config.uploads_path - ), "/_synapse/admin": admin_resource, } ) @@ -157,14 +154,10 @@ def start(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - ss = MediaRepositoryServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ss, config, use_worker_options=True) diff --git a/synapse/app/pusher.py b/synapse/app/pusher.py index 01a5ffc363..e46b6ac598 100644 --- a/synapse/app/pusher.py +++ b/synapse/app/pusher.py @@ -33,10 +33,10 @@ from synapse.replication.slave.storage.account_data import SlavedAccountDataStor from synapse.replication.slave.storage.events import SlavedEventStore from synapse.replication.slave.storage.pushers import SlavedPusherStore from synapse.replication.slave.storage.receipts import SlavedReceiptsStore +from synapse.replication.slave.storage.room import RoomStore from synapse.replication.tcp.client import ReplicationClientHandler from synapse.server import HomeServer from synapse.storage import DataStore -from synapse.storage.engines import create_engine from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.versionstring import get_version_string @@ -45,7 +45,11 @@ logger = logging.getLogger("synapse.app.pusher") class PusherSlaveStore( - SlavedEventStore, SlavedPusherStore, SlavedReceiptsStore, SlavedAccountDataStore + SlavedEventStore, + SlavedPusherStore, + SlavedReceiptsStore, + SlavedAccountDataStore, + RoomStore, ): update_pusher_last_stream_ordering_and_success = __func__( DataStore.update_pusher_last_stream_ordering_and_success @@ -137,9 +141,8 @@ class PusherReplicationHandler(ReplicationClientHandler): self.pusher_pool = hs.get_pusherpool() - @defer.inlineCallbacks - def on_rdata(self, stream_name, token, rows): - yield super(PusherReplicationHandler, self).on_rdata(stream_name, token, rows) + async def on_rdata(self, stream_name, token, rows): + await super(PusherReplicationHandler, self).on_rdata(stream_name, token, rows) run_in_background(self.poke_pushers, stream_name, token, rows) @defer.inlineCallbacks @@ -198,14 +201,10 @@ def start(config_options): # Force the pushers to start since they will be disabled in the main config config.start_pushers = True - database_engine = create_engine(config.database_config) - ps = PusherServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ps, config, use_worker_options=True) diff --git a/synapse/app/synchrotron.py b/synapse/app/synchrotron.py index b14da09f47..3218da07bd 100644 --- a/synapse/app/synchrotron.py +++ b/synapse/app/synchrotron.py @@ -48,14 +48,13 @@ from synapse.replication.slave.storage.receipts import SlavedReceiptsStore from synapse.replication.slave.storage.registration import SlavedRegistrationStore from synapse.replication.slave.storage.room import RoomStore from synapse.replication.tcp.client import ReplicationClientHandler -from synapse.replication.tcp.streams.events import EventsStreamEventRow +from synapse.replication.tcp.streams.events import EventsStreamEventRow, EventsStreamRow from synapse.rest.client.v1 import events from synapse.rest.client.v1.initial_sync import InitialSyncRestServlet from synapse.rest.client.v1.room import RoomInitialSyncRestServlet from synapse.rest.client.v2_alpha import sync from synapse.server import HomeServer from synapse.storage.data_stores.main.presence import UserPresenceState -from synapse.storage.engines import create_engine from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole from synapse.util.stringutils import random_string @@ -151,7 +150,7 @@ class SynchrotronPresence(object): def set_state(self, user, state, ignore_status_msg=False): # TODO Hows this supposed to work? - pass + return defer.succeed(None) get_states = __func__(PresenceHandler.get_states) get_state = __func__(PresenceHandler.get_state) @@ -359,9 +358,8 @@ class SyncReplicationHandler(ReplicationClientHandler): self.presence_handler = hs.get_presence_handler() self.notifier = hs.get_notifier() - @defer.inlineCallbacks - def on_rdata(self, stream_name, token, rows): - yield super(SyncReplicationHandler, self).on_rdata(stream_name, token, rows) + async def on_rdata(self, stream_name, token, rows): + await super(SyncReplicationHandler, self).on_rdata(stream_name, token, rows) run_in_background(self.process_and_notify, stream_name, token, rows) def get_streams_to_replicate(self): @@ -372,8 +370,7 @@ class SyncReplicationHandler(ReplicationClientHandler): def get_currently_syncing_users(self): return self.presence_handler.get_currently_syncing_users() - @defer.inlineCallbacks - def process_and_notify(self, stream_name, token, rows): + async def process_and_notify(self, stream_name, token, rows): try: if stream_name == "events": # We shouldn't get multiple rows per token for events stream, so @@ -381,7 +378,14 @@ class SyncReplicationHandler(ReplicationClientHandler): for row in rows: if row.type != EventsStreamEventRow.TypeId: continue - event = yield self.store.get_event(row.data.event_id) + assert isinstance(row, EventsStreamRow) + + event = await self.store.get_event( + row.data.event_id, allow_rejected=True + ) + if event.rejected_reason: + continue + extra_users = () if event.type == EventTypes.Member: extra_users = (event.state_key,) @@ -413,11 +417,11 @@ class SyncReplicationHandler(ReplicationClientHandler): elif stream_name == "device_lists": all_room_ids = set() for row in rows: - room_ids = yield self.store.get_rooms_for_user(row.user_id) + room_ids = await self.store.get_rooms_for_user(row.user_id) all_room_ids.update(room_ids) self.notifier.on_new_event("device_list_key", token, rooms=all_room_ids) elif stream_name == "presence": - yield self.presence_handler.process_replication_rows(token, rows) + await self.presence_handler.process_replication_rows(token, rows) elif stream_name == "receipts": self.notifier.on_new_event( "groups_key", token, users=[row.user_id for row in rows] @@ -437,14 +441,10 @@ def start(config_options): synapse.events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - ss = SynchrotronServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, application_service_handler=SynchrotronApplicationService(), ) diff --git a/synapse/app/user_dir.py b/synapse/app/user_dir.py index 6cb100319f..ba536d6f04 100644 --- a/synapse/app/user_dir.py +++ b/synapse/app/user_dir.py @@ -43,7 +43,7 @@ from synapse.replication.tcp.streams.events import ( from synapse.rest.client.v2_alpha import user_directory from synapse.server import HomeServer from synapse.storage.data_stores.main.user_directory import UserDirectoryStore -from synapse.storage.engines import create_engine +from synapse.storage.database import Database from synapse.util.caches.stream_change_cache import StreamChangeCache from synapse.util.httpresourcetree import create_resource_tree from synapse.util.manhole import manhole @@ -60,11 +60,11 @@ class UserDirectorySlaveStore( UserDirectoryStore, BaseSlavedStore, ): - def __init__(self, db_conn, hs): - super(UserDirectorySlaveStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(UserDirectorySlaveStore, self).__init__(database, db_conn, hs) events_max = self._stream_id_gen.get_current_token() - curr_state_delta_prefill, min_curr_state_delta_id = self._get_cache_dict( + curr_state_delta_prefill, min_curr_state_delta_id = self.db.get_cache_dict( db_conn, "current_state_delta_stream", entity_column="room_id", @@ -172,9 +172,8 @@ class UserDirectoryReplicationHandler(ReplicationClientHandler): super(UserDirectoryReplicationHandler, self).__init__(hs.get_datastore()) self.user_directory = hs.get_user_directory_handler() - @defer.inlineCallbacks - def on_rdata(self, stream_name, token, rows): - yield super(UserDirectoryReplicationHandler, self).on_rdata( + async def on_rdata(self, stream_name, token, rows): + await super(UserDirectoryReplicationHandler, self).on_rdata( stream_name, token, rows ) if stream_name == EventsStream.NAME: @@ -199,8 +198,6 @@ def start(config_options): events.USE_FROZEN_DICTS = config.use_frozen_dicts - database_engine = create_engine(config.database_config) - if config.update_user_directory: sys.stderr.write( "\nThe update_user_directory must be disabled in the main synapse process" @@ -215,10 +212,8 @@ def start(config_options): ss = UserDirectoryServer( config.server_name, - db_config=config.database_config, config=config, version_string="Synapse/" + get_version_string(synapse), - database_engine=database_engine, ) setup_logging(ss, config, use_worker_options=True) diff --git a/synapse/appservice/api.py b/synapse/appservice/api.py index 3e25bf5747..57174da021 100644 --- a/synapse/appservice/api.py +++ b/synapse/appservice/api.py @@ -185,7 +185,7 @@ class ApplicationServiceApi(SimpleHttpClient): if not _is_valid_3pe_metadata(info): logger.warning( - "query_3pe_protocol to %s did not return a" " valid result", uri + "query_3pe_protocol to %s did not return a valid result", uri ) return None diff --git a/synapse/config/appservice.py b/synapse/config/appservice.py index e77d3387ff..ca43e96bd1 100644 --- a/synapse/config/appservice.py +++ b/synapse/config/appservice.py @@ -134,7 +134,7 @@ def _load_appservice(hostname, as_info, config_filename): for regex_obj in as_info["namespaces"][ns]: if not isinstance(regex_obj, dict): raise ValueError( - "Expected namespace entry in %s to be an object," " but got %s", + "Expected namespace entry in %s to be an object, but got %s", ns, regex_obj, ) diff --git a/synapse/config/database.py b/synapse/config/database.py index 0e2509f0b1..219b32f670 100644 --- a/synapse/config/database.py +++ b/synapse/config/database.py @@ -12,12 +12,45 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +import logging import os from textwrap import indent import yaml -from ._base import Config +from synapse.config._base import Config, ConfigError + +logger = logging.getLogger(__name__) + + +class DatabaseConnectionConfig: + """Contains the connection config for a particular database. + + Args: + name: A label for the database, used for logging. + db_config: The config for a particular database, as per `database` + section of main config. Has three fields: `name` for database + module name, `args` for the args to give to the database + connector, and optional `data_stores` that is a list of stores to + provision on this database (defaulting to all). + """ + + def __init__(self, name: str, db_config: dict): + if db_config["name"] not in ("sqlite3", "psycopg2"): + raise ConfigError("Unsupported database type %r" % (db_config["name"],)) + + if db_config["name"] == "sqlite3": + db_config.setdefault("args", {}).update( + {"cp_min": 1, "cp_max": 1, "check_same_thread": False} + ) + + data_stores = db_config.get("data_stores") + if data_stores is None: + data_stores = ["main", "state"] + + self.name = name + self.config = db_config + self.data_stores = data_stores class DatabaseConfig(Config): @@ -26,22 +59,43 @@ class DatabaseConfig(Config): def read_config(self, config, **kwargs): self.event_cache_size = self.parse_size(config.get("event_cache_size", "10K")) - self.database_config = config.get("database") + # We *experimentally* support specifying multiple databases via the + # `databases` key. This is a map from a label to database config in the + # same format as the `database` config option, plus an extra + # `data_stores` key to specify which data store goes where. For example: + # + # databases: + # master: + # name: psycopg2 + # data_stores: ["main"] + # args: {} + # state: + # name: psycopg2 + # data_stores: ["state"] + # args: {} - if self.database_config is None: - self.database_config = {"name": "sqlite3", "args": {}} + multi_database_config = config.get("databases") + database_config = config.get("database") + + if multi_database_config and database_config: + raise ConfigError("Can't specify both 'database' and 'datbases' in config") + + if multi_database_config: + if config.get("database_path"): + raise ConfigError("Can't specify 'database_path' with 'databases'") + + self.databases = [ + DatabaseConnectionConfig(name, db_conf) + for name, db_conf in multi_database_config.items() + ] - name = self.database_config.get("name", None) - if name == "psycopg2": - pass - elif name == "sqlite3": - self.database_config.setdefault("args", {}).update( - {"cp_min": 1, "cp_max": 1, "check_same_thread": False} - ) else: - raise RuntimeError("Unsupported database type '%s'" % (name,)) + if database_config is None: + database_config = {"name": "sqlite3", "args": {}} + + self.databases = [DatabaseConnectionConfig("master", database_config)] - self.set_databasepath(config.get("database_path")) + self.set_databasepath(config.get("database_path")) def generate_config_section(self, data_dir_path, database_conf, **kwargs): if not database_conf: @@ -76,11 +130,24 @@ class DatabaseConfig(Config): self.set_databasepath(args.database_path) def set_databasepath(self, database_path): + if database_path is None: + return + if database_path != ":memory:": database_path = self.abspath(database_path) - if self.database_config.get("name", None) == "sqlite3": - if database_path is not None: - self.database_config["args"]["database"] = database_path + + # We only support setting a database path if we have a single sqlite3 + # database. + if len(self.databases) != 1: + raise ConfigError("Cannot specify 'database_path' with multiple databases") + + database = self.get_single_database() + if database.config["name"] != "sqlite3": + # We don't raise here as we haven't done so before for this case. + logger.warn("Ignoring 'database_path' for non-sqlite3 database") + return + + database.config["args"]["database"] = database_path @staticmethod def add_arguments(parser): @@ -91,3 +158,11 @@ class DatabaseConfig(Config): metavar="SQLITE_DATABASE_PATH", help="The path to a sqlite database to use.", ) + + def get_single_database(self) -> DatabaseConnectionConfig: + """Returns the database if there is only one, useful for e.g. tests + """ + if len(self.databases) != 1: + raise Exception("More than one database exists") + + return self.databases[0] diff --git a/synapse/config/emailconfig.py b/synapse/config/emailconfig.py index 43fad0bf8b..35756bed87 100644 --- a/synapse/config/emailconfig.py +++ b/synapse/config/emailconfig.py @@ -21,6 +21,7 @@ from __future__ import print_function import email.utils import os from enum import Enum +from typing import Optional import pkg_resources @@ -101,7 +102,7 @@ class EmailConfig(Config): # both in RegistrationConfig and here. We should factor this bit out self.account_threepid_delegate_email = self.trusted_third_party_id_servers[ 0 - ] + ] # type: Optional[str] self.using_identity_server_from_trusted_list = True else: raise ConfigError( @@ -146,6 +147,8 @@ class EmailConfig(Config): if k not in email_config: missing.append("email." + k) + # public_baseurl is required to build password reset and validation links that + # will be emailed to users if config.get("public_baseurl") is None: missing.append("public_baseurl") @@ -305,8 +308,23 @@ class EmailConfig(Config): # smtp_user: "exampleusername" # smtp_pass: "examplepassword" # require_transport_security: false + # + # # notif_from defines the "From" address to use when sending emails. + # # It must be set if email sending is enabled. + # # + # # The placeholder '%(app)s' will be replaced by the application name, + # # which is normally 'app_name' (below), but may be overridden by the + # # Matrix client application. + # # + # # Note that the placeholder must be written '%(app)s', including the + # # trailing 's'. + # # # notif_from: "Your Friendly %(app)s homeserver <noreply@example.com>" - # app_name: Matrix + # + # # app_name defines the default value for '%(app)s' in notif_from. It + # # defaults to 'Matrix'. + # # + # #app_name: my_branded_matrix_server # # # Enable email notifications by default # # diff --git a/synapse/config/key.py b/synapse/config/key.py index 52ff1b2621..066e7838c3 100644 --- a/synapse/config/key.py +++ b/synapse/config/key.py @@ -108,7 +108,7 @@ class KeyConfig(Config): self.signing_key = self.read_signing_keys(signing_key_path, "signing_key") self.old_signing_keys = self.read_old_signing_keys( - config.get("old_signing_keys", {}) + config.get("old_signing_keys") ) self.key_refresh_interval = self.parse_duration( config.get("key_refresh_interval", "1d") @@ -199,14 +199,19 @@ class KeyConfig(Config): signing_key_path: "%(base_key_name)s.signing.key" # The keys that the server used to sign messages with but won't use - # to sign new messages. E.g. it has lost its private key + # to sign new messages. # - #old_signing_keys: - # "ed25519:auto": - # # Base64 encoded public key - # key: "The public part of your old signing key." - # # Millisecond POSIX timestamp when the key expired. - # expired_ts: 123456789123 + old_signing_keys: + # For each key, `key` should be the base64-encoded public key, and + # `expired_ts`should be the time (in milliseconds since the unix epoch) that + # it was last used. + # + # It is possible to build an entry from an old signing.key file using the + # `export_signing_key` script which is provided with synapse. + # + # For example: + # + #"ed25519:id": { key: "base64string", expired_ts: 123456789123 } # How long key response published by this server is valid for. # Used to set the valid_until_ts in /key/v2 APIs. @@ -290,6 +295,8 @@ class KeyConfig(Config): raise ConfigError("Error reading %s: %s" % (name, str(e))) def read_old_signing_keys(self, old_signing_keys): + if old_signing_keys is None: + return {} keys = {} for key_id, key_data in old_signing_keys.items(): if is_signing_algorithm_supported(key_id): diff --git a/synapse/config/logger.py b/synapse/config/logger.py index 75bb904718..a25c70e928 100644 --- a/synapse/config/logger.py +++ b/synapse/config/logger.py @@ -12,7 +12,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. - +import argparse import logging import logging.config import os @@ -37,10 +37,17 @@ from synapse.logging._structured import ( from synapse.logging.context import LoggingContextFilter from synapse.util.versionstring import get_version_string -from ._base import Config +from ._base import Config, ConfigError DEFAULT_LOG_CONFIG = Template( - """ + """\ +# Log configuration for Synapse. +# +# This is a YAML file containing a standard Python logging configuration +# dictionary. See [1] for details on the valid settings. +# +# [1]: https://docs.python.org/3.7/library/logging.config.html#configuration-dictionary-schema + version: 1 formatters: @@ -81,11 +88,18 @@ disable_existing_loggers: false """ ) +LOG_FILE_ERROR = """\ +Support for the log_file configuration option and --log-file command-line option was +removed in Synapse 1.3.0. You should instead set up a separate log configuration file. +""" + class LoggingConfig(Config): section = "logging" def read_config(self, config, **kwargs): + if config.get("log_file"): + raise ConfigError(LOG_FILE_ERROR) self.log_config = self.abspath(config.get("log_config")) self.no_redirect_stdio = config.get("no_redirect_stdio", False) @@ -106,6 +120,8 @@ class LoggingConfig(Config): def read_arguments(self, args): if args.no_redirect_stdio is not None: self.no_redirect_stdio = args.no_redirect_stdio + if args.log_file is not None: + raise ConfigError(LOG_FILE_ERROR) @staticmethod def add_arguments(parser): @@ -118,6 +134,10 @@ class LoggingConfig(Config): help="Do not redirect stdout/stderr to the log", ) + logging_group.add_argument( + "-f", "--log-file", dest="log_file", help=argparse.SUPPRESS, + ) + def generate_files(self, config, config_dir_path): log_config = config.get("log_config") if log_config and not os.path.exists(log_config): diff --git a/synapse/config/ratelimiting.py b/synapse/config/ratelimiting.py index 947f653e03..4a3bfc4354 100644 --- a/synapse/config/ratelimiting.py +++ b/synapse/config/ratelimiting.py @@ -83,10 +83,9 @@ class RatelimitConfig(Config): ) rc_admin_redaction = config.get("rc_admin_redaction") + self.rc_admin_redaction = None if rc_admin_redaction: self.rc_admin_redaction = RateLimitConfig(rc_admin_redaction) - else: - self.rc_admin_redaction = None def generate_config_section(self, **kwargs): return """\ diff --git a/synapse/config/registration.py b/synapse/config/registration.py index 1f6dac69da..ee9614c5f7 100644 --- a/synapse/config/registration.py +++ b/synapse/config/registration.py @@ -106,6 +106,13 @@ class RegistrationConfig(Config): account_threepid_delegates = config.get("account_threepid_delegates") or {} self.account_threepid_delegate_email = account_threepid_delegates.get("email") self.account_threepid_delegate_msisdn = account_threepid_delegates.get("msisdn") + if self.account_threepid_delegate_msisdn and not self.public_baseurl: + raise ConfigError( + "The configuration option `public_baseurl` is required if " + "`account_threepid_delegate.msisdn` is set, such that " + "clients know where to submit validation tokens to. Please " + "configure `public_baseurl`." + ) self.default_identity_server = config.get("default_identity_server") self.allow_guest_access = config.get("allow_guest_access", False) diff --git a/synapse/config/repository.py b/synapse/config/repository.py index d0205e14b9..7d2dd27fd0 100644 --- a/synapse/config/repository.py +++ b/synapse/config/repository.py @@ -156,7 +156,6 @@ class ContentRepositoryConfig(Config): (provider_class, parsed_config, wrapper_config) ) - self.uploads_path = self.ensure_directory(config.get("uploads_path", "uploads")) self.dynamic_thumbnails = config.get("dynamic_thumbnails", False) self.thumbnail_requirements = parse_thumbnail_requirements( config.get("thumbnail_sizes", DEFAULT_THUMBNAIL_SIZES) @@ -231,10 +230,6 @@ class ContentRepositoryConfig(Config): # config: # directory: /mnt/some/other/directory - # Directory where in-progress uploads are stored. - # - uploads_path: "%(uploads_path)s" - # The largest allowed upload size in bytes # #max_upload_size: 10M diff --git a/synapse/config/room_directory.py b/synapse/config/room_directory.py index 7c9f05bde4..7ac7699676 100644 --- a/synapse/config/room_directory.py +++ b/synapse/config/room_directory.py @@ -170,7 +170,7 @@ class _RoomDirectoryRule(object): self.action = action else: raise ConfigError( - "%s rules can only have action of 'allow'" " or 'deny'" % (option_name,) + "%s rules can only have action of 'allow' or 'deny'" % (option_name,) ) self._alias_matches_all = alias == "*" diff --git a/synapse/config/saml2_config.py b/synapse/config/saml2_config.py index c5ea2d43a1..b91414aa35 100644 --- a/synapse/config/saml2_config.py +++ b/synapse/config/saml2_config.py @@ -14,17 +14,19 @@ # See the License for the specific language governing permissions and # limitations under the License. -import re +import logging from synapse.python_dependencies import DependencyException, check_requirements -from synapse.types import ( - map_username_to_mxid_localpart, - mxid_localpart_allowed_characters, -) -from synapse.util.module_loader import load_python_module +from synapse.util.module_loader import load_module, load_python_module from ._base import Config, ConfigError +logger = logging.getLogger(__name__) + +DEFAULT_USER_MAPPING_PROVIDER = ( + "synapse.handlers.saml_handler.DefaultSamlMappingProvider" +) + def _dict_merge(merge_dict, into_dict): """Do a deep merge of two dicts @@ -75,15 +77,69 @@ class SAML2Config(Config): self.saml2_enabled = True - self.saml2_mxid_source_attribute = saml2_config.get( - "mxid_source_attribute", "uid" - ) - self.saml2_grandfathered_mxid_source_attribute = saml2_config.get( "grandfathered_mxid_source_attribute", "uid" ) - saml2_config_dict = self._default_saml_config_dict() + # user_mapping_provider may be None if the key is present but has no value + ump_dict = saml2_config.get("user_mapping_provider") or {} + + # Use the default user mapping provider if not set + ump_dict.setdefault("module", DEFAULT_USER_MAPPING_PROVIDER) + + # Ensure a config is present + ump_dict["config"] = ump_dict.get("config") or {} + + if ump_dict["module"] == DEFAULT_USER_MAPPING_PROVIDER: + # Load deprecated options for use by the default module + old_mxid_source_attribute = saml2_config.get("mxid_source_attribute") + if old_mxid_source_attribute: + logger.warning( + "The config option saml2_config.mxid_source_attribute is deprecated. " + "Please use saml2_config.user_mapping_provider.config" + ".mxid_source_attribute instead." + ) + ump_dict["config"]["mxid_source_attribute"] = old_mxid_source_attribute + + old_mxid_mapping = saml2_config.get("mxid_mapping") + if old_mxid_mapping: + logger.warning( + "The config option saml2_config.mxid_mapping is deprecated. Please " + "use saml2_config.user_mapping_provider.config.mxid_mapping instead." + ) + ump_dict["config"]["mxid_mapping"] = old_mxid_mapping + + # Retrieve an instance of the module's class + # Pass the config dictionary to the module for processing + ( + self.saml2_user_mapping_provider_class, + self.saml2_user_mapping_provider_config, + ) = load_module(ump_dict) + + # Ensure loaded user mapping module has defined all necessary methods + # Note parse_config() is already checked during the call to load_module + required_methods = [ + "get_saml_attributes", + "saml_response_to_user_attributes", + ] + missing_methods = [ + method + for method in required_methods + if not hasattr(self.saml2_user_mapping_provider_class, method) + ] + if missing_methods: + raise ConfigError( + "Class specified by saml2_config." + "user_mapping_provider.module is missing required " + "methods: %s" % (", ".join(missing_methods),) + ) + + # Get the desired saml auth response attributes from the module + saml2_config_dict = self._default_saml_config_dict( + *self.saml2_user_mapping_provider_class.get_saml_attributes( + self.saml2_user_mapping_provider_config + ) + ) _dict_merge( merge_dict=saml2_config.get("sp_config", {}), into_dict=saml2_config_dict ) @@ -103,22 +159,27 @@ class SAML2Config(Config): saml2_config.get("saml_session_lifetime", "5m") ) - mapping = saml2_config.get("mxid_mapping", "hexencode") - try: - self.saml2_mxid_mapper = MXID_MAPPER_MAP[mapping] - except KeyError: - raise ConfigError("%s is not a known mxid_mapping" % (mapping,)) - - def _default_saml_config_dict(self): + def _default_saml_config_dict( + self, required_attributes: set, optional_attributes: set + ): + """Generate a configuration dictionary with required and optional attributes that + will be needed to process new user registration + + Args: + required_attributes: SAML auth response attributes that are + necessary to function + optional_attributes: SAML auth response attributes that can be used to add + additional information to Synapse user accounts, but are not required + + Returns: + dict: A SAML configuration dictionary + """ import saml2 public_baseurl = self.public_baseurl if public_baseurl is None: raise ConfigError("saml2_config requires a public_baseurl to be set") - required_attributes = {"uid", self.saml2_mxid_source_attribute} - - optional_attributes = {"displayName"} if self.saml2_grandfathered_mxid_source_attribute: optional_attributes.add(self.saml2_grandfathered_mxid_source_attribute) optional_attributes -= required_attributes @@ -207,33 +268,58 @@ class SAML2Config(Config): # #config_path: "%(config_dir_path)s/sp_conf.py" - # the lifetime of a SAML session. This defines how long a user has to + # The lifetime of a SAML session. This defines how long a user has to # complete the authentication process, if allow_unsolicited is unset. # The default is 5 minutes. # #saml_session_lifetime: 5m - # The SAML attribute (after mapping via the attribute maps) to use to derive - # the Matrix ID from. 'uid' by default. + # An external module can be provided here as a custom solution to + # mapping attributes returned from a saml provider onto a matrix user. # - #mxid_source_attribute: displayName - - # The mapping system to use for mapping the saml attribute onto a matrix ID. - # Options include: - # * 'hexencode' (which maps unpermitted characters to '=xx') - # * 'dotreplace' (which replaces unpermitted characters with '.'). - # The default is 'hexencode'. - # - #mxid_mapping: dotreplace - - # In previous versions of synapse, the mapping from SAML attribute to MXID was - # always calculated dynamically rather than stored in a table. For backwards- - # compatibility, we will look for user_ids matching such a pattern before - # creating a new account. + user_mapping_provider: + # The custom module's class. Uncomment to use a custom module. + # + #module: mapping_provider.SamlMappingProvider + + # Custom configuration values for the module. Below options are + # intended for the built-in provider, they should be changed if + # using a custom module. This section will be passed as a Python + # dictionary to the module's `parse_config` method. + # + config: + # The SAML attribute (after mapping via the attribute maps) to use + # to derive the Matrix ID from. 'uid' by default. + # + # Note: This used to be configured by the + # saml2_config.mxid_source_attribute option. If that is still + # defined, its value will be used instead. + # + #mxid_source_attribute: displayName + + # The mapping system to use for mapping the saml attribute onto a + # matrix ID. + # + # Options include: + # * 'hexencode' (which maps unpermitted characters to '=xx') + # * 'dotreplace' (which replaces unpermitted characters with + # '.'). + # The default is 'hexencode'. + # + # Note: This used to be configured by the + # saml2_config.mxid_mapping option. If that is still defined, its + # value will be used instead. + # + #mxid_mapping: dotreplace + + # In previous versions of synapse, the mapping from SAML attribute to + # MXID was always calculated dynamically rather than stored in a + # table. For backwards- compatibility, we will look for user_ids + # matching such a pattern before creating a new account. # # This setting controls the SAML attribute which will be used for this - # backwards-compatibility lookup. Typically it should be 'uid', but if the - # attribute maps are changed, it may be necessary to change it. + # backwards-compatibility lookup. Typically it should be 'uid', but if + # the attribute maps are changed, it may be necessary to change it. # # The default is 'uid'. # @@ -241,23 +327,3 @@ class SAML2Config(Config): """ % { "config_dir_path": config_dir_path } - - -DOT_REPLACE_PATTERN = re.compile( - ("[^%s]" % (re.escape("".join(mxid_localpart_allowed_characters)),)) -) - - -def dot_replace_for_mxid(username: str) -> str: - username = username.lower() - username = DOT_REPLACE_PATTERN.sub(".", username) - - # regular mxids aren't allowed to start with an underscore either - username = re.sub("^_", "", username) - return username - - -MXID_MAPPER_MAP = { - "hexencode": map_username_to_mxid_localpart, - "dotreplace": dot_replace_for_mxid, -} diff --git a/synapse/config/server.py b/synapse/config/server.py index 00d01c43af..9ac112233b 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -19,7 +19,7 @@ import logging import os.path import re from textwrap import indent -from typing import List +from typing import Dict, List, Optional import attr import yaml @@ -102,6 +102,12 @@ class ServerConfig(Config): "require_auth_for_profile_requests", False ) + # Whether to require sharing a room with a user to retrieve their + # profile data + self.limit_profile_requests_to_users_who_share_rooms = config.get( + "limit_profile_requests_to_users_who_share_rooms", False, + ) + if "restrict_public_rooms_to_local_users" in config and ( "allow_public_rooms_without_auth" in config or "allow_public_rooms_over_federation" in config @@ -118,15 +124,16 @@ class ServerConfig(Config): self.allow_public_rooms_without_auth = False self.allow_public_rooms_over_federation = False else: - # If set to 'False', requires authentication to access the server's public - # rooms directory through the client API. Defaults to 'True'. + # If set to 'true', removes the need for authentication to access the server's + # public rooms directory through the client API, meaning that anyone can + # query the room directory. Defaults to 'false'. self.allow_public_rooms_without_auth = config.get( - "allow_public_rooms_without_auth", True + "allow_public_rooms_without_auth", False ) - # If set to 'False', forbids any other homeserver to fetch the server's public - # rooms directory via federation. Defaults to 'True'. + # If set to 'true', allows any other homeserver to fetch the server's public + # rooms directory via federation. Defaults to 'false'. self.allow_public_rooms_over_federation = config.get( - "allow_public_rooms_over_federation", True + "allow_public_rooms_over_federation", False ) default_room_version = config.get("default_room_version", DEFAULT_ROOM_VERSION) @@ -199,7 +206,7 @@ class ServerConfig(Config): self.admin_contact = config.get("admin_contact", None) # FIXME: federation_domain_whitelist needs sytests - self.federation_domain_whitelist = None + self.federation_domain_whitelist = None # type: Optional[dict] federation_domain_whitelist = config.get("federation_domain_whitelist", None) if federation_domain_whitelist is not None: @@ -223,7 +230,7 @@ class ServerConfig(Config): self.federation_ip_range_blacklist.update(["0.0.0.0", "::"]) except Exception as e: raise ConfigError( - "Invalid range(s) provided in " "federation_ip_range_blacklist: %s" % e + "Invalid range(s) provided in federation_ip_range_blacklist: %s" % e ) if self.public_baseurl is not None: @@ -246,6 +253,124 @@ class ServerConfig(Config): # events with profile information that differ from the target's global profile. self.allow_per_room_profiles = config.get("allow_per_room_profiles", True) + retention_config = config.get("retention") + if retention_config is None: + retention_config = {} + + self.retention_enabled = retention_config.get("enabled", False) + + retention_default_policy = retention_config.get("default_policy") + + if retention_default_policy is not None: + self.retention_default_min_lifetime = retention_default_policy.get( + "min_lifetime" + ) + if self.retention_default_min_lifetime is not None: + self.retention_default_min_lifetime = self.parse_duration( + self.retention_default_min_lifetime + ) + + self.retention_default_max_lifetime = retention_default_policy.get( + "max_lifetime" + ) + if self.retention_default_max_lifetime is not None: + self.retention_default_max_lifetime = self.parse_duration( + self.retention_default_max_lifetime + ) + + if ( + self.retention_default_min_lifetime is not None + and self.retention_default_max_lifetime is not None + and ( + self.retention_default_min_lifetime + > self.retention_default_max_lifetime + ) + ): + raise ConfigError( + "The default retention policy's 'min_lifetime' can not be greater" + " than its 'max_lifetime'" + ) + else: + self.retention_default_min_lifetime = None + self.retention_default_max_lifetime = None + + self.retention_allowed_lifetime_min = retention_config.get( + "allowed_lifetime_min" + ) + if self.retention_allowed_lifetime_min is not None: + self.retention_allowed_lifetime_min = self.parse_duration( + self.retention_allowed_lifetime_min + ) + + self.retention_allowed_lifetime_max = retention_config.get( + "allowed_lifetime_max" + ) + if self.retention_allowed_lifetime_max is not None: + self.retention_allowed_lifetime_max = self.parse_duration( + self.retention_allowed_lifetime_max + ) + + if ( + self.retention_allowed_lifetime_min is not None + and self.retention_allowed_lifetime_max is not None + and self.retention_allowed_lifetime_min + > self.retention_allowed_lifetime_max + ): + raise ConfigError( + "Invalid retention policy limits: 'allowed_lifetime_min' can not be" + " greater than 'allowed_lifetime_max'" + ) + + self.retention_purge_jobs = [] # type: List[Dict[str, Optional[int]]] + for purge_job_config in retention_config.get("purge_jobs", []): + interval_config = purge_job_config.get("interval") + + if interval_config is None: + raise ConfigError( + "A retention policy's purge jobs configuration must have the" + " 'interval' key set." + ) + + interval = self.parse_duration(interval_config) + + shortest_max_lifetime = purge_job_config.get("shortest_max_lifetime") + + if shortest_max_lifetime is not None: + shortest_max_lifetime = self.parse_duration(shortest_max_lifetime) + + longest_max_lifetime = purge_job_config.get("longest_max_lifetime") + + if longest_max_lifetime is not None: + longest_max_lifetime = self.parse_duration(longest_max_lifetime) + + if ( + shortest_max_lifetime is not None + and longest_max_lifetime is not None + and shortest_max_lifetime > longest_max_lifetime + ): + raise ConfigError( + "A retention policy's purge jobs configuration's" + " 'shortest_max_lifetime' value can not be greater than its" + " 'longest_max_lifetime' value." + ) + + self.retention_purge_jobs.append( + { + "interval": interval, + "shortest_max_lifetime": shortest_max_lifetime, + "longest_max_lifetime": longest_max_lifetime, + } + ) + + if not self.retention_purge_jobs: + self.retention_purge_jobs = [ + { + "interval": self.parse_duration("1d"), + "shortest_max_lifetime": None, + "longest_max_lifetime": None, + } + ] + self.listeners = [] # type: List[dict] for listener in config.get("listeners", []): if not isinstance(listener.get("port", None), int): @@ -372,6 +497,8 @@ class ServerConfig(Config): "cleanup_extremities_with_dummy_events", True ) + self.enable_ephemeral_messages = config.get("enable_ephemeral_messages", False) + def has_tls_listener(self) -> bool: return any(l["tls"] for l in self.listeners) @@ -500,15 +627,23 @@ class ServerConfig(Config): # #require_auth_for_profile_requests: true - # If set to 'false', requires authentication to access the server's public rooms - # directory through the client API. Defaults to 'true'. + # Uncomment to require a user to share a room with another user in order + # to retrieve their profile information. Only checked on Client-Server + # requests. Profile requests from other servers should be checked by the + # requesting server. Defaults to 'false'. # - #allow_public_rooms_without_auth: false + #limit_profile_requests_to_users_who_share_rooms: true - # If set to 'false', forbids any other homeserver to fetch the server's public - # rooms directory via federation. Defaults to 'true'. + # If set to 'true', removes the need for authentication to access the server's + # public rooms directory through the client API, meaning that anyone can + # query the room directory. Defaults to 'false'. # - #allow_public_rooms_over_federation: false + #allow_public_rooms_without_auth: true + + # If set to 'true', allows any other homeserver to fetch the server's public + # rooms directory via federation. Defaults to 'false'. + # + #allow_public_rooms_over_federation: true # The default room version for newly created rooms. # @@ -761,6 +896,69 @@ class ServerConfig(Config): # Defaults to `28d`. Set to `null` to disable clearing out of old rows. # #user_ips_max_age: 14d + + # Message retention policy at the server level. + # + # Room admins and mods can define a retention period for their rooms using the + # 'm.room.retention' state event, and server admins can cap this period by setting + # the 'allowed_lifetime_min' and 'allowed_lifetime_max' config options. + # + # If this feature is enabled, Synapse will regularly look for and purge events + # which are older than the room's maximum retention period. Synapse will also + # filter events received over federation so that events that should have been + # purged are ignored and not stored again. + # + retention: + # The message retention policies feature is disabled by default. Uncomment the + # following line to enable it. + # + #enabled: true + + # Default retention policy. If set, Synapse will apply it to rooms that lack the + # 'm.room.retention' state event. Currently, the value of 'min_lifetime' doesn't + # matter much because Synapse doesn't take it into account yet. + # + #default_policy: + # min_lifetime: 1d + # max_lifetime: 1y + + # Retention policy limits. If set, a user won't be able to send a + # 'm.room.retention' event which features a 'min_lifetime' or a 'max_lifetime' + # that's not within this range. This is especially useful in closed federations, + # in which server admins can make sure every federating server applies the same + # rules. + # + #allowed_lifetime_min: 1d + #allowed_lifetime_max: 1y + + # Server admins can define the settings of the background jobs purging the + # events which lifetime has expired under the 'purge_jobs' section. + # + # If no configuration is provided, a single job will be set up to delete expired + # events in every room daily. + # + # Each job's configuration defines which range of message lifetimes the job + # takes care of. For example, if 'shortest_max_lifetime' is '2d' and + # 'longest_max_lifetime' is '3d', the job will handle purging expired events in + # rooms whose state defines a 'max_lifetime' that's both higher than 2 days, and + # lower than or equal to 3 days. Both the minimum and the maximum value of a + # range are optional, e.g. a job with no 'shortest_max_lifetime' and a + # 'longest_max_lifetime' of '3d' will handle every room with a retention policy + # which 'max_lifetime' is lower than or equal to three days. + # + # The rationale for this per-job configuration is that some rooms might have a + # retention policy with a low 'max_lifetime', where history needs to be purged + # of outdated messages on a more frequent basis than for the rest of the rooms + # (e.g. every 12h), but not want that purge to be performed by a job that's + # iterating over every room it knows, which could be heavy on the server. + # + #purge_jobs: + # - shortest_max_lifetime: 1d + # longest_max_lifetime: 3d + # interval: 12h + # - shortest_max_lifetime: 3d + # longest_max_lifetime: 1y + # interval: 1d """ % locals() ) @@ -787,14 +985,14 @@ class ServerConfig(Config): "--print-pidfile", action="store_true", default=None, - help="Print the path to the pidfile just" " before daemonizing", + help="Print the path to the pidfile just before daemonizing", ) server_group.add_argument( "--manhole", metavar="PORT", dest="manhole", type=int, - help="Turn on the twisted telnet manhole" " service on the given port.", + help="Turn on the twisted telnet manhole service on the given port.", ) diff --git a/synapse/crypto/event_signing.py b/synapse/crypto/event_signing.py index ccaa8a9920..e65bd61d97 100644 --- a/synapse/crypto/event_signing.py +++ b/synapse/crypto/event_signing.py @@ -14,7 +14,7 @@ # See the License for the specific language governing permissions and # limitations under the License. - +import collections.abc import hashlib import logging @@ -40,8 +40,11 @@ def check_event_content_hash(event, hash_algorithm=hashlib.sha256): # some malformed events lack a 'hashes'. Protect against it being missing # or a weird type by basically treating it the same as an unhashed event. hashes = event.get("hashes") - if not isinstance(hashes, dict): - raise SynapseError(400, "Malformed 'hashes'", Codes.UNAUTHORIZED) + # nb it might be a frozendict or a dict + if not isinstance(hashes, collections.abc.Mapping): + raise SynapseError( + 400, "Malformed 'hashes': %s" % (type(hashes),), Codes.UNAUTHORIZED + ) if name not in hashes: raise SynapseError( diff --git a/synapse/crypto/keyring.py b/synapse/crypto/keyring.py index 7cfad192e8..6fe5a6a26a 100644 --- a/synapse/crypto/keyring.py +++ b/synapse/crypto/keyring.py @@ -511,17 +511,18 @@ class BaseV2KeyFetcher(object): server_name = response_json["server_name"] verified = False for key_id in response_json["signatures"].get(server_name, {}): - # each of the keys used for the signature must be present in the response - # json. key = verify_keys.get(key_id) if not key: - raise KeyLookupError( - "Key response is signed by key id %s:%s but that key is not " - "present in the response" % (server_name, key_id) - ) + # the key may not be present in verify_keys if: + # * we got the key from the notary server, and: + # * the key belongs to the notary server, and: + # * the notary server is using a different key to sign notary + # responses. + continue verify_signed_json(response_json, server_name, key.verify_key) verified = True + break if not verified: raise KeyLookupError( diff --git a/synapse/event_auth.py b/synapse/event_auth.py index ec3243b27b..1033e5e121 100644 --- a/synapse/event_auth.py +++ b/synapse/event_auth.py @@ -14,6 +14,7 @@ # limitations under the License. import logging +from typing import Set, Tuple from canonicaljson import encode_canonical_json from signedjson.key import decode_verify_key_bytes @@ -42,12 +43,26 @@ def check(room_version, event, auth_events, do_sig_check=True, do_size_check=Tru Returns: if the auth checks pass. """ + assert isinstance(auth_events, dict) + if do_size_check: _check_size_limits(event) if not hasattr(event, "room_id"): raise AuthError(500, "Event has no room_id: %s" % event) + room_id = event.room_id + + # I'm not really expecting to get auth events in the wrong room, but let's + # sanity-check it + for auth_event in auth_events.values(): + if auth_event.room_id != room_id: + raise Exception( + "During auth for event %s in room %s, found event %s in the state " + "which is in room %s" + % (event.event_id, room_id, auth_event.event_id, auth_event.room_id) + ) + if do_sig_check: sender_domain = get_domain_from_id(event.sender) @@ -74,12 +89,6 @@ def check(room_version, event, auth_events, do_sig_check=True, do_size_check=Tru if not event.signatures.get(event_id_domain): raise AuthError(403, "Event not signed by sending server") - if auth_events is None: - # Oh, we don't know what the state of the room was, so we - # are trusting that this is allowed (at least for now) - logger.warning("Trusting event: %s", event.event_id) - return - if event.type == EventTypes.Create: sender_domain = get_domain_from_id(event.sender) room_id_domain = get_domain_from_id(event.room_id) @@ -625,7 +634,7 @@ def get_public_keys(invite_event): return public_keys -def auth_types_for_event(event): +def auth_types_for_event(event) -> Set[Tuple[str]]: """Given an event, return a list of (EventType, StateKey) that may be needed to auth the event. The returned list may be a superset of what would actually be required depending on the full state of the room. @@ -634,20 +643,20 @@ def auth_types_for_event(event): actually auth the event. """ if event.type == EventTypes.Create: - return [] + return set() - auth_types = [ + auth_types = { (EventTypes.PowerLevels, ""), (EventTypes.Member, event.sender), (EventTypes.Create, ""), - ] + } if event.type == EventTypes.Member: membership = event.content["membership"] if membership in [Membership.JOIN, Membership.INVITE]: - auth_types.append((EventTypes.JoinRules, "")) + auth_types.add((EventTypes.JoinRules, "")) - auth_types.append((EventTypes.Member, event.state_key)) + auth_types.add((EventTypes.Member, event.state_key)) if membership == Membership.INVITE: if "third_party_invite" in event.content: @@ -655,6 +664,6 @@ def auth_types_for_event(event): EventTypes.ThirdPartyInvite, event.content["third_party_invite"]["signed"]["token"], ) - auth_types.append(key) + auth_types.add(key) return auth_types diff --git a/synapse/events/snapshot.py b/synapse/events/snapshot.py index 64e898f40c..9ea85e93e6 100644 --- a/synapse/events/snapshot.py +++ b/synapse/events/snapshot.py @@ -12,7 +12,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -from typing import Dict, Optional, Tuple, Union +from typing import Optional, Union from six import iteritems @@ -23,6 +23,7 @@ from twisted.internet import defer from synapse.appservice import ApplicationService from synapse.logging.context import make_deferred_yieldable, run_in_background +from synapse.types import StateMap @attr.s(slots=True) @@ -106,13 +107,11 @@ class EventContext: _state_group = attr.ib(default=None, type=Optional[int]) state_group_before_event = attr.ib(default=None, type=Optional[int]) prev_group = attr.ib(default=None, type=Optional[int]) - delta_ids = attr.ib(default=None, type=Optional[Dict[Tuple[str, str], str]]) + delta_ids = attr.ib(default=None, type=Optional[StateMap[str]]) app_service = attr.ib(default=None, type=Optional[ApplicationService]) - _current_state_ids = attr.ib( - default=None, type=Optional[Dict[Tuple[str, str], str]] - ) - _prev_state_ids = attr.ib(default=None, type=Optional[Dict[Tuple[str, str], str]]) + _current_state_ids = attr.ib(default=None, type=Optional[StateMap[str]]) + _prev_state_ids = attr.ib(default=None, type=Optional[StateMap[str]]) @staticmethod def with_state( @@ -149,7 +148,7 @@ class EventContext: # the prev_state_ids, so if we're a state event we include the event # id that we replaced in the state. if event.is_state(): - prev_state_ids = yield self.get_prev_state_ids(store) + prev_state_ids = yield self.get_prev_state_ids() prev_state_id = prev_state_ids.get((event.type, event.state_key)) else: prev_state_id = None @@ -167,12 +166,13 @@ class EventContext: } @staticmethod - def deserialize(store, input): + def deserialize(storage, input): """Converts a dict that was produced by `serialize` back into a EventContext. Args: - store (DataStore): Used to convert AS ID to AS object + storage (Storage): Used to convert AS ID to AS object and fetch + state. input (dict): A dict produced by `serialize` Returns: @@ -181,6 +181,7 @@ class EventContext: context = _AsyncEventContextImpl( # We use the state_group and prev_state_id stuff to pull the # current_state_ids out of the DB and construct prev_state_ids. + storage=storage, prev_state_id=input["prev_state_id"], event_type=input["event_type"], event_state_key=input["event_state_key"], @@ -193,7 +194,7 @@ class EventContext: app_service_id = input["app_service_id"] if app_service_id: - context.app_service = store.get_app_service_by_id(app_service_id) + context.app_service = storage.main.get_app_service_by_id(app_service_id) return context @@ -216,7 +217,7 @@ class EventContext: return self._state_group @defer.inlineCallbacks - def get_current_state_ids(self, store): + def get_current_state_ids(self): """ Gets the room state map, including this event - ie, the state in ``state_group`` @@ -234,11 +235,11 @@ class EventContext: if self.rejected: raise RuntimeError("Attempt to access state_ids of rejected event") - yield self._ensure_fetched(store) + yield self._ensure_fetched() return self._current_state_ids @defer.inlineCallbacks - def get_prev_state_ids(self, store): + def get_prev_state_ids(self): """ Gets the room state map, excluding this event. @@ -250,7 +251,7 @@ class EventContext: Maps a (type, state_key) to the event ID of the state event matching this tuple. """ - yield self._ensure_fetched(store) + yield self._ensure_fetched() return self._prev_state_ids def get_cached_current_state_ids(self): @@ -270,7 +271,7 @@ class EventContext: return self._current_state_ids - def _ensure_fetched(self, store): + def _ensure_fetched(self): return defer.succeed(None) @@ -282,6 +283,8 @@ class _AsyncEventContextImpl(EventContext): Attributes: + _storage (Storage) + _fetching_state_deferred (Deferred|None): Resolves when *_state_ids have been calculated. None if we haven't started calculating yet @@ -295,28 +298,30 @@ class _AsyncEventContextImpl(EventContext): that was replaced. """ + # This needs to have a default as we're inheriting + _storage = attr.ib(default=None) _prev_state_id = attr.ib(default=None) _event_type = attr.ib(default=None) _event_state_key = attr.ib(default=None) _fetching_state_deferred = attr.ib(default=None) - def _ensure_fetched(self, store): + def _ensure_fetched(self): if not self._fetching_state_deferred: - self._fetching_state_deferred = run_in_background( - self._fill_out_state, store - ) + self._fetching_state_deferred = run_in_background(self._fill_out_state) return make_deferred_yieldable(self._fetching_state_deferred) @defer.inlineCallbacks - def _fill_out_state(self, store): + def _fill_out_state(self): """Called to populate the _current_state_ids and _prev_state_ids attributes by loading from the database. """ if self.state_group is None: return - self._current_state_ids = yield store.get_state_ids_for_group(self.state_group) + self._current_state_ids = yield self._storage.state.get_state_ids_for_group( + self.state_group + ) if self._prev_state_id and self._event_state_key is not None: self._prev_state_ids = dict(self._current_state_ids) diff --git a/synapse/events/third_party_rules.py b/synapse/events/third_party_rules.py index 714a9b1579..86f7e5f8aa 100644 --- a/synapse/events/third_party_rules.py +++ b/synapse/events/third_party_rules.py @@ -53,7 +53,7 @@ class ThirdPartyEventRules(object): if self.third_party_rules is None: return True - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() # Retrieve the state events from the database. state_events = {} diff --git a/synapse/events/validator.py b/synapse/events/validator.py index 272426e105..9b90c9ce04 100644 --- a/synapse/events/validator.py +++ b/synapse/events/validator.py @@ -13,7 +13,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from six import string_types +from six import integer_types, string_types from synapse.api.constants import MAX_ALIAS_LENGTH, EventTypes, Membership from synapse.api.errors import Codes, SynapseError @@ -22,11 +22,12 @@ from synapse.types import EventID, RoomID, UserID class EventValidator(object): - def validate_new(self, event): + def validate_new(self, event, config): """Validates the event has roughly the right format Args: - event (FrozenEvent) + event (FrozenEvent): The event to validate. + config (Config): The homeserver's configuration. """ self.validate_builder(event) @@ -67,6 +68,99 @@ class EventValidator(object): Codes.INVALID_PARAM, ) + if event.type == EventTypes.Retention: + self._validate_retention(event, config) + + def _validate_retention(self, event, config): + """Checks that an event that defines the retention policy for a room respects the + boundaries imposed by the server's administrator. + + Args: + event (FrozenEvent): The event to validate. + config (Config): The homeserver's configuration. + """ + min_lifetime = event.content.get("min_lifetime") + max_lifetime = event.content.get("max_lifetime") + + if min_lifetime is not None: + if not isinstance(min_lifetime, integer_types): + raise SynapseError( + code=400, + msg="'min_lifetime' must be an integer", + errcode=Codes.BAD_JSON, + ) + + if ( + config.retention_allowed_lifetime_min is not None + and min_lifetime < config.retention_allowed_lifetime_min + ): + raise SynapseError( + code=400, + msg=( + "'min_lifetime' can't be lower than the minimum allowed" + " value enforced by the server's administrator" + ), + errcode=Codes.BAD_JSON, + ) + + if ( + config.retention_allowed_lifetime_max is not None + and min_lifetime > config.retention_allowed_lifetime_max + ): + raise SynapseError( + code=400, + msg=( + "'min_lifetime' can't be greater than the maximum allowed" + " value enforced by the server's administrator" + ), + errcode=Codes.BAD_JSON, + ) + + if max_lifetime is not None: + if not isinstance(max_lifetime, integer_types): + raise SynapseError( + code=400, + msg="'max_lifetime' must be an integer", + errcode=Codes.BAD_JSON, + ) + + if ( + config.retention_allowed_lifetime_min is not None + and max_lifetime < config.retention_allowed_lifetime_min + ): + raise SynapseError( + code=400, + msg=( + "'max_lifetime' can't be lower than the minimum allowed value" + " enforced by the server's administrator" + ), + errcode=Codes.BAD_JSON, + ) + + if ( + config.retention_allowed_lifetime_max is not None + and max_lifetime > config.retention_allowed_lifetime_max + ): + raise SynapseError( + code=400, + msg=( + "'max_lifetime' can't be greater than the maximum allowed" + " value enforced by the server's administrator" + ), + errcode=Codes.BAD_JSON, + ) + + if ( + min_lifetime is not None + and max_lifetime is not None + and min_lifetime > max_lifetime + ): + raise SynapseError( + code=400, + msg="'min_lifetime' can't be greater than 'max_lifetime", + errcode=Codes.BAD_JSON, + ) + def validate_builder(self, event): """Validates that the builder/event has roughly the right format. Only checks values that we expect a proto event to have, rather than all the diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py index 27f6aff004..af652a7659 100644 --- a/synapse/federation/federation_client.py +++ b/synapse/federation/federation_client.py @@ -18,8 +18,6 @@ import copy import itertools import logging -from six.moves import range - from prometheus_client import Counter from twisted.internet import defer @@ -39,7 +37,7 @@ from synapse.api.room_versions import ( ) from synapse.events import builder, room_version_to_event_format from synapse.federation.federation_base import FederationBase, event_from_pdu_json -from synapse.logging.context import make_deferred_yieldable, run_in_background +from synapse.logging.context import make_deferred_yieldable from synapse.logging.utils import log_function from synapse.util import unwrapFirstError from synapse.util.caches.expiringcache import ExpiringCache @@ -310,162 +308,26 @@ class FederationClient(FederationBase): return signed_pdu @defer.inlineCallbacks - @log_function - def get_state_for_room(self, destination, room_id, event_id): - """Requests all of the room state at a given event from a remote homeserver. - - Args: - destination (str): The remote homeserver to query for the state. - room_id (str): The id of the room we're interested in. - event_id (str): The id of the event we want the state at. + def get_room_state_ids(self, destination: str, room_id: str, event_id: str): + """Calls the /state_ids endpoint to fetch the state at a particular point + in the room, and the auth events for the given event Returns: - Deferred[Tuple[List[EventBase], List[EventBase]]]: - A list of events in the state, and a list of events in the auth chain - for the given event. + Tuple[List[str], List[str]]: a tuple of (state event_ids, auth event_ids) """ - try: - # First we try and ask for just the IDs, as thats far quicker if - # we have most of the state and auth_chain already. - # However, this may 404 if the other side has an old synapse. - result = yield self.transport_layer.get_room_state_ids( - destination, room_id, event_id=event_id - ) - - state_event_ids = result["pdu_ids"] - auth_event_ids = result.get("auth_chain_ids", []) - - fetched_events, failed_to_fetch = yield self.get_events_from_store_or_dest( - destination, room_id, set(state_event_ids + auth_event_ids) - ) - - if failed_to_fetch: - logger.warning( - "Failed to fetch missing state/auth events for %s: %s", - room_id, - failed_to_fetch, - ) - - event_map = {ev.event_id: ev for ev in fetched_events} - - pdus = [event_map[e_id] for e_id in state_event_ids if e_id in event_map] - auth_chain = [ - event_map[e_id] for e_id in auth_event_ids if e_id in event_map - ] - - auth_chain.sort(key=lambda e: e.depth) - - return pdus, auth_chain - except HttpResponseException as e: - if e.code == 400 or e.code == 404: - logger.info("Failed to use get_room_state_ids API, falling back") - else: - raise e - - result = yield self.transport_layer.get_room_state( + result = yield self.transport_layer.get_room_state_ids( destination, room_id, event_id=event_id ) - room_version = yield self.store.get_room_version(room_id) - format_ver = room_version_to_event_format(room_version) - - pdus = [ - event_from_pdu_json(p, format_ver, outlier=True) for p in result["pdus"] - ] - - auth_chain = [ - event_from_pdu_json(p, format_ver, outlier=True) - for p in result.get("auth_chain", []) - ] - - seen_events = yield self.store.get_events( - [ev.event_id for ev in itertools.chain(pdus, auth_chain)] - ) - - signed_pdus = yield self._check_sigs_and_hash_and_fetch( - destination, - [p for p in pdus if p.event_id not in seen_events], - outlier=True, - room_version=room_version, - ) - signed_pdus.extend( - seen_events[p.event_id] for p in pdus if p.event_id in seen_events - ) - - signed_auth = yield self._check_sigs_and_hash_and_fetch( - destination, - [p for p in auth_chain if p.event_id not in seen_events], - outlier=True, - room_version=room_version, - ) - signed_auth.extend( - seen_events[p.event_id] for p in auth_chain if p.event_id in seen_events - ) - - signed_auth.sort(key=lambda e: e.depth) - - return signed_pdus, signed_auth - - @defer.inlineCallbacks - def get_events_from_store_or_dest(self, destination, room_id, event_ids): - """Fetch events from a remote destination, checking if we already have them. - - Args: - destination (str) - room_id (str) - event_ids (list) - - Returns: - Deferred: A deferred resolving to a 2-tuple where the first is a list of - events and the second is a list of event ids that we failed to fetch. - """ - seen_events = yield self.store.get_events(event_ids, allow_rejected=True) - signed_events = list(seen_events.values()) - - failed_to_fetch = set() - - missing_events = set(event_ids) - for k in seen_events: - missing_events.discard(k) - - if not missing_events: - return signed_events, failed_to_fetch - - logger.debug( - "Fetching unknown state/auth events %s for room %s", - missing_events, - event_ids, - ) - - room_version = yield self.store.get_room_version(room_id) - - batch_size = 20 - missing_events = list(missing_events) - for i in range(0, len(missing_events), batch_size): - batch = set(missing_events[i : i + batch_size]) - - deferreds = [ - run_in_background( - self.get_pdu, - destinations=[destination], - event_id=e_id, - room_version=room_version, - ) - for e_id in batch - ] - - res = yield make_deferred_yieldable( - defer.DeferredList(deferreds, consumeErrors=True) - ) - for success, result in res: - if success and result: - signed_events.append(result) - batch.discard(result.event_id) + state_event_ids = result["pdu_ids"] + auth_event_ids = result.get("auth_chain_ids", []) - # We removed all events we successfully fetched from `batch` - failed_to_fetch.update(batch) + if not isinstance(state_event_ids, list) or not isinstance( + auth_event_ids, list + ): + raise Exception("invalid response from /state_ids") - return signed_events, failed_to_fetch + return state_event_ids, auth_event_ids @defer.inlineCallbacks @log_function @@ -664,13 +526,7 @@ class FederationClient(FederationBase): @defer.inlineCallbacks def send_request(destination): - time_now = self._clock.time_msec() - _, content = yield self.transport_layer.send_join( - destination=destination, - room_id=pdu.room_id, - event_id=pdu.event_id, - content=pdu.get_pdu_json(time_now), - ) + content = yield self._do_send_join(destination, pdu) logger.debug("Got content: %s", content) @@ -738,6 +594,44 @@ class FederationClient(FederationBase): return self._try_destination_list("send_join", destinations, send_request) @defer.inlineCallbacks + def _do_send_join(self, destination, pdu): + time_now = self._clock.time_msec() + + try: + content = yield self.transport_layer.send_join_v2( + destination=destination, + room_id=pdu.room_id, + event_id=pdu.event_id, + content=pdu.get_pdu_json(time_now), + ) + + return content + except HttpResponseException as e: + if e.code in [400, 404]: + err = e.to_synapse_error() + + # If we receive an error response that isn't a generic error, or an + # unrecognised endpoint error, we assume that the remote understands + # the v2 invite API and this is a legitimate error. + if err.errcode not in [Codes.UNKNOWN, Codes.UNRECOGNIZED]: + raise err + else: + raise e.to_synapse_error() + + logger.debug("Couldn't send_join with the v2 API, falling back to the v1 API") + + resp = yield self.transport_layer.send_join_v1( + destination=destination, + room_id=pdu.room_id, + event_id=pdu.event_id, + content=pdu.get_pdu_json(time_now), + ) + + # We expect the v1 API to respond with [200, content], so we only return the + # content. + return resp[1] + + @defer.inlineCallbacks def send_invite(self, destination, room_id, event_id, pdu): room_version = yield self.store.get_room_version(room_id) @@ -846,18 +740,50 @@ class FederationClient(FederationBase): @defer.inlineCallbacks def send_request(destination): - time_now = self._clock.time_msec() - _, content = yield self.transport_layer.send_leave( + content = yield self._do_send_leave(destination, pdu) + + logger.debug("Got content: %s", content) + return None + + return self._try_destination_list("send_leave", destinations, send_request) + + @defer.inlineCallbacks + def _do_send_leave(self, destination, pdu): + time_now = self._clock.time_msec() + + try: + content = yield self.transport_layer.send_leave_v2( destination=destination, room_id=pdu.room_id, event_id=pdu.event_id, content=pdu.get_pdu_json(time_now), ) - logger.debug("Got content: %s", content) - return None + return content + except HttpResponseException as e: + if e.code in [400, 404]: + err = e.to_synapse_error() - return self._try_destination_list("send_leave", destinations, send_request) + # If we receive an error response that isn't a generic error, or an + # unrecognised endpoint error, we assume that the remote understands + # the v2 invite API and this is a legitimate error. + if err.errcode not in [Codes.UNKNOWN, Codes.UNRECOGNIZED]: + raise err + else: + raise e.to_synapse_error() + + logger.debug("Couldn't send_leave with the v2 API, falling back to the v1 API") + + resp = yield self.transport_layer.send_leave_v1( + destination=destination, + room_id=pdu.room_id, + event_id=pdu.event_id, + content=pdu.get_pdu_json(time_now), + ) + + # We expect the v1 API to respond with [200, content], so we only return the + # content. + return resp[1] def get_public_rooms( self, diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index d942d77a72..8eddb3bf2c 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -1,6 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2015, 2016 OpenMarket Ltd # Copyright 2018 New Vector Ltd +# Copyright 2019 Matrix.org Federation 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. @@ -14,6 +15,7 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging +from typing import Dict import six from six import iteritems @@ -21,6 +23,7 @@ from six import iteritems from canonicaljson import json from prometheus_client import Counter +from twisted.internet import defer from twisted.internet.abstract import isIPAddress from twisted.python import failure @@ -40,7 +43,11 @@ from synapse.federation.federation_base import FederationBase, event_from_pdu_js from synapse.federation.persistence import TransactionActions from synapse.federation.units import Edu, Transaction from synapse.http.endpoint import parse_server_name -from synapse.logging.context import nested_logging_context +from synapse.logging.context import ( + make_deferred_yieldable, + nested_logging_context, + run_in_background, +) from synapse.logging.opentracing import log_kv, start_active_span_from_edu, trace from synapse.logging.utils import log_function from synapse.replication.http.federation import ( @@ -48,7 +55,7 @@ from synapse.replication.http.federation import ( ReplicationGetQueryRestServlet, ) from synapse.types import get_domain_from_id -from synapse.util import glob_to_regex +from synapse.util import glob_to_regex, unwrapFirstError from synapse.util.async_helpers import Linearizer, concurrently_execute from synapse.util.caches.response_cache import ResponseCache @@ -73,6 +80,7 @@ class FederationServer(FederationBase): self.auth = hs.get_auth() self.handler = hs.get_handlers().federation_handler + self.state = hs.get_state_handler() self._server_linearizer = Linearizer("fed_server") self._transaction_linearizer = Linearizer("fed_txn_handler") @@ -158,6 +166,43 @@ class FederationServer(FederationBase): ) return 400, response + # We process PDUs and EDUs in parallel. This is important as we don't + # want to block things like to device messages from reaching clients + # behind the potentially expensive handling of PDUs. + pdu_results, _ = await make_deferred_yieldable( + defer.gatherResults( + [ + run_in_background( + self._handle_pdus_in_txn, origin, transaction, request_time + ), + run_in_background(self._handle_edus_in_txn, origin, transaction), + ], + consumeErrors=True, + ).addErrback(unwrapFirstError) + ) + + response = {"pdus": pdu_results} + + logger.debug("Returning: %s", str(response)) + + await self.transaction_actions.set_response(origin, transaction, 200, response) + return 200, response + + async def _handle_pdus_in_txn( + self, origin: str, transaction: Transaction, request_time: int + ) -> Dict[str, dict]: + """Process the PDUs in a received transaction. + + Args: + origin: the server making the request + transaction: incoming transaction + request_time: timestamp that the HTTP request arrived at + + Returns: + A map from event ID of a processed PDU to any errors we should + report back to the sending server. + """ + received_pdus_counter.inc(len(transaction.pdus)) origin_host, _ = parse_server_name(origin) @@ -248,25 +293,25 @@ class FederationServer(FederationBase): process_pdus_for_room, pdus_by_room.keys(), TRANSACTION_CONCURRENCY_LIMIT ) - if hasattr(transaction, "edus"): - for edu in (Edu(**x) for x in transaction.edus): - await self.received_edu(origin, edu.edu_type, edu.content) + return pdu_results - response = {"pdus": pdu_results} + async def _handle_edus_in_txn(self, origin: str, transaction: Transaction): + """Process the EDUs in a received transaction. + """ - logger.debug("Returning: %s", str(response)) + async def _process_edu(edu_dict): + received_edus_counter.inc() - await self.transaction_actions.set_response(origin, transaction, 200, response) - return 200, response + edu = Edu(**edu_dict) + await self.registry.on_edu(edu.edu_type, origin, edu.content) - async def received_edu(self, origin, edu_type, content): - received_edus_counter.inc() - await self.registry.on_edu(edu_type, origin, content) + await concurrently_execute( + _process_edu, + getattr(transaction, "edus", []), + TRANSACTION_CONCURRENCY_LIMIT, + ) async def on_context_state_request(self, origin, room_id, event_id): - if not event_id: - raise NotImplementedError("Specify an event") - origin_host, _ = parse_server_name(origin) await self.check_server_matches_acl(origin_host, room_id) @@ -280,13 +325,18 @@ class FederationServer(FederationBase): # - but that's non-trivial to get right, and anyway somewhat defeats # the point of the linearizer. with (await self._server_linearizer.queue((origin, room_id))): - resp = await self._state_resp_cache.wrap( - (room_id, event_id), - self._on_context_state_request_compute, - room_id, - event_id, + resp = dict( + await self._state_resp_cache.wrap( + (room_id, event_id), + self._on_context_state_request_compute, + room_id, + event_id, + ) ) + room_version = await self.store.get_room_version(room_id) + resp["room_version"] = room_version + return 200, resp async def on_state_ids_request(self, origin, room_id, event_id): @@ -306,7 +356,11 @@ class FederationServer(FederationBase): return 200, {"pdu_ids": state_ids, "auth_chain_ids": auth_chain_ids} async def _on_context_state_request_compute(self, room_id, event_id): - pdus = await self.handler.get_state_for_pdu(room_id, event_id) + if event_id: + pdus = await self.handler.get_state_for_pdu(room_id, event_id) + else: + pdus = (await self.state.get_current_state(room_id)).values() + auth_chain = await self.store.get_auth_chain([pdu.event_id for pdu in pdus]) return { @@ -376,15 +430,10 @@ class FederationServer(FederationBase): res_pdus = await self.handler.on_send_join_request(origin, pdu) time_now = self._clock.time_msec() - return ( - 200, - { - "state": [p.get_pdu_json(time_now) for p in res_pdus["state"]], - "auth_chain": [ - p.get_pdu_json(time_now) for p in res_pdus["auth_chain"] - ], - }, - ) + return { + "state": [p.get_pdu_json(time_now) for p in res_pdus["state"]], + "auth_chain": [p.get_pdu_json(time_now) for p in res_pdus["auth_chain"]], + } async def on_make_leave_request(self, origin, room_id, user_id): origin_host, _ = parse_server_name(origin) @@ -411,7 +460,7 @@ class FederationServer(FederationBase): pdu = await self._check_sigs_and_hash(room_version, pdu) await self.handler.on_send_leave_request(origin, pdu) - return 200, {} + return {} async def on_event_auth(self, origin, room_id, event_id): with (await self._server_linearizer.queue((origin, room_id))): diff --git a/synapse/federation/persistence.py b/synapse/federation/persistence.py index 44edcabed4..d68b4bd670 100644 --- a/synapse/federation/persistence.py +++ b/synapse/federation/persistence.py @@ -44,7 +44,7 @@ class TransactionActions(object): response code and response body. """ if not transaction.transaction_id: - raise RuntimeError("Cannot persist a transaction with no " "transaction_id") + raise RuntimeError("Cannot persist a transaction with no transaction_id") return self.store.get_received_txn_response(transaction.transaction_id, origin) @@ -56,7 +56,7 @@ class TransactionActions(object): Deferred """ if not transaction.transaction_id: - raise RuntimeError("Cannot persist a transaction with no " "transaction_id") + raise RuntimeError("Cannot persist a transaction with no transaction_id") return self.store.set_received_txn_response( transaction.transaction_id, origin, code, response diff --git a/synapse/federation/send_queue.py b/synapse/federation/send_queue.py index ced4925a98..174f6e42be 100644 --- a/synapse/federation/send_queue.py +++ b/synapse/federation/send_queue.py @@ -259,7 +259,9 @@ class FederationRemoteSendQueue(object): def federation_ack(self, token): self._clear_queue_before_pos(token) - def get_replication_rows(self, from_token, to_token, limit, federation_ack=None): + async def get_replication_rows( + self, from_token, to_token, limit, federation_ack=None + ): """Get rows to be sent over federation between the two tokens Args: diff --git a/synapse/federation/sender/__init__.py b/synapse/federation/sender/__init__.py index 2b2ee8612a..4ebb0e8bc0 100644 --- a/synapse/federation/sender/__init__.py +++ b/synapse/federation/sender/__init__.py @@ -49,7 +49,7 @@ sent_pdus_destination_dist_count = Counter( sent_pdus_destination_dist_total = Counter( "synapse_federation_client_sent_pdu_destinations:total", - "" "Total number of PDUs queued for sending across all destinations", + "Total number of PDUs queued for sending across all destinations", ) diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py index a5b36b1827..5012aaea35 100644 --- a/synapse/federation/sender/per_destination_queue.py +++ b/synapse/federation/sender/per_destination_queue.py @@ -31,6 +31,7 @@ from synapse.handlers.presence import format_user_presence_state from synapse.metrics import sent_transactions_counter from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage.presence import UserPresenceState +from synapse.types import StateMap from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter # This is defined in the Matrix spec and enforced by the receiver. @@ -77,7 +78,7 @@ class PerDestinationQueue(object): # Pending EDUs by their "key". Keyed EDUs are EDUs that get clobbered # based on their key (e.g. typing events by room_id) # Map of (edu_type, key) -> Edu - self._pending_edus_keyed = {} # type: dict[tuple[str, str], Edu] + self._pending_edus_keyed = {} # type: StateMap[Edu] # Map of user_id -> UserPresenceState of pending presence to be sent to this # destination diff --git a/synapse/federation/sender/transaction_manager.py b/synapse/federation/sender/transaction_manager.py index 67b3e1ab6e..5fed626d5b 100644 --- a/synapse/federation/sender/transaction_manager.py +++ b/synapse/federation/sender/transaction_manager.py @@ -84,7 +84,7 @@ class TransactionManager(object): txn_id = str(self._next_txn_id) logger.debug( - "TX [%s] {%s} Attempting new transaction" " (pdus: %d, edus: %d)", + "TX [%s] {%s} Attempting new transaction (pdus: %d, edus: %d)", destination, txn_id, len(pdus), @@ -103,7 +103,7 @@ class TransactionManager(object): self._next_txn_id += 1 logger.info( - "TX [%s] {%s} Sending transaction [%s]," " (PDUs: %d, EDUs: %d)", + "TX [%s] {%s} Sending transaction [%s], (PDUs: %d, EDUs: %d)", destination, txn_id, transaction.transaction_id, diff --git a/synapse/federation/transport/client.py b/synapse/federation/transport/client.py index dc95ab2113..198257414b 100644 --- a/synapse/federation/transport/client.py +++ b/synapse/federation/transport/client.py @@ -39,30 +39,6 @@ class TransportLayerClient(object): self.client = hs.get_http_client() @log_function - def get_room_state(self, destination, room_id, event_id): - """ Requests all state for a given room from the given server at the - given event. - - Args: - destination (str): The host name of the remote homeserver we want - to get the state from. - context (str): The name of the context we want the state of - event_id (str): The event we want the context at. - - Returns: - Deferred: Results in a dict received from the remote homeserver. - """ - logger.debug("get_room_state dest=%s, room=%s", destination, room_id) - - path = _create_v1_path("/state/%s", room_id) - return self.client.get_json( - destination, - path=path, - args={"event_id": event_id}, - try_trailing_slash_on_400=True, - ) - - @log_function def get_room_state_ids(self, destination, room_id, event_id): """ Requests all state for a given room from the given server at the given event. Returns the state's event_id's @@ -267,7 +243,7 @@ class TransportLayerClient(object): @defer.inlineCallbacks @log_function - def send_join(self, destination, room_id, event_id, content): + def send_join_v1(self, destination, room_id, event_id, content): path = _create_v1_path("/send_join/%s/%s", room_id, event_id) response = yield self.client.put_json( @@ -278,7 +254,18 @@ class TransportLayerClient(object): @defer.inlineCallbacks @log_function - def send_leave(self, destination, room_id, event_id, content): + def send_join_v2(self, destination, room_id, event_id, content): + path = _create_v2_path("/send_join/%s/%s", room_id, event_id) + + response = yield self.client.put_json( + destination=destination, path=path, data=content + ) + + return response + + @defer.inlineCallbacks + @log_function + def send_leave_v1(self, destination, room_id, event_id, content): path = _create_v1_path("/send_leave/%s/%s", room_id, event_id) response = yield self.client.put_json( @@ -296,6 +283,24 @@ class TransportLayerClient(object): @defer.inlineCallbacks @log_function + def send_leave_v2(self, destination, room_id, event_id, content): + path = _create_v2_path("/send_leave/%s/%s", room_id, event_id) + + response = yield self.client.put_json( + destination=destination, + path=path, + data=content, + # we want to do our best to send this through. The problem is + # that if it fails, we won't retry it later, so if the remote + # server was just having a momentary blip, the room will be out of + # sync. + ignore_backoff=True, + ) + + return response + + @defer.inlineCallbacks + @log_function def send_invite_v1(self, destination, room_id, event_id, content): path = _create_v1_path("/invite/%s/%s", room_id, event_id) diff --git a/synapse/federation/transport/server.py b/synapse/federation/transport/server.py index 09baa9c57d..b4cbf23394 100644 --- a/synapse/federation/transport/server.py +++ b/synapse/federation/transport/server.py @@ -421,7 +421,7 @@ class FederationEventServlet(BaseFederationServlet): return await self.handler.on_pdu_request(origin, event_id) -class FederationStateServlet(BaseFederationServlet): +class FederationStateV1Servlet(BaseFederationServlet): PATH = "/state/(?P<context>[^/]*)/?" # This is when someone asks for all data for a given context. @@ -429,7 +429,7 @@ class FederationStateServlet(BaseFederationServlet): return await self.handler.on_context_state_request( origin, context, - parse_string_from_args(query, "event_id", None, required=True), + parse_string_from_args(query, "event_id", None, required=False), ) @@ -506,11 +506,21 @@ class FederationMakeLeaveServlet(BaseFederationServlet): return 200, content -class FederationSendLeaveServlet(BaseFederationServlet): +class FederationV1SendLeaveServlet(BaseFederationServlet): PATH = "/send_leave/(?P<room_id>[^/]*)/(?P<event_id>[^/]*)" async def on_PUT(self, origin, content, query, room_id, event_id): content = await self.handler.on_send_leave_request(origin, content, room_id) + return 200, (200, content) + + +class FederationV2SendLeaveServlet(BaseFederationServlet): + PATH = "/send_leave/(?P<room_id>[^/]*)/(?P<event_id>[^/]*)" + + PREFIX = FEDERATION_V2_PREFIX + + async def on_PUT(self, origin, content, query, room_id, event_id): + content = await self.handler.on_send_leave_request(origin, content, room_id) return 200, content @@ -521,13 +531,25 @@ class FederationEventAuthServlet(BaseFederationServlet): return await self.handler.on_event_auth(origin, context, event_id) -class FederationSendJoinServlet(BaseFederationServlet): +class FederationV1SendJoinServlet(BaseFederationServlet): PATH = "/send_join/(?P<context>[^/]*)/(?P<event_id>[^/]*)" async def on_PUT(self, origin, content, query, context, event_id): # TODO(paul): assert that context/event_id parsed from path actually # match those given in content content = await self.handler.on_send_join_request(origin, content, context) + return 200, (200, content) + + +class FederationV2SendJoinServlet(BaseFederationServlet): + PATH = "/send_join/(?P<context>[^/]*)/(?P<event_id>[^/]*)" + + PREFIX = FEDERATION_V2_PREFIX + + async def on_PUT(self, origin, content, query, context, event_id): + # TODO(paul): assert that context/event_id parsed from path actually + # match those given in content + content = await self.handler.on_send_join_request(origin, content, context) return 200, content @@ -1360,15 +1382,17 @@ class RoomComplexityServlet(BaseFederationServlet): FEDERATION_SERVLET_CLASSES = ( FederationSendServlet, FederationEventServlet, - FederationStateServlet, + FederationStateV1Servlet, FederationStateIdsServlet, FederationBackfillServlet, FederationQueryServlet, FederationMakeJoinServlet, FederationMakeLeaveServlet, FederationEventServlet, - FederationSendJoinServlet, - FederationSendLeaveServlet, + FederationV1SendJoinServlet, + FederationV2SendJoinServlet, + FederationV1SendLeaveServlet, + FederationV2SendLeaveServlet, FederationV1InviteServlet, FederationV2InviteServlet, FederationQueryAuthServlet, diff --git a/synapse/groups/groups_server.py b/synapse/groups/groups_server.py index 29e8ffc295..0ec9be3cb5 100644 --- a/synapse/groups/groups_server.py +++ b/synapse/groups/groups_server.py @@ -773,6 +773,11 @@ class GroupsServerHandler(object): if not self.hs.is_mine_id(user_id): yield self.store.maybe_delete_remote_profile_cache(user_id) + # Delete group if the last user has left + users = yield self.store.get_users_in_group(group_id, include_private=True) + if not users: + yield self.store.delete_group(group_id) + return {} @defer.inlineCallbacks diff --git a/synapse/handlers/_base.py b/synapse/handlers/_base.py index d15c6282fb..51413d910e 100644 --- a/synapse/handlers/_base.py +++ b/synapse/handlers/_base.py @@ -134,7 +134,7 @@ class BaseHandler(object): guest_access = event.content.get("guest_access", "forbidden") if guest_access != "can_join": if context: - current_state_ids = yield context.get_current_state_ids(self.store) + current_state_ids = yield context.get_current_state_ids() current_state = yield self.store.get_events( list(current_state_ids.values()) ) diff --git a/synapse/handlers/account_data.py b/synapse/handlers/account_data.py index 2d7e6df6e4..a8d3fbc6de 100644 --- a/synapse/handlers/account_data.py +++ b/synapse/handlers/account_data.py @@ -13,8 +13,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -from twisted.internet import defer - class AccountDataEventSource(object): def __init__(self, hs): @@ -23,15 +21,14 @@ class AccountDataEventSource(object): def get_current_key(self, direction="f"): return self.store.get_max_account_data_stream_id() - @defer.inlineCallbacks - def get_new_events(self, user, from_key, **kwargs): + async def get_new_events(self, user, from_key, **kwargs): user_id = user.to_string() last_stream_id = from_key - current_stream_id = yield self.store.get_max_account_data_stream_id() + current_stream_id = self.store.get_max_account_data_stream_id() results = [] - tags = yield self.store.get_updated_tags(user_id, last_stream_id) + tags = await self.store.get_updated_tags(user_id, last_stream_id) for room_id, room_tags in tags.items(): results.append( @@ -41,7 +38,7 @@ class AccountDataEventSource(object): ( account_data, room_account_data, - ) = yield self.store.get_updated_account_data_for_user(user_id, last_stream_id) + ) = await self.store.get_updated_account_data_for_user(user_id, last_stream_id) for account_data_type, content in account_data.items(): results.append({"type": account_data_type, "content": content}) @@ -53,7 +50,3 @@ class AccountDataEventSource(object): ) return results, current_stream_id - - @defer.inlineCallbacks - def get_pagination_rows(self, user, config, key): - return [], config.to_id diff --git a/synapse/handlers/account_validity.py b/synapse/handlers/account_validity.py index d04e0fe576..829f52eca1 100644 --- a/synapse/handlers/account_validity.py +++ b/synapse/handlers/account_validity.py @@ -18,8 +18,7 @@ import email.utils import logging from email.mime.multipart import MIMEMultipart from email.mime.text import MIMEText - -from twisted.internet import defer +from typing import List from synapse.api.errors import StoreError from synapse.logging.context import make_deferred_yieldable @@ -78,42 +77,39 @@ class AccountValidityHandler(object): # run as a background process to make sure that the database transactions # have a logcontext to report to return run_as_background_process( - "send_renewals", self.send_renewal_emails + "send_renewals", self._send_renewal_emails ) self.clock.looping_call(send_emails, 30 * 60 * 1000) - @defer.inlineCallbacks - def send_renewal_emails(self): + async def _send_renewal_emails(self): """Gets the list of users whose account is expiring in the amount of time configured in the ``renew_at`` parameter from the ``account_validity`` configuration, and sends renewal emails to all of these users as long as they have an email 3PID attached to their account. """ - expiring_users = yield self.store.get_users_expiring_soon() + expiring_users = await self.store.get_users_expiring_soon() if expiring_users: for user in expiring_users: - yield self._send_renewal_email( + await self._send_renewal_email( user_id=user["user_id"], expiration_ts=user["expiration_ts_ms"] ) - @defer.inlineCallbacks - def send_renewal_email_to_user(self, user_id): - expiration_ts = yield self.store.get_expiration_ts_for_user(user_id) - yield self._send_renewal_email(user_id, expiration_ts) + async def send_renewal_email_to_user(self, user_id: str): + expiration_ts = await self.store.get_expiration_ts_for_user(user_id) + await self._send_renewal_email(user_id, expiration_ts) - @defer.inlineCallbacks - def _send_renewal_email(self, user_id, expiration_ts): + async def _send_renewal_email(self, user_id: str, expiration_ts: int): """Sends out a renewal email to every email address attached to the given user with a unique link allowing them to renew their account. Args: - user_id (str): ID of the user to send email(s) to. - expiration_ts (int): Timestamp in milliseconds for the expiration date of + user_id: ID of the user to send email(s) to. + expiration_ts: Timestamp in milliseconds for the expiration date of this user's account (used in the email templates). """ - addresses = yield self._get_email_addresses_for_user(user_id) + addresses = await self._get_email_addresses_for_user(user_id) # Stop right here if the user doesn't have at least one email address. # In this case, they will have to ask their server admin to renew their @@ -125,7 +121,7 @@ class AccountValidityHandler(object): return try: - user_display_name = yield self.store.get_profile_displayname( + user_display_name = await self.store.get_profile_displayname( UserID.from_string(user_id).localpart ) if user_display_name is None: @@ -133,7 +129,7 @@ class AccountValidityHandler(object): except StoreError: user_display_name = user_id - renewal_token = yield self._get_renewal_token(user_id) + renewal_token = await self._get_renewal_token(user_id) url = "%s_matrix/client/unstable/account_validity/renew?token=%s" % ( self.hs.config.public_baseurl, renewal_token, @@ -165,7 +161,7 @@ class AccountValidityHandler(object): logger.info("Sending renewal email to %s", address) - yield make_deferred_yieldable( + await make_deferred_yieldable( self.sendmail( self.hs.config.email_smtp_host, self._raw_from, @@ -180,19 +176,18 @@ class AccountValidityHandler(object): ) ) - yield self.store.set_renewal_mail_status(user_id=user_id, email_sent=True) + await self.store.set_renewal_mail_status(user_id=user_id, email_sent=True) - @defer.inlineCallbacks - def _get_email_addresses_for_user(self, user_id): + async def _get_email_addresses_for_user(self, user_id: str) -> List[str]: """Retrieve the list of email addresses attached to a user's account. Args: - user_id (str): ID of the user to lookup email addresses for. + user_id: ID of the user to lookup email addresses for. Returns: - defer.Deferred[list[str]]: Email addresses for this account. + Email addresses for this account. """ - threepids = yield self.store.user_get_threepids(user_id) + threepids = await self.store.user_get_threepids(user_id) addresses = [] for threepid in threepids: @@ -201,16 +196,15 @@ class AccountValidityHandler(object): return addresses - @defer.inlineCallbacks - def _get_renewal_token(self, user_id): + async def _get_renewal_token(self, user_id: str) -> str: """Generates a 32-byte long random string that will be inserted into the user's renewal email's unique link, then saves it into the database. Args: - user_id (str): ID of the user to generate a string for. + user_id: ID of the user to generate a string for. Returns: - defer.Deferred[str]: The generated string. + The generated string. Raises: StoreError(500): Couldn't generate a unique string after 5 attempts. @@ -219,52 +213,52 @@ class AccountValidityHandler(object): while attempts < 5: try: renewal_token = stringutils.random_string(32) - yield self.store.set_renewal_token_for_user(user_id, renewal_token) + await self.store.set_renewal_token_for_user(user_id, renewal_token) return renewal_token except StoreError: attempts += 1 raise StoreError(500, "Couldn't generate a unique string as refresh string.") - @defer.inlineCallbacks - def renew_account(self, renewal_token): + async def renew_account(self, renewal_token: str) -> bool: """Renews the account attached to a given renewal token by pushing back the expiration date by the current validity period in the server's configuration. Args: - renewal_token (str): Token sent with the renewal request. + renewal_token: Token sent with the renewal request. Returns: - bool: Whether the provided token is valid. + Whether the provided token is valid. """ try: - user_id = yield self.store.get_user_from_renewal_token(renewal_token) + user_id = await self.store.get_user_from_renewal_token(renewal_token) except StoreError: - defer.returnValue(False) + return False logger.debug("Renewing an account for user %s", user_id) - yield self.renew_account_for_user(user_id) + await self.renew_account_for_user(user_id) - defer.returnValue(True) + return True - @defer.inlineCallbacks - def renew_account_for_user(self, user_id, expiration_ts=None, email_sent=False): + async def renew_account_for_user( + self, user_id: str, expiration_ts: int = None, email_sent: bool = False + ) -> int: """Renews the account attached to a given user by pushing back the expiration date by the current validity period in the server's configuration. Args: - renewal_token (str): Token sent with the renewal request. - expiration_ts (int): New expiration date. Defaults to now + validity period. - email_sent (bool): Whether an email has been sent for this validity period. + renewal_token: Token sent with the renewal request. + expiration_ts: New expiration date. Defaults to now + validity period. + email_sen: Whether an email has been sent for this validity period. Defaults to False. Returns: - defer.Deferred[int]: New expiration date for this account, as a timestamp - in milliseconds since epoch. + New expiration date for this account, as a timestamp in + milliseconds since epoch. """ if expiration_ts is None: expiration_ts = self.clock.time_msec() + self._account_validity.period - yield self.store.set_account_validity_for_user( + await self.store.set_account_validity_for_user( user_id=user_id, expiration_ts=expiration_ts, email_sent=email_sent ) diff --git a/synapse/handlers/admin.py b/synapse/handlers/admin.py index 6407d56f8e..60a7c938bc 100644 --- a/synapse/handlers/admin.py +++ b/synapse/handlers/admin.py @@ -14,11 +14,11 @@ # limitations under the License. import logging - -from twisted.internet import defer +from typing import List from synapse.api.constants import Membership -from synapse.types import RoomStreamToken +from synapse.events import FrozenEvent +from synapse.types import RoomStreamToken, StateMap from synapse.visibility import filter_events_for_client from ._base import BaseHandler @@ -33,11 +33,10 @@ class AdminHandler(BaseHandler): self.storage = hs.get_storage() self.state_store = self.storage.state - @defer.inlineCallbacks - def get_whois(self, user): + async def get_whois(self, user): connections = [] - sessions = yield self.store.get_user_ip_and_agents(user) + sessions = await self.store.get_user_ip_and_agents(user) for session in sessions: connections.append( { @@ -54,37 +53,46 @@ class AdminHandler(BaseHandler): return ret - @defer.inlineCallbacks - def get_users(self): - """Function to reterive a list of users in users table. + async def get_user(self, user): + """Function to get user details""" + ret = await self.store.get_user_by_id(user.to_string()) + if ret: + profile = await self.store.get_profileinfo(user.localpart) + ret["displayname"] = profile.display_name + ret["avatar_url"] = profile.avatar_url + return ret + + async def get_users(self): + """Function to retrieve a list of users in users table. Args: Returns: defer.Deferred: resolves to list[dict[str, Any]] """ - ret = yield self.store.get_users() + ret = await self.store.get_users() return ret - @defer.inlineCallbacks - def get_users_paginate(self, order, start, limit): - """Function to reterive a paginated list of users from - users list. This will return a json object, which contains - list of users and the total number of users in users table. + async def get_users_paginate(self, start, limit, name, guests, deactivated): + """Function to retrieve a paginated list of users from + users list. This will return a json list of users. Args: - order (str): column name to order the select by this column start (int): start number to begin the query from - limit (int): number of rows to reterive + limit (int): number of rows to retrieve + name (string): filter for user names + guests (bool): whether to in include guest users + deactivated (bool): whether to include deactivated users Returns: - defer.Deferred: resolves to json object {list[dict[str, Any]], count} + defer.Deferred: resolves to json list[dict[str, Any]] """ - ret = yield self.store.get_users_paginate(order, start, limit) + ret = await self.store.get_users_paginate( + start, limit, name, guests, deactivated + ) return ret - @defer.inlineCallbacks - def search_users(self, term): + async def search_users(self, term): """Function to search users list for one or more users with the matched term. @@ -93,7 +101,7 @@ class AdminHandler(BaseHandler): Returns: defer.Deferred: resolves to list[dict[str, Any]] """ - ret = yield self.store.search_users(term) + ret = await self.store.search_users(term) return ret @@ -116,8 +124,7 @@ class AdminHandler(BaseHandler): """ return self.store.set_server_admin(user, admin) - @defer.inlineCallbacks - def export_user_data(self, user_id, writer): + async def export_user_data(self, user_id, writer): """Write all data we have on the user to the given writer. Args: @@ -129,7 +136,7 @@ class AdminHandler(BaseHandler): The returned value is that returned by `writer.finished()`. """ # Get all rooms the user is in or has been in - rooms = yield self.store.get_rooms_for_user_where_membership_is( + rooms = await self.store.get_rooms_for_local_user_where_membership_is( user_id, membership_list=( Membership.JOIN, @@ -142,7 +149,7 @@ class AdminHandler(BaseHandler): # We only try and fetch events for rooms the user has been in. If # they've been e.g. invited to a room without joining then we handle # those seperately. - rooms_user_has_been_in = yield self.store.get_rooms_user_has_been_in(user_id) + rooms_user_has_been_in = await self.store.get_rooms_user_has_been_in(user_id) for index, room in enumerate(rooms): room_id = room.room_id @@ -151,7 +158,7 @@ class AdminHandler(BaseHandler): "[%s] Handling room %s, %d/%d", user_id, room_id, index + 1, len(rooms) ) - forgotten = yield self.store.did_forget(user_id, room_id) + forgotten = await self.store.did_forget(user_id, room_id) if forgotten: logger.info("[%s] User forgot room %d, ignoring", user_id, room_id) continue @@ -163,7 +170,7 @@ class AdminHandler(BaseHandler): if room.membership == Membership.INVITE: event_id = room.event_id - invite = yield self.store.get_event(event_id, allow_none=True) + invite = await self.store.get_event(event_id, allow_none=True) if invite: invited_state = invite.unsigned["invite_room_state"] writer.write_invite(room_id, invite, invited_state) @@ -174,7 +181,7 @@ class AdminHandler(BaseHandler): # were joined. We estimate that point by looking at the # stream_ordering of the last membership if it wasn't a join. if room.membership == Membership.JOIN: - stream_ordering = yield self.store.get_room_max_stream_ordering() + stream_ordering = self.store.get_room_max_stream_ordering() else: stream_ordering = room.stream_ordering @@ -200,7 +207,7 @@ class AdminHandler(BaseHandler): # events that we have and then filtering, this isn't the most # efficient method perhaps but it does guarantee we get everything. while True: - events, _ = yield self.store.paginate_room_events( + events, _ = await self.store.paginate_room_events( room_id, from_key, to_key, limit=100, direction="f" ) if not events: @@ -208,7 +215,7 @@ class AdminHandler(BaseHandler): from_key = events[-1].internal_metadata.after - events = yield filter_events_for_client(self.storage, user_id, events) + events = await filter_events_for_client(self.storage, user_id, events) writer.write_events(room_id, events) @@ -244,7 +251,7 @@ class AdminHandler(BaseHandler): for event_id in extremities: if not event_to_unseen_prevs[event_id]: continue - state = yield self.state_store.get_state_for_event(event_id) + state = await self.state_store.get_state_for_event(event_id) writer.write_state(room_id, event_id, state) return writer.finished() @@ -254,35 +261,26 @@ class ExfiltrationWriter(object): """Interface used to specify how to write exported data. """ - def write_events(self, room_id, events): + def write_events(self, room_id: str, events: List[FrozenEvent]): """Write a batch of events for a room. - - Args: - room_id (str) - events (list[FrozenEvent]) """ pass - def write_state(self, room_id, event_id, state): + def write_state(self, room_id: str, event_id: str, state: StateMap[FrozenEvent]): """Write the state at the given event in the room. This only gets called for backward extremities rather than for each event. - - Args: - room_id (str) - event_id (str) - state (dict[tuple[str, str], FrozenEvent]) """ pass - def write_invite(self, room_id, event, state): + def write_invite(self, room_id: str, event: FrozenEvent, state: StateMap[dict]): """Write an invite for the room, with associated invite state. Args: - room_id (str) - event (FrozenEvent) - state (dict[tuple[str, str], dict]): A subset of the state at the + room_id + event + state: A subset of the state at the invite, with a subset of the event keys (type, state_key content and sender) """ diff --git a/synapse/handlers/deactivate_account.py b/synapse/handlers/deactivate_account.py index 63267a0a4c..2afb390a92 100644 --- a/synapse/handlers/deactivate_account.py +++ b/synapse/handlers/deactivate_account.py @@ -15,8 +15,6 @@ # limitations under the License. import logging -from twisted.internet import defer - from synapse.api.errors import SynapseError from synapse.metrics.background_process_metrics import run_as_background_process from synapse.types import UserID, create_requester @@ -46,8 +44,7 @@ class DeactivateAccountHandler(BaseHandler): self._account_validity_enabled = hs.config.account_validity.enabled - @defer.inlineCallbacks - def deactivate_account(self, user_id, erase_data, id_server=None): + async def deactivate_account(self, user_id, erase_data, id_server=None): """Deactivate a user's account Args: @@ -74,11 +71,11 @@ class DeactivateAccountHandler(BaseHandler): identity_server_supports_unbinding = True # Retrieve the 3PIDs this user has bound to an identity server - threepids = yield self.store.user_get_bound_threepids(user_id) + threepids = await self.store.user_get_bound_threepids(user_id) for threepid in threepids: try: - result = yield self._identity_handler.try_unbind_threepid( + result = await self._identity_handler.try_unbind_threepid( user_id, { "medium": threepid["medium"], @@ -91,30 +88,33 @@ class DeactivateAccountHandler(BaseHandler): # Do we want this to be a fatal error or should we carry on? logger.exception("Failed to remove threepid from ID server") raise SynapseError(400, "Failed to remove threepid from ID server") - yield self.store.user_delete_threepid( + await self.store.user_delete_threepid( user_id, threepid["medium"], threepid["address"] ) + # Remove all 3PIDs this user has bound to the homeserver + await self.store.user_delete_threepids(user_id) + # delete any devices belonging to the user, which will also # delete corresponding access tokens. - yield self._device_handler.delete_all_devices_for_user(user_id) + await self._device_handler.delete_all_devices_for_user(user_id) # then delete any remaining access tokens which weren't associated with # a device. - yield self._auth_handler.delete_access_tokens_for_user(user_id) + await self._auth_handler.delete_access_tokens_for_user(user_id) - yield self.store.user_set_password_hash(user_id, None) + await self.store.user_set_password_hash(user_id, None) # Add the user to a table of users pending deactivation (ie. # removal from all the rooms they're a member of) - yield self.store.add_user_pending_deactivation(user_id) + await self.store.add_user_pending_deactivation(user_id) # delete from user directory - yield self.user_directory_handler.handle_user_deactivated(user_id) + await self.user_directory_handler.handle_user_deactivated(user_id) # Mark the user as erased, if they asked for that if erase_data: logger.info("Marking %s as erased", user_id) - yield self.store.mark_user_erased(user_id) + await self.store.mark_user_erased(user_id) # Now start the process that goes through that list and # parts users from rooms (if it isn't already running) @@ -122,30 +122,29 @@ class DeactivateAccountHandler(BaseHandler): # Reject all pending invites for the user, so that the user doesn't show up in the # "invited" section of rooms' members list. - yield self._reject_pending_invites_for_user(user_id) + await self._reject_pending_invites_for_user(user_id) # Remove all information on the user from the account_validity table. if self._account_validity_enabled: - yield self.store.delete_account_validity_for_user(user_id) + await self.store.delete_account_validity_for_user(user_id) # Mark the user as deactivated. - yield self.store.set_user_deactivated_status(user_id, True) + await self.store.set_user_deactivated_status(user_id, True) return identity_server_supports_unbinding - @defer.inlineCallbacks - def _reject_pending_invites_for_user(self, user_id): + async def _reject_pending_invites_for_user(self, user_id): """Reject pending invites addressed to a given user ID. Args: user_id (str): The user ID to reject pending invites for. """ user = UserID.from_string(user_id) - pending_invites = yield self.store.get_invited_rooms_for_user(user_id) + pending_invites = await self.store.get_invited_rooms_for_local_user(user_id) for room in pending_invites: try: - yield self._room_member_handler.update_membership( + await self._room_member_handler.update_membership( create_requester(user), user, room.room_id, @@ -177,8 +176,7 @@ class DeactivateAccountHandler(BaseHandler): if not self._user_parter_running: run_as_background_process("user_parter_loop", self._user_parter_loop) - @defer.inlineCallbacks - def _user_parter_loop(self): + async def _user_parter_loop(self): """Loop that parts deactivated users from rooms Returns: @@ -188,19 +186,18 @@ class DeactivateAccountHandler(BaseHandler): logger.info("Starting user parter") try: while True: - user_id = yield self.store.get_user_pending_deactivation() + user_id = await self.store.get_user_pending_deactivation() if user_id is None: break logger.info("User parter parting %r", user_id) - yield self._part_user(user_id) - yield self.store.del_user_pending_deactivation(user_id) + await self._part_user(user_id) + await self.store.del_user_pending_deactivation(user_id) logger.info("User parter finished parting %r", user_id) logger.info("User parter finished: stopping") finally: self._user_parter_running = False - @defer.inlineCallbacks - def _part_user(self, user_id): + async def _part_user(self, user_id): """Causes the given user_id to leave all the rooms they're joined to Returns: @@ -208,11 +205,11 @@ class DeactivateAccountHandler(BaseHandler): """ user = UserID.from_string(user_id) - rooms_for_user = yield self.store.get_rooms_for_user(user_id) + rooms_for_user = await self.store.get_rooms_for_user(user_id) for room_id in rooms_for_user: logger.info("User parter parting %r from %r", user_id, room_id) try: - yield self._room_member_handler.update_membership( + await self._room_member_handler.update_membership( create_requester(user), user, room_id, diff --git a/synapse/handlers/directory.py b/synapse/handlers/directory.py index 69051101a6..a07d2f1a17 100644 --- a/synapse/handlers/directory.py +++ b/synapse/handlers/directory.py @@ -119,7 +119,7 @@ class DirectoryHandler(BaseHandler): if not service.is_interested_in_alias(room_alias.to_string()): raise SynapseError( 400, - "This application service has not reserved" " this kind of alias.", + "This application service has not reserved this kind of alias.", errcode=Codes.EXCLUSIVE, ) else: diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index f09a0b73c8..2d889364d4 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -30,6 +30,7 @@ from twisted.internet import defer from synapse.api.errors import CodeMessageException, Codes, NotFoundError, SynapseError from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.logging.opentracing import log_kv, set_tag, tag_args, trace +from synapse.replication.http.devices import ReplicationUserDevicesResyncRestServlet from synapse.types import ( UserID, get_domain_from_id, @@ -53,6 +54,12 @@ class E2eKeysHandler(object): self._edu_updater = SigningKeyEduUpdater(hs, self) + self._is_master = hs.config.worker_app is None + if not self._is_master: + self._user_device_resync_client = ReplicationUserDevicesResyncRestServlet.make_client( + hs + ) + federation_registry = hs.get_federation_registry() # FIXME: switch to m.signing_key_update when MSC1756 is merged into the spec @@ -191,9 +198,15 @@ class E2eKeysHandler(object): # probably be tracking their device lists. However, we haven't # done an initial sync on the device list so we do it now. try: - user_devices = yield self.device_handler.device_list_updater.user_device_resync( - user_id - ) + if self._is_master: + user_devices = yield self.device_handler.device_list_updater.user_device_resync( + user_id + ) + else: + user_devices = yield self._user_device_resync_client( + user_id=user_id + ) + user_devices = user_devices["devices"] for device in user_devices: results[user_id] = {device["device_id"]: device["keys"]} @@ -271,29 +284,26 @@ class E2eKeysHandler(object): self_signing_keys = {} user_signing_keys = {} - for user_id in query: - # XXX: consider changing the store functions to allow querying - # multiple users simultaneously. - key = yield self.store.get_e2e_cross_signing_key( - user_id, "master", from_user_id - ) - if key: - master_keys[user_id] = key + user_ids = list(query) - key = yield self.store.get_e2e_cross_signing_key( - user_id, "self_signing", from_user_id - ) - if key: - self_signing_keys[user_id] = key + keys = yield self.store.get_e2e_cross_signing_keys_bulk(user_ids, from_user_id) + for user_id, user_info in keys.items(): + if user_info is None: + continue + if "master" in user_info: + master_keys[user_id] = user_info["master"] + if "self_signing" in user_info: + self_signing_keys[user_id] = user_info["self_signing"] + + if ( + from_user_id in keys + and keys[from_user_id] is not None + and "user_signing" in keys[from_user_id] + ): # users can see other users' master and self-signing keys, but can # only see their own user-signing keys - if from_user_id == user_id: - key = yield self.store.get_e2e_cross_signing_key( - user_id, "user_signing", from_user_id - ) - if key: - user_signing_keys[user_id] = key + user_signing_keys[from_user_id] = keys[from_user_id]["user_signing"] return { "master_keys": master_keys, diff --git a/synapse/handlers/e2e_room_keys.py b/synapse/handlers/e2e_room_keys.py index 0cea445f0d..f1b4424a02 100644 --- a/synapse/handlers/e2e_room_keys.py +++ b/synapse/handlers/e2e_room_keys.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2017, 2018 New Vector Ltd +# Copyright 2019 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. @@ -103,14 +104,35 @@ class E2eRoomKeysHandler(object): rooms session_id(string): session ID to delete keys for, for None to delete keys for all sessions + Raises: + NotFoundError: if the backup version does not exist Returns: - A deferred of the deletion transaction + A dict containing the count and etag for the backup version """ # lock for consistency with uploading with (yield self._upload_linearizer.queue(user_id)): + # make sure the backup version exists + try: + version_info = yield self.store.get_e2e_room_keys_version_info( + user_id, version + ) + except StoreError as e: + if e.code == 404: + raise NotFoundError("Unknown backup version") + else: + raise + yield self.store.delete_e2e_room_keys(user_id, version, room_id, session_id) + version_etag = version_info["etag"] + 1 + yield self.store.update_e2e_room_keys_version( + user_id, version, None, version_etag + ) + + count = yield self.store.count_e2e_room_keys(user_id, version) + return {"etag": str(version_etag), "count": count} + @trace @defer.inlineCallbacks def upload_room_keys(self, user_id, version, room_keys): @@ -138,6 +160,9 @@ class E2eRoomKeysHandler(object): } } + Returns: + A dict containing the count and etag for the backup version + Raises: NotFoundError: if there are no versions defined RoomKeysVersionError: if the uploaded version is not the current version @@ -171,59 +196,62 @@ class E2eRoomKeysHandler(object): else: raise - # go through the room_keys. - # XXX: this should/could be done concurrently, given we're in a lock. + # Fetch any existing room keys for the sessions that have been + # submitted. Then compare them with the submitted keys. If the + # key is new, insert it; if the key should be updated, then update + # it; otherwise, drop it. + existing_keys = yield self.store.get_e2e_room_keys_multi( + user_id, version, room_keys["rooms"] + ) + to_insert = [] # batch the inserts together + changed = False # if anything has changed, we need to update the etag for room_id, room in iteritems(room_keys["rooms"]): - for session_id, session in iteritems(room["sessions"]): - yield self._upload_room_key( - user_id, version, room_id, session_id, session + for session_id, room_key in iteritems(room["sessions"]): + log_kv( + { + "message": "Trying to upload room key", + "room_id": room_id, + "session_id": session_id, + "user_id": user_id, + } ) - - @defer.inlineCallbacks - def _upload_room_key(self, user_id, version, room_id, session_id, room_key): - """Upload a given room_key for a given room and session into a given - version of the backup. Merges the key with any which might already exist. - - Args: - user_id(str): the user whose backup we're setting - version(str): the version ID of the backup we're updating - room_id(str): the ID of the room whose keys we're setting - session_id(str): the session whose room_key we're setting - room_key(dict): the room_key being set - """ - log_kv( - { - "message": "Trying to upload room key", - "room_id": room_id, - "session_id": session_id, - "user_id": user_id, - } - ) - # get the room_key for this particular row - current_room_key = None - try: - current_room_key = yield self.store.get_e2e_room_key( - user_id, version, room_id, session_id - ) - except StoreError as e: - if e.code == 404: - log_kv( - { - "message": "Room key not found.", - "room_id": room_id, - "user_id": user_id, - } + current_room_key = existing_keys.get(room_id, {}).get(session_id) + if current_room_key: + if self._should_replace_room_key(current_room_key, room_key): + log_kv({"message": "Replacing room key."}) + # updates are done one at a time in the DB, so send + # updates right away rather than batching them up, + # like we do with the inserts + yield self.store.update_e2e_room_key( + user_id, version, room_id, session_id, room_key + ) + changed = True + else: + log_kv({"message": "Not replacing room_key."}) + else: + log_kv( + { + "message": "Room key not found.", + "room_id": room_id, + "user_id": user_id, + } + ) + log_kv({"message": "Replacing room key."}) + to_insert.append((room_id, session_id, room_key)) + changed = True + + if len(to_insert): + yield self.store.add_e2e_room_keys(user_id, version, to_insert) + + version_etag = version_info["etag"] + if changed: + version_etag = version_etag + 1 + yield self.store.update_e2e_room_keys_version( + user_id, version, None, version_etag ) - else: - raise - if self._should_replace_room_key(current_room_key, room_key): - log_kv({"message": "Replacing room key."}) - yield self.store.set_e2e_room_key( - user_id, version, room_id, session_id, room_key - ) - else: - log_kv({"message": "Not replacing room_key."}) + count = yield self.store.count_e2e_room_keys(user_id, version) + return {"etag": str(version_etag), "count": count} @staticmethod def _should_replace_room_key(current_room_key, room_key): @@ -314,6 +342,8 @@ class E2eRoomKeysHandler(object): raise NotFoundError("Unknown backup version") else: raise + + res["count"] = yield self.store.count_e2e_room_keys(user_id, res["version"]) return res @trace diff --git a/synapse/handlers/events.py b/synapse/handlers/events.py index 45fe13c62f..ec18a42a68 100644 --- a/synapse/handlers/events.py +++ b/synapse/handlers/events.py @@ -16,8 +16,6 @@ import logging import random -from twisted.internet import defer - from synapse.api.constants import EventTypes, Membership from synapse.api.errors import AuthError, SynapseError from synapse.events import EventBase @@ -50,9 +48,8 @@ class EventStreamHandler(BaseHandler): self._server_notices_sender = hs.get_server_notices_sender() self._event_serializer = hs.get_event_client_serializer() - @defer.inlineCallbacks @log_function - def get_stream( + async def get_stream( self, auth_user_id, pagin_config, @@ -69,17 +66,17 @@ class EventStreamHandler(BaseHandler): """ if room_id: - blocked = yield self.store.is_room_blocked(room_id) + blocked = await self.store.is_room_blocked(room_id) if blocked: raise SynapseError(403, "This room has been blocked on this server") # send any outstanding server notices to the user. - yield self._server_notices_sender.on_user_syncing(auth_user_id) + await self._server_notices_sender.on_user_syncing(auth_user_id) auth_user = UserID.from_string(auth_user_id) presence_handler = self.hs.get_presence_handler() - context = yield presence_handler.user_syncing( + context = await presence_handler.user_syncing( auth_user_id, affect_presence=affect_presence ) with context: @@ -91,7 +88,7 @@ class EventStreamHandler(BaseHandler): # thundering herds on restart. timeout = random.randint(int(timeout * 0.9), int(timeout * 1.1)) - events, tokens = yield self.notifier.get_events_for( + events, tokens = await self.notifier.get_events_for( auth_user, pagin_config, timeout, @@ -112,14 +109,14 @@ class EventStreamHandler(BaseHandler): # Send down presence. if event.state_key == auth_user_id: # Send down presence for everyone in the room. - users = yield self.state.get_current_users_in_room( + users = await self.state.get_current_users_in_room( event.room_id ) - states = yield presence_handler.get_states(users, as_event=True) + states = await presence_handler.get_states(users, as_event=True) to_add.extend(states) else: - ev = yield presence_handler.get_state( + ev = await presence_handler.get_state( UserID.from_string(event.state_key), as_event=True ) to_add.append(ev) @@ -128,7 +125,7 @@ class EventStreamHandler(BaseHandler): time_now = self.clock.time_msec() - chunks = yield self._event_serializer.serialize_events( + chunks = await self._event_serializer.serialize_events( events, time_now, as_client_event=as_client_event, @@ -151,8 +148,7 @@ class EventHandler(BaseHandler): super(EventHandler, self).__init__(hs) self.storage = hs.get_storage() - @defer.inlineCallbacks - def get_event(self, user, room_id, event_id): + async def get_event(self, user, room_id, event_id): """Retrieve a single specified event. Args: @@ -167,15 +163,15 @@ class EventHandler(BaseHandler): AuthError if the user does not have the rights to inspect this event. """ - event = yield self.store.get_event(event_id, check_room_id=room_id) + event = await self.store.get_event(event_id, check_room_id=room_id) if not event: return None - users = yield self.store.get_users_in_room(event.room_id) + users = await self.store.get_users_in_room(event.room_id) is_peeking = user.to_string() not in users - filtered = yield filter_events_for_client( + filtered = await filter_events_for_client( self.storage, user.to_string(), [event], is_peeking=is_peeking ) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 0e904f2da0..d4f9a792fc 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -19,11 +19,13 @@ import itertools import logging +from typing import Dict, Iterable, List, Optional, Sequence, Tuple import six from six import iteritems, itervalues from six.moves import http_client, zip +import attr from signedjson.key import decode_verify_key_bytes from signedjson.sign import verify_signed_json from unpaddedbase64 import decode_base64 @@ -45,6 +47,7 @@ from synapse.api.errors import ( from synapse.api.room_versions import KNOWN_ROOM_VERSIONS, RoomVersions from synapse.crypto.event_signing import compute_event_signature from synapse.event_auth import auth_types_for_event +from synapse.events import EventBase from synapse.events.snapshot import EventContext from synapse.events.validator import EventValidator from synapse.logging.context import ( @@ -60,9 +63,9 @@ from synapse.replication.http.federation import ( ) from synapse.replication.http.membership import ReplicationUserJoinedLeftRoomRestServlet from synapse.state import StateResolutionStore, resolve_events_with_store -from synapse.types import UserID, get_domain_from_id -from synapse.util import unwrapFirstError -from synapse.util.async_helpers import Linearizer +from synapse.storage.data_stores.main.events_worker import EventRedactBehaviour +from synapse.types import StateMap, UserID, get_domain_from_id +from synapse.util.async_helpers import Linearizer, concurrently_execute from synapse.util.distributor import user_joined_room from synapse.util.retryutils import NotRetryingDestination from synapse.visibility import filter_events_for_server @@ -72,6 +75,23 @@ from ._base import BaseHandler logger = logging.getLogger(__name__) +@attr.s +class _NewEventInfo: + """Holds information about a received event, ready for passing to _handle_new_events + + Attributes: + event: the received event + + state: the state at that event + + auth_events: the auth_event map for that event + """ + + event = attr.ib(type=EventBase) + state = attr.ib(type=Optional[Sequence[EventBase]], default=None) + auth_events = attr.ib(type=Optional[StateMap[EventBase]], default=None) + + def shortstr(iterable, maxitems=5): """If iterable has maxitems or fewer, return the stringification of a list containing those items. @@ -121,6 +141,7 @@ class FederationHandler(BaseHandler): self.pusher_pool = hs.get_pusherpool() self.spam_checker = hs.get_spam_checker() self.event_creation_handler = hs.get_event_creation_handler() + self._message_handler = hs.get_message_handler() self._server_notices_mxid = hs.config.server_notices_mxid self.config = hs.config self.http_client = hs.get_simple_http_client() @@ -141,8 +162,9 @@ class FederationHandler(BaseHandler): self.third_party_event_rules = hs.get_third_party_event_rules() - @defer.inlineCallbacks - def on_receive_pdu(self, origin, pdu, sent_to_us_directly=False): + self._ephemeral_messages_enabled = hs.config.enable_ephemeral_messages + + async def on_receive_pdu(self, origin, pdu, sent_to_us_directly=False) -> None: """ Process a PDU received via a federation /send/ transaction, or via backfill of missing prev_events @@ -152,17 +174,15 @@ class FederationHandler(BaseHandler): pdu (FrozenEvent): received PDU sent_to_us_directly (bool): True if this event was pushed to us; False if we pulled it as the result of a missing prev_event. - - Returns (Deferred): completes with None """ room_id = pdu.room_id event_id = pdu.event_id - logger.info("[%s %s] handling received PDU: %s", room_id, event_id, pdu) + logger.info("handling received PDU: %s", pdu) # We reprocess pdus when we have seen them only as outliers - existing = yield self.store.get_event( + existing = await self.store.get_event( event_id, allow_none=True, allow_rejected=True ) @@ -206,7 +226,7 @@ class FederationHandler(BaseHandler): # # Note that if we were never in the room then we would have already # dropped the event, since we wouldn't know the room version. - is_in_room = yield self.auth.check_host_in_room(room_id, self.server_name) + is_in_room = await self.auth.check_host_in_room(room_id, self.server_name) if not is_in_room: logger.info( "[%s %s] Ignoring PDU from %s as we're not in the room", @@ -217,25 +237,24 @@ class FederationHandler(BaseHandler): return None state = None - auth_chain = [] # Get missing pdus if necessary. if not pdu.internal_metadata.is_outlier(): # We only backfill backwards to the min depth. - min_depth = yield self.get_min_depth_for_context(pdu.room_id) + min_depth = await self.get_min_depth_for_context(pdu.room_id) logger.debug("[%s %s] min_depth: %d", room_id, event_id, min_depth) prevs = set(pdu.prev_event_ids()) - seen = yield self.store.have_seen_events(prevs) + seen = await self.store.have_seen_events(prevs) - if min_depth and pdu.depth < min_depth: + if min_depth is not None and pdu.depth < min_depth: # This is so that we don't notify the user about this # message, to work around the fact that some events will # reference really really old events we really don't want to # send to the clients. pdu.internal_metadata.outlier = True - elif min_depth and pdu.depth > min_depth: + elif min_depth is not None and pdu.depth > min_depth: missing_prevs = prevs - seen if sent_to_us_directly and missing_prevs: # If we're missing stuff, ensure we only fetch stuff one @@ -247,7 +266,7 @@ class FederationHandler(BaseHandler): len(missing_prevs), shortstr(missing_prevs), ) - with (yield self._room_pdu_linearizer.queue(pdu.room_id)): + with (await self._room_pdu_linearizer.queue(pdu.room_id)): logger.info( "[%s %s] Acquired room lock to fetch %d missing prev_events", room_id, @@ -255,13 +274,19 @@ class FederationHandler(BaseHandler): len(missing_prevs), ) - yield self._get_missing_events_for_pdu( - origin, pdu, prevs, min_depth - ) + try: + await self._get_missing_events_for_pdu( + origin, pdu, prevs, min_depth + ) + except Exception as e: + raise Exception( + "Error fetching missing prev_events for %s: %s" + % (event_id, e) + ) # Update the set of things we've seen after trying to # fetch the missing stuff - seen = yield self.store.have_seen_events(prevs) + seen = await self.store.have_seen_events(prevs) if not prevs - seen: logger.info( @@ -269,14 +294,6 @@ class FederationHandler(BaseHandler): room_id, event_id, ) - elif missing_prevs: - logger.info( - "[%s %s] Not recursively fetching %d missing prev_events: %s", - room_id, - event_id, - len(missing_prevs), - shortstr(missing_prevs), - ) if prevs - seen: # We've still not been able to get all of the prev_events for this event. @@ -321,18 +338,21 @@ class FederationHandler(BaseHandler): affected=pdu.event_id, ) + logger.info( + "Event %s is missing prev_events: calculating state for a " + "backwards extremity", + event_id, + ) + # Calculate the state after each of the previous events, and # resolve them to find the correct state at the current event. - auth_chains = set() event_map = {event_id: pdu} try: # Get the state of the events we know about - ours = yield self.state_store.get_state_groups_ids(room_id, seen) + ours = await self.state_store.get_state_groups_ids(room_id, seen) # state_maps is a list of mappings from (type, state_key) to event_id - state_maps = list( - ours.values() - ) # type: list[dict[tuple[str, str], str]] + state_maps = list(ours.values()) # type: list[StateMap[str]] # we don't need this any more, let's delete it. del ours @@ -341,44 +361,17 @@ class FederationHandler(BaseHandler): # know about for p in prevs - seen: logger.info( - "[%s %s] Requesting state at missing prev_event %s", - room_id, - event_id, - p, + "Requesting state at missing prev_event %s", event_id, ) - room_version = yield self.store.get_room_version(room_id) - with nested_logging_context(p): # note that if any of the missing prevs share missing state or # auth events, the requests to fetch those events are deduped # by the get_pdu_cache in federation_client. - ( - remote_state, - got_auth_chain, - ) = yield self.federation_client.get_state_for_room( - origin, room_id, p - ) - - # we want the state *after* p; get_state_for_room returns the - # state *before* p. - remote_event = yield self.federation_client.get_pdu( - [origin], p, room_version, outlier=True + (remote_state, _,) = await self._get_state_for_room( + origin, room_id, p, include_event_in_state=True ) - if remote_event is None: - raise Exception( - "Unable to get missing prev_event %s" % (p,) - ) - - if remote_event.is_state(): - remote_state.append(remote_event) - - # XXX hrm I'm not convinced that duplicate events will compare - # for equality, so I'm not sure this does what the author - # hoped. - auth_chains.update(got_auth_chain) - remote_state_map = { (x.type, x.state_key): x.event_id for x in remote_state } @@ -387,7 +380,9 @@ class FederationHandler(BaseHandler): for x in remote_state: event_map[x.event_id] = x - state_map = yield resolve_events_with_store( + room_version = await self.store.get_room_version(room_id) + state_map = await resolve_events_with_store( + room_id, room_version, state_maps, event_map, @@ -399,15 +394,14 @@ class FederationHandler(BaseHandler): # First though we need to fetch all the events that are in # state_map, so we can build up the state below. - evs = yield self.store.get_events( + evs = await self.store.get_events( list(state_map.values()), get_prev_content=False, - check_redacted=False, + redact_behaviour=EventRedactBehaviour.AS_IS, ) event_map.update(evs) state = [event_map[e] for e in six.itervalues(state_map)] - auth_chain = list(auth_chains) except Exception: logger.warning( "[%s %s] Error attempting to resolve state at missing " @@ -423,12 +417,9 @@ class FederationHandler(BaseHandler): affected=event_id, ) - yield self._process_received_pdu( - origin, pdu, state=state, auth_chain=auth_chain - ) + await self._process_received_pdu(origin, pdu, state=state) - @defer.inlineCallbacks - def _get_missing_events_for_pdu(self, origin, pdu, prevs, min_depth): + async def _get_missing_events_for_pdu(self, origin, pdu, prevs, min_depth): """ Args: origin (str): Origin of the pdu. Will be called to get the missing events @@ -440,12 +431,12 @@ class FederationHandler(BaseHandler): room_id = pdu.room_id event_id = pdu.event_id - seen = yield self.store.have_seen_events(prevs) + seen = await self.store.have_seen_events(prevs) if not prevs - seen: return - latest = yield self.store.get_latest_event_ids_in_room(room_id) + latest = await self.store.get_latest_event_ids_in_room(room_id) # We add the prev events that we have seen to the latest # list to ensure the remote server doesn't give them to us @@ -509,7 +500,7 @@ class FederationHandler(BaseHandler): # All that said: Let's try increasing the timout to 60s and see what happens. try: - missing_events = yield self.federation_client.get_missing_events( + missing_events = await self.federation_client.get_missing_events( origin, room_id, earliest_events_ids=list(latest), @@ -548,7 +539,7 @@ class FederationHandler(BaseHandler): ) with nested_logging_context(ev.event_id): try: - yield self.on_receive_pdu(origin, ev, sent_to_us_directly=False) + await self.on_receive_pdu(origin, ev, sent_to_us_directly=False) except FederationError as e: if e.code == 403: logger.warning( @@ -560,61 +551,159 @@ class FederationHandler(BaseHandler): else: raise - @defer.inlineCallbacks - def _process_received_pdu(self, origin, event, state, auth_chain): - """ Called when we have a new pdu. We need to do auth checks and put it - through the StateHandler. + async def _get_state_for_room( + self, + destination: str, + room_id: str, + event_id: str, + include_event_in_state: bool = False, + ) -> Tuple[List[EventBase], List[EventBase]]: + """Requests all of the room state at a given event from a remote homeserver. + + Args: + destination: The remote homeserver to query for the state. + room_id: The id of the room we're interested in. + event_id: The id of the event we want the state at. + include_event_in_state: if true, the event itself will be included in the + returned state event list. + + Returns: + A list of events in the state, possibly including the event itself, and + a list of events in the auth chain for the given event. """ - room_id = event.room_id - event_id = event.event_id + ( + state_event_ids, + auth_event_ids, + ) = await self.federation_client.get_room_state_ids( + destination, room_id, event_id=event_id + ) - logger.debug("[%s %s] Processing event: %s", room_id, event_id, event) + desired_events = set(state_event_ids + auth_event_ids) + + if include_event_in_state: + desired_events.add(event_id) + + event_map = await self._get_events_from_store_or_dest( + destination, room_id, desired_events + ) + + failed_to_fetch = desired_events - event_map.keys() + if failed_to_fetch: + logger.warning( + "Failed to fetch missing state/auth events for %s %s", + event_id, + failed_to_fetch, + ) - event_ids = set() - if state: - event_ids |= {e.event_id for e in state} - if auth_chain: - event_ids |= {e.event_id for e in auth_chain} + remote_state = [ + event_map[e_id] for e_id in state_event_ids if e_id in event_map + ] - seen_ids = yield self.store.have_seen_events(event_ids) + if include_event_in_state: + remote_event = event_map.get(event_id) + if not remote_event: + raise Exception("Unable to get missing prev_event %s" % (event_id,)) + if remote_event.is_state() and remote_event.rejected_reason is None: + remote_state.append(remote_event) - if state and auth_chain is not None: - # If we have any state or auth_chain given to us by the replication - # layer, then we should handle them (if we haven't before.) + auth_chain = [event_map[e_id] for e_id in auth_event_ids if e_id in event_map] + auth_chain.sort(key=lambda e: e.depth) - event_infos = [] + return remote_state, auth_chain - for e in itertools.chain(auth_chain, state): - if e.event_id in seen_ids: - continue - e.internal_metadata.outlier = True - auth_ids = e.auth_event_ids() - auth = { - (e.type, e.state_key): e - for e in auth_chain - if e.event_id in auth_ids or e.type == EventTypes.Create - } - event_infos.append({"event": e, "auth_events": auth}) - seen_ids.add(e.event_id) + async def _get_events_from_store_or_dest( + self, destination: str, room_id: str, event_ids: Iterable[str] + ) -> Dict[str, EventBase]: + """Fetch events from a remote destination, checking if we already have them. - logger.info( - "[%s %s] persisting newly-received auth/state events %s", + Persists any events we don't already have as outliers. + + If we fail to fetch any of the events, a warning will be logged, and the event + will be omitted from the result. Likewise, any events which turn out not to + be in the given room. + + Returns: + map from event_id to event + """ + fetched_events = await self.store.get_events(event_ids, allow_rejected=True) + + missing_events = set(event_ids) - fetched_events.keys() + + if missing_events: + logger.debug( + "Fetching unknown state/auth events %s for room %s", + missing_events, + room_id, + ) + + await self._get_events_and_persist( + destination=destination, room_id=room_id, events=missing_events + ) + + # we need to make sure we re-load from the database to get the rejected + # state correct. + fetched_events.update( + (await self.store.get_events(missing_events, allow_rejected=True)) + ) + + # check for events which were in the wrong room. + # + # this can happen if a remote server claims that the state or + # auth_events at an event in room A are actually events in room B + + bad_events = list( + (event_id, event.room_id) + for event_id, event in fetched_events.items() + if event.room_id != room_id + ) + + for bad_event_id, bad_room_id in bad_events: + # This is a bogus situation, but since we may only discover it a long time + # after it happened, we try our best to carry on, by just omitting the + # bad events from the returned auth/state set. + logger.warning( + "Remote server %s claims event %s in room %s is an auth/state " + "event in room %s", + destination, + bad_event_id, + bad_room_id, room_id, - event_id, - [e["event"].event_id for e in event_infos], ) - yield self._handle_new_events(origin, event_infos) + + del fetched_events[bad_event_id] + + return fetched_events + + async def _process_received_pdu( + self, origin: str, event: EventBase, state: Optional[Iterable[EventBase]], + ): + """ Called when we have a new pdu. We need to do auth checks and put it + through the StateHandler. + + Args: + origin: server sending the event + + event: event to be persisted + + state: Normally None, but if we are handling a gap in the graph + (ie, we are missing one or more prev_events), the resolved state at the + event + """ + room_id = event.room_id + event_id = event.event_id + + logger.debug("[%s %s] Processing event: %s", room_id, event_id, event) try: - context = yield self._handle_new_event(origin, event, state=state) + context = await self._handle_new_event(origin, event, state=state) except AuthError as e: raise FederationError("ERROR", e.code, e.msg, affected=event.event_id) - room = yield self.store.get_room(room_id) + room = await self.store.get_room(room_id) if not room: try: - yield self.store.store_room( + await self.store.store_room( room_id=room_id, room_creator_user_id="", is_public=False ) except StoreError: @@ -627,11 +716,11 @@ class FederationHandler(BaseHandler): # changing their profile info. newly_joined = True - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = await context.get_prev_state_ids() prev_state_id = prev_state_ids.get((event.type, event.state_key)) if prev_state_id: - prev_state = yield self.store.get_event( + prev_state = await self.store.get_event( prev_state_id, allow_none=True ) if prev_state and prev_state.membership == Membership.JOIN: @@ -639,11 +728,10 @@ class FederationHandler(BaseHandler): if newly_joined: user = UserID.from_string(event.state_key) - yield self.user_joined_room(user, room_id) + await self.user_joined_room(user, room_id) @log_function - @defer.inlineCallbacks - def backfill(self, dest, room_id, limit, extremities): + async def backfill(self, dest, room_id, limit, extremities): """ Trigger a backfill request to `dest` for the given `room_id` This will attempt to get more events from the remote. If the other side @@ -660,9 +748,7 @@ class FederationHandler(BaseHandler): if dest == self.server_name: raise SynapseError(400, "Can't backfill from self.") - room_version = yield self.store.get_room_version(room_id) - - events = yield self.federation_client.backfill( + events = await self.federation_client.backfill( dest, room_id, limit=limit, extremities=extremities ) @@ -677,7 +763,7 @@ class FederationHandler(BaseHandler): # self._sanity_check_event(ev) # Don't bother processing events we already have. - seen_events = yield self.store.have_events_in_timeline( + seen_events = await self.store.have_events_in_timeline( set(e.event_id for e in events) ) @@ -690,6 +776,9 @@ class FederationHandler(BaseHandler): event_ids = set(e.event_id for e in events) + # build a list of events whose prev_events weren't in the batch. + # (XXX: this will include events whose prev_events we already have; that doesn't + # sound right?) edges = [ev.event_id for ev in events if set(ev.prev_event_ids()) - event_ids] logger.info("backfill: Got %d events with %d edges", len(events), len(edges)) @@ -700,8 +789,11 @@ class FederationHandler(BaseHandler): state_events = {} events_to_state = {} for e_id in edges: - state, auth = yield self.federation_client.get_state_for_room( - destination=dest, room_id=room_id, event_id=e_id + state, auth = await self._get_state_for_room( + destination=dest, + room_id=room_id, + event_id=e_id, + include_event_in_state=False, ) auth_events.update({a.event_id: a for a in auth}) auth_events.update({s.event_id: s for s in state}) @@ -718,95 +810,11 @@ class FederationHandler(BaseHandler): auth_events.update( {e_id: event_map[e_id] for e_id in required_auth if e_id in event_map} ) - missing_auth = required_auth - set(auth_events) - failed_to_fetch = set() - - # Try and fetch any missing auth events from both DB and remote servers. - # We repeatedly do this until we stop finding new auth events. - while missing_auth - failed_to_fetch: - logger.info("Missing auth for backfill: %r", missing_auth) - ret_events = yield self.store.get_events(missing_auth - failed_to_fetch) - auth_events.update(ret_events) - - required_auth.update( - a_id for event in ret_events.values() for a_id in event.auth_event_ids() - ) - missing_auth = required_auth - set(auth_events) - - if missing_auth - failed_to_fetch: - logger.info( - "Fetching missing auth for backfill: %r", - missing_auth - failed_to_fetch, - ) - - results = yield make_deferred_yieldable( - defer.gatherResults( - [ - run_in_background( - self.federation_client.get_pdu, - [dest], - event_id, - room_version=room_version, - outlier=True, - timeout=10000, - ) - for event_id in missing_auth - failed_to_fetch - ], - consumeErrors=True, - ) - ).addErrback(unwrapFirstError) - auth_events.update({a.event_id: a for a in results if a}) - required_auth.update( - a_id - for event in results - if event - for a_id in event.auth_event_ids() - ) - missing_auth = required_auth - set(auth_events) - - failed_to_fetch = missing_auth - set(auth_events) - - seen_events = yield self.store.have_seen_events( - set(auth_events.keys()) | set(state_events.keys()) - ) - # We now have a chunk of events plus associated state and auth chain to - # persist. We do the persistence in two steps: - # 1. Auth events and state get persisted as outliers, plus the - # backward extremities get persisted (as non-outliers). - # 2. The rest of the events in the chunk get persisted one by one, as - # each one depends on the previous event for its state. - # - # The important thing is that events in the chunk get persisted as - # non-outliers, including when those events are also in the state or - # auth chain. Caution must therefore be taken to ensure that they are - # not accidentally marked as outliers. - - # Step 1a: persist auth events that *don't* appear in the chunk ev_infos = [] - for a in auth_events.values(): - # We only want to persist auth events as outliers that we haven't - # seen and aren't about to persist as part of the backfilled chunk. - if a.event_id in seen_events or a.event_id in event_map: - continue - a.internal_metadata.outlier = True - ev_infos.append( - { - "event": a, - "auth_events": { - ( - auth_events[a_id].type, - auth_events[a_id].state_key, - ): auth_events[a_id] - for a_id in a.auth_event_ids() - if a_id in auth_events - }, - } - ) - - # Step 1b: persist the events in the chunk we fetched state for (i.e. - # the backwards extremities) as non-outliers. + # Step 1: persist the events in the chunk we fetched state for (i.e. + # the backwards extremities), with custom auth events and state for e_id in events_to_state: # For paranoia we ensure that these events are marked as # non-outliers @@ -814,10 +822,10 @@ class FederationHandler(BaseHandler): assert not ev.internal_metadata.is_outlier() ev_infos.append( - { - "event": ev, - "state": events_to_state[e_id], - "auth_events": { + _NewEventInfo( + event=ev, + state=events_to_state[e_id], + auth_events={ ( auth_events[a_id].type, auth_events[a_id].state_key, @@ -825,10 +833,10 @@ class FederationHandler(BaseHandler): for a_id in ev.auth_event_ids() if a_id in auth_events }, - } + ) ) - yield self._handle_new_events(dest, ev_infos, backfilled=True) + await self._handle_new_events(dest, ev_infos, backfilled=True) # Step 2: Persist the rest of the events in the chunk one by one events.sort(key=lambda e: e.depth) @@ -844,16 +852,15 @@ class FederationHandler(BaseHandler): # We store these one at a time since each event depends on the # previous to work out the state. # TODO: We can probably do something more clever here. - yield self._handle_new_event(dest, event, backfilled=True) + await self._handle_new_event(dest, event, backfilled=True) return events - @defer.inlineCallbacks - def maybe_backfill(self, room_id, current_depth): + async def maybe_backfill(self, room_id, current_depth): """Checks the database to see if we should backfill before paginating, and if so do. """ - extremities = yield self.store.get_oldest_events_with_depth_in_room(room_id) + extremities = await self.store.get_oldest_events_with_depth_in_room(room_id) if not extremities: logger.debug("Not backfilling as no extremeties found.") @@ -885,15 +892,17 @@ class FederationHandler(BaseHandler): # state *before* the event, ignoring the special casing certain event # types have. - forward_events = yield self.store.get_successor_events(list(extremities)) + forward_events = await self.store.get_successor_events(list(extremities)) - extremities_events = yield self.store.get_events( - forward_events, check_redacted=False, get_prev_content=False + extremities_events = await self.store.get_events( + forward_events, + redact_behaviour=EventRedactBehaviour.AS_IS, + get_prev_content=False, ) # We set `check_history_visibility_only` as we might otherwise get false # positives from users having been erased. - filtered_extremities = yield filter_events_for_server( + filtered_extremities = await filter_events_for_server( self.storage, self.server_name, list(extremities_events.values()), @@ -923,7 +932,7 @@ class FederationHandler(BaseHandler): # First we try hosts that are already in the room # TODO: HEURISTIC ALERT. - curr_state = yield self.state_handler.get_current_state(room_id) + curr_state = await self.state_handler.get_current_state(room_id) def get_domains_from_state(state): """Get joined domains from state @@ -962,12 +971,11 @@ class FederationHandler(BaseHandler): domain for domain, depth in curr_domains if domain != self.server_name ] - @defer.inlineCallbacks - def try_backfill(domains): + async def try_backfill(domains): # TODO: Should we try multiple of these at a time? for dom in domains: try: - yield self.backfill( + await self.backfill( dom, room_id, limit=100, extremities=extremities ) # If this succeeded then we probably already have the @@ -998,7 +1006,7 @@ class FederationHandler(BaseHandler): return False - success = yield try_backfill(likely_domains) + success = await try_backfill(likely_domains) if success: return True @@ -1012,7 +1020,7 @@ class FederationHandler(BaseHandler): logger.debug("calling resolve_state_groups in _maybe_backfill") resolve = preserve_fn(self.state_handler.resolve_state_groups_for_events) - states = yield make_deferred_yieldable( + states = await make_deferred_yieldable( defer.gatherResults( [resolve(room_id, [e]) for e in event_ids], consumeErrors=True ) @@ -1022,7 +1030,7 @@ class FederationHandler(BaseHandler): # event_ids. states = dict(zip(event_ids, [s.state for s in states])) - state_map = yield self.store.get_events( + state_map = await self.store.get_events( [e_id for ids in itervalues(states) for e_id in itervalues(ids)], get_prev_content=False, ) @@ -1038,7 +1046,7 @@ class FederationHandler(BaseHandler): for e_id, _ in sorted_extremeties_tuple: likely_domains = get_domains_from_state(states[e_id]) - success = yield try_backfill( + success = await try_backfill( [dom for dom, _ in likely_domains if dom not in tried_domains] ) if success: @@ -1048,6 +1056,56 @@ class FederationHandler(BaseHandler): return False + async def _get_events_and_persist( + self, destination: str, room_id: str, events: Iterable[str] + ): + """Fetch the given events from a server, and persist them as outliers. + + Logs a warning if we can't find the given event. + """ + + room_version = await self.store.get_room_version(room_id) + + event_infos = [] + + async def get_event(event_id: str): + with nested_logging_context(event_id): + try: + event = await self.federation_client.get_pdu( + [destination], event_id, room_version, outlier=True, + ) + if event is None: + logger.warning( + "Server %s didn't return event %s", destination, event_id, + ) + return + + # recursively fetch the auth events for this event + auth_events = await self._get_events_from_store_or_dest( + destination, room_id, event.auth_event_ids() + ) + auth = {} + for auth_event_id in event.auth_event_ids(): + ae = auth_events.get(auth_event_id) + if ae: + auth[(ae.type, ae.state_key)] = ae + + event_infos.append(_NewEventInfo(event, None, auth)) + + except Exception as e: + logger.warning( + "Error fetching missing state/auth event %s: %s %s", + event_id, + type(e), + e, + ) + + await concurrently_execute(get_event, events, 5) + + await self._handle_new_events( + destination, event_infos, + ) + def _sanity_check_event(self, ev): """ Do some early sanity checks of a received event @@ -1187,7 +1245,7 @@ class FederationHandler(BaseHandler): # Check whether this room is the result of an upgrade of a room we already know # about. If so, migrate over user information predecessor = yield self.store.get_room_predecessor(room_id) - if not predecessor: + if not predecessor or not isinstance(predecessor.get("room_id"), str): return old_room_id = predecessor["room_id"] logger.debug( @@ -1215,8 +1273,7 @@ class FederationHandler(BaseHandler): return True - @defer.inlineCallbacks - def _handle_queued_pdus(self, room_queue): + async def _handle_queued_pdus(self, room_queue): """Process PDUs which got queued up while we were busy send_joining. Args: @@ -1232,7 +1289,7 @@ class FederationHandler(BaseHandler): p.room_id, ) with nested_logging_context(p.event_id): - yield self.on_receive_pdu(origin, p, sent_to_us_directly=True) + await self.on_receive_pdu(origin, p, sent_to_us_directly=True) except Exception as e: logger.warning( "Error handling queued PDU %s from %s: %s", p.event_id, origin, e @@ -1362,7 +1419,7 @@ class FederationHandler(BaseHandler): user = UserID.from_string(event.state_key) yield self.user_joined_room(user, event.room_id) - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() state_ids = list(prev_state_ids.values()) auth_chain = yield self.store.get_auth_chain(state_ids) @@ -1428,9 +1485,9 @@ class FederationHandler(BaseHandler): return event @defer.inlineCallbacks - def do_remotely_reject_invite(self, target_hosts, room_id, user_id): + def do_remotely_reject_invite(self, target_hosts, room_id, user_id, content): origin, event, event_format_version = yield self._make_and_verify_event( - target_hosts, room_id, user_id, "leave" + target_hosts, room_id, user_id, "leave", content=content ) # Mark as outlier as we don't have any state for this event; we're not # even in the room. @@ -1710,7 +1767,12 @@ class FederationHandler(BaseHandler): return context @defer.inlineCallbacks - def _handle_new_events(self, origin, event_infos, backfilled=False): + def _handle_new_events( + self, + origin: str, + event_infos: Iterable[_NewEventInfo], + backfilled: bool = False, + ): """Creates the appropriate contexts and persists events. The events should not depend on one another, e.g. this should be used to persist a bunch of outliers, but not a chunk of individual events that depend @@ -1720,14 +1782,14 @@ class FederationHandler(BaseHandler): """ @defer.inlineCallbacks - def prep(ev_info): - event = ev_info["event"] + def prep(ev_info: _NewEventInfo): + event = ev_info.event with nested_logging_context(suffix=event.event_id): res = yield self._prep_event( origin, event, - state=ev_info.get("state"), - auth_events=ev_info.get("auth_events"), + state=ev_info.state, + auth_events=ev_info.auth_events, backfilled=backfilled, ) return res @@ -1741,7 +1803,7 @@ class FederationHandler(BaseHandler): yield self.persist_events_and_notify( [ - (ev_info["event"], context) + (ev_info.event, context) for ev_info, context in zip(event_infos, contexts) ], backfilled=backfilled, @@ -1843,7 +1905,14 @@ class FederationHandler(BaseHandler): yield self.persist_events_and_notify([(event, new_event_context)]) @defer.inlineCallbacks - def _prep_event(self, origin, event, state, auth_events, backfilled): + def _prep_event( + self, + origin: str, + event: EventBase, + state: Optional[Iterable[EventBase]], + auth_events: Optional[StateMap[EventBase]], + backfilled: bool, + ): """ Args: @@ -1851,7 +1920,7 @@ class FederationHandler(BaseHandler): event: state: auth_events: - backfilled (bool) + backfilled: Returns: Deferred, which resolves to synapse.events.snapshot.EventContext @@ -1859,7 +1928,7 @@ class FederationHandler(BaseHandler): context = yield self.state_handler.compute_event_context(event, old_state=state) if not auth_events: - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() auth_events_ids = yield self.auth.compute_auth_events( event, prev_state_ids, for_verification=True ) @@ -1887,15 +1956,16 @@ class FederationHandler(BaseHandler): return context @defer.inlineCallbacks - def _check_for_soft_fail(self, event, state, backfilled): + def _check_for_soft_fail( + self, event: EventBase, state: Optional[Iterable[EventBase]], backfilled: bool + ): """Checks if we should soft fail the event, if so marks the event as such. Args: - event (FrozenEvent) - state (dict|None): The state at the event if we don't have all the - event's prev events - backfilled (bool): Whether the event is from backfill + event + state: The state at the event if we don't have all the event's prev events + backfilled: Whether the event is from backfill Returns: Deferred @@ -2040,8 +2110,10 @@ class FederationHandler(BaseHandler): auth_events (dict[(str, str)->synapse.events.EventBase]): Map from (event_type, state_key) to event - What we expect the event's auth_events to be, based on the event's - position in the dag. I think? maybe?? + Normally, our calculated auth_events based on the state of the room + at the event's position in the DAG, though occasionally (eg if the + event is an outlier), may be the auth events claimed by the remote + server. Also NB that this function adds entries to it. Returns: @@ -2091,35 +2163,35 @@ class FederationHandler(BaseHandler): origin (str): event (synapse.events.EventBase): context (synapse.events.snapshot.EventContext): + auth_events (dict[(str, str)->synapse.events.EventBase]): + Map from (event_type, state_key) to event + + Normally, our calculated auth_events based on the state of the room + at the event's position in the DAG, though occasionally (eg if the + event is an outlier), may be the auth events claimed by the remote + server. + + Also NB that this function adds entries to it. Returns: defer.Deferred[EventContext]: updated context """ event_auth_events = set(event.auth_event_ids()) - if event.is_state(): - event_key = (event.type, event.state_key) - else: - event_key = None - - # if the event's auth_events refers to events which are not in our - # calculated auth_events, we need to fetch those events from somewhere. - # - # we start by fetching them from the store, and then try calling /event_auth/. + # missing_auth is the set of the event's auth_events which we don't yet have + # in auth_events. missing_auth = event_auth_events.difference( e.event_id for e in auth_events.values() ) + # if we have missing events, we need to fetch those events from somewhere. + # + # we start by checking if they are in the store, and then try calling /event_auth/. if missing_auth: - # TODO: can we use store.have_seen_events here instead? - have_events = yield self.store.get_seen_events_with_rejections(missing_auth) - logger.debug("Got events %s from store", have_events) - missing_auth.difference_update(have_events.keys()) - else: - have_events = {} - - have_events.update({e.event_id: "" for e in auth_events.values()}) + have_events = yield self.store.have_seen_events(missing_auth) + logger.debug("Events %s are in the store", have_events) + missing_auth.difference_update(have_events) if missing_auth: # If we don't have all the auth events, we need to get them. @@ -2165,19 +2237,18 @@ class FederationHandler(BaseHandler): except AuthError: pass - have_events = yield self.store.get_seen_events_with_rejections( - event.auth_event_ids() - ) except Exception: - # FIXME: logger.exception("Failed to get auth chain") if event.internal_metadata.is_outlier(): + # XXX: given that, for an outlier, we'll be working with the + # event's *claimed* auth events rather than those we calculated: + # (a) is there any point in this test, since different_auth below will + # obviously be empty + # (b) alternatively, why don't we do it earlier? logger.info("Skipping auth_event fetch for outlier") return context - # FIXME: Assumes we have and stored all the state for all the - # prev_events different_auth = event_auth_events.difference( e.event_id for e in auth_events.values() ) @@ -2191,53 +2262,58 @@ class FederationHandler(BaseHandler): different_auth, ) - room_version = yield self.store.get_room_version(event.room_id) + # XXX: currently this checks for redactions but I'm not convinced that is + # necessary? + different_events = yield self.store.get_events_as_list(different_auth) - different_events = yield make_deferred_yieldable( - defer.gatherResults( - [ - run_in_background( - self.store.get_event, d, allow_none=True, allow_rejected=False - ) - for d in different_auth - if d in have_events and not have_events[d] - ], - consumeErrors=True, - ) - ).addErrback(unwrapFirstError) + for d in different_events: + if d.room_id != event.room_id: + logger.warning( + "Event %s refers to auth_event %s which is in a different room", + event.event_id, + d.event_id, + ) - if different_events: - local_view = dict(auth_events) - remote_view = dict(auth_events) - remote_view.update( - {(d.type, d.state_key): d for d in different_events if d} - ) + # don't attempt to resolve the claimed auth events against our own + # in this case: just use our own auth events. + # + # XXX: should we reject the event in this case? It feels like we should, + # but then shouldn't we also do so if we've failed to fetch any of the + # auth events? + return context - new_state = yield self.state_handler.resolve_events( - room_version, - [list(local_view.values()), list(remote_view.values())], - event, - ) + # now we state-resolve between our own idea of the auth events, and the remote's + # idea of them. - logger.info( - "After state res: updating auth_events with new state %s", - { - (d.type, d.state_key): d.event_id - for d in new_state.values() - if auth_events.get((d.type, d.state_key)) != d - }, - ) + local_state = auth_events.values() + remote_auth_events = dict(auth_events) + remote_auth_events.update({(d.type, d.state_key): d for d in different_events}) + remote_state = remote_auth_events.values() - auth_events.update(new_state) + room_version = yield self.store.get_room_version(event.room_id) + new_state = yield self.state_handler.resolve_events( + room_version, (local_state, remote_state), event + ) - context = yield self._update_context_for_auth_events( - event, context, auth_events, event_key - ) + logger.info( + "After state res: updating auth_events with new state %s", + { + (d.type, d.state_key): d.event_id + for d in new_state.values() + if auth_events.get((d.type, d.state_key)) != d + }, + ) + + auth_events.update(new_state) + + context = yield self._update_context_for_auth_events( + event, context, auth_events + ) return context @defer.inlineCallbacks - def _update_context_for_auth_events(self, event, context, auth_events, event_key): + def _update_context_for_auth_events(self, event, context, auth_events): """Update the state_ids in an event context after auth event resolution, storing the changes as a new state group. @@ -2246,24 +2322,27 @@ class FederationHandler(BaseHandler): context (synapse.events.snapshot.EventContext): initial event context - auth_events (dict[(str, str)->str]): Events to update in the event + auth_events (dict[(str, str)->EventBase]): Events to update in the event context. - event_key ((str, str)): (type, state_key) for the current event. - this will not be included in the current_state in the context. - Returns: Deferred[EventContext]: new event context """ + # exclude the state key of the new event from the current_state in the context. + if event.is_state(): + event_key = (event.type, event.state_key) + else: + event_key = None state_updates = { k: a.event_id for k, a in iteritems(auth_events) if k != event_key } - current_state_ids = yield context.get_current_state_ids(self.store) + + current_state_ids = yield context.get_current_state_ids() current_state_ids = dict(current_state_ids) current_state_ids.update(state_updates) - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() prev_state_ids = dict(prev_state_ids) prev_state_ids.update({k: a.event_id for k, a in iteritems(auth_events)}) @@ -2459,7 +2538,7 @@ class FederationHandler(BaseHandler): room_version, event_dict, event, context ) - EventValidator().validate_new(event) + EventValidator().validate_new(event, self.config) # We need to tell the transaction queue to send this out, even # though the sender isn't a local user. @@ -2547,7 +2626,7 @@ class FederationHandler(BaseHandler): event.content["third_party_invite"]["signed"]["token"], ) original_invite = None - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() original_invite_id = prev_state_ids.get(key) if original_invite_id: original_invite = yield self.store.get_event( @@ -2574,7 +2653,7 @@ class FederationHandler(BaseHandler): event, context = yield self.event_creation_handler.create_new_client_event( builder=builder ) - EventValidator().validate_new(event) + EventValidator().validate_new(event, self.config) return (event, context) @defer.inlineCallbacks @@ -2595,7 +2674,7 @@ class FederationHandler(BaseHandler): signed = event.content["third_party_invite"]["signed"] token = signed["token"] - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() invite_event_id = prev_state_ids.get((EventTypes.ThirdPartyInvite, token)) invite_event = None @@ -2708,6 +2787,11 @@ class FederationHandler(BaseHandler): event_and_contexts, backfilled=backfilled ) + if self._ephemeral_messages_enabled: + for (event, context) in event_and_contexts: + # If there's an expiry timestamp on the event, schedule its expiry. + self._message_handler.maybe_schedule_expiry(event) + if not backfilled: # Never notify for backfilled events for event, _ in event_and_contexts: yield self._notify_persisted_event(event, max_stream_id) @@ -2764,7 +2848,7 @@ class FederationHandler(BaseHandler): room_id=room_id, user_id=user.to_string(), change="joined" ) else: - return user_joined_room(self.distributor, user, room_id) + return defer.succeed(user_joined_room(self.distributor, user, room_id)) @defer.inlineCallbacks def get_room_complexity(self, remote_room_hosts, room_id): diff --git a/synapse/handlers/groups_local.py b/synapse/handlers/groups_local.py index 92fecbfc44..319565510f 100644 --- a/synapse/handlers/groups_local.py +++ b/synapse/handlers/groups_local.py @@ -130,6 +130,8 @@ class GroupsLocalHandler(object): res = yield self.transport_client.get_group_summary( get_domain_from_id(group_id), group_id, requester_user_id ) + except HttpResponseException as e: + raise e.to_synapse_error() except RequestSendFailed: raise SynapseError(502, "Failed to contact group server") @@ -190,6 +192,8 @@ class GroupsLocalHandler(object): res = yield self.transport_client.create_group( get_domain_from_id(group_id), group_id, user_id, content ) + except HttpResponseException as e: + raise e.to_synapse_error() except RequestSendFailed: raise SynapseError(502, "Failed to contact group server") @@ -231,6 +235,8 @@ class GroupsLocalHandler(object): res = yield self.transport_client.get_users_in_group( get_domain_from_id(group_id), group_id, requester_user_id ) + except HttpResponseException as e: + raise e.to_synapse_error() except RequestSendFailed: raise SynapseError(502, "Failed to contact group server") @@ -271,6 +277,8 @@ class GroupsLocalHandler(object): res = yield self.transport_client.join_group( get_domain_from_id(group_id), group_id, user_id, content ) + except HttpResponseException as e: + raise e.to_synapse_error() except RequestSendFailed: raise SynapseError(502, "Failed to contact group server") @@ -315,6 +323,8 @@ class GroupsLocalHandler(object): res = yield self.transport_client.accept_group_invite( get_domain_from_id(group_id), group_id, user_id, content ) + except HttpResponseException as e: + raise e.to_synapse_error() except RequestSendFailed: raise SynapseError(502, "Failed to contact group server") @@ -361,6 +371,8 @@ class GroupsLocalHandler(object): requester_user_id, content, ) + except HttpResponseException as e: + raise e.to_synapse_error() except RequestSendFailed: raise SynapseError(502, "Failed to contact group server") @@ -424,6 +436,8 @@ class GroupsLocalHandler(object): user_id, content, ) + except HttpResponseException as e: + raise e.to_synapse_error() except RequestSendFailed: raise SynapseError(502, "Failed to contact group server") @@ -460,6 +474,8 @@ class GroupsLocalHandler(object): bulk_result = yield self.transport_client.bulk_get_publicised_groups( get_domain_from_id(user_id), [user_id] ) + except HttpResponseException as e: + raise e.to_synapse_error() except RequestSendFailed: raise SynapseError(502, "Failed to contact group server") diff --git a/synapse/handlers/initial_sync.py b/synapse/handlers/initial_sync.py index 81dce96f4b..2e6755f19c 100644 --- a/synapse/handlers/initial_sync.py +++ b/synapse/handlers/initial_sync.py @@ -26,7 +26,7 @@ from synapse.streams.config import PaginationConfig from synapse.types import StreamToken, UserID from synapse.util import unwrapFirstError from synapse.util.async_helpers import concurrently_execute -from synapse.util.caches.snapshot_cache import SnapshotCache +from synapse.util.caches.response_cache import ResponseCache from synapse.visibility import filter_events_for_client from ._base import BaseHandler @@ -41,7 +41,7 @@ class InitialSyncHandler(BaseHandler): self.state = hs.get_state_handler() self.clock = hs.get_clock() self.validator = EventValidator() - self.snapshot_cache = SnapshotCache() + self.snapshot_cache = ResponseCache(hs, "initial_sync_cache") self._event_serializer = hs.get_event_client_serializer() self.storage = hs.get_storage() self.state_store = self.storage.state @@ -79,21 +79,17 @@ class InitialSyncHandler(BaseHandler): as_client_event, include_archived, ) - now_ms = self.clock.time_msec() - result = self.snapshot_cache.get(now_ms, key) - if result is not None: - return result - return self.snapshot_cache.set( - now_ms, + return self.snapshot_cache.wrap( key, - self._snapshot_all_rooms( - user_id, pagin_config, as_client_event, include_archived - ), + self._snapshot_all_rooms, + user_id, + pagin_config, + as_client_event, + include_archived, ) - @defer.inlineCallbacks - def _snapshot_all_rooms( + async def _snapshot_all_rooms( self, user_id=None, pagin_config=None, @@ -105,7 +101,7 @@ class InitialSyncHandler(BaseHandler): if include_archived: memberships.append(Membership.LEAVE) - room_list = yield self.store.get_rooms_for_user_where_membership_is( + room_list = await self.store.get_rooms_for_local_user_where_membership_is( user_id=user_id, membership_list=memberships ) @@ -113,33 +109,32 @@ class InitialSyncHandler(BaseHandler): rooms_ret = [] - now_token = yield self.hs.get_event_sources().get_current_token() + now_token = await self.hs.get_event_sources().get_current_token() presence_stream = self.hs.get_event_sources().sources["presence"] pagination_config = PaginationConfig(from_token=now_token) - presence, _ = yield presence_stream.get_pagination_rows( + presence, _ = await presence_stream.get_pagination_rows( user, pagination_config.get_source_config("presence"), None ) receipt_stream = self.hs.get_event_sources().sources["receipt"] - receipt, _ = yield receipt_stream.get_pagination_rows( + receipt, _ = await receipt_stream.get_pagination_rows( user, pagination_config.get_source_config("receipt"), None ) - tags_by_room = yield self.store.get_tags_for_user(user_id) + tags_by_room = await self.store.get_tags_for_user(user_id) - account_data, account_data_by_room = yield self.store.get_account_data_for_user( + account_data, account_data_by_room = await self.store.get_account_data_for_user( user_id ) - public_room_ids = yield self.store.get_public_room_ids() + public_room_ids = await self.store.get_public_room_ids() limit = pagin_config.limit if limit is None: limit = 10 - @defer.inlineCallbacks - def handle_room(event): + async def handle_room(event): d = { "room_id": event.room_id, "membership": event.membership, @@ -152,8 +147,8 @@ class InitialSyncHandler(BaseHandler): time_now = self.clock.time_msec() d["inviter"] = event.sender - invite_event = yield self.store.get_event(event.event_id) - d["invite"] = yield self._event_serializer.serialize_event( + invite_event = await self.store.get_event(event.event_id) + d["invite"] = await self._event_serializer.serialize_event( invite_event, time_now, as_client_event ) @@ -177,7 +172,7 @@ class InitialSyncHandler(BaseHandler): lambda states: states[event.event_id] ) - (messages, token), current_state = yield make_deferred_yieldable( + (messages, token), current_state = await make_deferred_yieldable( defer.gatherResults( [ run_in_background( @@ -191,7 +186,7 @@ class InitialSyncHandler(BaseHandler): ) ).addErrback(unwrapFirstError) - messages = yield filter_events_for_client( + messages = await filter_events_for_client( self.storage, user_id, messages ) @@ -201,7 +196,7 @@ class InitialSyncHandler(BaseHandler): d["messages"] = { "chunk": ( - yield self._event_serializer.serialize_events( + await self._event_serializer.serialize_events( messages, time_now=time_now, as_client_event=as_client_event ) ), @@ -209,7 +204,7 @@ class InitialSyncHandler(BaseHandler): "end": end_token.to_string(), } - d["state"] = yield self._event_serializer.serialize_events( + d["state"] = await self._event_serializer.serialize_events( current_state.values(), time_now=time_now, as_client_event=as_client_event, @@ -232,7 +227,7 @@ class InitialSyncHandler(BaseHandler): except Exception: logger.exception("Failed to get snapshot") - yield concurrently_execute(handle_room, room_list, 10) + await concurrently_execute(handle_room, room_list, 10) account_data_events = [] for account_data_type, content in account_data.items(): @@ -256,8 +251,7 @@ class InitialSyncHandler(BaseHandler): return ret - @defer.inlineCallbacks - def room_initial_sync(self, requester, room_id, pagin_config=None): + async def room_initial_sync(self, requester, room_id, pagin_config=None): """Capture the a snapshot of a room. If user is currently a member of the room this will be what is currently in the room. If the user left the room this will be what was in the room when they left. @@ -274,32 +268,32 @@ class InitialSyncHandler(BaseHandler): A JSON serialisable dict with the snapshot of the room. """ - blocked = yield self.store.is_room_blocked(room_id) + blocked = await self.store.is_room_blocked(room_id) if blocked: raise SynapseError(403, "This room has been blocked on this server") user_id = requester.user.to_string() - membership, member_event_id = yield self._check_in_room_or_world_readable( + membership, member_event_id = await self._check_in_room_or_world_readable( room_id, user_id ) is_peeking = member_event_id is None if membership == Membership.JOIN: - result = yield self._room_initial_sync_joined( + result = await self._room_initial_sync_joined( user_id, room_id, pagin_config, membership, is_peeking ) elif membership == Membership.LEAVE: - result = yield self._room_initial_sync_parted( + result = await self._room_initial_sync_parted( user_id, room_id, pagin_config, membership, member_event_id, is_peeking ) account_data_events = [] - tags = yield self.store.get_tags_for_room(user_id, room_id) + tags = await self.store.get_tags_for_room(user_id, room_id) if tags: account_data_events.append({"type": "m.tag", "content": {"tags": tags}}) - account_data = yield self.store.get_account_data_for_room(user_id, room_id) + account_data = await self.store.get_account_data_for_room(user_id, room_id) for account_data_type, content in account_data.items(): account_data_events.append({"type": account_data_type, "content": content}) @@ -307,11 +301,10 @@ class InitialSyncHandler(BaseHandler): return result - @defer.inlineCallbacks - def _room_initial_sync_parted( + async def _room_initial_sync_parted( self, user_id, room_id, pagin_config, membership, member_event_id, is_peeking ): - room_state = yield self.state_store.get_state_for_events([member_event_id]) + room_state = await self.state_store.get_state_for_events([member_event_id]) room_state = room_state[member_event_id] @@ -319,13 +312,13 @@ class InitialSyncHandler(BaseHandler): if limit is None: limit = 10 - stream_token = yield self.store.get_stream_token_for_event(member_event_id) + stream_token = await self.store.get_stream_token_for_event(member_event_id) - messages, token = yield self.store.get_recent_events_for_room( + messages, token = await self.store.get_recent_events_for_room( room_id, limit=limit, end_token=stream_token ) - messages = yield filter_events_for_client( + messages = await filter_events_for_client( self.storage, user_id, messages, is_peeking=is_peeking ) @@ -339,13 +332,13 @@ class InitialSyncHandler(BaseHandler): "room_id": room_id, "messages": { "chunk": ( - yield self._event_serializer.serialize_events(messages, time_now) + await self._event_serializer.serialize_events(messages, time_now) ), "start": start_token.to_string(), "end": end_token.to_string(), }, "state": ( - yield self._event_serializer.serialize_events( + await self._event_serializer.serialize_events( room_state.values(), time_now ) ), @@ -353,19 +346,18 @@ class InitialSyncHandler(BaseHandler): "receipts": [], } - @defer.inlineCallbacks - def _room_initial_sync_joined( + async def _room_initial_sync_joined( self, user_id, room_id, pagin_config, membership, is_peeking ): - current_state = yield self.state.get_current_state(room_id=room_id) + current_state = await self.state.get_current_state(room_id=room_id) # TODO: These concurrently time_now = self.clock.time_msec() - state = yield self._event_serializer.serialize_events( + state = await self._event_serializer.serialize_events( current_state.values(), time_now ) - now_token = yield self.hs.get_event_sources().get_current_token() + now_token = await self.hs.get_event_sources().get_current_token() limit = pagin_config.limit if pagin_config else None if limit is None: @@ -380,28 +372,26 @@ class InitialSyncHandler(BaseHandler): presence_handler = self.hs.get_presence_handler() - @defer.inlineCallbacks - def get_presence(): + async def get_presence(): # If presence is disabled, return an empty list if not self.hs.config.use_presence: return [] - states = yield presence_handler.get_states( + states = await presence_handler.get_states( [m.user_id for m in room_members], as_event=True ) return states - @defer.inlineCallbacks - def get_receipts(): - receipts = yield self.store.get_linearized_receipts_for_room( + async def get_receipts(): + receipts = await self.store.get_linearized_receipts_for_room( room_id, to_key=now_token.receipt_key ) if not receipts: receipts = [] return receipts - presence, receipts, (messages, token) = yield make_deferred_yieldable( + presence, receipts, (messages, token) = await make_deferred_yieldable( defer.gatherResults( [ run_in_background(get_presence), @@ -417,7 +407,7 @@ class InitialSyncHandler(BaseHandler): ).addErrback(unwrapFirstError) ) - messages = yield filter_events_for_client( + messages = await filter_events_for_client( self.storage, user_id, messages, is_peeking=is_peeking ) @@ -430,7 +420,7 @@ class InitialSyncHandler(BaseHandler): "room_id": room_id, "messages": { "chunk": ( - yield self._event_serializer.serialize_events(messages, time_now) + await self._event_serializer.serialize_events(messages, time_now) ), "start": start_token.to_string(), "end": end_token.to_string(), @@ -444,18 +434,17 @@ class InitialSyncHandler(BaseHandler): return ret - @defer.inlineCallbacks - def _check_in_room_or_world_readable(self, room_id, user_id): + async def _check_in_room_or_world_readable(self, room_id, user_id): try: # check_user_was_in_room will return the most recent membership # event for the user if: # * The user is a non-guest user, and was ever in the room # * The user is a guest user, and has joined the room # else it will throw. - member_event = yield self.auth.check_user_was_in_room(room_id, user_id) + member_event = await self.auth.check_user_was_in_room(room_id, user_id) return member_event.membership, member_event.event_id except AuthError: - visibility = yield self.state_handler.get_current_state( + visibility = await self.state_handler.get_current_state( room_id, EventTypes.RoomHistoryVisibility, "" ) if ( diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index d682dc2b7a..8ea3aca2f4 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -15,6 +15,7 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging +from typing import Optional from six import iteritems, itervalues, string_types @@ -22,9 +23,16 @@ from canonicaljson import encode_canonical_json, json from twisted.internet import defer from twisted.internet.defer import succeed +from twisted.internet.interfaces import IDelayedCall from synapse import event_auth -from synapse.api.constants import EventTypes, Membership, RelationTypes, UserTypes +from synapse.api.constants import ( + EventContentFields, + EventTypes, + Membership, + RelationTypes, + UserTypes, +) from synapse.api.errors import ( AuthError, Codes, @@ -38,8 +46,9 @@ from synapse.events.validator import EventValidator from synapse.logging.context import run_in_background from synapse.metrics.background_process_metrics import run_as_background_process from synapse.replication.http.send_event import ReplicationSendEventRestServlet +from synapse.storage.data_stores.main.events_worker import EventRedactBehaviour from synapse.storage.state import StateFilter -from synapse.types import RoomAlias, UserID, create_requester +from synapse.types import Collection, RoomAlias, UserID, create_requester from synapse.util.async_helpers import Linearizer from synapse.util.frozenutils import frozendict_json_encoder from synapse.util.metrics import measure_func @@ -62,6 +71,17 @@ class MessageHandler(object): self.storage = hs.get_storage() self.state_store = self.storage.state self._event_serializer = hs.get_event_client_serializer() + self._ephemeral_events_enabled = hs.config.enable_ephemeral_messages + self._is_worker_app = bool(hs.config.worker_app) + + # The scheduled call to self._expire_event. None if no call is currently + # scheduled. + self._scheduled_expiry = None # type: Optional[IDelayedCall] + + if not hs.config.worker_app: + run_as_background_process( + "_schedule_next_expiry", self._schedule_next_expiry + ) @defer.inlineCallbacks def get_room_data( @@ -138,7 +158,7 @@ class MessageHandler(object): raise NotFoundError("Can't find event for token %s" % (at_token,)) visible_events = yield filter_events_for_client( - self.storage, user_id, last_events + self.storage, user_id, last_events, apply_retention_policies=False ) event = last_events[0] @@ -225,6 +245,100 @@ class MessageHandler(object): for user_id, profile in iteritems(users_with_profile) } + def maybe_schedule_expiry(self, event): + """Schedule the expiry of an event if there's not already one scheduled, + or if the one running is for an event that will expire after the provided + timestamp. + + This function needs to invalidate the event cache, which is only possible on + the master process, and therefore needs to be run on there. + + Args: + event (EventBase): The event to schedule the expiry of. + """ + assert not self._is_worker_app + + expiry_ts = event.content.get(EventContentFields.SELF_DESTRUCT_AFTER) + if not isinstance(expiry_ts, int) or event.is_state(): + return + + # _schedule_expiry_for_event won't actually schedule anything if there's already + # a task scheduled for a timestamp that's sooner than the provided one. + self._schedule_expiry_for_event(event.event_id, expiry_ts) + + @defer.inlineCallbacks + def _schedule_next_expiry(self): + """Retrieve the ID and the expiry timestamp of the next event to be expired, + and schedule an expiry task for it. + + If there's no event left to expire, set _expiry_scheduled to None so that a + future call to save_expiry_ts can schedule a new expiry task. + """ + # Try to get the expiry timestamp of the next event to expire. + res = yield self.store.get_next_event_to_expire() + if res: + event_id, expiry_ts = res + self._schedule_expiry_for_event(event_id, expiry_ts) + + def _schedule_expiry_for_event(self, event_id, expiry_ts): + """Schedule an expiry task for the provided event if there's not already one + scheduled at a timestamp that's sooner than the provided one. + + Args: + event_id (str): The ID of the event to expire. + expiry_ts (int): The timestamp at which to expire the event. + """ + if self._scheduled_expiry: + # If the provided timestamp refers to a time before the scheduled time of the + # next expiry task, cancel that task and reschedule it for this timestamp. + next_scheduled_expiry_ts = self._scheduled_expiry.getTime() * 1000 + if expiry_ts < next_scheduled_expiry_ts: + self._scheduled_expiry.cancel() + else: + return + + # Figure out how many seconds we need to wait before expiring the event. + now_ms = self.clock.time_msec() + delay = (expiry_ts - now_ms) / 1000 + + # callLater doesn't support negative delays, so trim the delay to 0 if we're + # in that case. + if delay < 0: + delay = 0 + + logger.info("Scheduling expiry for event %s in %.3fs", event_id, delay) + + self._scheduled_expiry = self.clock.call_later( + delay, + run_as_background_process, + "_expire_event", + self._expire_event, + event_id, + ) + + @defer.inlineCallbacks + def _expire_event(self, event_id): + """Retrieve and expire an event that needs to be expired from the database. + + If the event doesn't exist in the database, log it and delete the expiry date + from the database (so that we don't try to expire it again). + """ + assert self._ephemeral_events_enabled + + self._scheduled_expiry = None + + logger.info("Expiring event %s", event_id) + + try: + # Expire the event if we know about it. This function also deletes the expiry + # date from the database in the same database transaction. + yield self.store.expire_event(event_id) + except Exception as e: + logger.error("Could not expire event %s: %r", event_id, e) + + # Schedule the expiry of the next event to expire. + yield self._schedule_next_expiry() + # The duration (in ms) after which rooms should be removed # `_rooms_to_exclude_from_dummy_event_insertion` (with the effect that we will try @@ -250,6 +364,8 @@ class EventCreationHandler(object): self.config = hs.config self.require_membership_for_aliases = hs.config.require_membership_for_aliases + self.room_invite_state_types = self.hs.config.room_invite_state_types + self.send_event_to_master = ReplicationSendEventRestServlet.make_client(hs) # This is only used to get at ratelimit function, and maybe_kick_guest_users @@ -295,6 +411,10 @@ class EventCreationHandler(object): 5 * 60 * 1000, ) + self._message_handler = hs.get_message_handler() + + self._ephemeral_events_enabled = hs.config.enable_ephemeral_messages + @defer.inlineCallbacks def create_event( self, @@ -302,7 +422,7 @@ class EventCreationHandler(object): event_dict, token_id=None, txn_id=None, - prev_events_and_hashes=None, + prev_event_ids: Optional[Collection[str]] = None, require_consent=True, ): """ @@ -319,10 +439,9 @@ class EventCreationHandler(object): token_id (str) txn_id (str) - prev_events_and_hashes (list[(str, dict[str, str], int)]|None): + prev_event_ids: the forward extremities to use as the prev_events for the - new event. For each event, a tuple of (event_id, hashes, depth) - where *hashes* is a map from algorithm to hash. + new event. If None, they will be requested from the database. @@ -378,9 +497,7 @@ class EventCreationHandler(object): builder.internal_metadata.txn_id = txn_id event, context = yield self.create_new_client_event( - builder=builder, - requester=requester, - prev_events_and_hashes=prev_events_and_hashes, + builder=builder, requester=requester, prev_event_ids=prev_event_ids, ) # In an ideal world we wouldn't need the second part of this condition. However, @@ -395,7 +512,7 @@ class EventCreationHandler(object): # federation as well as those created locally. As of room v3, aliases events # can be created by users that are not in the room, therefore we have to # tolerate them in event_auth.check(). - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() prev_event_id = prev_state_ids.get((EventTypes.Member, event.sender)) prev_event = ( yield self.store.get_event(prev_event_id, allow_none=True) @@ -417,7 +534,7 @@ class EventCreationHandler(object): 403, "You must be in the room to create an alias for it" ) - self.validator.validate_new(event) + self.validator.validate_new(event, self.config) return (event, context) @@ -545,7 +662,7 @@ class EventCreationHandler(object): If so, returns the version of the event in context. Otherwise, returns None. """ - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() prev_event_id = prev_state_ids.get((event.type, event.state_key)) if not prev_event_id: return @@ -594,7 +711,7 @@ class EventCreationHandler(object): @measure_func("create_new_client_event") @defer.inlineCallbacks def create_new_client_event( - self, builder, requester=None, prev_events_and_hashes=None + self, builder, requester=None, prev_event_ids: Optional[Collection[str]] = None ): """Create a new event for a local client @@ -603,10 +720,9 @@ class EventCreationHandler(object): requester (synapse.types.Requester|None): - prev_events_and_hashes (list[(str, dict[str, str], int)]|None): + prev_event_ids: the forward extremities to use as the prev_events for the - new event. For each event, a tuple of (event_id, hashes, depth) - where *hashes* is a map from algorithm to hash. + new event. If None, they will be requested from the database. @@ -614,27 +730,20 @@ class EventCreationHandler(object): Deferred[(synapse.events.EventBase, synapse.events.snapshot.EventContext)] """ - if prev_events_and_hashes is not None: - assert len(prev_events_and_hashes) <= 10, ( + if prev_event_ids is not None: + assert len(prev_event_ids) <= 10, ( "Attempting to create an event with %i prev_events" - % (len(prev_events_and_hashes),) + % (len(prev_event_ids),) ) else: - prev_events_and_hashes = yield self.store.get_prev_events_for_room( - builder.room_id - ) - - prev_events = [ - (event_id, prev_hashes) - for event_id, prev_hashes, _ in prev_events_and_hashes - ] + prev_event_ids = yield self.store.get_prev_events_for_room(builder.room_id) - event = yield builder.build(prev_event_ids=[p for p, _ in prev_events]) + event = yield builder.build(prev_event_ids=prev_event_ids) context = yield self.state.compute_event_context(event) if requester: context.app_service = requester.app_service - self.validator.validate_new(event) + self.validator.validate_new(event, self.config) # If this event is an annotation then we check that that the sender # can't annotate the same way twice (e.g. stops users from liking an @@ -756,7 +865,7 @@ class EventCreationHandler(object): if event.type == EventTypes.Redaction: original_event = yield self.store.get_event( event.redacts, - check_redacted=False, + redact_behaviour=EventRedactBehaviour.AS_IS, get_prev_content=False, allow_rejected=False, allow_none=True, @@ -794,12 +903,12 @@ class EventCreationHandler(object): def is_inviter_member_event(e): return e.type == EventTypes.Member and e.sender == event.sender - current_state_ids = yield context.get_current_state_ids(self.store) + current_state_ids = yield context.get_current_state_ids() state_to_include_ids = [ e_id for k, e_id in iteritems(current_state_ids) - if k[0] in self.hs.config.room_invite_state_types + if k[0] in self.room_invite_state_types or k == (EventTypes.Member, event.sender) ] @@ -833,7 +942,7 @@ class EventCreationHandler(object): if event.type == EventTypes.Redaction: original_event = yield self.store.get_event( event.redacts, - check_redacted=False, + redact_behaviour=EventRedactBehaviour.AS_IS, get_prev_content=False, allow_rejected=False, allow_none=True, @@ -847,7 +956,7 @@ class EventCreationHandler(object): if original_event.room_id != event.room_id: raise SynapseError(400, "Cannot redact event from a different room") - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() auth_events_ids = yield self.auth.compute_auth_events( event, prev_state_ids, for_verification=True ) @@ -869,7 +978,7 @@ class EventCreationHandler(object): event.internal_metadata.recheck_redaction = False if event.type == EventTypes.Create: - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() if prev_state_ids: raise AuthError(403, "Changing the room create event is forbidden") @@ -877,6 +986,10 @@ class EventCreationHandler(object): event, context=context ) + if self._ephemeral_events_enabled: + # If there's an expiry timestamp on the event, schedule its expiry. + self._message_handler.maybe_schedule_expiry(event) + yield self.pusher_pool.on_new_notifications(event_stream_id, max_stream_id) def _notify(): @@ -918,9 +1031,7 @@ class EventCreationHandler(object): # For each room we need to find a joined member we can use to send # the dummy event with. - prev_events_and_hashes = yield self.store.get_prev_events_for_room(room_id) - - latest_event_ids = (event_id for (event_id, _, _) in prev_events_and_hashes) + latest_event_ids = yield self.store.get_prev_events_for_room(room_id) members = yield self.state.get_current_users_in_room( room_id, latest_event_ids=latest_event_ids @@ -939,7 +1050,7 @@ class EventCreationHandler(object): "room_id": room_id, "sender": user_id, }, - prev_events_and_hashes=prev_events_and_hashes, + prev_event_ids=latest_event_ids, ) event.internal_metadata.proactively_send = False diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index 260a4351ca..00a6afc963 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -15,12 +15,15 @@ # limitations under the License. import logging +from six import iteritems + from twisted.internet import defer from twisted.python.failure import Failure from synapse.api.constants import EventTypes, Membership from synapse.api.errors import SynapseError from synapse.logging.context import run_in_background +from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage.state import StateFilter from synapse.types import RoomStreamToken from synapse.util.async_helpers import ReadWriteLock @@ -80,6 +83,109 @@ class PaginationHandler(object): self._purges_by_id = {} self._event_serializer = hs.get_event_client_serializer() + self._retention_default_max_lifetime = hs.config.retention_default_max_lifetime + + if hs.config.retention_enabled: + # Run the purge jobs described in the configuration file. + for job in hs.config.retention_purge_jobs: + self.clock.looping_call( + run_as_background_process, + job["interval"], + "purge_history_for_rooms_in_range", + self.purge_history_for_rooms_in_range, + job["shortest_max_lifetime"], + job["longest_max_lifetime"], + ) + + @defer.inlineCallbacks + def purge_history_for_rooms_in_range(self, min_ms, max_ms): + """Purge outdated events from rooms within the given retention range. + + If a default retention policy is defined in the server's configuration and its + 'max_lifetime' is within this range, also targets rooms which don't have a + retention policy. + + Args: + min_ms (int|None): Duration in milliseconds that define the lower limit of + the range to handle (exclusive). If None, it means that the range has no + lower limit. + max_ms (int|None): Duration in milliseconds that define the upper limit of + the range to handle (inclusive). If None, it means that the range has no + upper limit. + """ + # We want the storage layer to to include rooms with no retention policy in its + # return value only if a default retention policy is defined in the server's + # configuration and that policy's 'max_lifetime' is either lower (or equal) than + # max_ms or higher than min_ms (or both). + if self._retention_default_max_lifetime is not None: + include_null = True + + if min_ms is not None and min_ms >= self._retention_default_max_lifetime: + # The default max_lifetime is lower than (or equal to) min_ms. + include_null = False + + if max_ms is not None and max_ms < self._retention_default_max_lifetime: + # The default max_lifetime is higher than max_ms. + include_null = False + else: + include_null = False + + rooms = yield self.store.get_rooms_for_retention_period_in_range( + min_ms, max_ms, include_null + ) + + for room_id, retention_policy in iteritems(rooms): + if room_id in self._purges_in_progress_by_room: + logger.warning( + "[purge] not purging room %s as there's an ongoing purge running" + " for this room", + room_id, + ) + continue + + max_lifetime = retention_policy["max_lifetime"] + + if max_lifetime is None: + # If max_lifetime is None, it means that include_null equals True, + # therefore we can safely assume that there is a default policy defined + # in the server's configuration. + max_lifetime = self._retention_default_max_lifetime + + # Figure out what token we should start purging at. + ts = self.clock.time_msec() - max_lifetime + + stream_ordering = yield self.store.find_first_stream_ordering_after_ts(ts) + + r = yield self.store.get_room_event_after_stream_ordering( + room_id, stream_ordering, + ) + if not r: + logger.warning( + "[purge] purging events not possible: No event found " + "(ts %i => stream_ordering %i)", + ts, + stream_ordering, + ) + continue + + (stream, topo, _event_id) = r + token = "t%d-%d" % (topo, stream) + + purge_id = random_string(16) + + self._purges_by_id[purge_id] = PurgeStatus() + + logger.info( + "Starting purging events in room %s (purge_id %s)" % (room_id, purge_id) + ) + + # We want to purge everything, including local events, and to run the purge in + # the background so that it's not blocking any other operation apart from + # other purges in the same room. + run_as_background_process( + "_purge_history", self._purge_history, purge_id, room_id, token, True, + ) + def start_purge_history(self, room_id, token, delete_local_events=False): """Start off a history purge on a room. @@ -174,8 +280,7 @@ class PaginationHandler(object): await self.storage.purge_events.purge_room(room_id) - @defer.inlineCallbacks - def get_messages( + async def get_messages( self, requester, room_id=None, @@ -201,7 +306,7 @@ class PaginationHandler(object): room_token = pagin_config.from_token.room_key else: pagin_config.from_token = ( - yield self.hs.get_event_sources().get_current_token_for_pagination() + await self.hs.get_event_sources().get_current_token_for_pagination() ) room_token = pagin_config.from_token.room_key @@ -213,11 +318,11 @@ class PaginationHandler(object): source_config = pagin_config.get_source_config("room") - with (yield self.pagination_lock.read(room_id)): + with (await self.pagination_lock.read(room_id)): ( membership, member_event_id, - ) = yield self.auth.check_in_room_or_world_readable(room_id, user_id) + ) = await self.auth.check_in_room_or_world_readable(room_id, user_id) if source_config.direction == "b": # if we're going backwards, we might need to backfill. This @@ -225,7 +330,7 @@ class PaginationHandler(object): if room_token.topological: max_topo = room_token.topological else: - max_topo = yield self.store.get_max_topological_token( + max_topo = await self.store.get_max_topological_token( room_id, room_token.stream ) @@ -233,18 +338,18 @@ class PaginationHandler(object): # If they have left the room then clamp the token to be before # they left the room, to save the effort of loading from the # database. - leave_token = yield self.store.get_topological_token_for_event( + leave_token = await self.store.get_topological_token_for_event( member_event_id ) leave_token = RoomStreamToken.parse(leave_token) if leave_token.topological < max_topo: source_config.from_key = str(leave_token) - yield self.hs.get_handlers().federation_handler.maybe_backfill( + await self.hs.get_handlers().federation_handler.maybe_backfill( room_id, max_topo ) - events, next_key = yield self.store.paginate_room_events( + events, next_key = await self.store.paginate_room_events( room_id=room_id, from_key=source_config.from_key, to_key=source_config.to_key, @@ -259,7 +364,7 @@ class PaginationHandler(object): if event_filter: events = event_filter.filter(events) - events = yield filter_events_for_client( + events = await filter_events_for_client( self.storage, user_id, events, is_peeking=(member_event_id is None) ) @@ -279,19 +384,19 @@ class PaginationHandler(object): (EventTypes.Member, event.sender) for event in events ) - state_ids = yield self.state_store.get_state_ids_for_event( + state_ids = await self.state_store.get_state_ids_for_event( events[0].event_id, state_filter=state_filter ) if state_ids: - state = yield self.store.get_events(list(state_ids.values())) + state = await self.store.get_events(list(state_ids.values())) state = state.values() time_now = self.clock.time_msec() chunk = { "chunk": ( - yield self._event_serializer.serialize_events( + await self._event_serializer.serialize_events( events, time_now, as_client_event=as_client_event ) ), @@ -300,7 +405,7 @@ class PaginationHandler(object): } if state: - chunk["state"] = yield self._event_serializer.serialize_events( + chunk["state"] = await self._event_serializer.serialize_events( state, time_now, as_client_event=as_client_event ) diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index eda15bc623..202aa9294f 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -95,12 +95,7 @@ assert LAST_ACTIVE_GRANULARITY < IDLE_TIMER class PresenceHandler(object): - def __init__(self, hs): - """ - - Args: - hs (synapse.server.HomeServer): - """ + def __init__(self, hs: "synapse.server.HomeServer"): self.hs = hs self.is_mine = hs.is_mine self.is_mine_id = hs.is_mine_id @@ -230,7 +225,7 @@ class PresenceHandler(object): is some spurious presence changes that will self-correct. """ # If the DB pool has already terminated, don't try updating - if not self.hs.get_db_pool().running: + if not self.store.db.is_running(): return logger.info( diff --git a/synapse/handlers/profile.py b/synapse/handlers/profile.py index 1e5a4613c9..f9579d69ee 100644 --- a/synapse/handlers/profile.py +++ b/synapse/handlers/profile.py @@ -295,12 +295,16 @@ class BaseProfileHandler(BaseHandler): be found to be in any room the server is in, and therefore the query is denied. """ + # Implementation of MSC1301: don't allow looking up profiles if the # requester isn't in the same room as the target. We expect requester to # be None when this function is called outside of a profile query, e.g. # when building a membership event. In this case, we must allow the # lookup. - if not self.hs.config.require_auth_for_profile_requests or not requester: + if ( + not self.hs.config.limit_profile_requests_to_users_who_share_rooms + or not requester + ): return # Always allow the user to query their own profile. diff --git a/synapse/handlers/register.py b/synapse/handlers/register.py index 95806af41e..7ffc194f0c 100644 --- a/synapse/handlers/register.py +++ b/synapse/handlers/register.py @@ -20,13 +20,7 @@ from twisted.internet import defer from synapse import types from synapse.api.constants import MAX_USERID_LENGTH, LoginType -from synapse.api.errors import ( - AuthError, - Codes, - ConsentNotGivenError, - RegistrationError, - SynapseError, -) +from synapse.api.errors import AuthError, Codes, ConsentNotGivenError, SynapseError from synapse.config.server import is_threepid_reserved from synapse.http.servlet import assert_params_in_dict from synapse.replication.http.login import RegisterDeviceReplicationServlet @@ -165,7 +159,7 @@ class RegistrationHandler(BaseHandler): Returns: Deferred[str]: user_id Raises: - RegistrationError if there was a problem registering. + SynapseError if there was a problem registering. """ yield self.check_registration_ratelimit(address) @@ -174,7 +168,7 @@ class RegistrationHandler(BaseHandler): if password: password_hash = yield self._auth_handler.hash(password) - if localpart: + if localpart is not None: yield self.check_username(localpart, guest_access_token=guest_access_token) was_guest = guest_access_token is not None @@ -182,7 +176,7 @@ class RegistrationHandler(BaseHandler): if not was_guest: try: int(localpart) - raise RegistrationError( + raise SynapseError( 400, "Numeric user IDs are reserved for guest users." ) except ValueError: @@ -266,7 +260,7 @@ class RegistrationHandler(BaseHandler): } # Bind email to new account - yield self._register_email_threepid(user_id, threepid_dict, None, False) + yield self._register_email_threepid(user_id, threepid_dict, None) return user_id diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index e92b2eafd5..9f50196ea7 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -1,6 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2014 - 2016 OpenMarket Ltd -# Copyright 2018 New Vector Ltd +# Copyright 2018-2019 New Vector Ltd +# Copyright 2019 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -15,6 +16,7 @@ # limitations under the License. """Contains functions for performing events on rooms.""" + import itertools import logging import math @@ -30,7 +32,15 @@ from synapse.api.errors import AuthError, Codes, NotFoundError, StoreError, Syna from synapse.api.room_versions import KNOWN_ROOM_VERSIONS from synapse.http.endpoint import parse_and_validate_server_name from synapse.storage.state import StateFilter -from synapse.types import RoomAlias, RoomID, RoomStreamToken, StreamToken, UserID +from synapse.types import ( + Requester, + RoomAlias, + RoomID, + RoomStreamToken, + StateMap, + StreamToken, + UserID, +) from synapse.util import stringutils from synapse.util.async_helpers import Linearizer from synapse.util.caches.response_cache import ResponseCache @@ -183,7 +193,7 @@ class RoomCreationHandler(BaseHandler): requester, tombstone_event, tombstone_context ) - old_room_state = yield tombstone_context.get_current_state_ids(self.store) + old_room_state = yield tombstone_context.get_current_state_ids() # update any aliases yield self._move_aliases_to_new_room( @@ -198,22 +208,26 @@ class RoomCreationHandler(BaseHandler): # finally, shut down the PLs in the old room, and update them in the new # room. yield self._update_upgraded_room_pls( - requester, old_room_id, new_room_id, old_room_state + requester, old_room_id, new_room_id, old_room_state, ) return new_room_id @defer.inlineCallbacks def _update_upgraded_room_pls( - self, requester, old_room_id, new_room_id, old_room_state + self, + requester: Requester, + old_room_id: str, + new_room_id: str, + old_room_state: StateMap[str], ): """Send updated power levels in both rooms after an upgrade Args: - requester (synapse.types.Requester): the user requesting the upgrade - old_room_id (unicode): the id of the room to be replaced - new_room_id (unicode): the id of the replacement room - old_room_state (dict[tuple[str, str], str]): the state map for the old room + requester: the user requesting the upgrade + old_room_id: the id of the room to be replaced + new_room_id: the id of the replacement room + old_room_state: the state map for the old room Returns: Deferred @@ -270,7 +284,7 @@ class RoomCreationHandler(BaseHandler): except AuthError as e: logger.warning("Unable to update PLs in old room: %s", e) - logger.info("Setting correct PLs in new room") + logger.info("Setting correct PLs in new room to %s", old_room_pl_state.content) yield self.event_creation_handler.create_and_send_nonmember_event( requester, { @@ -298,7 +312,7 @@ class RoomCreationHandler(BaseHandler): tombstone_event_id (unicode|str): the ID of the tombstone event in the old room. Returns: - Deferred[None] + Deferred """ user_id = requester.user.to_string() @@ -333,6 +347,7 @@ class RoomCreationHandler(BaseHandler): (EventTypes.Encryption, ""), (EventTypes.ServerACL, ""), (EventTypes.RelatedGroups, ""), + (EventTypes.PowerLevels, ""), ) old_room_state_ids = yield self.store.get_filtered_current_state_ids( @@ -346,6 +361,36 @@ class RoomCreationHandler(BaseHandler): if old_event: initial_state[k] = old_event.content + # Resolve the minimum power level required to send any state event + # We will give the upgrading user this power level temporarily (if necessary) such that + # they are able to copy all of the state events over, then revert them back to their + # original power level afterwards in _update_upgraded_room_pls + + # Copy over user power levels now as this will not be possible with >100PL users once + # the room has been created + + power_levels = initial_state[(EventTypes.PowerLevels, "")] + + # Calculate the minimum power level needed to clone the room + event_power_levels = power_levels.get("events", {}) + state_default = power_levels.get("state_default", 0) + ban = power_levels.get("ban") + needed_power_level = max(state_default, ban, max(event_power_levels.values())) + + # Raise the requester's power level in the new room if necessary + current_power_level = power_levels["users"][user_id] + if current_power_level < needed_power_level: + # make sure we copy the event content rather than overwriting it. + # note that if frozen_dicts are enabled, `power_levels` will be a frozen + # dict so we can't just copy.deepcopy it. + + new_power_levels = {k: v for k, v in power_levels.items() if k != "users"} + new_power_levels["users"] = { + k: v for k, v in power_levels.get("users", {}).items() if k != user_id + } + new_power_levels["users"][user_id] = needed_power_level + initial_state[(EventTypes.PowerLevels, "")] = new_power_levels + yield self._send_events_for_new_room( requester, new_room_id, @@ -706,7 +751,7 @@ class RoomCreationHandler(BaseHandler): initial_state, creation_content, room_alias=None, - power_level_content_override=None, + power_level_content_override=None, # Doesn't apply when initial state has power level state event content creator_join_profile=None, ): def create(etype, content, **kwargs): @@ -874,9 +919,16 @@ class RoomContextHandler(object): room_id, event_id, before_limit, after_limit, event_filter ) + if event_filter: + results["events_before"] = event_filter.filter(results["events_before"]) + results["events_after"] = event_filter.filter(results["events_after"]) + results["events_before"] = yield filter_evts(results["events_before"]) results["events_after"] = yield filter_evts(results["events_after"]) - results["event"] = event + # filter_evts can return a pruned event in case the user is allowed to see that + # there's something there but not see the content, so use the event that's in + # `filtered` rather than the event we retrieved from the datastore. + results["event"] = filtered[0] if results["events_after"]: last_event_id = results["events_after"][-1].event_id @@ -902,7 +954,12 @@ class RoomContextHandler(object): state = yield self.state_store.get_state_for_events( [last_event_id], state_filter=state_filter ) - results["state"] = list(state[last_event_id].values()) + + state_events = list(state[last_event_id].values()) + if event_filter: + state_events = event_filter.filter(state_events) + + results["state"] = yield filter_evts(state_events) # We use a dummy token here as we only care about the room portion of # the token, which we replace. @@ -972,15 +1029,3 @@ class RoomEventSource(object): def get_current_key_for_room(self, room_id): return self.store.get_room_events_max_id(room_id) - - @defer.inlineCallbacks - def get_pagination_rows(self, user, config, key): - events, next_key = yield self.store.paginate_room_events( - room_id=key, - from_key=config.from_key, - to_key=config.to_key, - direction=config.direction, - limit=config.limit, - ) - - return (events, next_key) diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index 6cfee4b361..15e8aa5249 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -25,7 +25,7 @@ from twisted.internet import defer from synapse import types from synapse.api.constants import EventTypes, Membership from synapse.api.errors import AuthError, Codes, SynapseError -from synapse.types import RoomID, UserID +from synapse.types import Collection, RoomID, UserID from synapse.util.async_helpers import Linearizer from synapse.util.distributor import user_joined_room, user_left_room @@ -94,7 +94,9 @@ class RoomMemberHandler(object): raise NotImplementedError() @abc.abstractmethod - def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target): + def _remote_reject_invite( + self, requester, remote_room_hosts, room_id, target, content + ): """Attempt to reject an invite for a room this server is not in. If we fail to do so we locally mark the invite as rejected. @@ -104,6 +106,7 @@ class RoomMemberHandler(object): reject invite room_id (str) target (UserID): The user rejecting the invite + content (dict): The content for the rejection event Returns: Deferred[dict]: A dictionary to be returned to the client, may @@ -146,7 +149,7 @@ class RoomMemberHandler(object): target, room_id, membership, - prev_events_and_hashes, + prev_event_ids: Collection[str], txn_id=None, ratelimit=True, content=None, @@ -174,7 +177,7 @@ class RoomMemberHandler(object): }, token_id=requester.access_token_id, txn_id=txn_id, - prev_events_and_hashes=prev_events_and_hashes, + prev_event_ids=prev_event_ids, require_consent=require_consent, ) @@ -190,7 +193,7 @@ class RoomMemberHandler(object): requester, event, context, extra_users=[target], ratelimit=ratelimit ) - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() prev_member_event_id = prev_state_ids.get((EventTypes.Member, user_id), None) @@ -367,8 +370,7 @@ class RoomMemberHandler(object): if block_invite: raise SynapseError(403, "Invites have been disabled on this server") - prev_events_and_hashes = yield self.store.get_prev_events_for_room(room_id) - latest_event_ids = (event_id for (event_id, _, _) in prev_events_and_hashes) + latest_event_ids = yield self.store.get_prev_events_for_room(room_id) current_state_ids = yield self.state_handler.get_current_state_ids( room_id, latest_event_ids=latest_event_ids @@ -471,7 +473,7 @@ class RoomMemberHandler(object): # send the rejection to the inviter's HS. remote_room_hosts = remote_room_hosts + [inviter.domain] res = yield self._remote_reject_invite( - requester, remote_room_hosts, room_id, target + requester, remote_room_hosts, room_id, target, content, ) return res @@ -482,7 +484,7 @@ class RoomMemberHandler(object): membership=effective_membership_state, txn_id=txn_id, ratelimit=ratelimit, - prev_events_and_hashes=prev_events_and_hashes, + prev_event_ids=latest_event_ids, content=content, require_consent=require_consent, ) @@ -504,6 +506,8 @@ class RoomMemberHandler(object): Returns: Deferred """ + logger.info("Transferring room state from %s to %s", old_room_id, room_id) + # Find all local users that were in the old room and copy over each user's state users = yield self.store.get_users_in_room(old_room_id) yield self.copy_user_state_on_room_upgrade(old_room_id, room_id, users) @@ -598,7 +602,7 @@ class RoomMemberHandler(object): if prev_event is not None: return - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() if event.membership == Membership.JOIN: if requester.is_guest: guest_can_join = yield self._can_guest_join(prev_state_ids) @@ -686,7 +690,7 @@ class RoomMemberHandler(object): @defer.inlineCallbacks def _get_inviter(self, user_id, room_id): - invite = yield self.store.get_invite_for_user_in_room( + invite = yield self.store.get_invite_for_local_user_in_room( user_id=user_id, room_id=room_id ) if invite: @@ -971,13 +975,15 @@ class RoomMemberMasterHandler(RoomMemberHandler): ) @defer.inlineCallbacks - def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target): + def _remote_reject_invite( + self, requester, remote_room_hosts, room_id, target, content + ): """Implements RoomMemberHandler._remote_reject_invite """ fed_handler = self.federation_handler try: ret = yield fed_handler.do_remotely_reject_invite( - remote_room_hosts, room_id, target.to_string() + remote_room_hosts, room_id, target.to_string(), content=content, ) return ret except Exception as e: diff --git a/synapse/handlers/room_member_worker.py b/synapse/handlers/room_member_worker.py index 75e96ae1a2..69be86893b 100644 --- a/synapse/handlers/room_member_worker.py +++ b/synapse/handlers/room_member_worker.py @@ -55,7 +55,9 @@ class RoomMemberWorkerHandler(RoomMemberHandler): return ret - def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target): + def _remote_reject_invite( + self, requester, remote_room_hosts, room_id, target, content + ): """Implements RoomMemberHandler._remote_reject_invite """ return self._remote_reject_client( @@ -63,6 +65,7 @@ class RoomMemberWorkerHandler(RoomMemberHandler): remote_room_hosts=remote_room_hosts, room_id=room_id, user_id=target.to_string(), + content=content, ) def _user_joined_room(self, target, room_id): diff --git a/synapse/handlers/saml_handler.py b/synapse/handlers/saml_handler.py index cc9e6b9bd0..107f97032b 100644 --- a/synapse/handlers/saml_handler.py +++ b/synapse/handlers/saml_handler.py @@ -13,20 +13,37 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging +import re +from typing import Tuple import attr import saml2 +import saml2.response from saml2.client import Saml2Client from synapse.api.errors import SynapseError +from synapse.config import ConfigError from synapse.http.servlet import parse_string +from synapse.module_api import ModuleApi from synapse.rest.client.v1.login import SSOAuthHandler -from synapse.types import UserID, map_username_to_mxid_localpart +from synapse.types import ( + UserID, + map_username_to_mxid_localpart, + mxid_localpart_allowed_characters, +) from synapse.util.async_helpers import Linearizer logger = logging.getLogger(__name__) +@attr.s +class Saml2SessionData: + """Data we track about SAML2 sessions""" + + # time the session was created, in milliseconds + creation_time = attr.ib() + + class SamlHandler: def __init__(self, hs): self._saml_client = Saml2Client(hs.config.saml2_sp_config) @@ -37,11 +54,15 @@ class SamlHandler: self._datastore = hs.get_datastore() self._hostname = hs.hostname self._saml2_session_lifetime = hs.config.saml2_session_lifetime - self._mxid_source_attribute = hs.config.saml2_mxid_source_attribute self._grandfathered_mxid_source_attribute = ( hs.config.saml2_grandfathered_mxid_source_attribute ) - self._mxid_mapper = hs.config.saml2_mxid_mapper + + # plugin to do custom mapping from saml response to mxid + self._user_mapping_provider = hs.config.saml2_user_mapping_provider_class( + hs.config.saml2_user_mapping_provider_config, + ModuleApi(hs, hs.get_auth_handler()), + ) # identifier for the external_ids table self._auth_provider_id = "saml" @@ -93,10 +114,10 @@ class SamlHandler: # the dict. self.expire_sessions() - user_id = await self._map_saml_response_to_user(resp_bytes) + user_id = await self._map_saml_response_to_user(resp_bytes, relay_state) self._sso_auth_handler.complete_sso_login(user_id, request, relay_state) - async def _map_saml_response_to_user(self, resp_bytes): + async def _map_saml_response_to_user(self, resp_bytes, client_redirect_url): try: saml2_auth = self._saml_client.parse_authn_request_response( resp_bytes, @@ -118,22 +139,10 @@ class SamlHandler: remote_user_id = saml2_auth.ava["uid"][0] except KeyError: logger.warning("SAML2 response lacks a 'uid' attestation") - raise SynapseError(400, "uid not in SAML2 response") - - try: - mxid_source = saml2_auth.ava[self._mxid_source_attribute][0] - except KeyError: - logger.warning( - "SAML2 response lacks a '%s' attestation", self._mxid_source_attribute - ) - raise SynapseError( - 400, "%s not in SAML2 response" % (self._mxid_source_attribute,) - ) + raise SynapseError(400, "'uid' not in SAML2 response") self._outstanding_requests_dict.pop(saml2_auth.in_response_to, None) - displayName = saml2_auth.ava.get("displayName", [None])[0] - with (await self._mapping_lock.queue(self._auth_provider_id)): # first of all, check if we already have a mapping for this user logger.info( @@ -173,22 +182,48 @@ class SamlHandler: ) return registered_user_id - # figure out a new mxid for this user - base_mxid_localpart = self._mxid_mapper(mxid_source) + # Map saml response to user attributes using the configured mapping provider + for i in range(1000): + attribute_dict = self._user_mapping_provider.saml_response_to_user_attributes( + saml2_auth, i, client_redirect_url=client_redirect_url, + ) - suffix = 0 - while True: - localpart = base_mxid_localpart + (str(suffix) if suffix else "") + logger.debug( + "Retrieved SAML attributes from user mapping provider: %s " + "(attempt %d)", + attribute_dict, + i, + ) + + localpart = attribute_dict.get("mxid_localpart") + if not localpart: + logger.error( + "SAML mapping provider plugin did not return a " + "mxid_localpart object" + ) + raise SynapseError(500, "Error parsing SAML2 response") + + displayname = attribute_dict.get("displayname") + + # Check if this mxid already exists if not await self._datastore.get_users_by_id_case_insensitive( UserID(localpart, self._hostname).to_string() ): + # This mxid is free break - suffix += 1 - logger.info("Allocating mxid for new user with localpart %s", localpart) + else: + # Unable to generate a username in 1000 iterations + # Break and return error to the user + raise SynapseError( + 500, "Unable to generate a Matrix ID from the SAML response" + ) + + logger.info("Mapped SAML user to local part %s", localpart) registered_user_id = await self._registration_handler.register_user( - localpart=localpart, default_display_name=displayName + localpart=localpart, default_display_name=displayname ) + await self._datastore.record_user_external_id( self._auth_provider_id, remote_user_id, registered_user_id ) @@ -205,9 +240,126 @@ class SamlHandler: del self._outstanding_requests_dict[reqid] +DOT_REPLACE_PATTERN = re.compile( + ("[^%s]" % (re.escape("".join(mxid_localpart_allowed_characters)),)) +) + + +def dot_replace_for_mxid(username: str) -> str: + username = username.lower() + username = DOT_REPLACE_PATTERN.sub(".", username) + + # regular mxids aren't allowed to start with an underscore either + username = re.sub("^_", "", username) + return username + + +MXID_MAPPER_MAP = { + "hexencode": map_username_to_mxid_localpart, + "dotreplace": dot_replace_for_mxid, +} + + @attr.s -class Saml2SessionData: - """Data we track about SAML2 sessions""" +class SamlConfig(object): + mxid_source_attribute = attr.ib() + mxid_mapper = attr.ib() - # time the session was created, in milliseconds - creation_time = attr.ib() + +class DefaultSamlMappingProvider(object): + __version__ = "0.0.1" + + def __init__(self, parsed_config: SamlConfig, module_api: ModuleApi): + """The default SAML user mapping provider + + Args: + parsed_config: Module configuration + module_api: module api proxy + """ + self._mxid_source_attribute = parsed_config.mxid_source_attribute + self._mxid_mapper = parsed_config.mxid_mapper + + def saml_response_to_user_attributes( + self, + saml_response: saml2.response.AuthnResponse, + failures: int, + client_redirect_url: str, + ) -> dict: + """Maps some text from a SAML response to attributes of a new user + + Args: + saml_response: A SAML auth response object + + failures: How many times a call to this function with this + saml_response has resulted in a failure + + client_redirect_url: where the client wants to redirect to + + Returns: + dict: A dict containing new user attributes. Possible keys: + * mxid_localpart (str): Required. The localpart of the user's mxid + * displayname (str): The displayname of the user + """ + try: + mxid_source = saml_response.ava[self._mxid_source_attribute][0] + except KeyError: + logger.warning( + "SAML2 response lacks a '%s' attestation", self._mxid_source_attribute, + ) + raise SynapseError( + 400, "%s not in SAML2 response" % (self._mxid_source_attribute,) + ) + + # Use the configured mapper for this mxid_source + base_mxid_localpart = self._mxid_mapper(mxid_source) + + # Append suffix integer if last call to this function failed to produce + # a usable mxid + localpart = base_mxid_localpart + (str(failures) if failures else "") + + # Retrieve the display name from the saml response + # If displayname is None, the mxid_localpart will be used instead + displayname = saml_response.ava.get("displayName", [None])[0] + + return { + "mxid_localpart": localpart, + "displayname": displayname, + } + + @staticmethod + def parse_config(config: dict) -> SamlConfig: + """Parse the dict provided by the homeserver's config + Args: + config: A dictionary containing configuration options for this provider + Returns: + SamlConfig: A custom config object for this module + """ + # Parse config options and use defaults where necessary + mxid_source_attribute = config.get("mxid_source_attribute", "uid") + mapping_type = config.get("mxid_mapping", "hexencode") + + # Retrieve the associating mapping function + try: + mxid_mapper = MXID_MAPPER_MAP[mapping_type] + except KeyError: + raise ConfigError( + "saml2_config.user_mapping_provider.config: '%s' is not a valid " + "mxid_mapping value" % (mapping_type,) + ) + + return SamlConfig(mxid_source_attribute, mxid_mapper) + + @staticmethod + def get_saml_attributes(config: SamlConfig) -> Tuple[set, set]: + """Returns the required attributes of a SAML + + Args: + config: A SamlConfig object containing configuration params for this provider + + Returns: + tuple[set,set]: The first set equates to the saml auth response + attributes that are required for the module to function, whereas the + second set consists of those attributes which can be used if + available, but are not necessary + """ + return {"uid", config.mxid_source_attribute}, {"displayName"} diff --git a/synapse/handlers/search.py b/synapse/handlers/search.py index 56ed262a1f..110097eab9 100644 --- a/synapse/handlers/search.py +++ b/synapse/handlers/search.py @@ -21,7 +21,7 @@ from unpaddedbase64 import decode_base64, encode_base64 from twisted.internet import defer from synapse.api.constants import EventTypes, Membership -from synapse.api.errors import SynapseError +from synapse.api.errors import NotFoundError, SynapseError from synapse.api.filtering import Filter from synapse.storage.state import StateFilter from synapse.visibility import filter_events_for_client @@ -37,6 +37,7 @@ class SearchHandler(BaseHandler): self._event_serializer = hs.get_event_client_serializer() self.storage = hs.get_storage() self.state_store = self.storage.state + self.auth = hs.get_auth() @defer.inlineCallbacks def get_old_rooms_from_upgraded_room(self, room_id): @@ -53,23 +54,38 @@ class SearchHandler(BaseHandler): room_id (str): id of the room to search through. Returns: - Deferred[iterable[unicode]]: predecessor room ids + Deferred[iterable[str]]: predecessor room ids """ historical_room_ids = [] - while True: - predecessor = yield self.store.get_room_predecessor(room_id) + # The initial room must have been known for us to get this far + predecessor = yield self.store.get_room_predecessor(room_id) - # If no predecessor, assume we've hit a dead end + while True: if not predecessor: + # We have reached the end of the chain of predecessors + break + + if not isinstance(predecessor.get("room_id"), str): + # This predecessor object is malformed. Exit here + break + + predecessor_room_id = predecessor["room_id"] + + # Don't add it to the list until we have checked that we are in the room + try: + next_predecessor_room = yield self.store.get_room_predecessor( + predecessor_room_id + ) + except NotFoundError: + # The predecessor is not a known room, so we are done here break - # Add predecessor's room ID - historical_room_ids.append(predecessor["room_id"]) + historical_room_ids.append(predecessor_room_id) - # Scan through the old room for further predecessors - room_id = predecessor["room_id"] + # And repeat + predecessor = next_predecessor_room return historical_room_ids @@ -163,7 +179,7 @@ class SearchHandler(BaseHandler): search_filter = Filter(filter_dict) # TODO: Search through left rooms too - rooms = yield self.store.get_rooms_for_user_where_membership_is( + rooms = yield self.store.get_rooms_for_local_user_where_membership_is( user.to_string(), membership_list=[Membership.JOIN], # membership_list=[Membership.JOIN, Membership.LEAVE, Membership.Ban], diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index b536d410e5..cd95f85e3f 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -22,8 +22,6 @@ from six import iteritems, itervalues from prometheus_client import Counter -from twisted.internet import defer - from synapse.api.constants import EventTypes, Membership from synapse.logging.context import LoggingContext from synapse.push.clientformat import format_push_rules_for_user @@ -241,8 +239,7 @@ class SyncHandler(object): expiry_ms=LAZY_LOADED_MEMBERS_CACHE_MAX_AGE, ) - @defer.inlineCallbacks - def wait_for_sync_for_user( + async def wait_for_sync_for_user( self, sync_config, since_token=None, timeout=0, full_state=False ): """Get the sync for a client if we have new data for it now. Otherwise @@ -255,9 +252,9 @@ class SyncHandler(object): # not been exceeded (if not part of the group by this point, almost certain # auth_blocking will occur) user_id = sync_config.user.to_string() - yield self.auth.check_auth_blocking(user_id) + await self.auth.check_auth_blocking(user_id) - res = yield self.response_cache.wrap( + res = await self.response_cache.wrap( sync_config.request_key, self._wait_for_sync_for_user, sync_config, @@ -267,8 +264,9 @@ class SyncHandler(object): ) return res - @defer.inlineCallbacks - def _wait_for_sync_for_user(self, sync_config, since_token, timeout, full_state): + async def _wait_for_sync_for_user( + self, sync_config, since_token, timeout, full_state + ): if since_token is None: sync_type = "initial_sync" elif full_state: @@ -283,7 +281,7 @@ class SyncHandler(object): 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. - result = yield self.current_sync_for_user( + result = await self.current_sync_for_user( sync_config, since_token, full_state=full_state ) else: @@ -291,7 +289,7 @@ class SyncHandler(object): def current_sync_callback(before_token, after_token): return self.current_sync_for_user(sync_config, since_token) - result = yield self.notifier.wait_for_events( + result = await self.notifier.wait_for_events( sync_config.user.to_string(), timeout, current_sync_callback, @@ -314,15 +312,13 @@ class SyncHandler(object): """ return self.generate_sync_result(sync_config, since_token, full_state) - @defer.inlineCallbacks - def push_rules_for_user(self, user): + async def push_rules_for_user(self, user): user_id = user.to_string() - rules = yield self.store.get_push_rules_for_user(user_id) + rules = await self.store.get_push_rules_for_user(user_id) rules = format_push_rules_for_user(user, rules) return rules - @defer.inlineCallbacks - def ephemeral_by_room(self, sync_result_builder, now_token, since_token=None): + async def ephemeral_by_room(self, sync_result_builder, now_token, since_token=None): """Get the ephemeral events for each room the user is in Args: sync_result_builder(SyncResultBuilder) @@ -343,7 +339,7 @@ class SyncHandler(object): room_ids = sync_result_builder.joined_room_ids typing_source = self.event_sources.sources["typing"] - typing, typing_key = yield typing_source.get_new_events( + typing, typing_key = await typing_source.get_new_events( user=sync_config.user, from_key=typing_key, limit=sync_config.filter_collection.ephemeral_limit(), @@ -365,7 +361,7 @@ class SyncHandler(object): receipt_key = since_token.receipt_key if since_token else "0" receipt_source = self.event_sources.sources["receipt"] - receipts, receipt_key = yield receipt_source.get_new_events( + receipts, receipt_key = await receipt_source.get_new_events( user=sync_config.user, from_key=receipt_key, limit=sync_config.filter_collection.ephemeral_limit(), @@ -382,8 +378,7 @@ class SyncHandler(object): return now_token, ephemeral_by_room - @defer.inlineCallbacks - def _load_filtered_recents( + async def _load_filtered_recents( self, room_id, sync_config, @@ -415,10 +410,10 @@ class SyncHandler(object): # ensure that we always include current state in the timeline current_state_ids = frozenset() if any(e.is_state() for e in recents): - current_state_ids = yield self.state.get_current_state_ids(room_id) + current_state_ids = await self.state.get_current_state_ids(room_id) current_state_ids = frozenset(itervalues(current_state_ids)) - recents = yield filter_events_for_client( + recents = await filter_events_for_client( self.storage, sync_config.user.to_string(), recents, @@ -449,14 +444,14 @@ class SyncHandler(object): # Otherwise, we want to return the last N events in the room # in toplogical ordering. if since_key: - events, end_key = yield self.store.get_room_events_stream_for_room( + events, end_key = await self.store.get_room_events_stream_for_room( room_id, limit=load_limit + 1, from_key=since_key, to_key=end_key, ) else: - events, end_key = yield self.store.get_recent_events_for_room( + events, end_key = await self.store.get_recent_events_for_room( room_id, limit=load_limit + 1, end_token=end_key ) loaded_recents = sync_config.filter_collection.filter_room_timeline( @@ -468,10 +463,10 @@ class SyncHandler(object): # ensure that we always include current state in the timeline current_state_ids = frozenset() if any(e.is_state() for e in loaded_recents): - current_state_ids = yield self.state.get_current_state_ids(room_id) + current_state_ids = await self.state.get_current_state_ids(room_id) current_state_ids = frozenset(itervalues(current_state_ids)) - loaded_recents = yield filter_events_for_client( + loaded_recents = await filter_events_for_client( self.storage, sync_config.user.to_string(), loaded_recents, @@ -498,8 +493,7 @@ class SyncHandler(object): limited=limited or newly_joined_room, ) - @defer.inlineCallbacks - def get_state_after_event(self, event, state_filter=StateFilter.all()): + async def get_state_after_event(self, event, state_filter=StateFilter.all()): """ Get the room state after the given event @@ -511,7 +505,7 @@ class SyncHandler(object): Returns: A Deferred map from ((type, state_key)->Event) """ - state_ids = yield self.state_store.get_state_ids_for_event( + state_ids = await self.state_store.get_state_ids_for_event( event.event_id, state_filter=state_filter ) if event.is_state(): @@ -519,8 +513,9 @@ class SyncHandler(object): state_ids[(event.type, event.state_key)] = event.event_id return state_ids - @defer.inlineCallbacks - def get_state_at(self, room_id, stream_position, state_filter=StateFilter.all()): + async def get_state_at( + self, room_id, stream_position, state_filter=StateFilter.all() + ): """ Get the room state at a particular stream position Args: @@ -536,13 +531,13 @@ class SyncHandler(object): # get_recent_events_for_room operates by topo ordering. This therefore # does not reliably give you the state at the given stream position. # (https://github.com/matrix-org/synapse/issues/3305) - last_events, _ = yield self.store.get_recent_events_for_room( + last_events, _ = await self.store.get_recent_events_for_room( room_id, end_token=stream_position.room_key, limit=1 ) if last_events: last_event = last_events[-1] - state = yield self.get_state_after_event( + state = await self.get_state_after_event( last_event, state_filter=state_filter ) @@ -551,8 +546,7 @@ class SyncHandler(object): state = {} return state - @defer.inlineCallbacks - def compute_summary(self, room_id, sync_config, batch, state, now_token): + async def compute_summary(self, room_id, sync_config, batch, state, now_token): """ Works out a room summary block for this room, summarising the number of joined members in the room, and providing the 'hero' members if the room has no name so clients can consistently name rooms. Also adds @@ -574,7 +568,7 @@ class SyncHandler(object): # FIXME: we could/should get this from room_stats when matthew/stats lands # FIXME: this promulgates https://github.com/matrix-org/synapse/issues/3305 - last_events, _ = yield self.store.get_recent_event_ids_for_room( + last_events, _ = await self.store.get_recent_event_ids_for_room( room_id, end_token=now_token.room_key, limit=1 ) @@ -582,7 +576,7 @@ class SyncHandler(object): return None last_event = last_events[-1] - state_ids = yield self.state_store.get_state_ids_for_event( + state_ids = await self.state_store.get_state_ids_for_event( last_event.event_id, state_filter=StateFilter.from_types( [(EventTypes.Name, ""), (EventTypes.CanonicalAlias, "")] @@ -590,7 +584,7 @@ class SyncHandler(object): ) # this is heavily cached, thus: fast. - details = yield self.store.get_room_summary(room_id) + details = await self.store.get_room_summary(room_id) name_id = state_ids.get((EventTypes.Name, "")) canonical_alias_id = state_ids.get((EventTypes.CanonicalAlias, "")) @@ -608,12 +602,12 @@ class SyncHandler(object): # calculating heroes. Empty strings are falsey, so we check # for the "name" value and default to an empty string. if name_id: - name = yield self.store.get_event(name_id, allow_none=True) + name = await self.store.get_event(name_id, allow_none=True) if name and name.content.get("name"): return summary if canonical_alias_id: - canonical_alias = yield self.store.get_event( + canonical_alias = await self.store.get_event( canonical_alias_id, allow_none=True ) if canonical_alias and canonical_alias.content.get("alias"): @@ -678,7 +672,7 @@ class SyncHandler(object): ) ] - missing_hero_state = yield self.store.get_events(missing_hero_event_ids) + missing_hero_state = await self.store.get_events(missing_hero_event_ids) missing_hero_state = missing_hero_state.values() for s in missing_hero_state: @@ -697,8 +691,7 @@ class SyncHandler(object): logger.debug("found LruCache for %r", cache_key) return cache - @defer.inlineCallbacks - def compute_state_delta( + async def compute_state_delta( self, room_id, batch, sync_config, since_token, now_token, full_state ): """ Works out the difference in state between the start of the timeline @@ -759,16 +752,16 @@ class SyncHandler(object): if full_state: if batch: - current_state_ids = yield self.state_store.get_state_ids_for_event( + current_state_ids = await self.state_store.get_state_ids_for_event( batch.events[-1].event_id, state_filter=state_filter ) - state_ids = yield self.state_store.get_state_ids_for_event( + state_ids = await self.state_store.get_state_ids_for_event( batch.events[0].event_id, state_filter=state_filter ) else: - current_state_ids = yield self.get_state_at( + current_state_ids = await self.get_state_at( room_id, stream_position=now_token, state_filter=state_filter ) @@ -783,13 +776,13 @@ class SyncHandler(object): ) elif batch.limited: if batch: - state_at_timeline_start = yield self.state_store.get_state_ids_for_event( + state_at_timeline_start = await self.state_store.get_state_ids_for_event( batch.events[0].event_id, state_filter=state_filter ) else: # We can get here if the user has ignored the senders of all # the recent events. - state_at_timeline_start = yield self.get_state_at( + state_at_timeline_start = await self.get_state_at( room_id, stream_position=now_token, state_filter=state_filter ) @@ -807,19 +800,19 @@ class SyncHandler(object): # about them). state_filter = StateFilter.all() - state_at_previous_sync = yield self.get_state_at( + state_at_previous_sync = await self.get_state_at( room_id, stream_position=since_token, state_filter=state_filter ) if batch: - current_state_ids = yield self.state_store.get_state_ids_for_event( + current_state_ids = await self.state_store.get_state_ids_for_event( batch.events[-1].event_id, state_filter=state_filter ) else: # Its not clear how we get here, but empirically we do # (#5407). Logging has been added elsewhere to try and # figure out where this state comes from. - current_state_ids = yield self.get_state_at( + current_state_ids = await self.get_state_at( room_id, stream_position=now_token, state_filter=state_filter ) @@ -843,7 +836,7 @@ class SyncHandler(object): # So we fish out all the member events corresponding to the # timeline here, and then dedupe any redundant ones below. - state_ids = yield self.state_store.get_state_ids_for_event( + state_ids = await self.state_store.get_state_ids_for_event( batch.events[0].event_id, # we only want members! state_filter=StateFilter.from_types( @@ -883,7 +876,7 @@ class SyncHandler(object): state = {} if state_ids: - state = yield self.store.get_events(list(state_ids.values())) + state = await self.store.get_events(list(state_ids.values())) return { (e.type, e.state_key): e @@ -892,10 +885,9 @@ class SyncHandler(object): ) } - @defer.inlineCallbacks - def unread_notifs_for_room_id(self, room_id, sync_config): + async def unread_notifs_for_room_id(self, room_id, sync_config): with Measure(self.clock, "unread_notifs_for_room_id"): - last_unread_event_id = yield self.store.get_last_receipt_event_id_for_user( + last_unread_event_id = await self.store.get_last_receipt_event_id_for_user( user_id=sync_config.user.to_string(), room_id=room_id, receipt_type="m.read", @@ -903,7 +895,7 @@ class SyncHandler(object): notifs = [] if last_unread_event_id: - notifs = yield self.store.get_unread_event_push_actions_by_room_for_user( + notifs = await self.store.get_unread_event_push_actions_by_room_for_user( room_id, sync_config.user.to_string(), last_unread_event_id ) return notifs @@ -912,8 +904,9 @@ class SyncHandler(object): # count is whatever it was last time. return None - @defer.inlineCallbacks - def generate_sync_result(self, sync_config, since_token=None, full_state=False): + async def generate_sync_result( + self, sync_config, since_token=None, full_state=False + ): """Generates a sync result. Args: @@ -928,7 +921,7 @@ class SyncHandler(object): # this is due to some of the underlying streams not supporting the ability # to query up to a given point. # Always use the `now_token` in `SyncResultBuilder` - now_token = yield self.event_sources.get_current_token() + now_token = await self.event_sources.get_current_token() logger.info( "Calculating sync response for %r between %s and %s", @@ -944,10 +937,9 @@ class SyncHandler(object): # See https://github.com/matrix-org/matrix-doc/issues/1144 raise NotImplementedError() else: - joined_room_ids = yield self.get_rooms_for_user_at( + joined_room_ids = await self.get_rooms_for_user_at( user_id, now_token.room_stream_id ) - sync_result_builder = SyncResultBuilder( sync_config, full_state, @@ -956,11 +948,11 @@ class SyncHandler(object): joined_room_ids=joined_room_ids, ) - account_data_by_room = yield self._generate_sync_entry_for_account_data( + account_data_by_room = await self._generate_sync_entry_for_account_data( sync_result_builder ) - res = yield self._generate_sync_entry_for_rooms( + res = await self._generate_sync_entry_for_rooms( sync_result_builder, account_data_by_room ) newly_joined_rooms, newly_joined_or_invited_users, _, _ = res @@ -970,13 +962,13 @@ class SyncHandler(object): since_token is None and sync_config.filter_collection.blocks_all_presence() ) if self.hs_config.use_presence and not block_all_presence_data: - yield self._generate_sync_entry_for_presence( + await self._generate_sync_entry_for_presence( sync_result_builder, newly_joined_rooms, newly_joined_or_invited_users ) - yield self._generate_sync_entry_for_to_device(sync_result_builder) + await self._generate_sync_entry_for_to_device(sync_result_builder) - device_lists = yield self._generate_sync_entry_for_device_list( + device_lists = await self._generate_sync_entry_for_device_list( sync_result_builder, newly_joined_rooms=newly_joined_rooms, newly_joined_or_invited_users=newly_joined_or_invited_users, @@ -987,11 +979,11 @@ class SyncHandler(object): device_id = sync_config.device_id one_time_key_counts = {} if device_id: - one_time_key_counts = yield self.store.count_e2e_one_time_keys( + one_time_key_counts = await self.store.count_e2e_one_time_keys( user_id, device_id ) - yield self._generate_sync_entry_for_groups(sync_result_builder) + await self._generate_sync_entry_for_groups(sync_result_builder) # debug for https://github.com/matrix-org/synapse/issues/4422 for joined_room in sync_result_builder.joined: @@ -1015,18 +1007,17 @@ class SyncHandler(object): ) @measure_func("_generate_sync_entry_for_groups") - @defer.inlineCallbacks - def _generate_sync_entry_for_groups(self, sync_result_builder): + async def _generate_sync_entry_for_groups(self, sync_result_builder): user_id = sync_result_builder.sync_config.user.to_string() since_token = sync_result_builder.since_token now_token = sync_result_builder.now_token if since_token and since_token.groups_key: - results = yield self.store.get_groups_changes_for_user( + results = await self.store.get_groups_changes_for_user( user_id, since_token.groups_key, now_token.groups_key ) else: - results = yield self.store.get_all_groups_for_user( + results = await self.store.get_all_groups_for_user( user_id, now_token.groups_key ) @@ -1059,8 +1050,7 @@ class SyncHandler(object): ) @measure_func("_generate_sync_entry_for_device_list") - @defer.inlineCallbacks - def _generate_sync_entry_for_device_list( + async def _generate_sync_entry_for_device_list( self, sync_result_builder, newly_joined_rooms, @@ -1108,32 +1098,32 @@ class SyncHandler(object): # room with by looking at all users that have left a room plus users # that were in a room we've left. - users_who_share_room = yield self.store.get_users_who_share_room_with_user( + users_who_share_room = await self.store.get_users_who_share_room_with_user( user_id ) # Step 1a, check for changes in devices of users we share a room with - users_that_have_changed = yield self.store.get_users_whose_devices_changed( + users_that_have_changed = await self.store.get_users_whose_devices_changed( since_token.device_list_key, users_who_share_room ) # Step 1b, check for newly joined rooms for room_id in newly_joined_rooms: - joined_users = yield self.state.get_current_users_in_room(room_id) + joined_users = await self.state.get_current_users_in_room(room_id) newly_joined_or_invited_users.update(joined_users) # TODO: Check that these users are actually new, i.e. either they # weren't in the previous sync *or* they left and rejoined. users_that_have_changed.update(newly_joined_or_invited_users) - user_signatures_changed = yield self.store.get_users_whose_signatures_changed( + user_signatures_changed = await self.store.get_users_whose_signatures_changed( user_id, since_token.device_list_key ) users_that_have_changed.update(user_signatures_changed) # Now find users that we no longer track for room_id in newly_left_rooms: - left_users = yield self.state.get_current_users_in_room(room_id) + left_users = await self.state.get_current_users_in_room(room_id) newly_left_users.update(left_users) # Remove any users that we still share a room with. @@ -1143,8 +1133,7 @@ class SyncHandler(object): else: return DeviceLists(changed=[], left=[]) - @defer.inlineCallbacks - def _generate_sync_entry_for_to_device(self, sync_result_builder): + async def _generate_sync_entry_for_to_device(self, sync_result_builder): """Generates the portion of the sync response. Populates `sync_result_builder` with the result. @@ -1165,14 +1154,14 @@ class SyncHandler(object): # We only delete messages when a new message comes in, but that's # fine so long as we delete them at some point. - deleted = yield self.store.delete_messages_for_device( + deleted = await self.store.delete_messages_for_device( user_id, device_id, since_stream_id ) logger.debug( "Deleted %d to-device messages up to %d", deleted, since_stream_id ) - messages, stream_id = yield self.store.get_new_messages_for_device( + messages, stream_id = await self.store.get_new_messages_for_device( user_id, device_id, since_stream_id, now_token.to_device_key ) @@ -1190,8 +1179,7 @@ class SyncHandler(object): else: sync_result_builder.to_device = [] - @defer.inlineCallbacks - def _generate_sync_entry_for_account_data(self, sync_result_builder): + async def _generate_sync_entry_for_account_data(self, sync_result_builder): """Generates the account data portion of the sync response. Populates `sync_result_builder` with the result. @@ -1209,25 +1197,25 @@ class SyncHandler(object): ( account_data, account_data_by_room, - ) = yield self.store.get_updated_account_data_for_user( + ) = await self.store.get_updated_account_data_for_user( user_id, since_token.account_data_key ) - push_rules_changed = yield self.store.have_push_rules_changed_for_user( + push_rules_changed = await self.store.have_push_rules_changed_for_user( user_id, int(since_token.push_rules_key) ) if push_rules_changed: - account_data["m.push_rules"] = yield self.push_rules_for_user( + account_data["m.push_rules"] = await self.push_rules_for_user( sync_config.user ) else: ( account_data, account_data_by_room, - ) = yield self.store.get_account_data_for_user(sync_config.user.to_string()) + ) = await self.store.get_account_data_for_user(sync_config.user.to_string()) - account_data["m.push_rules"] = yield self.push_rules_for_user( + account_data["m.push_rules"] = await self.push_rules_for_user( sync_config.user ) @@ -1242,8 +1230,7 @@ class SyncHandler(object): return account_data_by_room - @defer.inlineCallbacks - def _generate_sync_entry_for_presence( + async def _generate_sync_entry_for_presence( self, sync_result_builder, newly_joined_rooms, newly_joined_or_invited_users ): """Generates the presence portion of the sync response. Populates the @@ -1271,7 +1258,7 @@ class SyncHandler(object): presence_key = None include_offline = False - presence, presence_key = yield presence_source.get_new_events( + presence, presence_key = await presence_source.get_new_events( user=user, from_key=presence_key, is_guest=sync_config.is_guest, @@ -1283,12 +1270,12 @@ class SyncHandler(object): extra_users_ids = set(newly_joined_or_invited_users) for room_id in newly_joined_rooms: - users = yield self.state.get_current_users_in_room(room_id) + users = await self.state.get_current_users_in_room(room_id) extra_users_ids.update(users) extra_users_ids.discard(user.to_string()) if extra_users_ids: - states = yield self.presence_handler.get_states(extra_users_ids) + states = await self.presence_handler.get_states(extra_users_ids) presence.extend(states) # Deduplicate the presence entries so that there's at most one per user @@ -1298,8 +1285,9 @@ class SyncHandler(object): sync_result_builder.presence = presence - @defer.inlineCallbacks - def _generate_sync_entry_for_rooms(self, sync_result_builder, account_data_by_room): + async def _generate_sync_entry_for_rooms( + self, sync_result_builder, account_data_by_room + ): """Generates the rooms portion of the sync response. Populates the `sync_result_builder` with the result. @@ -1321,7 +1309,7 @@ class SyncHandler(object): if block_all_room_ephemeral: ephemeral_by_room = {} else: - now_token, ephemeral_by_room = yield self.ephemeral_by_room( + now_token, ephemeral_by_room = await self.ephemeral_by_room( sync_result_builder, now_token=sync_result_builder.now_token, since_token=sync_result_builder.since_token, @@ -1333,16 +1321,16 @@ class SyncHandler(object): since_token = sync_result_builder.since_token if not sync_result_builder.full_state: if since_token and not ephemeral_by_room and not account_data_by_room: - have_changed = yield self._have_rooms_changed(sync_result_builder) + have_changed = await self._have_rooms_changed(sync_result_builder) if not have_changed: - tags_by_room = yield self.store.get_updated_tags( + tags_by_room = await self.store.get_updated_tags( user_id, since_token.account_data_key ) if not tags_by_room: logger.debug("no-oping sync") return [], [], [], [] - ignored_account_data = yield self.store.get_global_account_data_by_type_for_user( + ignored_account_data = await self.store.get_global_account_data_by_type_for_user( "m.ignored_user_list", user_id=user_id ) @@ -1352,18 +1340,18 @@ class SyncHandler(object): ignored_users = frozenset() if since_token: - res = yield self._get_rooms_changed(sync_result_builder, ignored_users) + res = await self._get_rooms_changed(sync_result_builder, ignored_users) room_entries, invited, newly_joined_rooms, newly_left_rooms = res - tags_by_room = yield self.store.get_updated_tags( + tags_by_room = await self.store.get_updated_tags( user_id, since_token.account_data_key ) else: - res = yield self._get_all_rooms(sync_result_builder, ignored_users) + res = await self._get_all_rooms(sync_result_builder, ignored_users) room_entries, invited, newly_joined_rooms = res newly_left_rooms = [] - tags_by_room = yield self.store.get_tags_for_user(user_id) + tags_by_room = await self.store.get_tags_for_user(user_id) def handle_room_entries(room_entry): return self._generate_room_entry( @@ -1376,7 +1364,7 @@ class SyncHandler(object): always_include=sync_result_builder.full_state, ) - yield concurrently_execute(handle_room_entries, room_entries, 10) + await concurrently_execute(handle_room_entries, room_entries, 10) sync_result_builder.invited.extend(invited) @@ -1410,8 +1398,7 @@ class SyncHandler(object): newly_left_users, ) - @defer.inlineCallbacks - def _have_rooms_changed(self, sync_result_builder): + async def _have_rooms_changed(self, sync_result_builder): """Returns whether there may be any new events that should be sent down the sync. Returns True if there are. """ @@ -1422,7 +1409,7 @@ class SyncHandler(object): assert since_token # Get a list of membership change events that have happened. - rooms_changed = yield self.store.get_membership_changes_for_user( + rooms_changed = await self.store.get_membership_changes_for_user( user_id, since_token.room_key, now_token.room_key ) @@ -1435,8 +1422,7 @@ class SyncHandler(object): return True return False - @defer.inlineCallbacks - def _get_rooms_changed(self, sync_result_builder, ignored_users): + async def _get_rooms_changed(self, sync_result_builder, ignored_users): """Gets the the changes that have happened since the last sync. Args: @@ -1461,7 +1447,7 @@ class SyncHandler(object): assert since_token # Get a list of membership change events that have happened. - rooms_changed = yield self.store.get_membership_changes_for_user( + rooms_changed = await self.store.get_membership_changes_for_user( user_id, since_token.room_key, now_token.room_key ) @@ -1499,11 +1485,11 @@ class SyncHandler(object): continue if room_id in sync_result_builder.joined_room_ids or has_join: - old_state_ids = yield self.get_state_at(room_id, since_token) + old_state_ids = await self.get_state_at(room_id, since_token) old_mem_ev_id = old_state_ids.get((EventTypes.Member, user_id), None) old_mem_ev = None if old_mem_ev_id: - old_mem_ev = yield self.store.get_event( + old_mem_ev = await self.store.get_event( old_mem_ev_id, allow_none=True ) @@ -1536,13 +1522,13 @@ class SyncHandler(object): newly_left_rooms.append(room_id) else: if not old_state_ids: - old_state_ids = yield self.get_state_at(room_id, since_token) + old_state_ids = await self.get_state_at(room_id, since_token) old_mem_ev_id = old_state_ids.get( (EventTypes.Member, user_id), None ) old_mem_ev = None if old_mem_ev_id: - old_mem_ev = yield self.store.get_event( + old_mem_ev = await self.store.get_event( old_mem_ev_id, allow_none=True ) if old_mem_ev and old_mem_ev.membership == Membership.JOIN: @@ -1566,7 +1552,7 @@ class SyncHandler(object): if leave_events: leave_event = leave_events[-1] - leave_stream_token = yield self.store.get_stream_token_for_event( + leave_stream_token = await self.store.get_stream_token_for_event( leave_event.event_id ) leave_token = since_token.copy_and_replace( @@ -1603,7 +1589,7 @@ class SyncHandler(object): timeline_limit = sync_config.filter_collection.timeline_limit() # Get all events for rooms we're currently joined to. - room_to_events = yield self.store.get_room_events_stream_for_rooms( + room_to_events = await self.store.get_room_events_stream_for_rooms( room_ids=sync_result_builder.joined_room_ids, from_key=since_token.room_key, to_key=now_token.room_key, @@ -1652,8 +1638,7 @@ class SyncHandler(object): return room_entries, invited, newly_joined_rooms, newly_left_rooms - @defer.inlineCallbacks - def _get_all_rooms(self, sync_result_builder, ignored_users): + async def _get_all_rooms(self, sync_result_builder, ignored_users): """Returns entries for all rooms for the user. Args: @@ -1677,7 +1662,7 @@ class SyncHandler(object): Membership.BAN, ) - room_list = yield self.store.get_rooms_for_user_where_membership_is( + room_list = await self.store.get_rooms_for_local_user_where_membership_is( user_id=user_id, membership_list=membership_list ) @@ -1700,7 +1685,7 @@ class SyncHandler(object): elif event.membership == Membership.INVITE: if event.sender in ignored_users: continue - invite = yield self.store.get_event(event.event_id) + invite = await self.store.get_event(event.event_id) invited.append(InvitedSyncResult(room_id=event.room_id, invite=invite)) elif event.membership in (Membership.LEAVE, Membership.BAN): # Always send down rooms we were banned or kicked from. @@ -1726,8 +1711,7 @@ class SyncHandler(object): return room_entries, invited, [] - @defer.inlineCallbacks - def _generate_room_entry( + async def _generate_room_entry( self, sync_result_builder, ignored_users, @@ -1769,7 +1753,7 @@ class SyncHandler(object): since_token = room_builder.since_token upto_token = room_builder.upto_token - batch = yield self._load_filtered_recents( + batch = await self._load_filtered_recents( room_id, sync_config, now_token=upto_token, @@ -1796,7 +1780,7 @@ class SyncHandler(object): # tag was added by synapse e.g. for server notice rooms. if full_state: user_id = sync_result_builder.sync_config.user.to_string() - tags = yield self.store.get_tags_for_room(user_id, room_id) + tags = await self.store.get_tags_for_room(user_id, room_id) # If there aren't any tags, don't send the empty tags list down # sync @@ -1821,7 +1805,7 @@ class SyncHandler(object): ): return - state = yield self.compute_state_delta( + state = await self.compute_state_delta( room_id, batch, sync_config, since_token, now_token, full_state=full_state ) @@ -1844,7 +1828,7 @@ class SyncHandler(object): ) or since_token is None ): - summary = yield self.compute_summary( + summary = await self.compute_summary( room_id, sync_config, batch, state, now_token ) @@ -1861,7 +1845,7 @@ class SyncHandler(object): ) if room_sync or always_include: - notifs = yield self.unread_notifs_for_room_id(room_id, sync_config) + notifs = await self.unread_notifs_for_room_id(room_id, sync_config) if notifs is not None: unread_notifications["notification_count"] = notifs["notify_count"] @@ -1887,8 +1871,7 @@ class SyncHandler(object): else: raise Exception("Unrecognized rtype: %r", room_builder.rtype) - @defer.inlineCallbacks - def get_rooms_for_user_at(self, user_id, stream_ordering): + async def get_rooms_for_user_at(self, user_id, stream_ordering): """Get set of joined rooms for a user at the given stream ordering. The stream ordering *must* be recent, otherwise this may throw an @@ -1903,7 +1886,7 @@ class SyncHandler(object): Deferred[frozenset[str]]: Set of room_ids the user is in at given stream_ordering. """ - joined_rooms = yield self.store.get_rooms_for_user_with_stream_ordering(user_id) + joined_rooms = await self.store.get_rooms_for_user_with_stream_ordering(user_id) joined_room_ids = set() @@ -1921,10 +1904,10 @@ class SyncHandler(object): logger.info("User joined room after current token: %s", room_id) - extrems = yield self.store.get_forward_extremeties_for_room( + extrems = await self.store.get_forward_extremeties_for_room( room_id, stream_ordering ) - users_in_room = yield self.state.get_current_users_in_room(room_id, extrems) + users_in_room = await self.state.get_current_users_in_room(room_id, extrems) if user_id in users_in_room: joined_room_ids.add(room_id) diff --git a/synapse/handlers/typing.py b/synapse/handlers/typing.py index 856337b7e2..d5ca9cb07b 100644 --- a/synapse/handlers/typing.py +++ b/synapse/handlers/typing.py @@ -257,7 +257,7 @@ class TypingHandler(object): "typing_key", self._latest_room_serial, rooms=[member.room_id] ) - def get_all_typing_updates(self, last_id, current_id): + async def get_all_typing_updates(self, last_id, current_id): if last_id == current_id: return [] @@ -313,10 +313,7 @@ class TypingNotificationEventSource(object): events.append(self._make_event_for(room_id)) - return events, handler._latest_room_serial + return defer.succeed((events, handler._latest_room_serial)) def get_current_key(self): return self.get_typing_handler()._latest_room_serial - - def get_pagination_rows(self, user, pagination_config, key): - return [], pagination_config.from_key diff --git a/synapse/http/server.py b/synapse/http/server.py index 943d12c907..04bc2385a2 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -14,8 +14,8 @@ # See the License for the specific language governing permissions and # limitations under the License. -import cgi import collections +import html import http.client import logging import types @@ -36,6 +36,7 @@ import synapse.metrics from synapse.api.errors import ( CodeMessageException, Codes, + RedirectException, SynapseError, UnrecognizedRequestError, ) @@ -153,14 +154,18 @@ def _return_html_error(f, request): Args: f (twisted.python.failure.Failure): - request (twisted.web.iweb.IRequest): + request (twisted.web.server.Request): """ if f.check(CodeMessageException): cme = f.value code = cme.code msg = cme.msg - if isinstance(cme, SynapseError): + if isinstance(cme, RedirectException): + logger.info("%s redirect to %s", request, cme.location) + request.setHeader(b"location", cme.location) + request.cookies.extend(cme.cookies) + elif isinstance(cme, SynapseError): logger.info("%s SynapseError: %s - %s", request, code, msg) else: logger.error( @@ -178,7 +183,7 @@ def _return_html_error(f, request): exc_info=(f.type, f.value, f.getTracebackObject()), ) - body = HTML_ERROR_TEMPLATE.format(code=code, msg=cgi.escape(msg)).encode("utf-8") + body = HTML_ERROR_TEMPLATE.format(code=code, msg=html.escape(msg)).encode("utf-8") request.setResponseCode(code) request.setHeader(b"Content-Type", b"text/html; charset=utf-8") request.setHeader(b"Content-Length", b"%i" % (len(body),)) diff --git a/synapse/http/servlet.py b/synapse/http/servlet.py index e9a5e46ced..13fcb408a6 100644 --- a/synapse/http/servlet.py +++ b/synapse/http/servlet.py @@ -96,7 +96,7 @@ def parse_boolean_from_args(args, name, default=None, required=False): return {b"true": True, b"false": False}[args[name][0]] except Exception: message = ( - "Boolean query parameter %r must be one of" " ['true', 'false']" + "Boolean query parameter %r must be one of ['true', 'false']" ) % (name,) raise SynapseError(400, message) else: diff --git a/synapse/http/site.py b/synapse/http/site.py index ff8184a3d0..911251c0bc 100644 --- a/synapse/http/site.py +++ b/synapse/http/site.py @@ -47,9 +47,9 @@ class SynapseRequest(Request): logcontext(LoggingContext) : the log context for this request """ - def __init__(self, site, channel, *args, **kw): + def __init__(self, channel, *args, **kw): Request.__init__(self, channel, *args, **kw) - self.site = site + self.site = channel.site self._channel = channel # this is used by the tests self.authenticated_entity = None self.start_time = 0 @@ -88,7 +88,7 @@ class SynapseRequest(Request): def get_redacted_uri(self): uri = self.uri if isinstance(uri, bytes): - uri = self.uri.decode("ascii") + uri = self.uri.decode("ascii", errors="replace") return redact_uri(uri) def get_method(self): @@ -331,18 +331,6 @@ class XForwardedForRequest(SynapseRequest): ) -class SynapseRequestFactory(object): - def __init__(self, site, x_forwarded_for): - self.site = site - self.x_forwarded_for = x_forwarded_for - - def __call__(self, *args, **kwargs): - if self.x_forwarded_for: - return XForwardedForRequest(self.site, *args, **kwargs) - else: - return SynapseRequest(self.site, *args, **kwargs) - - class SynapseSite(Site): """ Subclass of a twisted http Site that does access logging with python's @@ -364,7 +352,7 @@ class SynapseSite(Site): self.site_tag = site_tag proxied = config.get("x_forwarded", False) - self.requestFactory = SynapseRequestFactory(self, proxied) + self.requestFactory = XForwardedForRequest if proxied else SynapseRequest self.access_logger = logging.getLogger(logger_name) self.server_version_string = server_version_string.encode("ascii") diff --git a/synapse/logging/_structured.py b/synapse/logging/_structured.py index 334ddaf39a..ffa7b20ca8 100644 --- a/synapse/logging/_structured.py +++ b/synapse/logging/_structured.py @@ -261,6 +261,18 @@ def parse_drain_configs( ) +class StoppableLogPublisher(LogPublisher): + """ + A log publisher that can tell its observers to shut down any external + communications. + """ + + def stop(self): + for obs in self._observers: + if hasattr(obs, "stop"): + obs.stop() + + def setup_structured_logging( hs, config, @@ -336,7 +348,7 @@ def setup_structured_logging( # We should never get here, but, just in case, throw an error. raise ConfigError("%s drain type cannot be configured" % (observer.type,)) - publisher = LogPublisher(*observers) + publisher = StoppableLogPublisher(*observers) log_filter = LogLevelFilterPredicate() for namespace, namespace_config in log_config.get( diff --git a/synapse/logging/_terse_json.py b/synapse/logging/_terse_json.py index 76ce7d8808..c0b9384189 100644 --- a/synapse/logging/_terse_json.py +++ b/synapse/logging/_terse_json.py @@ -17,25 +17,29 @@ Log formatters that output terse JSON. """ +import json import sys +import traceback from collections import deque from ipaddress import IPv4Address, IPv6Address, ip_address from math import floor -from typing import IO +from typing import IO, Optional import attr -from simplejson import dumps from zope.interface import implementer from twisted.application.internet import ClientService +from twisted.internet.defer import Deferred from twisted.internet.endpoints import ( HostnameEndpoint, TCP4ClientEndpoint, TCP6ClientEndpoint, ) +from twisted.internet.interfaces import IPushProducer, ITransport from twisted.internet.protocol import Factory, Protocol from twisted.logger import FileLogObserver, ILogObserver, Logger -from twisted.python.failure import Failure + +_encoder = json.JSONEncoder(ensure_ascii=False, separators=(",", ":")) def flatten_event(event: dict, metadata: dict, include_time: bool = False): @@ -141,12 +145,50 @@ def TerseJSONToConsoleLogObserver(outFile: IO[str], metadata: dict) -> FileLogOb def formatEvent(_event: dict) -> str: flattened = flatten_event(_event, metadata) - return dumps(flattened, ensure_ascii=False, separators=(",", ":")) + "\n" + return _encoder.encode(flattened) + "\n" return FileLogObserver(outFile, formatEvent) @attr.s +@implementer(IPushProducer) +class LogProducer(object): + """ + An IPushProducer that writes logs from its buffer to its transport when it + is resumed. + + Args: + buffer: Log buffer to read logs from. + transport: Transport to write to. + """ + + transport = attr.ib(type=ITransport) + _buffer = attr.ib(type=deque) + _paused = attr.ib(default=False, type=bool, init=False) + + def pauseProducing(self): + self._paused = True + + def stopProducing(self): + self._paused = True + self._buffer = deque() + + def resumeProducing(self): + self._paused = False + + while self._paused is False and (self._buffer and self.transport.connected): + try: + event = self._buffer.popleft() + self.transport.write(_encoder.encode(event).encode("utf8")) + self.transport.write(b"\n") + except Exception: + # Something has gone wrong writing to the transport -- log it + # and break out of the while. + traceback.print_exc(file=sys.__stderr__) + break + + +@attr.s @implementer(ILogObserver) class TerseJSONToTCPLogObserver(object): """ @@ -165,8 +207,9 @@ class TerseJSONToTCPLogObserver(object): metadata = attr.ib(type=dict) maximum_buffer = attr.ib(type=int) _buffer = attr.ib(default=attr.Factory(deque), type=deque) - _writer = attr.ib(default=None) + _connection_waiter = attr.ib(default=None, type=Optional[Deferred]) _logger = attr.ib(default=attr.Factory(Logger)) + _producer = attr.ib(default=None, type=Optional[LogProducer]) def start(self) -> None: @@ -187,38 +230,44 @@ class TerseJSONToTCPLogObserver(object): factory = Factory.forProtocol(Protocol) self._service = ClientService(endpoint, factory, clock=self.hs.get_reactor()) self._service.startService() + self._connect() - def _write_loop(self) -> None: + def stop(self): + self._service.stopService() + + def _connect(self) -> None: """ - Implement the write loop. + Triggers an attempt to connect then write to the remote if not already writing. """ - if self._writer: + if self._connection_waiter: return - self._writer = self._service.whenConnected() + self._connection_waiter = self._service.whenConnected(failAfterFailures=1) + + @self._connection_waiter.addErrback + def fail(r): + r.printTraceback(file=sys.__stderr__) + self._connection_waiter = None + self._connect() - @self._writer.addBoth + @self._connection_waiter.addCallback def writer(r): - if isinstance(r, Failure): - r.printTraceback(file=sys.__stderr__) - self._writer = None - self.hs.get_reactor().callLater(1, self._write_loop) + # We have a connection. If we already have a producer, and its + # transport is the same, just trigger a resumeProducing. + if self._producer and r.transport is self._producer.transport: + self._producer.resumeProducing() + self._connection_waiter = None return - try: - for event in self._buffer: - r.transport.write( - dumps(event, ensure_ascii=False, separators=(",", ":")).encode( - "utf8" - ) - ) - r.transport.write(b"\n") - self._buffer.clear() - except Exception as e: - sys.__stderr__.write("Failed writing out logs with %s\n" % (str(e),)) - - self._writer = False - self.hs.get_reactor().callLater(1, self._write_loop) + # If the producer is still producing, stop it. + if self._producer: + self._producer.stopProducing() + + # Make a new producer and start it. + self._producer = LogProducer(buffer=self._buffer, transport=r.transport) + r.transport.registerProducer(self._producer, True) + self._producer.resumeProducing() + self._connection_waiter = None def _handle_pressure(self) -> None: """ @@ -277,4 +326,4 @@ class TerseJSONToTCPLogObserver(object): self._logger.failure("Failed clearing backpressure") # Try and write immediately. - self._write_loop() + self._connect() diff --git a/synapse/logging/context.py b/synapse/logging/context.py index 2c1fb9ddac..1b940842f6 100644 --- a/synapse/logging/context.py +++ b/synapse/logging/context.py @@ -23,6 +23,7 @@ them. See doc/log_contexts.rst for details on how this works. """ +import inspect import logging import threading import types @@ -404,6 +405,9 @@ class LoggingContext(object): """ current = get_thread_resource_usage() + # Indicate to mypy that we know that self.usage_start is None. + assert self.usage_start is not None + utime_delta = current.ru_utime - self.usage_start.ru_utime stime_delta = current.ru_stime - self.usage_start.ru_stime @@ -567,6 +571,9 @@ def run_in_background(f, *args, **kwargs): yield or await on (for instance because you want to pass it to deferred.gatherResults()). + If f returns a Coroutine object, it will be wrapped into a Deferred (which will have + the side effect of executing the coroutine). + Note that if you completely discard the result, you should make sure that `f` doesn't raise any deferred exceptions, otherwise a scary-looking CRITICAL error about an unhandled error will be logged without much @@ -612,7 +619,8 @@ def run_in_background(f, *args, **kwargs): def make_deferred_yieldable(deferred): - """Given a deferred, make it follow the Synapse logcontext rules: + """Given a deferred (or coroutine), make it follow the Synapse logcontext + rules: If the deferred has completed (or is not actually a Deferred), essentially does nothing (just returns another completed deferred with the @@ -624,6 +632,13 @@ def make_deferred_yieldable(deferred): (This is more-or-less the opposite operation to run_in_background.) """ + if inspect.isawaitable(deferred): + # If we're given a coroutine we convert it to a deferred so that we + # run it and find out if it immediately finishes, it it does then we + # don't need to fiddle with log contexts at all and can return + # immediately. + deferred = defer.ensureDeferred(deferred) + if not isinstance(deferred, defer.Deferred): return deferred diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py index 735b882363..d680ee95e1 100644 --- a/synapse/module_api/__init__.py +++ b/synapse/module_api/__init__.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2017 New Vector Ltd +# Copyright 2020 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. @@ -16,18 +17,26 @@ import logging from twisted.internet import defer +from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.types import UserID +""" +This package defines the 'stable' API which can be used by extension modules which +are loaded into Synapse. +""" + +__all__ = ["errors", "make_deferred_yieldable", "run_in_background", "ModuleApi"] + logger = logging.getLogger(__name__) class ModuleApi(object): - """A proxy object that gets passed to password auth providers so they + """A proxy object that gets passed to various plugin modules so they can register new users etc if necessary. """ def __init__(self, hs, auth_handler): - self.hs = hs + self._hs = hs self._store = hs.get_datastore() self._auth = hs.get_auth() @@ -64,7 +73,7 @@ class ModuleApi(object): """ if username.startswith("@"): return username - return UserID(username, self.hs.hostname).to_string() + return UserID(username, self._hs.hostname).to_string() def check_user_exists(self, user_id): """Check if user exists. @@ -111,10 +120,14 @@ class ModuleApi(object): displayname (str|None): The displayname of the new user. emails (List[str]): Emails to bind to the new user. + Raises: + SynapseError if there is an error performing the registration. Check the + 'errcode' property for more information on the reason for failure + Returns: Deferred[str]: user_id """ - return self.hs.get_registration_handler().register_user( + return self._hs.get_registration_handler().register_user( localpart=localpart, default_display_name=displayname, bind_emails=emails ) @@ -131,12 +144,34 @@ class ModuleApi(object): Returns: defer.Deferred[tuple[str, str]]: Tuple of device ID and access token """ - return self.hs.get_registration_handler().register_device( + return self._hs.get_registration_handler().register_device( user_id=user_id, device_id=device_id, initial_display_name=initial_display_name, ) + def record_user_external_id( + self, auth_provider_id: str, remote_user_id: str, registered_user_id: str + ) -> defer.Deferred: + """Record a mapping from an external user id to a mxid + + Args: + auth_provider: identifier for the remote auth provider + external_id: id on that system + user_id: complete mxid that it is mapped to + """ + return self._store.record_user_external_id( + auth_provider_id, remote_user_id, registered_user_id + ) + + def generate_short_term_login_token( + self, user_id: str, duration_in_ms: int = (2 * 60 * 1000) + ) -> str: + """Generate a login token suitable for m.login.token authentication""" + return self._hs.get_macaroon_generator().generate_short_term_login_token( + user_id, duration_in_ms + ) + @defer.inlineCallbacks def invalidate_access_token(self, access_token): """Invalidate an access token for a user @@ -157,7 +192,7 @@ class ModuleApi(object): user_id = user_info["user"].to_string() if device_id: # delete the device, which will also delete its access tokens - yield self.hs.get_device_handler().delete_device(user_id, device_id) + yield self._hs.get_device_handler().delete_device(user_id, device_id) else: # no associated device. Just delete the access token. yield self._auth_handler.delete_access_token(access_token) @@ -175,4 +210,4 @@ class ModuleApi(object): Returns: Deferred[object]: result of func """ - return self._store.runInteraction(desc, func, *args, **kwargs) + return self._store.db.runInteraction(desc, func, *args, **kwargs) diff --git a/synapse/module_api/errors.py b/synapse/module_api/errors.py new file mode 100644 index 0000000000..b15441772c --- /dev/null +++ b/synapse/module_api/errors.py @@ -0,0 +1,18 @@ +# -*- coding: utf-8 -*- +# Copyright 2020 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. + +"""Exception types which are exposed as part of the stable module API""" + +from synapse.api.errors import RedirectException, SynapseError # noqa: F401 diff --git a/synapse/notifier.py b/synapse/notifier.py index af161a81d7..5f5f765bea 100644 --- a/synapse/notifier.py +++ b/synapse/notifier.py @@ -304,8 +304,7 @@ class Notifier(object): without waking up any of the normal user event streams""" self.notify_replication() - @defer.inlineCallbacks - def wait_for_events( + async def wait_for_events( self, user_id, timeout, callback, room_ids=None, from_token=StreamToken.START ): """Wait until the callback returns a non empty response or the @@ -313,9 +312,9 @@ class Notifier(object): """ user_stream = self.user_to_user_stream.get(user_id) if user_stream is None: - current_token = yield self.event_sources.get_current_token() + current_token = await self.event_sources.get_current_token() if room_ids is None: - room_ids = yield self.store.get_rooms_for_user(user_id) + room_ids = await self.store.get_rooms_for_user(user_id) user_stream = _NotifierUserStream( user_id=user_id, rooms=room_ids, @@ -344,11 +343,11 @@ class Notifier(object): self.hs.get_reactor(), ) with PreserveLoggingContext(): - yield listener.deferred + await listener.deferred current_token = user_stream.current_token - result = yield callback(prev_token, current_token) + result = await callback(prev_token, current_token) if result: break @@ -364,12 +363,11 @@ class Notifier(object): # This happened if there was no timeout or if the timeout had # already expired. current_token = user_stream.current_token - result = yield callback(prev_token, current_token) + result = await callback(prev_token, current_token) return result - @defer.inlineCallbacks - def get_events_for( + async def get_events_for( self, user, pagination_config, @@ -391,15 +389,14 @@ class Notifier(object): """ from_token = pagination_config.from_token if not from_token: - from_token = yield self.event_sources.get_current_token() + from_token = await self.event_sources.get_current_token() limit = pagination_config.limit - room_ids, is_joined = yield self._get_room_ids(user, explicit_room_id) + room_ids, is_joined = await self._get_room_ids(user, explicit_room_id) is_peeking = not is_joined - @defer.inlineCallbacks - def check_for_updates(before_token, after_token): + async def check_for_updates(before_token, after_token): if not after_token.is_after(before_token): return EventStreamResult([], (from_token, from_token)) @@ -415,7 +412,7 @@ class Notifier(object): if only_keys and name not in only_keys: continue - new_events, new_key = yield source.get_new_events( + new_events, new_key = await source.get_new_events( user=user, from_key=getattr(from_token, keyname), limit=limit, @@ -425,7 +422,7 @@ class Notifier(object): ) if name == "room": - new_events = yield filter_events_for_client( + new_events = await filter_events_for_client( self.storage, user.to_string(), new_events, @@ -461,7 +458,7 @@ class Notifier(object): user_id_for_stream, ) - result = yield self.wait_for_events( + result = await self.wait_for_events( user_id_for_stream, timeout, check_for_updates, diff --git a/synapse/push/bulk_push_rule_evaluator.py b/synapse/push/bulk_push_rule_evaluator.py index 1400f51691..823ebdaa60 100644 --- a/synapse/push/bulk_push_rule_evaluator.py +++ b/synapse/push/bulk_push_rule_evaluator.py @@ -118,7 +118,7 @@ class BulkPushRuleEvaluator(object): @defer.inlineCallbacks def _get_power_levels_and_sender_level(self, event, context): - prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_state_ids = yield context.get_prev_state_ids() pl_event_id = prev_state_ids.get(POWER_KEY) if pl_event_id: # fastpath: if there's a power level event, that's all we need, and @@ -306,7 +306,7 @@ class RulesForRoom(object): push_rules_delta_state_cache_metric.inc_hits() else: - current_state_ids = yield context.get_current_state_ids(self.store) + current_state_ids = yield context.get_current_state_ids() push_rules_delta_state_cache_metric.inc_misses() push_rules_state_size_counter.inc(len(current_state_ids)) @@ -388,15 +388,7 @@ class RulesForRoom(object): """ sequence = self.sequence - rows = yield self.store._simple_select_many_batch( - table="room_memberships", - column="event_id", - iterable=member_event_ids.values(), - retcols=("user_id", "membership", "event_id"), - keyvalues={}, - batch_size=500, - desc="_get_rules_for_member_event_ids", - ) + rows = yield self.store.get_membership_from_event_ids(member_event_ids.values()) members = {row["event_id"]: (row["user_id"], row["membership"]) for row in rows} diff --git a/synapse/push/httppusher.py b/synapse/push/httppusher.py index e994037be6..d0879b0490 100644 --- a/synapse/push/httppusher.py +++ b/synapse/push/httppusher.py @@ -246,7 +246,7 @@ class HttpPusher(object): # fixed, we don't suddenly deliver a load # of old notifications. logger.warning( - "Giving up on a notification to user %s, " "pushkey %s", + "Giving up on a notification to user %s, pushkey %s", self.user_id, self.pushkey, ) @@ -299,8 +299,7 @@ class HttpPusher(object): # for sanity, we only remove the pushkey if it # was the one we actually sent... logger.warning( - ("Ignoring rejected pushkey %s because we" " didn't send it"), - pk, + ("Ignoring rejected pushkey %s because we didn't send it"), pk, ) else: logger.info("Pushkey %s was rejected: removing", pk) diff --git a/synapse/push/mailer.py b/synapse/push/mailer.py index 1d15a06a58..b13b646bfd 100644 --- a/synapse/push/mailer.py +++ b/synapse/push/mailer.py @@ -43,7 +43,7 @@ logger = logging.getLogger(__name__) MESSAGE_FROM_PERSON_IN_ROOM = ( - "You have a message on %(app)s from %(person)s " "in the %(room)s room..." + "You have a message on %(app)s from %(person)s in the %(room)s room..." ) MESSAGE_FROM_PERSON = "You have a message on %(app)s from %(person)s..." MESSAGES_FROM_PERSON = "You have messages on %(app)s from %(person)s..." @@ -55,7 +55,7 @@ MESSAGES_FROM_PERSON_AND_OTHERS = ( "You have messages on %(app)s from %(person)s and others..." ) INVITE_FROM_PERSON_TO_ROOM = ( - "%(person)s has invited you to join the " "%(room)s room on %(app)s..." + "%(person)s has invited you to join the %(room)s room on %(app)s..." ) INVITE_FROM_PERSON = "%(person)s has invited you to chat on %(app)s..." diff --git a/synapse/push/push_tools.py b/synapse/push/push_tools.py index de5c101a58..5dae4648c0 100644 --- a/synapse/push/push_tools.py +++ b/synapse/push/push_tools.py @@ -21,7 +21,7 @@ from synapse.storage import Storage @defer.inlineCallbacks def get_badge_count(store, user_id): - invites = yield store.get_invited_rooms_for_user(user_id) + invites = yield store.get_invited_rooms_for_local_user(user_id) joins = yield store.get_rooms_for_user(user_id) my_receipts_by_room = yield store.get_receipts_for_user(user_id, "m.read") diff --git a/synapse/push/pusher.py b/synapse/push/pusher.py index f277aeb131..8ad0bf5936 100644 --- a/synapse/push/pusher.py +++ b/synapse/push/pusher.py @@ -80,9 +80,11 @@ class PusherFactory(object): return EmailPusher(self.hs, pusherdict, mailer) def _app_name_from_pusherdict(self, pusherdict): - if "data" in pusherdict and "brand" in pusherdict["data"]: - app_name = pusherdict["data"]["brand"] - else: - app_name = self.config.email_app_name + data = pusherdict["data"] - return app_name + if isinstance(data, dict): + brand = data.get("brand") + if isinstance(brand, str): + return brand + + return self.config.email_app_name diff --git a/synapse/push/pusherpool.py b/synapse/push/pusherpool.py index 0f6992202d..b9dca5bc63 100644 --- a/synapse/push/pusherpool.py +++ b/synapse/push/pusherpool.py @@ -232,7 +232,6 @@ class PusherPool: Deferred """ pushers = yield self.store.get_all_pushers() - logger.info("Starting %d pushers", len(pushers)) # Stagger starting up the pushers so we don't completely drown the # process on start up. @@ -245,7 +244,7 @@ class PusherPool: """Start the given pusher Args: - pusherdict (dict): + pusherdict (dict): dict with the values pulled from the db table Returns: Deferred[EmailPusher|HttpPusher] @@ -254,7 +253,8 @@ class PusherPool: p = self.pusher_factory.create_pusher(pusherdict) except PusherConfigException as e: logger.warning( - "Pusher incorrectly configured user=%s, appid=%s, pushkey=%s: %s", + "Pusher incorrectly configured id=%i, user=%s, appid=%s, pushkey=%s: %s", + pusherdict["id"], pusherdict.get("user_name"), pusherdict.get("app_id"), pusherdict.get("pushkey"), @@ -262,7 +262,9 @@ class PusherPool: ) return except Exception: - logger.exception("Couldn't start a pusher: caught Exception") + logger.exception( + "Couldn't start pusher id %i: caught Exception", pusherdict["id"], + ) return if not p: diff --git a/synapse/replication/http/__init__.py b/synapse/replication/http/__init__.py index 81b85352b1..28dbc6fcba 100644 --- a/synapse/replication/http/__init__.py +++ b/synapse/replication/http/__init__.py @@ -14,7 +14,14 @@ # limitations under the License. from synapse.http.server import JsonResource -from synapse.replication.http import federation, login, membership, register, send_event +from synapse.replication.http import ( + devices, + federation, + login, + membership, + register, + send_event, +) REPLICATION_PREFIX = "/_synapse/replication" @@ -30,3 +37,4 @@ class ReplicationRestResource(JsonResource): federation.register_servlets(hs, self) login.register_servlets(hs, self) register.register_servlets(hs, self) + devices.register_servlets(hs, self) diff --git a/synapse/replication/http/_base.py b/synapse/replication/http/_base.py index c8056b0c0c..444eb7b7f4 100644 --- a/synapse/replication/http/_base.py +++ b/synapse/replication/http/_base.py @@ -16,6 +16,7 @@ import abc import logging import re +from typing import Dict, List, Tuple from six import raise_from from six.moves import urllib @@ -78,9 +79,8 @@ class ReplicationEndpoint(object): __metaclass__ = abc.ABCMeta - NAME = abc.abstractproperty() - PATH_ARGS = abc.abstractproperty() - + NAME = abc.abstractproperty() # type: str # type: ignore + PATH_ARGS = abc.abstractproperty() # type: Tuple[str, ...] # type: ignore METHOD = "POST" CACHE = True RETRY_ON_TIMEOUT = True @@ -171,7 +171,7 @@ class ReplicationEndpoint(object): # have a good idea that the request has either succeeded or failed on # the master, and so whether we should clean up or not. while True: - headers = {} + headers = {} # type: Dict[bytes, List[bytes]] inject_active_span_byte_dict(headers, None, check_destination=False) try: result = yield request_func(uri, data, headers=headers) @@ -207,7 +207,7 @@ class ReplicationEndpoint(object): method = self.METHOD if self.CACHE: - handler = self._cached_handler + handler = self._cached_handler # type: ignore url_args.append("txn_id") args = "/".join("(?P<%s>[^/]+)" % (arg,) for arg in url_args) diff --git a/synapse/replication/http/devices.py b/synapse/replication/http/devices.py new file mode 100644 index 0000000000..e32aac0a25 --- /dev/null +++ b/synapse/replication/http/devices.py @@ -0,0 +1,73 @@ +# -*- coding: utf-8 -*- +# Copyright 2018 New Vector Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging + +from synapse.replication.http._base import ReplicationEndpoint + +logger = logging.getLogger(__name__) + + +class ReplicationUserDevicesResyncRestServlet(ReplicationEndpoint): + """Ask master to resync the device list for a user by contacting their + server. + + This must happen on master so that the results can be correctly cached in + the database and streamed to workers. + + Request format: + + POST /_synapse/replication/user_device_resync/:user_id + + {} + + Response is equivalent to ` /_matrix/federation/v1/user/devices/:user_id` + response, e.g.: + + { + "user_id": "@alice:example.org", + "devices": [ + { + "device_id": "JLAFKJWSCS", + "keys": { ... }, + "device_display_name": "Alice's Mobile Phone" + } + ] + } + """ + + NAME = "user_device_resync" + PATH_ARGS = ("user_id",) + CACHE = False + + def __init__(self, hs): + super(ReplicationUserDevicesResyncRestServlet, self).__init__(hs) + + self.device_list_updater = hs.get_device_handler().device_list_updater + self.store = hs.get_datastore() + self.clock = hs.get_clock() + + @staticmethod + def _serialize_payload(user_id): + return {} + + async def _handle_request(self, request, user_id): + user_devices = await self.device_list_updater.user_device_resync(user_id) + + return 200, user_devices + + +def register_servlets(hs, http_server): + ReplicationUserDevicesResyncRestServlet(hs).register(http_server) diff --git a/synapse/replication/http/federation.py b/synapse/replication/http/federation.py index 9af4e7e173..49a3251372 100644 --- a/synapse/replication/http/federation.py +++ b/synapse/replication/http/federation.py @@ -51,6 +51,7 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): super(ReplicationFederationSendEventsRestServlet, self).__init__(hs) self.store = hs.get_datastore() + self.storage = hs.get_storage() self.clock = hs.get_clock() self.federation_handler = hs.get_handlers().federation_handler @@ -100,7 +101,9 @@ class ReplicationFederationSendEventsRestServlet(ReplicationEndpoint): EventType = event_type_from_format_version(format_ver) event = EventType(event_dict, internal_metadata, rejected_reason) - context = EventContext.deserialize(self.store, event_payload["context"]) + context = EventContext.deserialize( + self.storage, event_payload["context"] + ) event_and_contexts.append((event, context)) diff --git a/synapse/replication/http/membership.py b/synapse/replication/http/membership.py index cc1f249740..3577611fd7 100644 --- a/synapse/replication/http/membership.py +++ b/synapse/replication/http/membership.py @@ -93,6 +93,7 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint): { "requester": ..., "remote_room_hosts": [...], + "content": { ... } } """ @@ -107,7 +108,7 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint): self.clock = hs.get_clock() @staticmethod - def _serialize_payload(requester, room_id, user_id, remote_room_hosts): + def _serialize_payload(requester, room_id, user_id, remote_room_hosts, content): """ Args: requester(Requester) @@ -118,12 +119,14 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint): return { "requester": requester.serialize(), "remote_room_hosts": remote_room_hosts, + "content": content, } async def _handle_request(self, request, room_id, user_id): content = parse_json_object_from_request(request) remote_room_hosts = content["remote_room_hosts"] + event_content = content["content"] requester = Requester.deserialize(self.store, content["requester"]) @@ -134,7 +137,7 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint): try: event = await self.federation_handler.do_remotely_reject_invite( - remote_room_hosts, room_id, user_id + remote_room_hosts, room_id, user_id, event_content, ) ret = event.get_pdu_json() except Exception as e: diff --git a/synapse/replication/http/send_event.py b/synapse/replication/http/send_event.py index 9bafd60b14..84b92f16ad 100644 --- a/synapse/replication/http/send_event.py +++ b/synapse/replication/http/send_event.py @@ -54,6 +54,7 @@ class ReplicationSendEventRestServlet(ReplicationEndpoint): self.event_creation_handler = hs.get_event_creation_handler() self.store = hs.get_datastore() + self.storage = hs.get_storage() self.clock = hs.get_clock() @staticmethod @@ -100,7 +101,7 @@ class ReplicationSendEventRestServlet(ReplicationEndpoint): event = EventType(event_dict, internal_metadata, rejected_reason) requester = Requester.deserialize(self.store, content["requester"]) - context = EventContext.deserialize(self.store, content["context"]) + context = EventContext.deserialize(self.storage, content["context"]) ratelimit = content["ratelimit"] extra_users = [UserID.from_string(u) for u in content["extra_users"]] diff --git a/synapse/replication/slave/storage/_base.py b/synapse/replication/slave/storage/_base.py index 456bc005a0..704282c800 100644 --- a/synapse/replication/slave/storage/_base.py +++ b/synapse/replication/slave/storage/_base.py @@ -14,11 +14,13 @@ # limitations under the License. import logging -from typing import Dict +from typing import Dict, Optional import six -from synapse.storage._base import _CURRENT_STATE_CACHE_NAME, SQLBaseStore +from synapse.storage._base import SQLBaseStore +from synapse.storage.data_stores.main.cache import CURRENT_STATE_CACHE_NAME +from synapse.storage.database import Database from synapse.storage.engines import PostgresEngine from ._slaved_id_tracker import SlavedIdTracker @@ -34,12 +36,12 @@ def __func__(inp): class BaseSlavedStore(SQLBaseStore): - def __init__(self, db_conn, hs): - super(BaseSlavedStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(BaseSlavedStore, self).__init__(database, db_conn, hs) if isinstance(self.database_engine, PostgresEngine): self._cache_id_gen = SlavedIdTracker( db_conn, "cache_invalidation_stream", "stream_id" - ) + ) # type: Optional[SlavedIdTracker] else: self._cache_id_gen = None @@ -60,9 +62,10 @@ class BaseSlavedStore(SQLBaseStore): def process_replication_rows(self, stream_name, token, rows): if stream_name == "caches": - self._cache_id_gen.advance(token) + if self._cache_id_gen: + self._cache_id_gen.advance(token) for row in rows: - if row.cache_func == _CURRENT_STATE_CACHE_NAME: + if row.cache_func == CURRENT_STATE_CACHE_NAME: room_id = row.keys[0] members_changed = set(row.keys[1:]) self._invalidate_state_caches(room_id, members_changed) diff --git a/synapse/replication/slave/storage/account_data.py b/synapse/replication/slave/storage/account_data.py index bc2f6a12ae..ebe94909cb 100644 --- a/synapse/replication/slave/storage/account_data.py +++ b/synapse/replication/slave/storage/account_data.py @@ -18,15 +18,16 @@ from synapse.replication.slave.storage._base import BaseSlavedStore from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker from synapse.storage.data_stores.main.account_data import AccountDataWorkerStore from synapse.storage.data_stores.main.tags import TagsWorkerStore +from synapse.storage.database import Database class SlavedAccountDataStore(TagsWorkerStore, AccountDataWorkerStore, BaseSlavedStore): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): self._account_data_id_gen = SlavedIdTracker( db_conn, "account_data_max_stream_id", "stream_id" ) - super(SlavedAccountDataStore, self).__init__(db_conn, hs) + super(SlavedAccountDataStore, self).__init__(database, db_conn, hs) def get_max_account_data_stream_id(self): return self._account_data_id_gen.get_current_token() diff --git a/synapse/replication/slave/storage/client_ips.py b/synapse/replication/slave/storage/client_ips.py index 268014f980..f26258e677 100644 --- a/synapse/replication/slave/storage/client_ips.py +++ b/synapse/replication/slave/storage/client_ips.py @@ -14,14 +14,16 @@ # limitations under the License. from synapse.storage.data_stores.main.client_ips import LAST_SEEN_GRANULARITY +from synapse.storage.database import Database +from synapse.util.caches import CACHE_SIZE_FACTOR from synapse.util.caches.descriptors import Cache from ._base import BaseSlavedStore class SlavedClientIpStore(BaseSlavedStore): - def __init__(self, db_conn, hs): - super(SlavedClientIpStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SlavedClientIpStore, self).__init__(database, db_conn, hs) self.client_ip_last_seen = Cache( name="client_ip_last_seen", keylen=4, max_entries=50000 diff --git a/synapse/replication/slave/storage/deviceinbox.py b/synapse/replication/slave/storage/deviceinbox.py index 9fb6c5c6ff..0c237c6e0f 100644 --- a/synapse/replication/slave/storage/deviceinbox.py +++ b/synapse/replication/slave/storage/deviceinbox.py @@ -16,13 +16,14 @@ from synapse.replication.slave.storage._base import BaseSlavedStore from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker from synapse.storage.data_stores.main.deviceinbox import DeviceInboxWorkerStore +from synapse.storage.database import Database from synapse.util.caches.expiringcache import ExpiringCache from synapse.util.caches.stream_change_cache import StreamChangeCache class SlavedDeviceInboxStore(DeviceInboxWorkerStore, BaseSlavedStore): - def __init__(self, db_conn, hs): - super(SlavedDeviceInboxStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SlavedDeviceInboxStore, self).__init__(database, db_conn, hs) self._device_inbox_id_gen = SlavedIdTracker( db_conn, "device_max_stream_id", "stream_id" ) diff --git a/synapse/replication/slave/storage/devices.py b/synapse/replication/slave/storage/devices.py index de50748c30..dc625e0d7a 100644 --- a/synapse/replication/slave/storage/devices.py +++ b/synapse/replication/slave/storage/devices.py @@ -18,12 +18,13 @@ from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker from synapse.replication.tcp.streams._base import DeviceListsStream, UserSignatureStream from synapse.storage.data_stores.main.devices import DeviceWorkerStore from synapse.storage.data_stores.main.end_to_end_keys import EndToEndKeyWorkerStore +from synapse.storage.database import Database from synapse.util.caches.stream_change_cache import StreamChangeCache class SlavedDeviceStore(EndToEndKeyWorkerStore, DeviceWorkerStore, BaseSlavedStore): - def __init__(self, db_conn, hs): - super(SlavedDeviceStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SlavedDeviceStore, self).__init__(database, db_conn, hs) self.hs = hs diff --git a/synapse/replication/slave/storage/events.py b/synapse/replication/slave/storage/events.py index d0a0eaf75b..3aa6cb8b96 100644 --- a/synapse/replication/slave/storage/events.py +++ b/synapse/replication/slave/storage/events.py @@ -31,6 +31,7 @@ from synapse.storage.data_stores.main.signatures import SignatureWorkerStore from synapse.storage.data_stores.main.state import StateGroupWorkerStore from synapse.storage.data_stores.main.stream import StreamWorkerStore from synapse.storage.data_stores.main.user_erasure_store import UserErasureWorkerStore +from synapse.storage.database import Database from ._base import BaseSlavedStore from ._slaved_id_tracker import SlavedIdTracker @@ -59,13 +60,13 @@ class SlavedEventStore( RelationsWorkerStore, BaseSlavedStore, ): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): self._stream_id_gen = SlavedIdTracker(db_conn, "events", "stream_ordering") self._backfill_id_gen = SlavedIdTracker( db_conn, "events", "stream_ordering", step=-1 ) - super(SlavedEventStore, self).__init__(db_conn, hs) + super(SlavedEventStore, self).__init__(database, db_conn, hs) # Cached functions can't be accessed through a class instance so we need # to reach inside the __dict__ to extract them. @@ -151,7 +152,7 @@ class SlavedEventStore( if etype == EventTypes.Member: self._membership_stream_cache.entity_has_changed(state_key, stream_ordering) - self.get_invited_rooms_for_user.invalidate((state_key,)) + self.get_invited_rooms_for_local_user.invalidate((state_key,)) if relates_to: self.get_relations_for_event.invalidate_many((relates_to,)) diff --git a/synapse/replication/slave/storage/filtering.py b/synapse/replication/slave/storage/filtering.py index 5c84ebd125..bcb0688954 100644 --- a/synapse/replication/slave/storage/filtering.py +++ b/synapse/replication/slave/storage/filtering.py @@ -14,13 +14,14 @@ # limitations under the License. from synapse.storage.data_stores.main.filtering import FilteringStore +from synapse.storage.database import Database from ._base import BaseSlavedStore class SlavedFilteringStore(BaseSlavedStore): - def __init__(self, db_conn, hs): - super(SlavedFilteringStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SlavedFilteringStore, self).__init__(database, db_conn, hs) # Filters are immutable so this cache doesn't need to be expired get_user_filter = FilteringStore.__dict__["get_user_filter"] diff --git a/synapse/replication/slave/storage/groups.py b/synapse/replication/slave/storage/groups.py index 28a46edd28..69a4ae42f9 100644 --- a/synapse/replication/slave/storage/groups.py +++ b/synapse/replication/slave/storage/groups.py @@ -14,6 +14,7 @@ # limitations under the License. from synapse.storage import DataStore +from synapse.storage.database import Database from synapse.util.caches.stream_change_cache import StreamChangeCache from ._base import BaseSlavedStore, __func__ @@ -21,8 +22,8 @@ from ._slaved_id_tracker import SlavedIdTracker class SlavedGroupServerStore(BaseSlavedStore): - def __init__(self, db_conn, hs): - super(SlavedGroupServerStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SlavedGroupServerStore, self).__init__(database, db_conn, hs) self.hs = hs diff --git a/synapse/replication/slave/storage/presence.py b/synapse/replication/slave/storage/presence.py index 747ced0c84..ad8f0c15a9 100644 --- a/synapse/replication/slave/storage/presence.py +++ b/synapse/replication/slave/storage/presence.py @@ -15,6 +15,7 @@ from synapse.storage import DataStore from synapse.storage.data_stores.main.presence import PresenceStore +from synapse.storage.database import Database from synapse.util.caches.stream_change_cache import StreamChangeCache from ._base import BaseSlavedStore, __func__ @@ -22,13 +23,13 @@ from ._slaved_id_tracker import SlavedIdTracker class SlavedPresenceStore(BaseSlavedStore): - def __init__(self, db_conn, hs): - super(SlavedPresenceStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SlavedPresenceStore, self).__init__(database, db_conn, hs) self._presence_id_gen = SlavedIdTracker(db_conn, "presence_stream", "stream_id") self._presence_on_startup = self._get_active_presence(db_conn) - self.presence_stream_cache = self.presence_stream_cache = StreamChangeCache( + self.presence_stream_cache = StreamChangeCache( "PresenceStreamChangeCache", self._presence_id_gen.get_current_token() ) diff --git a/synapse/replication/slave/storage/push_rule.py b/synapse/replication/slave/storage/push_rule.py index 3655f05e54..eebd5a1fb6 100644 --- a/synapse/replication/slave/storage/push_rule.py +++ b/synapse/replication/slave/storage/push_rule.py @@ -15,17 +15,18 @@ # limitations under the License. from synapse.storage.data_stores.main.push_rule import PushRulesWorkerStore +from synapse.storage.database import Database from ._slaved_id_tracker import SlavedIdTracker from .events import SlavedEventStore class SlavedPushRuleStore(SlavedEventStore, PushRulesWorkerStore): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): self._push_rules_stream_id_gen = SlavedIdTracker( db_conn, "push_rules_stream", "stream_id" ) - super(SlavedPushRuleStore, self).__init__(db_conn, hs) + super(SlavedPushRuleStore, self).__init__(database, db_conn, hs) def get_push_rules_stream_token(self): return ( diff --git a/synapse/replication/slave/storage/pushers.py b/synapse/replication/slave/storage/pushers.py index b4331d0799..f22c2d44a3 100644 --- a/synapse/replication/slave/storage/pushers.py +++ b/synapse/replication/slave/storage/pushers.py @@ -15,14 +15,15 @@ # limitations under the License. from synapse.storage.data_stores.main.pusher import PusherWorkerStore +from synapse.storage.database import Database from ._base import BaseSlavedStore from ._slaved_id_tracker import SlavedIdTracker class SlavedPusherStore(PusherWorkerStore, BaseSlavedStore): - def __init__(self, db_conn, hs): - super(SlavedPusherStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SlavedPusherStore, self).__init__(database, db_conn, hs) self._pushers_id_gen = SlavedIdTracker( db_conn, "pushers", "id", extra_tables=[("deleted_pushers", "stream_id")] ) diff --git a/synapse/replication/slave/storage/receipts.py b/synapse/replication/slave/storage/receipts.py index 43d823c601..d40dc6e1f5 100644 --- a/synapse/replication/slave/storage/receipts.py +++ b/synapse/replication/slave/storage/receipts.py @@ -15,6 +15,7 @@ # limitations under the License. from synapse.storage.data_stores.main.receipts import ReceiptsWorkerStore +from synapse.storage.database import Database from ._base import BaseSlavedStore from ._slaved_id_tracker import SlavedIdTracker @@ -29,14 +30,14 @@ from ._slaved_id_tracker import SlavedIdTracker class SlavedReceiptsStore(ReceiptsWorkerStore, BaseSlavedStore): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): # We instantiate this first as the ReceiptsWorkerStore constructor # needs to be able to call get_max_receipt_stream_id self._receipts_id_gen = SlavedIdTracker( db_conn, "receipts_linearized", "stream_id" ) - super(SlavedReceiptsStore, self).__init__(db_conn, hs) + super(SlavedReceiptsStore, self).__init__(database, db_conn, hs) def get_max_receipt_stream_id(self): return self._receipts_id_gen.get_current_token() diff --git a/synapse/replication/slave/storage/room.py b/synapse/replication/slave/storage/room.py index d9ad386b28..3a20f45316 100644 --- a/synapse/replication/slave/storage/room.py +++ b/synapse/replication/slave/storage/room.py @@ -14,14 +14,15 @@ # limitations under the License. from synapse.storage.data_stores.main.room import RoomWorkerStore +from synapse.storage.database import Database from ._base import BaseSlavedStore from ._slaved_id_tracker import SlavedIdTracker class RoomStore(RoomWorkerStore, BaseSlavedStore): - def __init__(self, db_conn, hs): - super(RoomStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(RoomStore, self).__init__(database, db_conn, hs) self._public_room_id_gen = SlavedIdTracker( db_conn, "public_room_list_stream", "stream_id" ) diff --git a/synapse/replication/tcp/client.py b/synapse/replication/tcp/client.py index fead78388c..52a0aefe68 100644 --- a/synapse/replication/tcp/client.py +++ b/synapse/replication/tcp/client.py @@ -16,7 +16,7 @@ """ import logging -from typing import Dict +from typing import Dict, List, Optional from twisted.internet import defer from twisted.internet.protocol import ReconnectingClientFactory @@ -28,6 +28,7 @@ from synapse.replication.tcp.protocol import ( ) from .commands import ( + Command, FederationAckCommand, InvalidateCacheCommand, RemovePusherCommand, @@ -46,7 +47,8 @@ class ReplicationClientFactory(ReconnectingClientFactory): is required. """ - maxDelay = 30 # Try at least once every N seconds + initialDelay = 0.1 + maxDelay = 1 # Try at least once every N seconds def __init__(self, hs, client_name, handler: AbstractReplicationClientHandler): self.client_name = client_name @@ -88,15 +90,15 @@ class ReplicationClientHandler(AbstractReplicationClientHandler): # Any pending commands to be sent once a new connection has been # established - self.pending_commands = [] + self.pending_commands = [] # type: List[Command] # Map from string -> deferred, to wake up when receiveing a SYNC with # the given string. # Used for tests. - self.awaiting_syncs = {} + self.awaiting_syncs = {} # type: Dict[str, defer.Deferred] # The factory used to create connections. - self.factory = None + self.factory = None # type: Optional[ReplicationClientFactory] def start_replication(self, hs): """Helper method to start a replication connection to the remote server @@ -108,7 +110,7 @@ class ReplicationClientHandler(AbstractReplicationClientHandler): port = hs.config.worker_replication_port hs.get_reactor().connectTCP(host, port, self.factory) - def on_rdata(self, stream_name, token, rows): + async def on_rdata(self, stream_name, token, rows): """Called to handle a batch of replication data with a given stream token. By default this just pokes the slave store. Can be overridden in subclasses to @@ -119,20 +121,17 @@ class ReplicationClientHandler(AbstractReplicationClientHandler): token (int): stream token for this batch of rows rows (list): a list of Stream.ROW_TYPE objects as returned by Stream.parse_row. - - Returns: - Deferred|None """ logger.debug("Received rdata %s -> %s", stream_name, token) - return self.store.process_replication_rows(stream_name, token, rows) + self.store.process_replication_rows(stream_name, token, rows) - def on_position(self, stream_name, token): + async def on_position(self, stream_name, token): """Called when we get new position data. By default this just pokes the slave store. Can be overriden in subclasses to handle more. """ - return self.store.process_replication_rows(stream_name, token, []) + self.store.process_replication_rows(stream_name, token, []) def on_sync(self, data): """When we received a SYNC we wake up any deferreds that were waiting @@ -234,4 +233,5 @@ class ReplicationClientHandler(AbstractReplicationClientHandler): # We don't reset the delay any earlier as otherwise if there is a # problem during start up we'll end up tight looping connecting to the # server. - self.factory.resetDelay() + if self.factory: + self.factory.resetDelay() diff --git a/synapse/replication/tcp/commands.py b/synapse/replication/tcp/commands.py index 0ff2a7199f..cbb36b9acf 100644 --- a/synapse/replication/tcp/commands.py +++ b/synapse/replication/tcp/commands.py @@ -20,15 +20,16 @@ allowed to be sent by which side. import logging import platform +from typing import Tuple, Type if platform.python_implementation() == "PyPy": import json _json_encoder = json.JSONEncoder() else: - import simplejson as json + import simplejson as json # type: ignore[no-redef] # noqa: F821 - _json_encoder = json.JSONEncoder(namedtuple_as_object=False) + _json_encoder = json.JSONEncoder(namedtuple_as_object=False) # type: ignore[call-arg] # noqa: F821 logger = logging.getLogger(__name__) @@ -44,7 +45,7 @@ class Command(object): The default implementation creates a command of form `<NAME> <data>` """ - NAME = None + NAME = None # type: str def __init__(self, data): self.data = data @@ -386,25 +387,24 @@ class UserIpCommand(Command): ) +_COMMANDS = ( + ServerCommand, + RdataCommand, + PositionCommand, + ErrorCommand, + PingCommand, + NameCommand, + ReplicateCommand, + UserSyncCommand, + FederationAckCommand, + SyncCommand, + RemovePusherCommand, + InvalidateCacheCommand, + UserIpCommand, +) # type: Tuple[Type[Command], ...] + # Map of command name to command type. -COMMAND_MAP = { - cmd.NAME: cmd - for cmd in ( - ServerCommand, - RdataCommand, - PositionCommand, - ErrorCommand, - PingCommand, - NameCommand, - ReplicateCommand, - UserSyncCommand, - FederationAckCommand, - SyncCommand, - RemovePusherCommand, - InvalidateCacheCommand, - UserIpCommand, - ) -} +COMMAND_MAP = {cmd.NAME: cmd for cmd in _COMMANDS} # The commands the server is allowed to send VALID_SERVER_COMMANDS = ( diff --git a/synapse/replication/tcp/protocol.py b/synapse/replication/tcp/protocol.py index afaf002fe6..5f4bdf84d2 100644 --- a/synapse/replication/tcp/protocol.py +++ b/synapse/replication/tcp/protocol.py @@ -53,6 +53,7 @@ import fcntl import logging import struct from collections import defaultdict +from typing import Any, DefaultDict, Dict, List, Set, Tuple from six import iteritems, iterkeys @@ -65,13 +66,11 @@ from twisted.python.failure import Failure from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.metrics import LaterGauge from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.util import Clock -from synapse.util.stringutils import random_string - -from .commands import ( +from synapse.replication.tcp.commands import ( COMMAND_MAP, VALID_CLIENT_COMMANDS, VALID_SERVER_COMMANDS, + Command, ErrorCommand, NameCommand, PingCommand, @@ -82,7 +81,10 @@ from .commands import ( SyncCommand, UserSyncCommand, ) -from .streams import STREAMS_MAP +from synapse.replication.tcp.streams import STREAMS_MAP +from synapse.types import Collection +from synapse.util import Clock +from synapse.util.stringutils import random_string connection_close_counter = Counter( "synapse_replication_tcp_protocol_close_reason", "", ["reason_type"] @@ -124,8 +126,11 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver): delimiter = b"\n" - VALID_INBOUND_COMMANDS = [] # Valid commands we expect to receive - VALID_OUTBOUND_COMMANDS = [] # Valid commans we can send + # Valid commands we expect to receive + VALID_INBOUND_COMMANDS = [] # type: Collection[str] + + # Valid commands we can send + VALID_OUTBOUND_COMMANDS = [] # type: Collection[str] max_line_buffer = 10000 @@ -144,13 +149,13 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver): self.conn_id = random_string(5) # To dedupe in case of name clashes. # List of pending commands to send once we've established the connection - self.pending_commands = [] + self.pending_commands = [] # type: List[Command] # The LoopingCall for sending pings. self._send_ping_loop = None - self.inbound_commands_counter = defaultdict(int) - self.outbound_commands_counter = defaultdict(int) + self.inbound_commands_counter = defaultdict(int) # type: DefaultDict[str, int] + self.outbound_commands_counter = defaultdict(int) # type: DefaultDict[str, int] def connectionMade(self): logger.info("[%s] Connection established", self.id()) @@ -235,19 +240,16 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver): "replication-" + cmd.get_logcontext_id(), self.handle_command, cmd ) - def handle_command(self, cmd): + async def handle_command(self, cmd: Command): """Handle a command we have received over the replication stream. - By default delegates to on_<COMMAND> + By default delegates to on_<COMMAND>, which should return an awaitable. Args: - cmd (synapse.replication.tcp.commands.Command): received command - - Returns: - Deferred + cmd: received command """ handler = getattr(self, "on_%s" % (cmd.NAME,)) - return handler(cmd) + await handler(cmd) def close(self): logger.warning("[%s] Closing connection", self.id()) @@ -320,10 +322,10 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver): for cmd in pending: self.send_command(cmd) - def on_PING(self, line): + async def on_PING(self, line): self.received_ping = True - def on_ERROR(self, cmd): + async def on_ERROR(self, cmd): logger.error("[%s] Remote reported error: %r", self.id(), cmd.data) def pauseProducing(self): @@ -409,30 +411,30 @@ class ServerReplicationStreamProtocol(BaseReplicationStreamProtocol): self.streamer = streamer # The streams the client has subscribed to and is up to date with - self.replication_streams = set() + self.replication_streams = set() # type: Set[str] # The streams the client is currently subscribing to. - self.connecting_streams = set() + self.connecting_streams = set() # type: Set[str] # Map from stream name to list of updates to send once we've finished # subscribing the client to the stream. - self.pending_rdata = {} + self.pending_rdata = {} # type: Dict[str, List[Tuple[int, Any]]] def connectionMade(self): self.send_command(ServerCommand(self.server_name)) BaseReplicationStreamProtocol.connectionMade(self) self.streamer.new_connection(self) - def on_NAME(self, cmd): + async def on_NAME(self, cmd): logger.info("[%s] Renamed to %r", self.id(), cmd.data) self.name = cmd.data - def on_USER_SYNC(self, cmd): - return self.streamer.on_user_sync( + async def on_USER_SYNC(self, cmd): + await self.streamer.on_user_sync( self.conn_id, cmd.user_id, cmd.is_syncing, cmd.last_sync_ms ) - def on_REPLICATE(self, cmd): + async def on_REPLICATE(self, cmd): stream_name = cmd.stream_name token = cmd.token @@ -443,23 +445,23 @@ class ServerReplicationStreamProtocol(BaseReplicationStreamProtocol): for stream in iterkeys(self.streamer.streams_by_name) ] - return make_deferred_yieldable( + await make_deferred_yieldable( defer.gatherResults(deferreds, consumeErrors=True) ) else: - return self.subscribe_to_stream(stream_name, token) + await self.subscribe_to_stream(stream_name, token) - def on_FEDERATION_ACK(self, cmd): - return self.streamer.federation_ack(cmd.token) + async def on_FEDERATION_ACK(self, cmd): + self.streamer.federation_ack(cmd.token) - def on_REMOVE_PUSHER(self, cmd): - return self.streamer.on_remove_pusher(cmd.app_id, cmd.push_key, cmd.user_id) + async def on_REMOVE_PUSHER(self, cmd): + await self.streamer.on_remove_pusher(cmd.app_id, cmd.push_key, cmd.user_id) - def on_INVALIDATE_CACHE(self, cmd): - return self.streamer.on_invalidate_cache(cmd.cache_func, cmd.keys) + async def on_INVALIDATE_CACHE(self, cmd): + self.streamer.on_invalidate_cache(cmd.cache_func, cmd.keys) - def on_USER_IP(self, cmd): - return self.streamer.on_user_ip( + async def on_USER_IP(self, cmd): + self.streamer.on_user_ip( cmd.user_id, cmd.access_token, cmd.ip, @@ -468,8 +470,7 @@ class ServerReplicationStreamProtocol(BaseReplicationStreamProtocol): cmd.last_seen, ) - @defer.inlineCallbacks - def subscribe_to_stream(self, stream_name, token): + async def subscribe_to_stream(self, stream_name, token): """Subscribe the remote to a stream. This invloves checking if they've missed anything and sending those @@ -481,7 +482,7 @@ class ServerReplicationStreamProtocol(BaseReplicationStreamProtocol): try: # Get missing updates - updates, current_token = yield self.streamer.get_stream_updates( + updates, current_token = await self.streamer.get_stream_updates( stream_name, token ) @@ -566,7 +567,7 @@ class AbstractReplicationClientHandler(metaclass=abc.ABCMeta): """ @abc.abstractmethod - def on_rdata(self, stream_name, token, rows): + async def on_rdata(self, stream_name, token, rows): """Called to handle a batch of replication data with a given stream token. Args: @@ -574,14 +575,11 @@ class AbstractReplicationClientHandler(metaclass=abc.ABCMeta): token (int): stream token for this batch of rows rows (list): a list of Stream.ROW_TYPE objects as returned by Stream.parse_row. - - Returns: - Deferred|None """ raise NotImplementedError() @abc.abstractmethod - def on_position(self, stream_name, token): + async def on_position(self, stream_name, token): """Called when we get new position data.""" raise NotImplementedError() @@ -642,11 +640,11 @@ class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol): # Set of stream names that have been subscribe to, but haven't yet # caught up with. This is used to track when the client has been fully # connected to the remote. - self.streams_connecting = set() + self.streams_connecting = set() # type: Set[str] # Map of stream to batched updates. See RdataCommand for info on how # batching works. - self.pending_batches = {} + self.pending_batches = {} # type: Dict[str, Any] def connectionMade(self): self.send_command(NameCommand(self.client_name)) @@ -670,12 +668,12 @@ class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol): if not self.streams_connecting: self.handler.finished_connecting() - def on_SERVER(self, cmd): + async def on_SERVER(self, cmd): if cmd.data != self.server_name: logger.error("[%s] Connected to wrong remote: %r", self.id(), cmd.data) self.send_error("Wrong remote") - def on_RDATA(self, cmd): + async def on_RDATA(self, cmd): stream_name = cmd.stream_name inbound_rdata_count.labels(stream_name).inc() @@ -695,19 +693,19 @@ class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol): # Check if this is the last of a batch of updates rows = self.pending_batches.pop(stream_name, []) rows.append(row) - return self.handler.on_rdata(stream_name, cmd.token, rows) + await self.handler.on_rdata(stream_name, cmd.token, rows) - def on_POSITION(self, cmd): + async def on_POSITION(self, cmd): # When we get a `POSITION` command it means we've finished getting # missing updates for the given stream, and are now up to date. self.streams_connecting.discard(cmd.stream_name) if not self.streams_connecting: self.handler.finished_connecting() - return self.handler.on_position(cmd.stream_name, cmd.token) + await self.handler.on_position(cmd.stream_name, cmd.token) - def on_SYNC(self, cmd): - return self.handler.on_sync(cmd.data) + async def on_SYNC(self, cmd): + self.handler.on_sync(cmd.data) def replicate(self, stream_name, token): """Send the subscription request to the server @@ -766,7 +764,7 @@ def transport_kernel_read_buffer_size(protocol, read=True): op = SIOCINQ else: op = SIOCOUTQ - size = struct.unpack("I", fcntl.ioctl(fileno, op, "\0\0\0\0"))[0] + size = struct.unpack("I", fcntl.ioctl(fileno, op, b"\0\0\0\0"))[0] return size return 0 diff --git a/synapse/replication/tcp/resource.py b/synapse/replication/tcp/resource.py index d1e98428bc..b1752e88cd 100644 --- a/synapse/replication/tcp/resource.py +++ b/synapse/replication/tcp/resource.py @@ -17,12 +17,12 @@ import logging import random +from typing import List from six import itervalues from prometheus_client import Counter -from twisted.internet import defer from twisted.internet.protocol import Factory from synapse.metrics import LaterGauge @@ -79,7 +79,7 @@ class ReplicationStreamer(object): self._replication_torture_level = hs.config.replication_torture_level # Current connections. - self.connections = [] + self.connections = [] # type: List[ServerReplicationStreamProtocol] LaterGauge( "synapse_replication_tcp_resource_total_connections", @@ -154,8 +154,7 @@ class ReplicationStreamer(object): run_as_background_process("replication_notifier", self._run_notifier_loop) - @defer.inlineCallbacks - def _run_notifier_loop(self): + async def _run_notifier_loop(self): self.is_looping = True try: @@ -184,7 +183,7 @@ class ReplicationStreamer(object): continue if self._replication_torture_level: - yield self.clock.sleep( + await self.clock.sleep( self._replication_torture_level / 1000.0 ) @@ -195,7 +194,7 @@ class ReplicationStreamer(object): stream.upto_token, ) try: - updates, current_token = yield stream.get_updates() + updates, current_token = await stream.get_updates() except Exception: logger.info("Failed to handle stream %s", stream.NAME) raise @@ -232,7 +231,7 @@ class ReplicationStreamer(object): self.is_looping = False @measure_func("repl.get_stream_updates") - def get_stream_updates(self, stream_name, token): + async def get_stream_updates(self, stream_name, token): """For a given stream get all updates since token. This is called when a client first subscribes to a stream. """ @@ -240,7 +239,7 @@ class ReplicationStreamer(object): if not stream: raise Exception("unknown stream %s", stream_name) - return stream.get_updates_since(token) + return await stream.get_updates_since(token) @measure_func("repl.federation_ack") def federation_ack(self, token): @@ -251,22 +250,20 @@ class ReplicationStreamer(object): self.federation_sender.federation_ack(token) @measure_func("repl.on_user_sync") - @defer.inlineCallbacks - def on_user_sync(self, conn_id, user_id, is_syncing, last_sync_ms): + async def on_user_sync(self, conn_id, user_id, is_syncing, last_sync_ms): """A client has started/stopped syncing on a worker. """ user_sync_counter.inc() - yield self.presence_handler.update_external_syncs_row( + await self.presence_handler.update_external_syncs_row( conn_id, user_id, is_syncing, last_sync_ms ) @measure_func("repl.on_remove_pusher") - @defer.inlineCallbacks - def on_remove_pusher(self, app_id, push_key, user_id): + async def on_remove_pusher(self, app_id, push_key, user_id): """A client has asked us to remove a pusher """ remove_pusher_counter.inc() - yield self.store.delete_pusher_by_app_id_pushkey_user_id( + await self.store.delete_pusher_by_app_id_pushkey_user_id( app_id=app_id, pushkey=push_key, user_id=user_id ) @@ -280,15 +277,16 @@ class ReplicationStreamer(object): getattr(self.store, cache_func).invalidate(tuple(keys)) @measure_func("repl.on_user_ip") - @defer.inlineCallbacks - def on_user_ip(self, user_id, access_token, ip, user_agent, device_id, last_seen): + async def on_user_ip( + self, user_id, access_token, ip, user_agent, device_id, last_seen + ): """The client saw a user request """ user_ip_cache_counter.inc() - yield self.store.insert_client_ip( + await self.store.insert_client_ip( user_id, access_token, ip, user_agent, device_id, last_seen ) - yield self._server_notices_sender.on_user_ip(user_id) + await self._server_notices_sender.on_user_ip(user_id) def send_sync_to_all_connections(self, data): """Sends a SYNC command to all clients. diff --git a/synapse/replication/tcp/streams/_base.py b/synapse/replication/tcp/streams/_base.py index 9e45429d49..e03e77199b 100644 --- a/synapse/replication/tcp/streams/_base.py +++ b/synapse/replication/tcp/streams/_base.py @@ -14,12 +14,10 @@ # See the License for the specific language governing permissions and # limitations under the License. - import itertools import logging from collections import namedtuple - -from twisted.internet import defer +from typing import Any logger = logging.getLogger(__name__) @@ -88,8 +86,7 @@ TagAccountDataStreamRow = namedtuple( "TagAccountDataStreamRow", ("user_id", "room_id", "data") # str # str # dict ) AccountDataStreamRow = namedtuple( - "AccountDataStream", - ("user_id", "room_id", "data_type", "data"), # str # str # str # dict + "AccountDataStream", ("user_id", "room_id", "data_type") # str # str # str ) GroupsStreamRow = namedtuple( "GroupsStreamRow", @@ -105,8 +102,9 @@ class Stream(object): time it was called up until the point `advance_current_token` was called. """ - NAME = None # The name of the stream - ROW_TYPE = None # The type of the row. Used by the default impl of parse_row. + NAME = None # type: str # The name of the stream + # The type of the row. Used by the default impl of parse_row. + ROW_TYPE = None # type: Any _LIMITED = True # Whether the update function takes a limit @classmethod @@ -144,8 +142,7 @@ class Stream(object): self.upto_token = self.current_token() self.last_token = self.upto_token - @defer.inlineCallbacks - def get_updates(self): + async def get_updates(self): """Gets all updates since the last time this function was called (or since the stream was constructed if it hadn't been called before), until the `upto_token` @@ -156,13 +153,12 @@ class Stream(object): list of ``(token, row)`` entries. ``row`` will be json-serialised and sent over the replication steam. """ - updates, current_token = yield self.get_updates_since(self.last_token) + updates, current_token = await self.get_updates_since(self.last_token) self.last_token = current_token return updates, current_token - @defer.inlineCallbacks - def get_updates_since(self, from_token): + async def get_updates_since(self, from_token): """Like get_updates except allows specifying from when we should stream updates @@ -182,15 +178,16 @@ class Stream(object): if from_token == current_token: return [], current_token + logger.info("get_updates_since: %s", self.__class__) if self._LIMITED: - rows = yield self.update_function( + rows = await self.update_function( from_token, current_token, limit=MAX_EVENTS_BEHIND + 1 ) # never turn more than MAX_EVENTS_BEHIND + 1 into updates. rows = itertools.islice(rows, MAX_EVENTS_BEHIND + 1) else: - rows = yield self.update_function(from_token, current_token) + rows = await self.update_function(from_token, current_token) updates = [(row[0], row[1:]) for row in rows] @@ -232,8 +229,8 @@ class BackfillStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_current_backfill_token - self.update_function = store.get_all_new_backfill_event_rows + self.current_token = store.get_current_backfill_token # type: ignore + self.update_function = store.get_all_new_backfill_event_rows # type: ignore super(BackfillStream, self).__init__(hs) @@ -247,8 +244,8 @@ class PresenceStream(Stream): store = hs.get_datastore() presence_handler = hs.get_presence_handler() - self.current_token = store.get_current_presence_token - self.update_function = presence_handler.get_all_presence_updates + self.current_token = store.get_current_presence_token # type: ignore + self.update_function = presence_handler.get_all_presence_updates # type: ignore super(PresenceStream, self).__init__(hs) @@ -261,8 +258,8 @@ class TypingStream(Stream): def __init__(self, hs): typing_handler = hs.get_typing_handler() - self.current_token = typing_handler.get_current_token - self.update_function = typing_handler.get_all_typing_updates + self.current_token = typing_handler.get_current_token # type: ignore + self.update_function = typing_handler.get_all_typing_updates # type: ignore super(TypingStream, self).__init__(hs) @@ -274,8 +271,8 @@ class ReceiptsStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_max_receipt_stream_id - self.update_function = store.get_all_updated_receipts + self.current_token = store.get_max_receipt_stream_id # type: ignore + self.update_function = store.get_all_updated_receipts # type: ignore super(ReceiptsStream, self).__init__(hs) @@ -295,9 +292,8 @@ class PushRulesStream(Stream): push_rules_token, _ = self.store.get_push_rules_stream_token() return push_rules_token - @defer.inlineCallbacks - def update_function(self, from_token, to_token, limit): - rows = yield self.store.get_all_push_rule_updates(from_token, to_token, limit) + async def update_function(self, from_token, to_token, limit): + rows = await self.store.get_all_push_rule_updates(from_token, to_token, limit) return [(row[0], row[2]) for row in rows] @@ -311,8 +307,8 @@ class PushersStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_pushers_stream_token - self.update_function = store.get_all_updated_pushers_rows + self.current_token = store.get_pushers_stream_token # type: ignore + self.update_function = store.get_all_updated_pushers_rows # type: ignore super(PushersStream, self).__init__(hs) @@ -328,8 +324,8 @@ class CachesStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_cache_stream_token - self.update_function = store.get_all_updated_caches + self.current_token = store.get_cache_stream_token # type: ignore + self.update_function = store.get_all_updated_caches # type: ignore super(CachesStream, self).__init__(hs) @@ -344,8 +340,8 @@ class PublicRoomsStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_current_public_room_stream_id - self.update_function = store.get_all_new_public_rooms + self.current_token = store.get_current_public_room_stream_id # type: ignore + self.update_function = store.get_all_new_public_rooms # type: ignore super(PublicRoomsStream, self).__init__(hs) @@ -361,8 +357,8 @@ class DeviceListsStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_device_stream_token - self.update_function = store.get_all_device_list_changes_for_remotes + self.current_token = store.get_device_stream_token # type: ignore + self.update_function = store.get_all_device_list_changes_for_remotes # type: ignore super(DeviceListsStream, self).__init__(hs) @@ -377,8 +373,8 @@ class ToDeviceStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_to_device_stream_token - self.update_function = store.get_all_new_device_messages + self.current_token = store.get_to_device_stream_token # type: ignore + self.update_function = store.get_all_new_device_messages # type: ignore super(ToDeviceStream, self).__init__(hs) @@ -393,8 +389,8 @@ class TagAccountDataStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_max_account_data_stream_id - self.update_function = store.get_all_updated_tags + self.current_token = store.get_max_account_data_stream_id # type: ignore + self.update_function = store.get_all_updated_tags # type: ignore super(TagAccountDataStream, self).__init__(hs) @@ -409,20 +405,19 @@ class AccountDataStream(Stream): def __init__(self, hs): self.store = hs.get_datastore() - self.current_token = self.store.get_max_account_data_stream_id + self.current_token = self.store.get_max_account_data_stream_id # type: ignore super(AccountDataStream, self).__init__(hs) - @defer.inlineCallbacks - def update_function(self, from_token, to_token, limit): - global_results, room_results = yield self.store.get_all_updated_account_data( + async def update_function(self, from_token, to_token, limit): + global_results, room_results = await self.store.get_all_updated_account_data( from_token, from_token, to_token, limit ) results = list(room_results) results.extend( - (stream_id, user_id, None, account_data_type, content) - for stream_id, user_id, account_data_type, content in global_results + (stream_id, user_id, None, account_data_type) + for stream_id, user_id, account_data_type in global_results ) return results @@ -435,8 +430,8 @@ class GroupServerStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_group_stream_token - self.update_function = store.get_all_groups_changes + self.current_token = store.get_group_stream_token # type: ignore + self.update_function = store.get_all_groups_changes # type: ignore super(GroupServerStream, self).__init__(hs) @@ -452,7 +447,7 @@ class UserSignatureStream(Stream): def __init__(self, hs): store = hs.get_datastore() - self.current_token = store.get_device_stream_token - self.update_function = store.get_all_user_signature_changes_for_remotes + self.current_token = store.get_device_stream_token # type: ignore + self.update_function = store.get_all_user_signature_changes_for_remotes # type: ignore super(UserSignatureStream, self).__init__(hs) diff --git a/synapse/replication/tcp/streams/events.py b/synapse/replication/tcp/streams/events.py index d97669c886..b3afabb8cd 100644 --- a/synapse/replication/tcp/streams/events.py +++ b/synapse/replication/tcp/streams/events.py @@ -13,12 +13,12 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. + import heapq +from typing import Tuple, Type import attr -from twisted.internet import defer - from ._base import Stream @@ -63,7 +63,8 @@ class BaseEventsStreamRow(object): Specifies how to identify, serialize and deserialize the different types. """ - TypeId = None # Unique string that ids the type. Must be overriden in sub classes. + # Unique string that ids the type. Must be overriden in sub classes. + TypeId = None # type: str @classmethod def from_data(cls, data): @@ -99,9 +100,12 @@ class EventsStreamCurrentStateRow(BaseEventsStreamRow): event_id = attr.ib() # str, optional -TypeToRow = { - Row.TypeId: Row for Row in (EventsStreamEventRow, EventsStreamCurrentStateRow) -} +_EventRows = ( + EventsStreamEventRow, + EventsStreamCurrentStateRow, +) # type: Tuple[Type[BaseEventsStreamRow], ...] + +TypeToRow = {Row.TypeId: Row for Row in _EventRows} class EventsStream(Stream): @@ -112,20 +116,19 @@ class EventsStream(Stream): def __init__(self, hs): self._store = hs.get_datastore() - self.current_token = self._store.get_current_events_token + self.current_token = self._store.get_current_events_token # type: ignore super(EventsStream, self).__init__(hs) - @defer.inlineCallbacks - def update_function(self, from_token, current_token, limit=None): - event_rows = yield self._store.get_all_new_forward_event_rows( + async def update_function(self, from_token, current_token, limit=None): + event_rows = await self._store.get_all_new_forward_event_rows( from_token, current_token, limit ) event_updates = ( (row[0], EventsStreamEventRow.TypeId, row[1:]) for row in event_rows ) - state_rows = yield self._store.get_all_updated_current_state_deltas( + state_rows = await self._store.get_all_updated_current_state_deltas( from_token, current_token, limit ) state_updates = ( diff --git a/synapse/replication/tcp/streams/federation.py b/synapse/replication/tcp/streams/federation.py index dc2484109d..615f3dc9ac 100644 --- a/synapse/replication/tcp/streams/federation.py +++ b/synapse/replication/tcp/streams/federation.py @@ -37,7 +37,7 @@ class FederationStream(Stream): def __init__(self, hs): federation_sender = hs.get_federation_sender() - self.current_token = federation_sender.get_current_token - self.update_function = federation_sender.get_replication_rows + self.current_token = federation_sender.get_current_token # type: ignore + self.update_function = federation_sender.get_replication_rows # type: ignore super(FederationStream, self).__init__(hs) diff --git a/synapse/rest/admin/__init__.py b/synapse/rest/admin/__init__.py index 68a59a3424..a10b4a9b72 100644 --- a/synapse/rest/admin/__init__.py +++ b/synapse/rest/admin/__init__.py @@ -34,12 +34,13 @@ from synapse.rest.admin.server_notice_servlet import SendServerNoticeServlet from synapse.rest.admin.users import ( AccountValidityRenewServlet, DeactivateAccountRestServlet, - GetUsersPaginatedRestServlet, ResetPasswordRestServlet, SearchUsersRestServlet, UserAdminServlet, UserRegisterServlet, + UserRestServletV2, UsersRestServlet, + UsersRestServletV2, WhoisRestServlet, ) from synapse.util.versionstring import get_version_string @@ -191,6 +192,8 @@ def register_servlets(hs, http_server): SendServerNoticeServlet(hs).register(http_server) VersionServlet(hs).register(http_server) UserAdminServlet(hs).register(http_server) + UserRestServletV2(hs).register(http_server) + UsersRestServletV2(hs).register(http_server) def register_servlets_for_client_rest_resource(hs, http_server): @@ -201,7 +204,6 @@ def register_servlets_for_client_rest_resource(hs, http_server): PurgeHistoryRestServlet(hs).register(http_server) UsersRestServlet(hs).register(http_server) ResetPasswordRestServlet(hs).register(http_server) - GetUsersPaginatedRestServlet(hs).register(http_server) SearchUsersRestServlet(hs).register(http_server) ShutdownRoomRestServlet(hs).register(http_server) UserRegisterServlet(hs).register(http_server) diff --git a/synapse/rest/admin/media.py b/synapse/rest/admin/media.py index fa833e54cf..3a445d6eed 100644 --- a/synapse/rest/admin/media.py +++ b/synapse/rest/admin/media.py @@ -32,16 +32,24 @@ class QuarantineMediaInRoom(RestServlet): this server. """ - PATTERNS = historical_admin_path_patterns("/quarantine_media/(?P<room_id>[^/]+)") + PATTERNS = ( + historical_admin_path_patterns("/room/(?P<room_id>[^/]+)/media/quarantine") + + + # This path kept around for legacy reasons + historical_admin_path_patterns("/quarantine_media/(?P<room_id>![^/]+)") + ) def __init__(self, hs): self.store = hs.get_datastore() self.auth = hs.get_auth() - async def on_POST(self, request, room_id): + async def on_POST(self, request, room_id: str): requester = await self.auth.get_user_by_req(request) await assert_user_is_admin(self.auth, requester.user) + logging.info("Quarantining room: %s", room_id) + + # Quarantine all media in this room num_quarantined = await self.store.quarantine_media_ids_in_room( room_id, requester.user.to_string() ) @@ -49,6 +57,60 @@ class QuarantineMediaInRoom(RestServlet): return 200, {"num_quarantined": num_quarantined} +class QuarantineMediaByUser(RestServlet): + """Quarantines all local media by a given user so that no one can download it via + this server. + """ + + PATTERNS = historical_admin_path_patterns( + "/user/(?P<user_id>[^/]+)/media/quarantine" + ) + + def __init__(self, hs): + self.store = hs.get_datastore() + self.auth = hs.get_auth() + + async def on_POST(self, request, user_id: str): + requester = await self.auth.get_user_by_req(request) + await assert_user_is_admin(self.auth, requester.user) + + logging.info("Quarantining local media by user: %s", user_id) + + # Quarantine all media this user has uploaded + num_quarantined = await self.store.quarantine_media_ids_by_user( + user_id, requester.user.to_string() + ) + + return 200, {"num_quarantined": num_quarantined} + + +class QuarantineMediaByID(RestServlet): + """Quarantines local or remote media by a given ID so that no one can download + it via this server. + """ + + PATTERNS = historical_admin_path_patterns( + "/media/quarantine/(?P<server_name>[^/]+)/(?P<media_id>[^/]+)" + ) + + def __init__(self, hs): + self.store = hs.get_datastore() + self.auth = hs.get_auth() + + async def on_POST(self, request, server_name: str, media_id: str): + requester = await self.auth.get_user_by_req(request) + await assert_user_is_admin(self.auth, requester.user) + + logging.info("Quarantining local media by ID: %s/%s", server_name, media_id) + + # Quarantine this media id + await self.store.quarantine_media_by_id( + server_name, media_id, requester.user.to_string() + ) + + return 200, {} + + class ListMediaInRoom(RestServlet): """Lists all of the media in a given room. """ @@ -94,4 +156,6 @@ def register_servlets_for_media_repo(hs, http_server): """ PurgeMediaCacheRestServlet(hs).register(http_server) QuarantineMediaInRoom(hs).register(http_server) + QuarantineMediaByID(hs).register(http_server) + QuarantineMediaByUser(hs).register(http_server) ListMediaInRoom(hs).register(http_server) diff --git a/synapse/rest/admin/users.py b/synapse/rest/admin/users.py index 58a83f93af..574cb90c74 100644 --- a/synapse/rest/admin/users.py +++ b/synapse/rest/admin/users.py @@ -25,6 +25,7 @@ from synapse.api.errors import Codes, SynapseError from synapse.http.servlet import ( RestServlet, assert_params_in_dict, + parse_boolean, parse_integer, parse_json_object_from_request, parse_string, @@ -59,73 +60,189 @@ class UsersRestServlet(RestServlet): return 200, ret -class GetUsersPaginatedRestServlet(RestServlet): - """Get request to get specific number of users from Synapse. +class UsersRestServletV2(RestServlet): + PATTERNS = (re.compile("^/_synapse/admin/v2/users$"),) + + """Get request to list all local users. This needs user to have administrator access in Synapse. - Example: - http://localhost:8008/_synapse/admin/v1/users_paginate/ - @admin:user?access_token=admin_access_token&start=0&limit=10 - Returns: - 200 OK with json object {list[dict[str, Any]], count} or empty object. - """ - PATTERNS = historical_admin_path_patterns( - "/users_paginate/(?P<target_user_id>[^/]*)" - ) + GET /_synapse/admin/v2/users?from=0&limit=10&guests=false + + returns: + 200 OK with list of users if success otherwise an error. + + The parameters `from` and `limit` are required only for pagination. + By default, a `limit` of 100 is used. + The parameter `user_id` can be used to filter by user id. + The parameter `guests` can be used to exclude guest users. + The parameter `deactivated` can be used to include deactivated users. + """ def __init__(self, hs): - self.store = hs.get_datastore() self.hs = hs self.auth = hs.get_auth() - self.handlers = hs.get_handlers() + self.admin_handler = hs.get_handlers().admin_handler - async def on_GET(self, request, target_user_id): - """Get request to get specific number of users from Synapse. - This needs user to have administrator access in Synapse. - """ + async def on_GET(self, request): await assert_requester_is_admin(self.auth, request) - target_user = UserID.from_string(target_user_id) + start = parse_integer(request, "from", default=0) + limit = parse_integer(request, "limit", default=100) + user_id = parse_string(request, "user_id", default=None) + guests = parse_boolean(request, "guests", default=True) + deactivated = parse_boolean(request, "deactivated", default=False) - if not self.hs.is_mine(target_user): - raise SynapseError(400, "Can only users a local user") + users = await self.admin_handler.get_users_paginate( + start, limit, user_id, guests, deactivated + ) + ret = {"users": users} + if len(users) >= limit: + ret["next_token"] = str(start + len(users)) - order = "name" # order by name in user table - start = parse_integer(request, "start", required=True) - limit = parse_integer(request, "limit", required=True) + return 200, ret - logger.info("limit: %s, start: %s", limit, start) - ret = await self.handlers.admin_handler.get_users_paginate(order, start, limit) - return 200, ret +class UserRestServletV2(RestServlet): + PATTERNS = (re.compile("^/_synapse/admin/v2/users/(?P<user_id>@[^/]+)$"),) - async def on_POST(self, request, target_user_id): - """Post request to get specific number of users from Synapse.. - This needs user to have administrator access in Synapse. - Example: - http://localhost:8008/_synapse/admin/v1/users_paginate/ - @admin:user?access_token=admin_access_token - JsonBodyToSend: - { - "start": "0", - "limit": "10 - } - Returns: - 200 OK with json object {list[dict[str, Any]], count} or empty object. - """ + """Get request to list user details. + This needs user to have administrator access in Synapse. + + GET /_synapse/admin/v2/users/<user_id> + + returns: + 200 OK with user details if success otherwise an error. + + Put request to allow an administrator to add or modify a user. + This needs user to have administrator access in Synapse. + We use PUT instead of POST since we already know the id of the user + object to create. POST could be used to create guests. + + PUT /_synapse/admin/v2/users/<user_id> + { + "password": "secret", + "displayname": "User" + } + + returns: + 201 OK with new user object if user was created or + 200 OK with modified user object if user was modified + otherwise an error. + """ + + def __init__(self, hs): + self.hs = hs + self.auth = hs.get_auth() + self.admin_handler = hs.get_handlers().admin_handler + self.profile_handler = hs.get_profile_handler() + self.set_password_handler = hs.get_set_password_handler() + self.deactivate_account_handler = hs.get_deactivate_account_handler() + self.registration_handler = hs.get_registration_handler() + + async def on_GET(self, request, user_id): await assert_requester_is_admin(self.auth, request) - UserID.from_string(target_user_id) - order = "name" # order by name in user table - params = parse_json_object_from_request(request) - assert_params_in_dict(params, ["limit", "start"]) - limit = params["limit"] - start = params["start"] - logger.info("limit: %s, start: %s", limit, start) + target_user = UserID.from_string(user_id) + if not self.hs.is_mine(target_user): + raise SynapseError(400, "Can only lookup local users") + + ret = await self.admin_handler.get_user(target_user) - ret = await self.handlers.admin_handler.get_users_paginate(order, start, limit) return 200, ret + async def on_PUT(self, request, user_id): + await assert_requester_is_admin(self.auth, request) + + target_user = UserID.from_string(user_id) + body = parse_json_object_from_request(request) + + if not self.hs.is_mine(target_user): + raise SynapseError(400, "This endpoint can only be used with local users") + + user = await self.admin_handler.get_user(target_user) + + if user: # modify user + requester = await self.auth.get_user_by_req(request) + + if "displayname" in body: + await self.profile_handler.set_displayname( + target_user, requester, body["displayname"], True + ) + + if "avatar_url" in body: + await self.profile_handler.set_avatar_url( + target_user, requester, body["avatar_url"], True + ) + + if "admin" in body: + set_admin_to = bool(body["admin"]) + if set_admin_to != user["admin"]: + auth_user = requester.user + if target_user == auth_user and not set_admin_to: + raise SynapseError(400, "You may not demote yourself.") + + await self.admin_handler.set_user_server_admin( + target_user, set_admin_to + ) + + if "password" in body: + if ( + not isinstance(body["password"], text_type) + or len(body["password"]) > 512 + ): + raise SynapseError(400, "Invalid password") + else: + new_password = body["password"] + await self._set_password_handler.set_password( + target_user, new_password, requester + ) + + if "deactivated" in body: + deactivate = bool(body["deactivated"]) + if deactivate and not user["deactivated"]: + result = await self.deactivate_account_handler.deactivate_account( + target_user.to_string(), False + ) + if not result: + raise SynapseError(500, "Could not deactivate user") + + user = await self.admin_handler.get_user(target_user) + return 200, user + + else: # create user + if "password" not in body: + raise SynapseError( + 400, "password must be specified", errcode=Codes.BAD_JSON + ) + elif ( + not isinstance(body["password"], text_type) + or len(body["password"]) > 512 + ): + raise SynapseError(400, "Invalid password") + + admin = body.get("admin", None) + user_type = body.get("user_type", None) + displayname = body.get("displayname", None) + + if user_type is not None and user_type not in UserTypes.ALL_USER_TYPES: + raise SynapseError(400, "Invalid user type") + + user_id = await self.registration_handler.register_user( + localpart=target_user.localpart, + password=body["password"], + admin=bool(admin), + default_display_name=displayname, + user_type=user_type, + ) + if "avatar_url" in body: + await self.profile_handler.set_avatar_url( + user_id, requester, body["avatar_url"], True + ) + + ret = await self.admin_handler.get_user(target_user) + + return 201, ret + class UserRegisterServlet(RestServlet): """ diff --git a/synapse/rest/client/v1/directory.py b/synapse/rest/client/v1/directory.py index 4ea3666874..5934b1fe8b 100644 --- a/synapse/rest/client/v1/directory.py +++ b/synapse/rest/client/v1/directory.py @@ -16,8 +16,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import ( AuthError, Codes, @@ -47,17 +45,15 @@ class ClientDirectoryServer(RestServlet): self.handlers = hs.get_handlers() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request, room_alias): + async def on_GET(self, request, room_alias): room_alias = RoomAlias.from_string(room_alias) dir_handler = self.handlers.directory_handler - res = yield dir_handler.get_association(room_alias) + res = await dir_handler.get_association(room_alias) return 200, res - @defer.inlineCallbacks - def on_PUT(self, request, room_alias): + async def on_PUT(self, request, room_alias): room_alias = RoomAlias.from_string(room_alias) content = parse_json_object_from_request(request) @@ -77,26 +73,25 @@ class ClientDirectoryServer(RestServlet): # TODO(erikj): Check types. - room = yield self.store.get_room(room_id) + room = await self.store.get_room(room_id) if room is None: raise SynapseError(400, "Room does not exist") - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) - yield self.handlers.directory_handler.create_association( + await self.handlers.directory_handler.create_association( requester, room_alias, room_id, servers ) return 200, {} - @defer.inlineCallbacks - def on_DELETE(self, request, room_alias): + async def on_DELETE(self, request, room_alias): dir_handler = self.handlers.directory_handler try: - service = yield self.auth.get_appservice_by_req(request) + service = await self.auth.get_appservice_by_req(request) room_alias = RoomAlias.from_string(room_alias) - yield dir_handler.delete_appservice_association(service, room_alias) + await dir_handler.delete_appservice_association(service, room_alias) logger.info( "Application service at %s deleted alias %s", service.url, @@ -107,12 +102,12 @@ class ClientDirectoryServer(RestServlet): # fallback to default user behaviour if they aren't an AS pass - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) user = requester.user room_alias = RoomAlias.from_string(room_alias) - yield dir_handler.delete_association(requester, room_alias) + await dir_handler.delete_association(requester, room_alias) logger.info( "User %s deleted alias %s", user.to_string(), room_alias.to_string() @@ -130,32 +125,29 @@ class ClientDirectoryListServer(RestServlet): self.handlers = hs.get_handlers() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request, room_id): - room = yield self.store.get_room(room_id) + async def on_GET(self, request, room_id): + room = await self.store.get_room(room_id) if room is None: raise NotFoundError("Unknown room") return 200, {"visibility": "public" if room["is_public"] else "private"} - @defer.inlineCallbacks - def on_PUT(self, request, room_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, room_id): + requester = await self.auth.get_user_by_req(request) content = parse_json_object_from_request(request) visibility = content.get("visibility", "public") - yield self.handlers.directory_handler.edit_published_room_list( + await self.handlers.directory_handler.edit_published_room_list( requester, room_id, visibility ) return 200, {} - @defer.inlineCallbacks - def on_DELETE(self, request, room_id): - requester = yield self.auth.get_user_by_req(request) + async def on_DELETE(self, request, room_id): + requester = await self.auth.get_user_by_req(request) - yield self.handlers.directory_handler.edit_published_room_list( + await self.handlers.directory_handler.edit_published_room_list( requester, room_id, "private" ) @@ -181,15 +173,14 @@ class ClientAppserviceDirectoryListServer(RestServlet): def on_DELETE(self, request, network_id, room_id): return self._edit(request, network_id, room_id, "private") - @defer.inlineCallbacks - def _edit(self, request, network_id, room_id, visibility): - requester = yield self.auth.get_user_by_req(request) + async def _edit(self, request, network_id, room_id, visibility): + requester = await self.auth.get_user_by_req(request) if not requester.app_service: raise AuthError( 403, "Only appservices can edit the appservice published room list" ) - yield self.handlers.directory_handler.edit_published_appservice_room_list( + await self.handlers.directory_handler.edit_published_appservice_room_list( requester.app_service.id, network_id, room_id, visibility ) diff --git a/synapse/rest/client/v1/events.py b/synapse/rest/client/v1/events.py index 6651b4cf07..4beb617733 100644 --- a/synapse/rest/client/v1/events.py +++ b/synapse/rest/client/v1/events.py @@ -16,8 +16,6 @@ """This module contains REST servlets to do with event streaming, /events.""" import logging -from twisted.internet import defer - from synapse.api.errors import SynapseError from synapse.http.servlet import RestServlet from synapse.rest.client.v2_alpha._base import client_patterns @@ -36,9 +34,8 @@ class EventStreamRestServlet(RestServlet): self.event_stream_handler = hs.get_event_stream_handler() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request, allow_guest=True) is_guest = requester.is_guest room_id = None if is_guest: @@ -57,7 +54,7 @@ class EventStreamRestServlet(RestServlet): as_client_event = b"raw" not in request.args - chunk = yield self.event_stream_handler.get_stream( + chunk = await self.event_stream_handler.get_stream( requester.user.to_string(), pagin_config, timeout=timeout, @@ -83,14 +80,13 @@ class EventRestServlet(RestServlet): self.event_handler = hs.get_event_handler() self._event_serializer = hs.get_event_client_serializer() - @defer.inlineCallbacks - def on_GET(self, request, event_id): - requester = yield self.auth.get_user_by_req(request) - event = yield self.event_handler.get_event(requester.user, None, event_id) + async def on_GET(self, request, event_id): + requester = await self.auth.get_user_by_req(request) + event = await self.event_handler.get_event(requester.user, None, event_id) time_now = self.clock.time_msec() if event: - event = yield self._event_serializer.serialize_event(event, time_now) + event = await self._event_serializer.serialize_event(event, time_now) return 200, event else: return 404, "Event not found." diff --git a/synapse/rest/client/v1/initial_sync.py b/synapse/rest/client/v1/initial_sync.py index 2da3cd7511..910b3b4eeb 100644 --- a/synapse/rest/client/v1/initial_sync.py +++ b/synapse/rest/client/v1/initial_sync.py @@ -13,7 +13,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -from twisted.internet import defer from synapse.http.servlet import RestServlet, parse_boolean from synapse.rest.client.v2_alpha._base import client_patterns @@ -29,13 +28,12 @@ class InitialSyncRestServlet(RestServlet): self.initial_sync_handler = hs.get_initial_sync_handler() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request) as_client_event = b"raw" not in request.args pagination_config = PaginationConfig.from_request(request) include_archived = parse_boolean(request, "archived", default=False) - content = yield self.initial_sync_handler.snapshot_all_rooms( + content = await self.initial_sync_handler.snapshot_all_rooms( user_id=requester.user.to_string(), pagin_config=pagination_config, as_client_event=as_client_event, diff --git a/synapse/rest/client/v1/login.py b/synapse/rest/client/v1/login.py index 19eb15003d..ff9c978fe7 100644 --- a/synapse/rest/client/v1/login.py +++ b/synapse/rest/client/v1/login.py @@ -18,7 +18,6 @@ import xml.etree.ElementTree as ET from six.moves import urllib -from twisted.internet import defer from twisted.web.client import PartialDownloadError from synapse.api.errors import Codes, LoginError, SynapseError @@ -130,8 +129,7 @@ class LoginRestServlet(RestServlet): def on_OPTIONS(self, request): return 200, {} - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): self._address_ratelimiter.ratelimit( request.getClientIP(), time_now_s=self.hs.clock.time(), @@ -145,11 +143,11 @@ class LoginRestServlet(RestServlet): if self.jwt_enabled and ( login_submission["type"] == LoginRestServlet.JWT_TYPE ): - result = yield self.do_jwt_login(login_submission) + result = await self.do_jwt_login(login_submission) elif login_submission["type"] == LoginRestServlet.TOKEN_TYPE: - result = yield self.do_token_login(login_submission) + result = await self.do_token_login(login_submission) else: - result = yield self._do_other_login(login_submission) + result = await self._do_other_login(login_submission) except KeyError: raise SynapseError(400, "Missing JSON keys.") @@ -158,8 +156,7 @@ class LoginRestServlet(RestServlet): result["well_known"] = well_known_data return 200, result - @defer.inlineCallbacks - def _do_other_login(self, login_submission): + async def _do_other_login(self, login_submission): """Handle non-token/saml/jwt logins Args: @@ -219,20 +216,20 @@ class LoginRestServlet(RestServlet): ( canonical_user_id, callback_3pid, - ) = yield self.auth_handler.check_password_provider_3pid( + ) = await self.auth_handler.check_password_provider_3pid( medium, address, login_submission["password"] ) if canonical_user_id: # Authentication through password provider and 3pid succeeded - result = yield self._complete_login( + result = await self._complete_login( canonical_user_id, login_submission, callback_3pid ) return result # No password providers were able to handle this 3pid # Check local store - user_id = yield self.hs.get_datastore().get_user_id_by_threepid( + user_id = await self.hs.get_datastore().get_user_id_by_threepid( medium, address ) if not user_id: @@ -280,7 +277,7 @@ class LoginRestServlet(RestServlet): ) try: - canonical_user_id, callback = yield self.auth_handler.validate_login( + canonical_user_id, callback = await self.auth_handler.validate_login( identifier["user"], login_submission ) except LoginError: @@ -297,13 +294,12 @@ class LoginRestServlet(RestServlet): ) raise - result = yield self._complete_login( + result = await self._complete_login( canonical_user_id, login_submission, callback ) return result - @defer.inlineCallbacks - def _complete_login( + async def _complete_login( self, user_id, login_submission, callback=None, create_non_existant_users=False ): """Called when we've successfully authed the user and now need to @@ -337,15 +333,15 @@ class LoginRestServlet(RestServlet): ) if create_non_existant_users: - user_id = yield self.auth_handler.check_user_exists(user_id) + user_id = await self.auth_handler.check_user_exists(user_id) if not user_id: - user_id = yield self.registration_handler.register_user( + user_id = await self.registration_handler.register_user( localpart=UserID.from_string(user_id).localpart ) device_id = login_submission.get("device_id") initial_display_name = login_submission.get("initial_device_display_name") - device_id, access_token = yield self.registration_handler.register_device( + device_id, access_token = await self.registration_handler.register_device( user_id, device_id, initial_display_name ) @@ -357,23 +353,21 @@ class LoginRestServlet(RestServlet): } if callback is not None: - yield callback(result) + await callback(result) return result - @defer.inlineCallbacks - def do_token_login(self, login_submission): + async def do_token_login(self, login_submission): token = login_submission["token"] auth_handler = self.auth_handler - user_id = yield auth_handler.validate_short_term_login_token_and_get_user_id( + user_id = await auth_handler.validate_short_term_login_token_and_get_user_id( token ) - result = yield self._complete_login(user_id, login_submission) + result = await self._complete_login(user_id, login_submission) return result - @defer.inlineCallbacks - def do_jwt_login(self, login_submission): + async def do_jwt_login(self, login_submission): token = login_submission.get("token", None) if token is None: raise LoginError( @@ -397,7 +391,7 @@ class LoginRestServlet(RestServlet): raise LoginError(401, "Invalid JWT", errcode=Codes.UNAUTHORIZED) user_id = UserID(user, self.hs.hostname).to_string() - result = yield self._complete_login( + result = await self._complete_login( user_id, login_submission, create_non_existant_users=True ) return result @@ -460,8 +454,7 @@ class CasTicketServlet(RestServlet): self._sso_auth_handler = SSOAuthHandler(hs) self._http_client = hs.get_proxied_http_client() - @defer.inlineCallbacks - def on_GET(self, request): + async def on_GET(self, request): client_redirect_url = parse_string(request, "redirectUrl", required=True) uri = self.cas_server_url + "/proxyValidate" args = { @@ -469,12 +462,12 @@ class CasTicketServlet(RestServlet): "service": self.cas_service_url, } try: - body = yield self._http_client.get_raw(uri, args) + body = await self._http_client.get_raw(uri, args) except PartialDownloadError as pde: # Twisted raises this error if the connection is closed, # even if that's being used old-http style to signal end-of-data body = pde.response - result = yield self.handle_cas_response(request, body, client_redirect_url) + result = await self.handle_cas_response(request, body, client_redirect_url) return result def handle_cas_response(self, request, cas_response_body, client_redirect_url): @@ -555,8 +548,7 @@ class SSOAuthHandler(object): self._registration_handler = hs.get_registration_handler() self._macaroon_gen = hs.get_macaroon_generator() - @defer.inlineCallbacks - def on_successful_auth( + async def on_successful_auth( self, username, request, client_redirect_url, user_display_name=None ): """Called once the user has successfully authenticated with the SSO. @@ -582,9 +574,9 @@ class SSOAuthHandler(object): """ localpart = map_username_to_mxid_localpart(username) user_id = UserID(localpart, self._hostname).to_string() - registered_user_id = yield self._auth_handler.check_user_exists(user_id) + registered_user_id = await self._auth_handler.check_user_exists(user_id) if not registered_user_id: - registered_user_id = yield self._registration_handler.register_user( + registered_user_id = await self._registration_handler.register_user( localpart=localpart, default_display_name=user_display_name ) diff --git a/synapse/rest/client/v1/logout.py b/synapse/rest/client/v1/logout.py index 4785a34d75..1cf3caf832 100644 --- a/synapse/rest/client/v1/logout.py +++ b/synapse/rest/client/v1/logout.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.http.servlet import RestServlet from synapse.rest.client.v2_alpha._base import client_patterns @@ -35,17 +33,16 @@ class LogoutRestServlet(RestServlet): def on_OPTIONS(self, request): return 200, {} - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request) if requester.device_id is None: # the acccess token wasn't associated with a device. # Just delete the access token access_token = self.auth.get_access_token_from_request(request) - yield self._auth_handler.delete_access_token(access_token) + await self._auth_handler.delete_access_token(access_token) else: - yield self._device_handler.delete_device( + await self._device_handler.delete_device( requester.user.to_string(), requester.device_id ) @@ -64,17 +61,16 @@ class LogoutAllRestServlet(RestServlet): def on_OPTIONS(self, request): return 200, {} - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() # first delete all of the user's devices - yield self._device_handler.delete_all_devices_for_user(user_id) + await self._device_handler.delete_all_devices_for_user(user_id) # .. and then delete any access tokens which weren't associated with # devices. - yield self._auth_handler.delete_access_tokens_for_user(user_id) + await self._auth_handler.delete_access_tokens_for_user(user_id) return 200, {} diff --git a/synapse/rest/client/v1/presence.py b/synapse/rest/client/v1/presence.py index 0153525cef..eec16f8ad8 100644 --- a/synapse/rest/client/v1/presence.py +++ b/synapse/rest/client/v1/presence.py @@ -19,8 +19,6 @@ import logging from six import string_types -from twisted.internet import defer - from synapse.api.errors import AuthError, SynapseError from synapse.handlers.presence import format_user_presence_state from synapse.http.servlet import RestServlet, parse_json_object_from_request @@ -40,27 +38,25 @@ class PresenceStatusRestServlet(RestServlet): self.clock = hs.get_clock() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request, user_id): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request, user_id): + requester = await self.auth.get_user_by_req(request) user = UserID.from_string(user_id) if requester.user != user: - allowed = yield self.presence_handler.is_visible( + allowed = await self.presence_handler.is_visible( observed_user=user, observer_user=requester.user ) if not allowed: raise AuthError(403, "You are not allowed to see their presence.") - state = yield self.presence_handler.get_state(target_user=user) + state = await self.presence_handler.get_state(target_user=user) state = format_user_presence_state(state, self.clock.time_msec()) return 200, state - @defer.inlineCallbacks - def on_PUT(self, request, user_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, user_id): + requester = await self.auth.get_user_by_req(request) user = UserID.from_string(user_id) if requester.user != user: @@ -86,7 +82,7 @@ class PresenceStatusRestServlet(RestServlet): raise SynapseError(400, "Unable to parse state") if self.hs.config.use_presence: - yield self.presence_handler.set_state(user, state) + await self.presence_handler.set_state(user, state) return 200, {} diff --git a/synapse/rest/client/v1/profile.py b/synapse/rest/client/v1/profile.py index bbce2e2b71..e7fe50ed72 100644 --- a/synapse/rest/client/v1/profile.py +++ b/synapse/rest/client/v1/profile.py @@ -14,8 +14,8 @@ # limitations under the License. """ This module contains REST servlets to do with profile: /profile/<paths> """ -from twisted.internet import defer +from synapse.api.errors import Codes, SynapseError from synapse.http.servlet import RestServlet, parse_json_object_from_request from synapse.rest.client.v2_alpha._base import client_patterns from synapse.types import UserID @@ -30,19 +30,18 @@ class ProfileDisplaynameRestServlet(RestServlet): self.profile_handler = hs.get_profile_handler() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request, user_id): + async def on_GET(self, request, user_id): requester_user = None if self.hs.config.require_auth_for_profile_requests: - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) requester_user = requester.user user = UserID.from_string(user_id) - yield self.profile_handler.check_profile_query_allowed(user, requester_user) + await self.profile_handler.check_profile_query_allowed(user, requester_user) - displayname = yield self.profile_handler.get_displayname(user) + displayname = await self.profile_handler.get_displayname(user) ret = {} if displayname is not None: @@ -50,11 +49,10 @@ class ProfileDisplaynameRestServlet(RestServlet): return 200, ret - @defer.inlineCallbacks - def on_PUT(self, request, user_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_PUT(self, request, user_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) user = UserID.from_string(user_id) - is_admin = yield self.auth.is_server_admin(requester.user) + is_admin = await self.auth.is_server_admin(requester.user) content = parse_json_object_from_request(request) @@ -63,7 +61,7 @@ class ProfileDisplaynameRestServlet(RestServlet): except Exception: return 400, "Unable to parse name" - yield self.profile_handler.set_displayname(user, requester, new_name, is_admin) + await self.profile_handler.set_displayname(user, requester, new_name, is_admin) return 200, {} @@ -80,19 +78,18 @@ class ProfileAvatarURLRestServlet(RestServlet): self.profile_handler = hs.get_profile_handler() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request, user_id): + async def on_GET(self, request, user_id): requester_user = None if self.hs.config.require_auth_for_profile_requests: - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) requester_user = requester.user user = UserID.from_string(user_id) - yield self.profile_handler.check_profile_query_allowed(user, requester_user) + await self.profile_handler.check_profile_query_allowed(user, requester_user) - avatar_url = yield self.profile_handler.get_avatar_url(user) + avatar_url = await self.profile_handler.get_avatar_url(user) ret = {} if avatar_url is not None: @@ -100,19 +97,22 @@ class ProfileAvatarURLRestServlet(RestServlet): return 200, ret - @defer.inlineCallbacks - def on_PUT(self, request, user_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, user_id): + requester = await self.auth.get_user_by_req(request) user = UserID.from_string(user_id) - is_admin = yield self.auth.is_server_admin(requester.user) + is_admin = await self.auth.is_server_admin(requester.user) content = parse_json_object_from_request(request) try: - new_name = content["avatar_url"] - except Exception: - return 400, "Unable to parse name" + new_avatar_url = content["avatar_url"] + except KeyError: + raise SynapseError( + 400, "Missing key 'avatar_url'", errcode=Codes.MISSING_PARAM + ) - yield self.profile_handler.set_avatar_url(user, requester, new_name, is_admin) + await self.profile_handler.set_avatar_url( + user, requester, new_avatar_url, is_admin + ) return 200, {} @@ -129,20 +129,19 @@ class ProfileRestServlet(RestServlet): self.profile_handler = hs.get_profile_handler() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request, user_id): + async def on_GET(self, request, user_id): requester_user = None if self.hs.config.require_auth_for_profile_requests: - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) requester_user = requester.user user = UserID.from_string(user_id) - yield self.profile_handler.check_profile_query_allowed(user, requester_user) + await self.profile_handler.check_profile_query_allowed(user, requester_user) - displayname = yield self.profile_handler.get_displayname(user) - avatar_url = yield self.profile_handler.get_avatar_url(user) + displayname = await self.profile_handler.get_displayname(user) + avatar_url = await self.profile_handler.get_avatar_url(user) ret = {} if displayname is not None: diff --git a/synapse/rest/client/v1/push_rule.py b/synapse/rest/client/v1/push_rule.py index 9f8c3d09e3..4f74600239 100644 --- a/synapse/rest/client/v1/push_rule.py +++ b/synapse/rest/client/v1/push_rule.py @@ -13,7 +13,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -from twisted.internet import defer from synapse.api.errors import ( NotFoundError, @@ -46,8 +45,7 @@ class PushRuleRestServlet(RestServlet): self.notifier = hs.get_notifier() self._is_worker = hs.config.worker_app is not None - @defer.inlineCallbacks - def on_PUT(self, request, path): + async def on_PUT(self, request, path): if self._is_worker: raise Exception("Cannot handle PUT /push_rules on worker") @@ -57,7 +55,7 @@ class PushRuleRestServlet(RestServlet): except InvalidRuleException as e: raise SynapseError(400, str(e)) - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) if "/" in spec["rule_id"] or "\\" in spec["rule_id"]: raise SynapseError(400, "rule_id may not contain slashes") @@ -67,7 +65,7 @@ class PushRuleRestServlet(RestServlet): user_id = requester.user.to_string() if "attr" in spec: - yield self.set_rule_attr(user_id, spec, content) + await self.set_rule_attr(user_id, spec, content) self.notify_user(user_id) return 200, {} @@ -91,7 +89,7 @@ class PushRuleRestServlet(RestServlet): after = _namespaced_rule_id(spec, after) try: - yield self.store.add_push_rule( + await self.store.add_push_rule( user_id=user_id, rule_id=_namespaced_rule_id_from_spec(spec), priority_class=priority_class, @@ -108,20 +106,19 @@ class PushRuleRestServlet(RestServlet): return 200, {} - @defer.inlineCallbacks - def on_DELETE(self, request, path): + async def on_DELETE(self, request, path): if self._is_worker: raise Exception("Cannot handle DELETE /push_rules on worker") spec = _rule_spec_from_path([x for x in path.split("/")]) - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() namespaced_rule_id = _namespaced_rule_id_from_spec(spec) try: - yield self.store.delete_push_rule(user_id, namespaced_rule_id) + await self.store.delete_push_rule(user_id, namespaced_rule_id) self.notify_user(user_id) return 200, {} except StoreError as e: @@ -130,15 +127,14 @@ class PushRuleRestServlet(RestServlet): else: raise - @defer.inlineCallbacks - def on_GET(self, request, path): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request, path): + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() # we build up the full structure and then decide which bits of it # to send which means doing unnecessary work sometimes but is # is probably not going to make a whole lot of difference - rules = yield self.store.get_push_rules_for_user(user_id) + rules = await self.store.get_push_rules_for_user(user_id) rules = format_push_rules_for_user(requester.user, rules) diff --git a/synapse/rest/client/v1/pusher.py b/synapse/rest/client/v1/pusher.py index 41660682d9..6f6b7aed6e 100644 --- a/synapse/rest/client/v1/pusher.py +++ b/synapse/rest/client/v1/pusher.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import Codes, StoreError, SynapseError from synapse.http.server import finish_request from synapse.http.servlet import ( @@ -30,6 +28,17 @@ from synapse.rest.client.v2_alpha._base import client_patterns logger = logging.getLogger(__name__) +ALLOWED_KEYS = { + "app_display_name", + "app_id", + "data", + "device_display_name", + "kind", + "lang", + "profile_tag", + "pushkey", +} + class PushersRestServlet(RestServlet): PATTERNS = client_patterns("/pushers$", v1=True) @@ -39,30 +48,17 @@ class PushersRestServlet(RestServlet): self.hs = hs self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request) user = requester.user - pushers = yield self.hs.get_datastore().get_pushers_by_user_id(user.to_string()) + pushers = await self.hs.get_datastore().get_pushers_by_user_id(user.to_string()) - allowed_keys = [ - "app_display_name", - "app_id", - "data", - "device_display_name", - "kind", - "lang", - "profile_tag", - "pushkey", - ] - - for p in pushers: - for k, v in list(p.items()): - if k not in allowed_keys: - del p[k] + filtered_pushers = list( + {k: v for k, v in p.items() if k in ALLOWED_KEYS} for p in pushers + ) - return 200, {"pushers": pushers} + return 200, {"pushers": filtered_pushers} def on_OPTIONS(self, _): return 200, {} @@ -78,9 +74,8 @@ class PushersSetRestServlet(RestServlet): self.notifier = hs.get_notifier() self.pusher_pool = self.hs.get_pusherpool() - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request) user = requester.user content = parse_json_object_from_request(request) @@ -91,7 +86,7 @@ class PushersSetRestServlet(RestServlet): and "kind" in content and content["kind"] is None ): - yield self.pusher_pool.remove_pusher( + await self.pusher_pool.remove_pusher( content["app_id"], content["pushkey"], user_id=user.to_string() ) return 200, {} @@ -117,14 +112,14 @@ class PushersSetRestServlet(RestServlet): append = content["append"] if not append: - yield self.pusher_pool.remove_pushers_by_app_id_and_pushkey_not_user( + await self.pusher_pool.remove_pushers_by_app_id_and_pushkey_not_user( app_id=content["app_id"], pushkey=content["pushkey"], not_user_id=user.to_string(), ) try: - yield self.pusher_pool.add_pusher( + await self.pusher_pool.add_pusher( user_id=user.to_string(), access_token=requester.access_token_id, kind=content["kind"], @@ -164,16 +159,15 @@ class PushersRemoveRestServlet(RestServlet): self.auth = hs.get_auth() self.pusher_pool = self.hs.get_pusherpool() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request, rights="delete_pusher") + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request, rights="delete_pusher") user = requester.user app_id = parse_string(request, "app_id", required=True) pushkey = parse_string(request, "pushkey", required=True) try: - yield self.pusher_pool.remove_pusher( + await self.pusher_pool.remove_pusher( app_id=app_id, pushkey=pushkey, user_id=user.to_string() ) except StoreError as se: diff --git a/synapse/rest/client/v1/room.py b/synapse/rest/client/v1/room.py index 86bbcc0eea..711d4ad304 100644 --- a/synapse/rest/client/v1/room.py +++ b/synapse/rest/client/v1/room.py @@ -714,7 +714,7 @@ class RoomMembershipRestServlet(TransactionRestServlet): target = UserID.from_string(content["user_id"]) event_content = None - if "reason" in content and membership_action in ["kick", "ban"]: + if "reason" in content: event_content = {"reason": content["reason"]} await self.room_member_handler.update_membership( diff --git a/synapse/rest/client/v1/voip.py b/synapse/rest/client/v1/voip.py index 2afdbb89e5..747d46eac2 100644 --- a/synapse/rest/client/v1/voip.py +++ b/synapse/rest/client/v1/voip.py @@ -17,8 +17,6 @@ import base64 import hashlib import hmac -from twisted.internet import defer - from synapse.http.servlet import RestServlet from synapse.rest.client.v2_alpha._base import client_patterns @@ -31,9 +29,8 @@ class VoipRestServlet(RestServlet): self.hs = hs self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req( + async def on_GET(self, request): + requester = await self.auth.get_user_by_req( request, self.hs.config.turn_allow_guests ) diff --git a/synapse/rest/client/v2_alpha/_base.py b/synapse/rest/client/v2_alpha/_base.py index 8250ae0ae1..2a3f4dd58f 100644 --- a/synapse/rest/client/v2_alpha/_base.py +++ b/synapse/rest/client/v2_alpha/_base.py @@ -78,7 +78,7 @@ def interactive_auth_handler(orig): """ def wrapped(*args, **kwargs): - res = defer.maybeDeferred(orig, *args, **kwargs) + res = defer.ensureDeferred(orig(*args, **kwargs)) res.addErrback(_catch_incomplete_interactive_auth) return res diff --git a/synapse/rest/client/v2_alpha/account.py b/synapse/rest/client/v2_alpha/account.py index f26eae794c..fc240f5cf8 100644 --- a/synapse/rest/client/v2_alpha/account.py +++ b/synapse/rest/client/v2_alpha/account.py @@ -18,8 +18,6 @@ import logging from six.moves import http_client -from twisted.internet import defer - from synapse.api.constants import LoginType from synapse.api.errors import Codes, SynapseError, ThreepidValidationError from synapse.config.emailconfig import ThreepidBehaviour @@ -67,8 +65,7 @@ class EmailPasswordRequestTokenRestServlet(RestServlet): template_text=template_text, ) - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): if self.config.threepid_behaviour_email == ThreepidBehaviour.OFF: if self.config.local_threepid_handling_disabled_due_to_email_config: logger.warning( @@ -95,7 +92,7 @@ class EmailPasswordRequestTokenRestServlet(RestServlet): Codes.THREEPID_DENIED, ) - existing_user_id = yield self.hs.get_datastore().get_user_id_by_threepid( + existing_user_id = await self.hs.get_datastore().get_user_id_by_threepid( "email", email ) @@ -106,7 +103,7 @@ class EmailPasswordRequestTokenRestServlet(RestServlet): assert self.hs.config.account_threepid_delegate_email # Have the configured identity server handle the request - ret = yield self.identity_handler.requestEmailToken( + ret = await self.identity_handler.requestEmailToken( self.hs.config.account_threepid_delegate_email, email, client_secret, @@ -115,7 +112,7 @@ class EmailPasswordRequestTokenRestServlet(RestServlet): ) else: # Send password reset emails from Synapse - sid = yield self.identity_handler.send_threepid_validation( + sid = await self.identity_handler.send_threepid_validation( email, client_secret, send_attempt, @@ -153,8 +150,7 @@ class PasswordResetSubmitTokenServlet(RestServlet): [self.config.email_password_reset_template_failure_html], ) - @defer.inlineCallbacks - def on_GET(self, request, medium): + async def on_GET(self, request, medium): # We currently only handle threepid token submissions for email if medium != "email": raise SynapseError( @@ -176,7 +172,7 @@ class PasswordResetSubmitTokenServlet(RestServlet): # Attempt to validate a 3PID session try: # Mark the session as valid - next_link = yield self.store.validate_threepid_session( + next_link = await self.store.validate_threepid_session( sid, client_secret, token, self.clock.time_msec() ) @@ -218,8 +214,7 @@ class PasswordRestServlet(RestServlet): self._set_password_handler = hs.get_set_password_handler() @interactive_auth_handler - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): body = parse_json_object_from_request(request) # there are two possibilities here. Either the user does not have an @@ -233,14 +228,14 @@ class PasswordRestServlet(RestServlet): # In the second case, we require a password to confirm their identity. if self.auth.has_access_token(request): - requester = yield self.auth.get_user_by_req(request) - params = yield self.auth_handler.validate_user_via_ui_auth( + requester = await self.auth.get_user_by_req(request) + params = await self.auth_handler.validate_user_via_ui_auth( requester, body, self.hs.get_ip_from_request(request) ) user_id = requester.user.to_string() else: requester = None - result, params, _ = yield self.auth_handler.check_auth( + result, params, _ = await self.auth_handler.check_auth( [[LoginType.EMAIL_IDENTITY]], body, self.hs.get_ip_from_request(request) ) @@ -254,7 +249,7 @@ class PasswordRestServlet(RestServlet): # (See add_threepid in synapse/handlers/auth.py) threepid["address"] = threepid["address"].lower() # if using email, we must know about the email they're authing with! - threepid_user_id = yield self.datastore.get_user_id_by_threepid( + threepid_user_id = await self.datastore.get_user_id_by_threepid( threepid["medium"], threepid["address"] ) if not threepid_user_id: @@ -267,7 +262,7 @@ class PasswordRestServlet(RestServlet): assert_params_in_dict(params, ["new_password"]) new_password = params["new_password"] - yield self._set_password_handler.set_password(user_id, new_password, requester) + await self._set_password_handler.set_password(user_id, new_password, requester) return 200, {} @@ -286,8 +281,7 @@ class DeactivateAccountRestServlet(RestServlet): self._deactivate_account_handler = hs.get_deactivate_account_handler() @interactive_auth_handler - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): body = parse_json_object_from_request(request) erase = body.get("erase", False) if not isinstance(erase, bool): @@ -297,19 +291,19 @@ class DeactivateAccountRestServlet(RestServlet): Codes.BAD_JSON, ) - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) # allow ASes to dectivate their own users if requester.app_service: - yield self._deactivate_account_handler.deactivate_account( + await self._deactivate_account_handler.deactivate_account( requester.user.to_string(), erase ) return 200, {} - yield self.auth_handler.validate_user_via_ui_auth( + await self.auth_handler.validate_user_via_ui_auth( requester, body, self.hs.get_ip_from_request(request) ) - result = yield self._deactivate_account_handler.deactivate_account( + result = await self._deactivate_account_handler.deactivate_account( requester.user.to_string(), erase, id_server=body.get("id_server") ) if result: @@ -346,8 +340,7 @@ class EmailThreepidRequestTokenRestServlet(RestServlet): template_text=template_text, ) - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): if self.config.threepid_behaviour_email == ThreepidBehaviour.OFF: if self.config.local_threepid_handling_disabled_due_to_email_config: logger.warning( @@ -371,7 +364,7 @@ class EmailThreepidRequestTokenRestServlet(RestServlet): Codes.THREEPID_DENIED, ) - existing_user_id = yield self.store.get_user_id_by_threepid( + existing_user_id = await self.store.get_user_id_by_threepid( "email", body["email"] ) @@ -382,7 +375,7 @@ class EmailThreepidRequestTokenRestServlet(RestServlet): assert self.hs.config.account_threepid_delegate_email # Have the configured identity server handle the request - ret = yield self.identity_handler.requestEmailToken( + ret = await self.identity_handler.requestEmailToken( self.hs.config.account_threepid_delegate_email, email, client_secret, @@ -391,7 +384,7 @@ class EmailThreepidRequestTokenRestServlet(RestServlet): ) else: # Send threepid validation emails from Synapse - sid = yield self.identity_handler.send_threepid_validation( + sid = await self.identity_handler.send_threepid_validation( email, client_secret, send_attempt, @@ -414,8 +407,7 @@ class MsisdnThreepidRequestTokenRestServlet(RestServlet): self.store = self.hs.get_datastore() self.identity_handler = hs.get_handlers().identity_handler - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): body = parse_json_object_from_request(request) assert_params_in_dict( body, ["client_secret", "country", "phone_number", "send_attempt"] @@ -435,7 +427,7 @@ class MsisdnThreepidRequestTokenRestServlet(RestServlet): Codes.THREEPID_DENIED, ) - existing_user_id = yield self.store.get_user_id_by_threepid("msisdn", msisdn) + existing_user_id = await self.store.get_user_id_by_threepid("msisdn", msisdn) if existing_user_id is not None: raise SynapseError(400, "MSISDN is already in use", Codes.THREEPID_IN_USE) @@ -450,7 +442,7 @@ class MsisdnThreepidRequestTokenRestServlet(RestServlet): "Adding phone numbers to user account is not supported by this homeserver", ) - ret = yield self.identity_handler.requestMsisdnToken( + ret = await self.identity_handler.requestMsisdnToken( self.hs.config.account_threepid_delegate_msisdn, country, phone_number, @@ -484,8 +476,7 @@ class AddThreepidEmailSubmitTokenServlet(RestServlet): [self.config.email_add_threepid_template_failure_html], ) - @defer.inlineCallbacks - def on_GET(self, request): + async def on_GET(self, request): if self.config.threepid_behaviour_email == ThreepidBehaviour.OFF: if self.config.local_threepid_handling_disabled_due_to_email_config: logger.warning( @@ -508,7 +499,7 @@ class AddThreepidEmailSubmitTokenServlet(RestServlet): # Attempt to validate a 3PID session try: # Mark the session as valid - next_link = yield self.store.validate_threepid_session( + next_link = await self.store.validate_threepid_session( sid, client_secret, token, self.clock.time_msec() ) @@ -558,8 +549,7 @@ class AddThreepidMsisdnSubmitTokenServlet(RestServlet): self.store = hs.get_datastore() self.identity_handler = hs.get_handlers().identity_handler - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): if not self.config.account_threepid_delegate_msisdn: raise SynapseError( 400, @@ -571,7 +561,7 @@ class AddThreepidMsisdnSubmitTokenServlet(RestServlet): assert_params_in_dict(body, ["client_secret", "sid", "token"]) # Proxy submit_token request to msisdn threepid delegate - response = yield self.identity_handler.proxy_msisdn_submit_token( + response = await self.identity_handler.proxy_msisdn_submit_token( self.config.account_threepid_delegate_msisdn, body["client_secret"], body["sid"], @@ -591,17 +581,15 @@ class ThreepidRestServlet(RestServlet): self.auth_handler = hs.get_auth_handler() self.datastore = self.hs.get_datastore() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request) - threepids = yield self.datastore.user_get_threepids(requester.user.to_string()) + threepids = await self.datastore.user_get_threepids(requester.user.to_string()) return 200, {"threepids": threepids} - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() body = parse_json_object_from_request(request) @@ -615,11 +603,11 @@ class ThreepidRestServlet(RestServlet): client_secret = threepid_creds["client_secret"] sid = threepid_creds["sid"] - validation_session = yield self.identity_handler.validate_threepid_session( + validation_session = await self.identity_handler.validate_threepid_session( client_secret, sid ) if validation_session: - yield self.auth_handler.add_threepid( + await self.auth_handler.add_threepid( user_id, validation_session["medium"], validation_session["address"], @@ -642,9 +630,9 @@ class ThreepidAddRestServlet(RestServlet): self.auth = hs.get_auth() self.auth_handler = hs.get_auth_handler() - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request) + @interactive_auth_handler + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() body = parse_json_object_from_request(request) @@ -652,11 +640,15 @@ class ThreepidAddRestServlet(RestServlet): client_secret = body["client_secret"] sid = body["sid"] - validation_session = yield self.identity_handler.validate_threepid_session( + await self.auth_handler.validate_user_via_ui_auth( + requester, body, self.hs.get_ip_from_request(request) + ) + + validation_session = await self.identity_handler.validate_threepid_session( client_secret, sid ) if validation_session: - yield self.auth_handler.add_threepid( + await self.auth_handler.add_threepid( user_id, validation_session["medium"], validation_session["address"], @@ -678,8 +670,7 @@ class ThreepidBindRestServlet(RestServlet): self.identity_handler = hs.get_handlers().identity_handler self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): body = parse_json_object_from_request(request) assert_params_in_dict(body, ["id_server", "sid", "client_secret"]) @@ -688,10 +679,10 @@ class ThreepidBindRestServlet(RestServlet): client_secret = body["client_secret"] id_access_token = body.get("id_access_token") # optional - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() - yield self.identity_handler.bind_threepid( + await self.identity_handler.bind_threepid( client_secret, sid, user_id, id_server, id_access_token ) @@ -708,12 +699,11 @@ class ThreepidUnbindRestServlet(RestServlet): self.auth = hs.get_auth() self.datastore = self.hs.get_datastore() - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): """Unbind the given 3pid from a specific identity server, or identity servers that are known to have this 3pid bound """ - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) body = parse_json_object_from_request(request) assert_params_in_dict(body, ["medium", "address"]) @@ -723,7 +713,7 @@ class ThreepidUnbindRestServlet(RestServlet): # Attempt to unbind the threepid from an identity server. If id_server is None, try to # unbind from all identity servers this threepid has been added to in the past - result = yield self.identity_handler.try_unbind_threepid( + result = await self.identity_handler.try_unbind_threepid( requester.user.to_string(), {"address": address, "medium": medium, "id_server": id_server}, ) @@ -738,16 +728,15 @@ class ThreepidDeleteRestServlet(RestServlet): self.auth = hs.get_auth() self.auth_handler = hs.get_auth_handler() - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): body = parse_json_object_from_request(request) assert_params_in_dict(body, ["medium", "address"]) - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() try: - ret = yield self.auth_handler.delete_threepid( + ret = await self.auth_handler.delete_threepid( user_id, body["medium"], body["address"], body.get("id_server") ) except Exception: @@ -772,9 +761,8 @@ class WhoamiRestServlet(RestServlet): super(WhoamiRestServlet, self).__init__() self.auth = hs.get_auth() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request) return 200, {"user_id": requester.user.to_string()} diff --git a/synapse/rest/client/v2_alpha/account_data.py b/synapse/rest/client/v2_alpha/account_data.py index f0db204ffa..64eb7fec3b 100644 --- a/synapse/rest/client/v2_alpha/account_data.py +++ b/synapse/rest/client/v2_alpha/account_data.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import AuthError, NotFoundError, SynapseError from synapse.http.servlet import RestServlet, parse_json_object_from_request @@ -41,15 +39,14 @@ class AccountDataServlet(RestServlet): self.store = hs.get_datastore() self.notifier = hs.get_notifier() - @defer.inlineCallbacks - def on_PUT(self, request, user_id, account_data_type): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, user_id, account_data_type): + requester = await self.auth.get_user_by_req(request) if user_id != requester.user.to_string(): raise AuthError(403, "Cannot add account data for other users.") body = parse_json_object_from_request(request) - max_id = yield self.store.add_account_data_for_user( + max_id = await self.store.add_account_data_for_user( user_id, account_data_type, body ) @@ -57,13 +54,12 @@ class AccountDataServlet(RestServlet): return 200, {} - @defer.inlineCallbacks - def on_GET(self, request, user_id, account_data_type): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request, user_id, account_data_type): + requester = await self.auth.get_user_by_req(request) if user_id != requester.user.to_string(): raise AuthError(403, "Cannot get account data for other users.") - event = yield self.store.get_global_account_data_by_type_for_user( + event = await self.store.get_global_account_data_by_type_for_user( account_data_type, user_id ) @@ -91,9 +87,8 @@ class RoomAccountDataServlet(RestServlet): self.store = hs.get_datastore() self.notifier = hs.get_notifier() - @defer.inlineCallbacks - def on_PUT(self, request, user_id, room_id, account_data_type): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, user_id, room_id, account_data_type): + requester = await self.auth.get_user_by_req(request) if user_id != requester.user.to_string(): raise AuthError(403, "Cannot add account data for other users.") @@ -106,7 +101,7 @@ class RoomAccountDataServlet(RestServlet): " Use /rooms/!roomId:server.name/read_markers", ) - max_id = yield self.store.add_account_data_to_room( + max_id = await self.store.add_account_data_to_room( user_id, room_id, account_data_type, body ) @@ -114,13 +109,12 @@ class RoomAccountDataServlet(RestServlet): return 200, {} - @defer.inlineCallbacks - def on_GET(self, request, user_id, room_id, account_data_type): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request, user_id, room_id, account_data_type): + requester = await self.auth.get_user_by_req(request) if user_id != requester.user.to_string(): raise AuthError(403, "Cannot get account data for other users.") - event = yield self.store.get_account_data_for_room_and_type( + event = await self.store.get_account_data_for_room_and_type( user_id, room_id, account_data_type ) diff --git a/synapse/rest/client/v2_alpha/account_validity.py b/synapse/rest/client/v2_alpha/account_validity.py index 33f6a23028..2f10fa64e2 100644 --- a/synapse/rest/client/v2_alpha/account_validity.py +++ b/synapse/rest/client/v2_alpha/account_validity.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import AuthError, SynapseError from synapse.http.server import finish_request from synapse.http.servlet import RestServlet @@ -45,13 +43,12 @@ class AccountValidityRenewServlet(RestServlet): self.success_html = hs.config.account_validity.account_renewed_html_content self.failure_html = hs.config.account_validity.invalid_token_html_content - @defer.inlineCallbacks - def on_GET(self, request): + async def on_GET(self, request): if b"token" not in request.args: raise SynapseError(400, "Missing renewal token") renewal_token = request.args[b"token"][0] - token_valid = yield self.account_activity_handler.renew_account( + token_valid = await self.account_activity_handler.renew_account( renewal_token.decode("utf8") ) @@ -67,7 +64,6 @@ class AccountValidityRenewServlet(RestServlet): request.setHeader(b"Content-Length", b"%d" % (len(response),)) request.write(response.encode("utf8")) finish_request(request) - defer.returnValue(None) class AccountValiditySendMailServlet(RestServlet): @@ -85,18 +81,17 @@ class AccountValiditySendMailServlet(RestServlet): self.auth = hs.get_auth() self.account_validity = self.hs.config.account_validity - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): if not self.account_validity.renew_by_email_enabled: raise AuthError( 403, "Account renewal via email is disabled on this server." ) - requester = yield self.auth.get_user_by_req(request, allow_expired=True) + requester = await self.auth.get_user_by_req(request, allow_expired=True) user_id = requester.user.to_string() - yield self.account_activity_handler.send_renewal_email_to_user(user_id) + await self.account_activity_handler.send_renewal_email_to_user(user_id) - defer.returnValue((200, {})) + return 200, {} def register_servlets(hs, http_server): diff --git a/synapse/rest/client/v2_alpha/auth.py b/synapse/rest/client/v2_alpha/auth.py index f21aff39e5..50e080673b 100644 --- a/synapse/rest/client/v2_alpha/auth.py +++ b/synapse/rest/client/v2_alpha/auth.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.constants import LoginType from synapse.api.errors import SynapseError from synapse.api.urls import CLIENT_API_PREFIX @@ -171,8 +169,7 @@ class AuthRestServlet(RestServlet): else: raise SynapseError(404, "Unknown auth stage type") - @defer.inlineCallbacks - def on_POST(self, request, stagetype): + async def on_POST(self, request, stagetype): session = parse_string(request, "session") if not session: @@ -186,7 +183,7 @@ class AuthRestServlet(RestServlet): authdict = {"response": response, "session": session} - success = yield self.auth_handler.add_oob_auth( + success = await self.auth_handler.add_oob_auth( LoginType.RECAPTCHA, authdict, self.hs.get_ip_from_request(request) ) @@ -209,13 +206,9 @@ class AuthRestServlet(RestServlet): return None elif stagetype == LoginType.TERMS: - if ("session" not in request.args or len(request.args["session"])) == 0: - raise SynapseError(400, "No session supplied") - - session = request.args["session"][0] authdict = {"session": session} - success = yield self.auth_handler.add_oob_auth( + success = await self.auth_handler.add_oob_auth( LoginType.TERMS, authdict, self.hs.get_ip_from_request(request) ) diff --git a/synapse/rest/client/v2_alpha/capabilities.py b/synapse/rest/client/v2_alpha/capabilities.py index acd58af193..fe9d019c44 100644 --- a/synapse/rest/client/v2_alpha/capabilities.py +++ b/synapse/rest/client/v2_alpha/capabilities.py @@ -14,8 +14,6 @@ # limitations under the License. import logging -from twisted.internet import defer - from synapse.api.room_versions import KNOWN_ROOM_VERSIONS from synapse.http.servlet import RestServlet @@ -40,10 +38,9 @@ class CapabilitiesRestServlet(RestServlet): self.auth = hs.get_auth() self.store = hs.get_datastore() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) - user = yield self.store.get_user_by_id(requester.user.to_string()) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request, allow_guest=True) + user = await self.store.get_user_by_id(requester.user.to_string()) change_password = bool(user["password_hash"]) response = { diff --git a/synapse/rest/client/v2_alpha/devices.py b/synapse/rest/client/v2_alpha/devices.py index 26d0235208..94ff73f384 100644 --- a/synapse/rest/client/v2_alpha/devices.py +++ b/synapse/rest/client/v2_alpha/devices.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api import errors from synapse.http.servlet import ( RestServlet, @@ -42,10 +40,9 @@ class DevicesRestServlet(RestServlet): self.auth = hs.get_auth() self.device_handler = hs.get_device_handler() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) - devices = yield self.device_handler.get_devices_by_user( + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request, allow_guest=True) + devices = await self.device_handler.get_devices_by_user( requester.user.to_string() ) return 200, {"devices": devices} @@ -67,9 +64,8 @@ class DeleteDevicesRestServlet(RestServlet): self.auth_handler = hs.get_auth_handler() @interactive_auth_handler - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request) try: body = parse_json_object_from_request(request) @@ -84,11 +80,11 @@ class DeleteDevicesRestServlet(RestServlet): assert_params_in_dict(body, ["devices"]) - yield self.auth_handler.validate_user_via_ui_auth( + await self.auth_handler.validate_user_via_ui_auth( requester, body, self.hs.get_ip_from_request(request) ) - yield self.device_handler.delete_devices( + await self.device_handler.delete_devices( requester.user.to_string(), body["devices"] ) return 200, {} @@ -108,18 +104,16 @@ class DeviceRestServlet(RestServlet): self.device_handler = hs.get_device_handler() self.auth_handler = hs.get_auth_handler() - @defer.inlineCallbacks - def on_GET(self, request, device_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) - device = yield self.device_handler.get_device( + async def on_GET(self, request, device_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) + device = await self.device_handler.get_device( requester.user.to_string(), device_id ) return 200, device @interactive_auth_handler - @defer.inlineCallbacks - def on_DELETE(self, request, device_id): - requester = yield self.auth.get_user_by_req(request) + async def on_DELETE(self, request, device_id): + requester = await self.auth.get_user_by_req(request) try: body = parse_json_object_from_request(request) @@ -132,19 +126,18 @@ class DeviceRestServlet(RestServlet): else: raise - yield self.auth_handler.validate_user_via_ui_auth( + await self.auth_handler.validate_user_via_ui_auth( requester, body, self.hs.get_ip_from_request(request) ) - yield self.device_handler.delete_device(requester.user.to_string(), device_id) + await self.device_handler.delete_device(requester.user.to_string(), device_id) return 200, {} - @defer.inlineCallbacks - def on_PUT(self, request, device_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_PUT(self, request, device_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) body = parse_json_object_from_request(request) - yield self.device_handler.update_device( + await self.device_handler.update_device( requester.user.to_string(), device_id, body ) return 200, {} diff --git a/synapse/rest/client/v2_alpha/filter.py b/synapse/rest/client/v2_alpha/filter.py index 17a8bc7366..b28da017cd 100644 --- a/synapse/rest/client/v2_alpha/filter.py +++ b/synapse/rest/client/v2_alpha/filter.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import AuthError, NotFoundError, StoreError, SynapseError from synapse.http.servlet import RestServlet, parse_json_object_from_request from synapse.types import UserID @@ -35,10 +33,9 @@ class GetFilterRestServlet(RestServlet): self.auth = hs.get_auth() self.filtering = hs.get_filtering() - @defer.inlineCallbacks - def on_GET(self, request, user_id, filter_id): + async def on_GET(self, request, user_id, filter_id): target_user = UserID.from_string(user_id) - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) if target_user != requester.user: raise AuthError(403, "Cannot get filters for other users") @@ -52,7 +49,7 @@ class GetFilterRestServlet(RestServlet): raise SynapseError(400, "Invalid filter_id") try: - filter_collection = yield self.filtering.get_user_filter( + filter_collection = await self.filtering.get_user_filter( user_localpart=target_user.localpart, filter_id=filter_id ) except StoreError as e: @@ -72,11 +69,10 @@ class CreateFilterRestServlet(RestServlet): self.auth = hs.get_auth() self.filtering = hs.get_filtering() - @defer.inlineCallbacks - def on_POST(self, request, user_id): + async def on_POST(self, request, user_id): target_user = UserID.from_string(user_id) - requester = yield self.auth.get_user_by_req(request) + requester = await self.auth.get_user_by_req(request) if target_user != requester.user: raise AuthError(403, "Cannot create filters for other users") @@ -87,7 +83,7 @@ class CreateFilterRestServlet(RestServlet): content = parse_json_object_from_request(request) set_timeline_upper_limit(content, self.hs.config.filter_timeline_limit) - filter_id = yield self.filtering.add_user_filter( + filter_id = await self.filtering.add_user_filter( user_localpart=target_user.localpart, user_filter=content ) diff --git a/synapse/rest/client/v2_alpha/groups.py b/synapse/rest/client/v2_alpha/groups.py index 999a0fa80c..d84a6d7e11 100644 --- a/synapse/rest/client/v2_alpha/groups.py +++ b/synapse/rest/client/v2_alpha/groups.py @@ -16,8 +16,6 @@ import logging -from twisted.internet import defer - from synapse.http.servlet import RestServlet, parse_json_object_from_request from synapse.types import GroupID @@ -38,24 +36,22 @@ class GroupServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, group_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - group_description = yield self.groups_handler.get_group_profile( + group_description = await self.groups_handler.get_group_profile( group_id, requester_user_id ) return 200, group_description - @defer.inlineCallbacks - def on_POST(self, request, group_id): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request, group_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - yield self.groups_handler.update_group_profile( + await self.groups_handler.update_group_profile( group_id, requester_user_id, content ) @@ -74,12 +70,11 @@ class GroupSummaryServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, group_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - get_group_summary = yield self.groups_handler.get_group_summary( + get_group_summary = await self.groups_handler.get_group_summary( group_id, requester_user_id ) @@ -106,13 +101,12 @@ class GroupSummaryRoomsCatServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id, category_id, room_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id, category_id, room_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - resp = yield self.groups_handler.update_group_summary_room( + resp = await self.groups_handler.update_group_summary_room( group_id, requester_user_id, room_id=room_id, @@ -122,12 +116,11 @@ class GroupSummaryRoomsCatServlet(RestServlet): return 200, resp - @defer.inlineCallbacks - def on_DELETE(self, request, group_id, category_id, room_id): - requester = yield self.auth.get_user_by_req(request) + async def on_DELETE(self, request, group_id, category_id, room_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() - resp = yield self.groups_handler.delete_group_summary_room( + resp = await self.groups_handler.delete_group_summary_room( group_id, requester_user_id, room_id=room_id, category_id=category_id ) @@ -148,35 +141,32 @@ class GroupCategoryServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id, category_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, group_id, category_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - category = yield self.groups_handler.get_group_category( + category = await self.groups_handler.get_group_category( group_id, requester_user_id, category_id=category_id ) return 200, category - @defer.inlineCallbacks - def on_PUT(self, request, group_id, category_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id, category_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - resp = yield self.groups_handler.update_group_category( + resp = await self.groups_handler.update_group_category( group_id, requester_user_id, category_id=category_id, content=content ) return 200, resp - @defer.inlineCallbacks - def on_DELETE(self, request, group_id, category_id): - requester = yield self.auth.get_user_by_req(request) + async def on_DELETE(self, request, group_id, category_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() - resp = yield self.groups_handler.delete_group_category( + resp = await self.groups_handler.delete_group_category( group_id, requester_user_id, category_id=category_id ) @@ -195,12 +185,11 @@ class GroupCategoriesServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, group_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - category = yield self.groups_handler.get_group_categories( + category = await self.groups_handler.get_group_categories( group_id, requester_user_id ) @@ -219,35 +208,32 @@ class GroupRoleServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id, role_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, group_id, role_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - category = yield self.groups_handler.get_group_role( + category = await self.groups_handler.get_group_role( group_id, requester_user_id, role_id=role_id ) return 200, category - @defer.inlineCallbacks - def on_PUT(self, request, group_id, role_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id, role_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - resp = yield self.groups_handler.update_group_role( + resp = await self.groups_handler.update_group_role( group_id, requester_user_id, role_id=role_id, content=content ) return 200, resp - @defer.inlineCallbacks - def on_DELETE(self, request, group_id, role_id): - requester = yield self.auth.get_user_by_req(request) + async def on_DELETE(self, request, group_id, role_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() - resp = yield self.groups_handler.delete_group_role( + resp = await self.groups_handler.delete_group_role( group_id, requester_user_id, role_id=role_id ) @@ -266,12 +252,11 @@ class GroupRolesServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, group_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - category = yield self.groups_handler.get_group_roles( + category = await self.groups_handler.get_group_roles( group_id, requester_user_id ) @@ -298,13 +283,12 @@ class GroupSummaryUsersRoleServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id, role_id, user_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id, role_id, user_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - resp = yield self.groups_handler.update_group_summary_user( + resp = await self.groups_handler.update_group_summary_user( group_id, requester_user_id, user_id=user_id, @@ -314,12 +298,11 @@ class GroupSummaryUsersRoleServlet(RestServlet): return 200, resp - @defer.inlineCallbacks - def on_DELETE(self, request, group_id, role_id, user_id): - requester = yield self.auth.get_user_by_req(request) + async def on_DELETE(self, request, group_id, role_id, user_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() - resp = yield self.groups_handler.delete_group_summary_user( + resp = await self.groups_handler.delete_group_summary_user( group_id, requester_user_id, user_id=user_id, role_id=role_id ) @@ -338,12 +321,11 @@ class GroupRoomServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, group_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - result = yield self.groups_handler.get_rooms_in_group( + result = await self.groups_handler.get_rooms_in_group( group_id, requester_user_id ) @@ -362,12 +344,11 @@ class GroupUsersServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, group_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - result = yield self.groups_handler.get_users_in_group( + result = await self.groups_handler.get_users_in_group( group_id, requester_user_id ) @@ -386,12 +367,11 @@ class GroupInvitedUsersServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, group_id): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request, group_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() - result = yield self.groups_handler.get_invited_users_in_group( + result = await self.groups_handler.get_invited_users_in_group( group_id, requester_user_id ) @@ -409,14 +389,13 @@ class GroupSettingJoinPolicyServlet(RestServlet): self.auth = hs.get_auth() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - result = yield self.groups_handler.set_group_join_policy( + result = await self.groups_handler.set_group_join_policy( group_id, requester_user_id, content ) @@ -436,9 +415,8 @@ class GroupCreateServlet(RestServlet): self.groups_handler = hs.get_groups_local_handler() self.server_name = hs.hostname - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() # TODO: Create group on remote server @@ -446,7 +424,7 @@ class GroupCreateServlet(RestServlet): localpart = content.pop("localpart") group_id = GroupID(localpart, self.server_name).to_string() - result = yield self.groups_handler.create_group( + result = await self.groups_handler.create_group( group_id, requester_user_id, content ) @@ -467,24 +445,22 @@ class GroupAdminRoomsServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id, room_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id, room_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - result = yield self.groups_handler.add_room_to_group( + result = await self.groups_handler.add_room_to_group( group_id, requester_user_id, room_id, content ) return 200, result - @defer.inlineCallbacks - def on_DELETE(self, request, group_id, room_id): - requester = yield self.auth.get_user_by_req(request) + async def on_DELETE(self, request, group_id, room_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() - result = yield self.groups_handler.remove_room_from_group( + result = await self.groups_handler.remove_room_from_group( group_id, requester_user_id, room_id ) @@ -506,13 +482,12 @@ class GroupAdminRoomsConfigServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id, room_id, config_key): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id, room_id, config_key): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - result = yield self.groups_handler.update_room_in_group( + result = await self.groups_handler.update_room_in_group( group_id, requester_user_id, room_id, config_key, content ) @@ -535,14 +510,13 @@ class GroupAdminUsersInviteServlet(RestServlet): self.store = hs.get_datastore() self.is_mine_id = hs.is_mine_id - @defer.inlineCallbacks - def on_PUT(self, request, group_id, user_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id, user_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) config = content.get("config", {}) - result = yield self.groups_handler.invite( + result = await self.groups_handler.invite( group_id, user_id, requester_user_id, config ) @@ -563,13 +537,12 @@ class GroupAdminUsersKickServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id, user_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id, user_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - result = yield self.groups_handler.remove_user_from_group( + result = await self.groups_handler.remove_user_from_group( group_id, user_id, requester_user_id, content ) @@ -588,13 +561,12 @@ class GroupSelfLeaveServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - result = yield self.groups_handler.remove_user_from_group( + result = await self.groups_handler.remove_user_from_group( group_id, requester_user_id, requester_user_id, content ) @@ -613,13 +585,12 @@ class GroupSelfJoinServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - result = yield self.groups_handler.join_group( + result = await self.groups_handler.join_group( group_id, requester_user_id, content ) @@ -638,13 +609,12 @@ class GroupSelfAcceptInviteServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_PUT(self, request, group_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) - result = yield self.groups_handler.accept_invite( + result = await self.groups_handler.accept_invite( group_id, requester_user_id, content ) @@ -663,14 +633,13 @@ class GroupSelfUpdatePublicityServlet(RestServlet): self.clock = hs.get_clock() self.store = hs.get_datastore() - @defer.inlineCallbacks - def on_PUT(self, request, group_id): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, group_id): + requester = await self.auth.get_user_by_req(request) requester_user_id = requester.user.to_string() content = parse_json_object_from_request(request) publicise = content["publicise"] - yield self.store.update_group_publicity(group_id, requester_user_id, publicise) + await self.store.update_group_publicity(group_id, requester_user_id, publicise) return 200, {} @@ -688,11 +657,10 @@ class PublicisedGroupsForUserServlet(RestServlet): self.store = hs.get_datastore() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request, user_id): - yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, user_id): + await self.auth.get_user_by_req(request, allow_guest=True) - result = yield self.groups_handler.get_publicised_groups_for_user(user_id) + result = await self.groups_handler.get_publicised_groups_for_user(user_id) return 200, result @@ -710,14 +678,13 @@ class PublicisedGroupsForUsersServlet(RestServlet): self.store = hs.get_datastore() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_POST(self, request): - yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_POST(self, request): + await self.auth.get_user_by_req(request, allow_guest=True) content = parse_json_object_from_request(request) user_ids = content["user_ids"] - result = yield self.groups_handler.bulk_get_publicised_groups(user_ids) + result = await self.groups_handler.bulk_get_publicised_groups(user_ids) return 200, result @@ -734,12 +701,11 @@ class GroupsForUserServlet(RestServlet): self.clock = hs.get_clock() self.groups_handler = hs.get_groups_local_handler() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request, allow_guest=True) requester_user_id = requester.user.to_string() - result = yield self.groups_handler.get_joined_groups(requester_user_id) + result = await self.groups_handler.get_joined_groups(requester_user_id) return 200, result diff --git a/synapse/rest/client/v2_alpha/keys.py b/synapse/rest/client/v2_alpha/keys.py index 341567ae21..f7ed4daf90 100644 --- a/synapse/rest/client/v2_alpha/keys.py +++ b/synapse/rest/client/v2_alpha/keys.py @@ -16,8 +16,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import SynapseError from synapse.http.servlet import ( RestServlet, @@ -71,9 +69,8 @@ class KeyUploadServlet(RestServlet): self.e2e_keys_handler = hs.get_e2e_keys_handler() @trace(opname="upload_keys") - @defer.inlineCallbacks - def on_POST(self, request, device_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_POST(self, request, device_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) user_id = requester.user.to_string() body = parse_json_object_from_request(request) @@ -103,7 +100,7 @@ class KeyUploadServlet(RestServlet): 400, "To upload keys, you must pass device_id when authenticating" ) - result = yield self.e2e_keys_handler.upload_keys_for_user( + result = await self.e2e_keys_handler.upload_keys_for_user( user_id, device_id, body ) return 200, result @@ -154,13 +151,12 @@ class KeyQueryServlet(RestServlet): self.auth = hs.get_auth() self.e2e_keys_handler = hs.get_e2e_keys_handler() - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request, allow_guest=True) user_id = requester.user.to_string() timeout = parse_integer(request, "timeout", 10 * 1000) body = parse_json_object_from_request(request) - result = yield self.e2e_keys_handler.query_devices(body, timeout, user_id) + result = await self.e2e_keys_handler.query_devices(body, timeout, user_id) return 200, result @@ -185,9 +181,8 @@ class KeyChangesServlet(RestServlet): self.auth = hs.get_auth() self.device_handler = hs.get_device_handler() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request, allow_guest=True) from_token_string = parse_string(request, "from") set_tag("from", from_token_string) @@ -200,7 +195,7 @@ class KeyChangesServlet(RestServlet): user_id = requester.user.to_string() - results = yield self.device_handler.get_user_ids_changed(user_id, from_token) + results = await self.device_handler.get_user_ids_changed(user_id, from_token) return 200, results @@ -231,12 +226,11 @@ class OneTimeKeyServlet(RestServlet): self.auth = hs.get_auth() self.e2e_keys_handler = hs.get_e2e_keys_handler() - @defer.inlineCallbacks - def on_POST(self, request): - yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_POST(self, request): + await self.auth.get_user_by_req(request, allow_guest=True) timeout = parse_integer(request, "timeout", 10 * 1000) body = parse_json_object_from_request(request) - result = yield self.e2e_keys_handler.claim_one_time_keys(body, timeout) + result = await self.e2e_keys_handler.claim_one_time_keys(body, timeout) return 200, result @@ -263,17 +257,16 @@ class SigningKeyUploadServlet(RestServlet): self.auth_handler = hs.get_auth_handler() @interactive_auth_handler - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() body = parse_json_object_from_request(request) - yield self.auth_handler.validate_user_via_ui_auth( + await self.auth_handler.validate_user_via_ui_auth( requester, body, self.hs.get_ip_from_request(request) ) - result = yield self.e2e_keys_handler.upload_signing_keys_for_user(user_id, body) + result = await self.e2e_keys_handler.upload_signing_keys_for_user(user_id, body) return 200, result @@ -315,13 +308,12 @@ class SignaturesUploadServlet(RestServlet): self.auth = hs.get_auth() self.e2e_keys_handler = hs.get_e2e_keys_handler() - @defer.inlineCallbacks - def on_POST(self, request): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_POST(self, request): + requester = await self.auth.get_user_by_req(request, allow_guest=True) user_id = requester.user.to_string() body = parse_json_object_from_request(request) - result = yield self.e2e_keys_handler.upload_signatures_for_device_keys( + result = await self.e2e_keys_handler.upload_signatures_for_device_keys( user_id, body ) return 200, result diff --git a/synapse/rest/client/v2_alpha/notifications.py b/synapse/rest/client/v2_alpha/notifications.py index 10c1ad5b07..aa911d75ee 100644 --- a/synapse/rest/client/v2_alpha/notifications.py +++ b/synapse/rest/client/v2_alpha/notifications.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.events.utils import format_event_for_client_v2_without_room_id from synapse.http.servlet import RestServlet, parse_integer, parse_string @@ -35,9 +33,8 @@ class NotificationsServlet(RestServlet): self.clock = hs.get_clock() self._event_serializer = hs.get_event_client_serializer() - @defer.inlineCallbacks - def on_GET(self, request): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request): + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() from_token = parse_string(request, "from", required=False) @@ -46,16 +43,16 @@ class NotificationsServlet(RestServlet): limit = min(limit, 500) - push_actions = yield self.store.get_push_actions_for_user( + push_actions = await self.store.get_push_actions_for_user( user_id, from_token, limit, only_highlight=(only == "highlight") ) - receipts_by_room = yield self.store.get_receipts_for_user_with_orderings( + receipts_by_room = await self.store.get_receipts_for_user_with_orderings( user_id, "m.read" ) notif_event_ids = [pa["event_id"] for pa in push_actions] - notif_events = yield self.store.get_events(notif_event_ids) + notif_events = await self.store.get_events(notif_event_ids) returned_push_actions = [] @@ -68,7 +65,7 @@ class NotificationsServlet(RestServlet): "actions": pa["actions"], "ts": pa["received_ts"], "event": ( - yield self._event_serializer.serialize_event( + await self._event_serializer.serialize_event( notif_events[pa["event_id"]], self.clock.time_msec(), event_format=format_event_for_client_v2_without_room_id, diff --git a/synapse/rest/client/v2_alpha/openid.py b/synapse/rest/client/v2_alpha/openid.py index b4925c0f59..6ae9a5a8e9 100644 --- a/synapse/rest/client/v2_alpha/openid.py +++ b/synapse/rest/client/v2_alpha/openid.py @@ -16,8 +16,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import AuthError from synapse.http.servlet import RestServlet, parse_json_object_from_request from synapse.util.stringutils import random_string @@ -68,9 +66,8 @@ class IdTokenServlet(RestServlet): self.clock = hs.get_clock() self.server_name = hs.config.server_name - @defer.inlineCallbacks - def on_POST(self, request, user_id): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request, user_id): + requester = await self.auth.get_user_by_req(request) if user_id != requester.user.to_string(): raise AuthError(403, "Cannot request tokens for other users.") @@ -81,7 +78,7 @@ class IdTokenServlet(RestServlet): token = random_string(24) ts_valid_until_ms = self.clock.time_msec() + self.EXPIRES_MS - yield self.store.insert_open_id_token(token, ts_valid_until_ms, user_id) + await self.store.insert_open_id_token(token, ts_valid_until_ms, user_id) return ( 200, diff --git a/synapse/rest/client/v2_alpha/register.py b/synapse/rest/client/v2_alpha/register.py index 91db923814..66de16a1fa 100644 --- a/synapse/rest/client/v2_alpha/register.py +++ b/synapse/rest/client/v2_alpha/register.py @@ -20,8 +20,6 @@ from typing import List, Union from six import string_types -from twisted.internet import defer - import synapse import synapse.types from synapse.api.constants import LoginType @@ -102,8 +100,7 @@ class EmailRegisterRequestTokenRestServlet(RestServlet): template_text=template_text, ) - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): if self.hs.config.threepid_behaviour_email == ThreepidBehaviour.OFF: if self.hs.config.local_threepid_handling_disabled_due_to_email_config: logger.warning( @@ -129,7 +126,7 @@ class EmailRegisterRequestTokenRestServlet(RestServlet): Codes.THREEPID_DENIED, ) - existing_user_id = yield self.hs.get_datastore().get_user_id_by_threepid( + existing_user_id = await self.hs.get_datastore().get_user_id_by_threepid( "email", body["email"] ) @@ -140,7 +137,7 @@ class EmailRegisterRequestTokenRestServlet(RestServlet): assert self.hs.config.account_threepid_delegate_email # Have the configured identity server handle the request - ret = yield self.identity_handler.requestEmailToken( + ret = await self.identity_handler.requestEmailToken( self.hs.config.account_threepid_delegate_email, email, client_secret, @@ -149,7 +146,7 @@ class EmailRegisterRequestTokenRestServlet(RestServlet): ) else: # Send registration emails from Synapse - sid = yield self.identity_handler.send_threepid_validation( + sid = await self.identity_handler.send_threepid_validation( email, client_secret, send_attempt, @@ -175,8 +172,7 @@ class MsisdnRegisterRequestTokenRestServlet(RestServlet): self.hs = hs self.identity_handler = hs.get_handlers().identity_handler - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): body = parse_json_object_from_request(request) assert_params_in_dict( @@ -197,7 +193,7 @@ class MsisdnRegisterRequestTokenRestServlet(RestServlet): Codes.THREEPID_DENIED, ) - existing_user_id = yield self.hs.get_datastore().get_user_id_by_threepid( + existing_user_id = await self.hs.get_datastore().get_user_id_by_threepid( "msisdn", msisdn ) @@ -215,7 +211,7 @@ class MsisdnRegisterRequestTokenRestServlet(RestServlet): 400, "Registration by phone number is not supported on this homeserver" ) - ret = yield self.identity_handler.requestMsisdnToken( + ret = await self.identity_handler.requestMsisdnToken( self.hs.config.account_threepid_delegate_msisdn, country, phone_number, @@ -258,8 +254,7 @@ class RegistrationSubmitTokenServlet(RestServlet): [self.config.email_registration_template_failure_html], ) - @defer.inlineCallbacks - def on_GET(self, request, medium): + async def on_GET(self, request, medium): if medium != "email": raise SynapseError( 400, "This medium is currently not supported for registration" @@ -280,7 +275,7 @@ class RegistrationSubmitTokenServlet(RestServlet): # Attempt to validate a 3PID session try: # Mark the session as valid - next_link = yield self.store.validate_threepid_session( + next_link = await self.store.validate_threepid_session( sid, client_secret, token, self.clock.time_msec() ) @@ -338,8 +333,7 @@ class UsernameAvailabilityRestServlet(RestServlet): ), ) - @defer.inlineCallbacks - def on_GET(self, request): + async def on_GET(self, request): if not self.hs.config.enable_registration: raise SynapseError( 403, "Registration has been disabled", errcode=Codes.FORBIDDEN @@ -347,11 +341,11 @@ class UsernameAvailabilityRestServlet(RestServlet): ip = self.hs.get_ip_from_request(request) with self.ratelimiter.ratelimit(ip) as wait_deferred: - yield wait_deferred + await wait_deferred username = parse_string(request, "username", required=True) - yield self.registration_handler.check_username(username) + await self.registration_handler.check_username(username) return 200, {"available": True} @@ -382,8 +376,7 @@ class RegisterRestServlet(RestServlet): ) @interactive_auth_handler - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): body = parse_json_object_from_request(request) client_addr = request.getClientIP() @@ -408,7 +401,7 @@ class RegisterRestServlet(RestServlet): kind = request.args[b"kind"][0] if kind == b"guest": - ret = yield self._do_guest_registration(body, address=client_addr) + ret = await self._do_guest_registration(body, address=client_addr) return ret elif kind != b"user": raise UnrecognizedRequestError( @@ -435,7 +428,7 @@ class RegisterRestServlet(RestServlet): appservice = None if self.auth.has_access_token(request): - appservice = yield self.auth.get_appservice_by_req(request) + appservice = await self.auth.get_appservice_by_req(request) # fork off as soon as possible for ASes which have completely # different registration flows to normal users @@ -455,7 +448,7 @@ class RegisterRestServlet(RestServlet): access_token = self.auth.get_access_token_from_request(request) if isinstance(desired_username, string_types): - result = yield self._do_appservice_registration( + result = await self._do_appservice_registration( desired_username, access_token, body ) return 200, result # we throw for non 200 responses @@ -495,13 +488,13 @@ class RegisterRestServlet(RestServlet): ) if desired_username is not None: - yield self.registration_handler.check_username( + await self.registration_handler.check_username( desired_username, guest_access_token=guest_access_token, assigned_user_id=registered_user_id, ) - auth_result, params, session_id = yield self.auth_handler.check_auth( + auth_result, params, session_id = await self.auth_handler.check_auth( self._registration_flows, body, self.hs.get_ip_from_request(request) ) @@ -557,7 +550,7 @@ class RegisterRestServlet(RestServlet): medium = auth_result[login_type]["medium"] address = auth_result[login_type]["address"] - existing_user_id = yield self.store.get_user_id_by_threepid( + existing_user_id = await self.store.get_user_id_by_threepid( medium, address ) @@ -568,7 +561,7 @@ class RegisterRestServlet(RestServlet): Codes.THREEPID_IN_USE, ) - registered_user_id = yield self.registration_handler.register_user( + registered_user_id = await self.registration_handler.register_user( localpart=desired_username, password=new_password, guest_access_token=guest_access_token, @@ -581,7 +574,7 @@ class RegisterRestServlet(RestServlet): if is_threepid_reserved( self.hs.config.mau_limits_reserved_threepids, threepid ): - yield self.store.upsert_monthly_active_user(registered_user_id) + await self.store.upsert_monthly_active_user(registered_user_id) # remember that we've now registered that user account, and with # what user ID (since the user may not have specified) @@ -591,12 +584,12 @@ class RegisterRestServlet(RestServlet): registered = True - return_dict = yield self._create_registration_details( + return_dict = await self._create_registration_details( registered_user_id, params ) if registered: - yield self.registration_handler.post_registration_actions( + await self.registration_handler.post_registration_actions( user_id=registered_user_id, auth_result=auth_result, access_token=return_dict.get("access_token"), @@ -607,15 +600,13 @@ class RegisterRestServlet(RestServlet): def on_OPTIONS(self, _): return 200, {} - @defer.inlineCallbacks - def _do_appservice_registration(self, username, as_token, body): - user_id = yield self.registration_handler.appservice_register( + async def _do_appservice_registration(self, username, as_token, body): + user_id = await self.registration_handler.appservice_register( username, as_token ) - return (yield self._create_registration_details(user_id, body)) + return await self._create_registration_details(user_id, body) - @defer.inlineCallbacks - def _create_registration_details(self, user_id, params): + async def _create_registration_details(self, user_id, params): """Complete registration of newly-registered user Allocates device_id if one was not given; also creates access_token. @@ -631,18 +622,17 @@ class RegisterRestServlet(RestServlet): if not params.get("inhibit_login", False): device_id = params.get("device_id") initial_display_name = params.get("initial_device_display_name") - device_id, access_token = yield self.registration_handler.register_device( + device_id, access_token = await self.registration_handler.register_device( user_id, device_id, initial_display_name, is_guest=False ) result.update({"access_token": access_token, "device_id": device_id}) return result - @defer.inlineCallbacks - def _do_guest_registration(self, params, address=None): + async def _do_guest_registration(self, params, address=None): if not self.hs.config.allow_guest_access: raise SynapseError(403, "Guest access is disabled") - user_id = yield self.registration_handler.register_user( + user_id = await self.registration_handler.register_user( make_guest=True, address=address ) @@ -650,7 +640,7 @@ class RegisterRestServlet(RestServlet): # we have nowhere to store it. device_id = synapse.api.auth.GUEST_DEVICE_ID initial_display_name = params.get("initial_device_display_name") - device_id, access_token = yield self.registration_handler.register_device( + device_id, access_token = await self.registration_handler.register_device( user_id, device_id, initial_display_name, is_guest=True ) diff --git a/synapse/rest/client/v2_alpha/relations.py b/synapse/rest/client/v2_alpha/relations.py index 040b37c504..9be9a34b91 100644 --- a/synapse/rest/client/v2_alpha/relations.py +++ b/synapse/rest/client/v2_alpha/relations.py @@ -21,8 +21,6 @@ any time to reflect changes in the MSC. import logging -from twisted.internet import defer - from synapse.api.constants import EventTypes, RelationTypes from synapse.api.errors import SynapseError from synapse.http.servlet import ( @@ -86,11 +84,10 @@ class RelationSendServlet(RestServlet): request, self.on_PUT_or_POST, request, *args, **kwargs ) - @defer.inlineCallbacks - def on_PUT_or_POST( + async def on_PUT_or_POST( self, request, room_id, parent_id, relation_type, event_type, txn_id=None ): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + requester = await self.auth.get_user_by_req(request, allow_guest=True) if event_type == EventTypes.Member: # Add relations to a membership is meaningless, so we just deny it @@ -114,7 +111,7 @@ class RelationSendServlet(RestServlet): "sender": requester.user.to_string(), } - event = yield self.event_creation_handler.create_and_send_nonmember_event( + event = await self.event_creation_handler.create_and_send_nonmember_event( requester, event_dict=event_dict, txn_id=txn_id ) @@ -140,17 +137,18 @@ class RelationPaginationServlet(RestServlet): self._event_serializer = hs.get_event_client_serializer() self.event_handler = hs.get_event_handler() - @defer.inlineCallbacks - def on_GET(self, request, room_id, parent_id, relation_type=None, event_type=None): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET( + self, request, room_id, parent_id, relation_type=None, event_type=None + ): + requester = await self.auth.get_user_by_req(request, allow_guest=True) - yield self.auth.check_in_room_or_world_readable( + await self.auth.check_in_room_or_world_readable( room_id, requester.user.to_string() ) # This gets the original event and checks that a) the event exists and # b) the user is allowed to view it. - event = yield self.event_handler.get_event(requester.user, room_id, parent_id) + event = await self.event_handler.get_event(requester.user, room_id, parent_id) limit = parse_integer(request, "limit", default=5) from_token = parse_string(request, "from") @@ -167,7 +165,7 @@ class RelationPaginationServlet(RestServlet): if to_token: to_token = RelationPaginationToken.from_string(to_token) - pagination_chunk = yield self.store.get_relations_for_event( + pagination_chunk = await self.store.get_relations_for_event( event_id=parent_id, relation_type=relation_type, event_type=event_type, @@ -176,7 +174,7 @@ class RelationPaginationServlet(RestServlet): to_token=to_token, ) - events = yield self.store.get_events_as_list( + events = await self.store.get_events_as_list( [c["event_id"] for c in pagination_chunk.chunk] ) @@ -184,13 +182,13 @@ class RelationPaginationServlet(RestServlet): # We set bundle_aggregations to False when retrieving the original # event because we want the content before relations were applied to # it. - original_event = yield self._event_serializer.serialize_event( + original_event = await self._event_serializer.serialize_event( event, now, bundle_aggregations=False ) # Similarly, we don't allow relations to be applied to relations, so we # return the original relations without any aggregations on top of them # here. - events = yield self._event_serializer.serialize_events( + events = await self._event_serializer.serialize_events( events, now, bundle_aggregations=False ) @@ -232,17 +230,18 @@ class RelationAggregationPaginationServlet(RestServlet): self.store = hs.get_datastore() self.event_handler = hs.get_event_handler() - @defer.inlineCallbacks - def on_GET(self, request, room_id, parent_id, relation_type=None, event_type=None): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET( + self, request, room_id, parent_id, relation_type=None, event_type=None + ): + requester = await self.auth.get_user_by_req(request, allow_guest=True) - yield self.auth.check_in_room_or_world_readable( + await self.auth.check_in_room_or_world_readable( room_id, requester.user.to_string() ) # This checks that a) the event exists and b) the user is allowed to # view it. - event = yield self.event_handler.get_event(requester.user, room_id, parent_id) + event = await self.event_handler.get_event(requester.user, room_id, parent_id) if relation_type not in (RelationTypes.ANNOTATION, None): raise SynapseError(400, "Relation type must be 'annotation'") @@ -262,7 +261,7 @@ class RelationAggregationPaginationServlet(RestServlet): if to_token: to_token = AggregationPaginationToken.from_string(to_token) - pagination_chunk = yield self.store.get_aggregation_groups_for_event( + pagination_chunk = await self.store.get_aggregation_groups_for_event( event_id=parent_id, event_type=event_type, limit=limit, @@ -311,17 +310,16 @@ class RelationAggregationGroupPaginationServlet(RestServlet): self._event_serializer = hs.get_event_client_serializer() self.event_handler = hs.get_event_handler() - @defer.inlineCallbacks - def on_GET(self, request, room_id, parent_id, relation_type, event_type, key): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, room_id, parent_id, relation_type, event_type, key): + requester = await self.auth.get_user_by_req(request, allow_guest=True) - yield self.auth.check_in_room_or_world_readable( + await self.auth.check_in_room_or_world_readable( room_id, requester.user.to_string() ) # This checks that a) the event exists and b) the user is allowed to # view it. - yield self.event_handler.get_event(requester.user, room_id, parent_id) + await self.event_handler.get_event(requester.user, room_id, parent_id) if relation_type != RelationTypes.ANNOTATION: raise SynapseError(400, "Relation type must be 'annotation'") @@ -336,7 +334,7 @@ class RelationAggregationGroupPaginationServlet(RestServlet): if to_token: to_token = RelationPaginationToken.from_string(to_token) - result = yield self.store.get_relations_for_event( + result = await self.store.get_relations_for_event( event_id=parent_id, relation_type=relation_type, event_type=event_type, @@ -346,12 +344,12 @@ class RelationAggregationGroupPaginationServlet(RestServlet): to_token=to_token, ) - events = yield self.store.get_events_as_list( + events = await self.store.get_events_as_list( [c["event_id"] for c in result.chunk] ) now = self.clock.time_msec() - events = yield self._event_serializer.serialize_events(events, now) + events = await self._event_serializer.serialize_events(events, now) return_value = result.to_dict() return_value["chunk"] = events diff --git a/synapse/rest/client/v2_alpha/report_event.py b/synapse/rest/client/v2_alpha/report_event.py index e7449864cd..f067b5edac 100644 --- a/synapse/rest/client/v2_alpha/report_event.py +++ b/synapse/rest/client/v2_alpha/report_event.py @@ -18,8 +18,6 @@ import logging from six import string_types from six.moves import http_client -from twisted.internet import defer - from synapse.api.errors import Codes, SynapseError from synapse.http.servlet import ( RestServlet, @@ -42,9 +40,8 @@ class ReportEventRestServlet(RestServlet): self.clock = hs.get_clock() self.store = hs.get_datastore() - @defer.inlineCallbacks - def on_POST(self, request, room_id, event_id): - requester = yield self.auth.get_user_by_req(request) + async def on_POST(self, request, room_id, event_id): + requester = await self.auth.get_user_by_req(request) user_id = requester.user.to_string() body = parse_json_object_from_request(request) @@ -63,7 +60,7 @@ class ReportEventRestServlet(RestServlet): Codes.BAD_JSON, ) - yield self.store.add_event_report( + await self.store.add_event_report( room_id=room_id, event_id=event_id, user_id=user_id, diff --git a/synapse/rest/client/v2_alpha/room_keys.py b/synapse/rest/client/v2_alpha/room_keys.py index d596786430..38952a1d27 100644 --- a/synapse/rest/client/v2_alpha/room_keys.py +++ b/synapse/rest/client/v2_alpha/room_keys.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import Codes, NotFoundError, SynapseError from synapse.http.servlet import ( RestServlet, @@ -43,8 +41,7 @@ class RoomKeysServlet(RestServlet): self.auth = hs.get_auth() self.e2e_room_keys_handler = hs.get_e2e_room_keys_handler() - @defer.inlineCallbacks - def on_PUT(self, request, room_id, session_id): + async def on_PUT(self, request, room_id, session_id): """ Uploads one or more encrypted E2E room keys for backup purposes. room_id: the ID of the room the keys are for (optional) @@ -123,7 +120,7 @@ class RoomKeysServlet(RestServlet): } } """ - requester = yield self.auth.get_user_by_req(request, allow_guest=False) + requester = await self.auth.get_user_by_req(request, allow_guest=False) user_id = requester.user.to_string() body = parse_json_object_from_request(request) version = parse_string(request, "version") @@ -134,11 +131,10 @@ class RoomKeysServlet(RestServlet): if room_id: body = {"rooms": {room_id: body}} - yield self.e2e_room_keys_handler.upload_room_keys(user_id, version, body) - return 200, {} + ret = await self.e2e_room_keys_handler.upload_room_keys(user_id, version, body) + return 200, ret - @defer.inlineCallbacks - def on_GET(self, request, room_id, session_id): + async def on_GET(self, request, room_id, session_id): """ Retrieves one or more encrypted E2E room keys for backup purposes. Symmetric with the PUT version of the API. @@ -190,11 +186,11 @@ class RoomKeysServlet(RestServlet): } } """ - requester = yield self.auth.get_user_by_req(request, allow_guest=False) + requester = await self.auth.get_user_by_req(request, allow_guest=False) user_id = requester.user.to_string() version = parse_string(request, "version") - room_keys = yield self.e2e_room_keys_handler.get_room_keys( + room_keys = await self.e2e_room_keys_handler.get_room_keys( user_id, version, room_id, session_id ) @@ -220,8 +216,7 @@ class RoomKeysServlet(RestServlet): return 200, room_keys - @defer.inlineCallbacks - def on_DELETE(self, request, room_id, session_id): + async def on_DELETE(self, request, room_id, session_id): """ Deletes one or more encrypted E2E room keys for a user for backup purposes. @@ -235,14 +230,14 @@ class RoomKeysServlet(RestServlet): the version must already have been created via the /change_secret API. """ - requester = yield self.auth.get_user_by_req(request, allow_guest=False) + requester = await self.auth.get_user_by_req(request, allow_guest=False) user_id = requester.user.to_string() version = parse_string(request, "version") - yield self.e2e_room_keys_handler.delete_room_keys( + ret = await self.e2e_room_keys_handler.delete_room_keys( user_id, version, room_id, session_id ) - return 200, {} + return 200, ret class RoomKeysNewVersionServlet(RestServlet): @@ -257,8 +252,7 @@ class RoomKeysNewVersionServlet(RestServlet): self.auth = hs.get_auth() self.e2e_room_keys_handler = hs.get_e2e_room_keys_handler() - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): """ Create a new backup version for this user's room_keys with the given info. The version is allocated by the server and returned to the user @@ -288,11 +282,11 @@ class RoomKeysNewVersionServlet(RestServlet): "version": 12345 } """ - requester = yield self.auth.get_user_by_req(request, allow_guest=False) + requester = await self.auth.get_user_by_req(request, allow_guest=False) user_id = requester.user.to_string() info = parse_json_object_from_request(request) - new_version = yield self.e2e_room_keys_handler.create_version(user_id, info) + new_version = await self.e2e_room_keys_handler.create_version(user_id, info) return 200, {"version": new_version} # we deliberately don't have a PUT /version, as these things really should @@ -311,8 +305,7 @@ class RoomKeysVersionServlet(RestServlet): self.auth = hs.get_auth() self.e2e_room_keys_handler = hs.get_e2e_room_keys_handler() - @defer.inlineCallbacks - def on_GET(self, request, version): + async def on_GET(self, request, version): """ Retrieve the version information about a given version of the user's room_keys backup. If the version part is missing, returns info about the @@ -330,18 +323,17 @@ class RoomKeysVersionServlet(RestServlet): "auth_data": "dGhpcyBzaG91bGQgYWN0dWFsbHkgYmUgZW5jcnlwdGVkIGpzb24K" } """ - requester = yield self.auth.get_user_by_req(request, allow_guest=False) + requester = await self.auth.get_user_by_req(request, allow_guest=False) user_id = requester.user.to_string() try: - info = yield self.e2e_room_keys_handler.get_version_info(user_id, version) + info = await self.e2e_room_keys_handler.get_version_info(user_id, version) except SynapseError as e: if e.code == 404: raise SynapseError(404, "No backup found", Codes.NOT_FOUND) return 200, info - @defer.inlineCallbacks - def on_DELETE(self, request, version): + async def on_DELETE(self, request, version): """ Delete the information about a given version of the user's room_keys backup. If the version part is missing, deletes the most @@ -354,14 +346,13 @@ class RoomKeysVersionServlet(RestServlet): if version is None: raise SynapseError(400, "No version specified to delete", Codes.NOT_FOUND) - requester = yield self.auth.get_user_by_req(request, allow_guest=False) + requester = await self.auth.get_user_by_req(request, allow_guest=False) user_id = requester.user.to_string() - yield self.e2e_room_keys_handler.delete_version(user_id, version) + await self.e2e_room_keys_handler.delete_version(user_id, version) return 200, {} - @defer.inlineCallbacks - def on_PUT(self, request, version): + async def on_PUT(self, request, version): """ Update the information about a given version of the user's room_keys backup. @@ -382,7 +373,7 @@ class RoomKeysVersionServlet(RestServlet): Content-Type: application/json {} """ - requester = yield self.auth.get_user_by_req(request, allow_guest=False) + requester = await self.auth.get_user_by_req(request, allow_guest=False) user_id = requester.user.to_string() info = parse_json_object_from_request(request) @@ -391,7 +382,7 @@ class RoomKeysVersionServlet(RestServlet): 400, "No version specified to update", Codes.MISSING_PARAM ) - yield self.e2e_room_keys_handler.update_version(user_id, version, info) + await self.e2e_room_keys_handler.update_version(user_id, version, info) return 200, {} diff --git a/synapse/rest/client/v2_alpha/room_upgrade_rest_servlet.py b/synapse/rest/client/v2_alpha/room_upgrade_rest_servlet.py index d2c3316eb7..ca97330797 100644 --- a/synapse/rest/client/v2_alpha/room_upgrade_rest_servlet.py +++ b/synapse/rest/client/v2_alpha/room_upgrade_rest_servlet.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import Codes, SynapseError from synapse.api.room_versions import KNOWN_ROOM_VERSIONS from synapse.http.servlet import ( @@ -59,9 +57,8 @@ class RoomUpgradeRestServlet(RestServlet): self._room_creation_handler = hs.get_room_creation_handler() self._auth = hs.get_auth() - @defer.inlineCallbacks - def on_POST(self, request, room_id): - requester = yield self._auth.get_user_by_req(request) + async def on_POST(self, request, room_id): + requester = await self._auth.get_user_by_req(request) content = parse_json_object_from_request(request) assert_params_in_dict(content, ("new_version",)) @@ -74,7 +71,7 @@ class RoomUpgradeRestServlet(RestServlet): Codes.UNSUPPORTED_ROOM_VERSION, ) - new_room_id = yield self._room_creation_handler.upgrade_room( + new_room_id = await self._room_creation_handler.upgrade_room( requester, room_id, new_version ) diff --git a/synapse/rest/client/v2_alpha/sendtodevice.py b/synapse/rest/client/v2_alpha/sendtodevice.py index d90e52ed1a..501b52fb6c 100644 --- a/synapse/rest/client/v2_alpha/sendtodevice.py +++ b/synapse/rest/client/v2_alpha/sendtodevice.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.http import servlet from synapse.http.servlet import parse_json_object_from_request from synapse.logging.opentracing import set_tag, trace @@ -51,15 +49,14 @@ class SendToDeviceRestServlet(servlet.RestServlet): request, self._put, request, message_type, txn_id ) - @defer.inlineCallbacks - def _put(self, request, message_type, txn_id): - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + async def _put(self, request, message_type, txn_id): + requester = await self.auth.get_user_by_req(request, allow_guest=True) content = parse_json_object_from_request(request) sender_user_id = requester.user.to_string() - yield self.device_message_handler.send_device_message( + await self.device_message_handler.send_device_message( sender_user_id, message_type, content["messages"] ) diff --git a/synapse/rest/client/v2_alpha/sync.py b/synapse/rest/client/v2_alpha/sync.py index ccd8b17b23..d8292ce29f 100644 --- a/synapse/rest/client/v2_alpha/sync.py +++ b/synapse/rest/client/v2_alpha/sync.py @@ -18,8 +18,6 @@ import logging from canonicaljson import json -from twisted.internet import defer - from synapse.api.constants import PresenceState from synapse.api.errors import Codes, StoreError, SynapseError from synapse.api.filtering import DEFAULT_FILTER_COLLECTION, FilterCollection @@ -87,8 +85,7 @@ class SyncRestServlet(RestServlet): self._server_notices_sender = hs.get_server_notices_sender() self._event_serializer = hs.get_event_client_serializer() - @defer.inlineCallbacks - def on_GET(self, request): + async def on_GET(self, request): if b"from" in request.args: # /events used to use 'from', but /sync uses 'since'. # Lets be helpful and whine if we see a 'from'. @@ -96,7 +93,7 @@ class SyncRestServlet(RestServlet): 400, "'from' is not a valid query parameter. Did you mean 'since'?" ) - requester = yield self.auth.get_user_by_req(request, allow_guest=True) + requester = await self.auth.get_user_by_req(request, allow_guest=True) user = requester.user device_id = requester.device_id @@ -138,7 +135,7 @@ class SyncRestServlet(RestServlet): filter_collection = FilterCollection(filter_object) else: try: - filter_collection = yield self.filtering.get_user_filter( + filter_collection = await self.filtering.get_user_filter( user.localpart, filter_id ) except StoreError as err: @@ -161,20 +158,20 @@ class SyncRestServlet(RestServlet): since_token = None # send any outstanding server notices to the user. - yield self._server_notices_sender.on_user_syncing(user.to_string()) + await self._server_notices_sender.on_user_syncing(user.to_string()) affect_presence = set_presence != PresenceState.OFFLINE if affect_presence: - yield self.presence_handler.set_state( + await self.presence_handler.set_state( user, {"presence": set_presence}, True ) - context = yield self.presence_handler.user_syncing( + context = await self.presence_handler.user_syncing( user.to_string(), affect_presence=affect_presence ) with context: - sync_result = yield self.sync_handler.wait_for_sync_for_user( + sync_result = await self.sync_handler.wait_for_sync_for_user( sync_config, since_token=since_token, timeout=timeout, @@ -182,14 +179,13 @@ class SyncRestServlet(RestServlet): ) time_now = self.clock.time_msec() - response_content = yield self.encode_response( + response_content = await self.encode_response( time_now, sync_result, requester.access_token_id, filter_collection ) return 200, response_content - @defer.inlineCallbacks - def encode_response(self, time_now, sync_result, access_token_id, filter): + async def encode_response(self, time_now, sync_result, access_token_id, filter): if filter.event_format == "client": event_formatter = format_event_for_client_v2_without_room_id elif filter.event_format == "federation": @@ -197,7 +193,7 @@ class SyncRestServlet(RestServlet): else: raise Exception("Unknown event format %s" % (filter.event_format,)) - joined = yield self.encode_joined( + joined = await self.encode_joined( sync_result.joined, time_now, access_token_id, @@ -205,11 +201,11 @@ class SyncRestServlet(RestServlet): event_formatter, ) - invited = yield self.encode_invited( + invited = await self.encode_invited( sync_result.invited, time_now, access_token_id, event_formatter ) - archived = yield self.encode_archived( + archived = await self.encode_archived( sync_result.archived, time_now, access_token_id, @@ -250,8 +246,9 @@ class SyncRestServlet(RestServlet): ] } - @defer.inlineCallbacks - def encode_joined(self, rooms, time_now, token_id, event_fields, event_formatter): + async def encode_joined( + self, rooms, time_now, token_id, event_fields, event_formatter + ): """ Encode the joined rooms in a sync result @@ -272,7 +269,7 @@ class SyncRestServlet(RestServlet): """ joined = {} for room in rooms: - joined[room.room_id] = yield self.encode_room( + joined[room.room_id] = await self.encode_room( room, time_now, token_id, @@ -283,8 +280,7 @@ class SyncRestServlet(RestServlet): return joined - @defer.inlineCallbacks - def encode_invited(self, rooms, time_now, token_id, event_formatter): + async def encode_invited(self, rooms, time_now, token_id, event_formatter): """ Encode the invited rooms in a sync result @@ -304,7 +300,7 @@ class SyncRestServlet(RestServlet): """ invited = {} for room in rooms: - invite = yield self._event_serializer.serialize_event( + invite = await self._event_serializer.serialize_event( room.invite, time_now, token_id=token_id, @@ -319,8 +315,9 @@ class SyncRestServlet(RestServlet): return invited - @defer.inlineCallbacks - def encode_archived(self, rooms, time_now, token_id, event_fields, event_formatter): + async def encode_archived( + self, rooms, time_now, token_id, event_fields, event_formatter + ): """ Encode the archived rooms in a sync result @@ -341,7 +338,7 @@ class SyncRestServlet(RestServlet): """ joined = {} for room in rooms: - joined[room.room_id] = yield self.encode_room( + joined[room.room_id] = await self.encode_room( room, time_now, token_id, @@ -352,8 +349,7 @@ class SyncRestServlet(RestServlet): return joined - @defer.inlineCallbacks - def encode_room( + async def encode_room( self, room, time_now, token_id, joined, only_fields, event_formatter ): """ @@ -401,8 +397,8 @@ class SyncRestServlet(RestServlet): event.room_id, ) - serialized_state = yield serialize(state_events) - serialized_timeline = yield serialize(timeline_events) + serialized_state = await serialize(state_events) + serialized_timeline = await serialize(timeline_events) account_data = room.account_data diff --git a/synapse/rest/client/v2_alpha/tags.py b/synapse/rest/client/v2_alpha/tags.py index 3b555669a0..a3f12e8a77 100644 --- a/synapse/rest/client/v2_alpha/tags.py +++ b/synapse/rest/client/v2_alpha/tags.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import AuthError from synapse.http.servlet import RestServlet, parse_json_object_from_request @@ -37,13 +35,12 @@ class TagListServlet(RestServlet): self.auth = hs.get_auth() self.store = hs.get_datastore() - @defer.inlineCallbacks - def on_GET(self, request, user_id, room_id): - requester = yield self.auth.get_user_by_req(request) + async def on_GET(self, request, user_id, room_id): + requester = await self.auth.get_user_by_req(request) if user_id != requester.user.to_string(): raise AuthError(403, "Cannot get tags for other users.") - tags = yield self.store.get_tags_for_room(user_id, room_id) + tags = await self.store.get_tags_for_room(user_id, room_id) return 200, {"tags": tags} @@ -64,27 +61,25 @@ class TagServlet(RestServlet): self.store = hs.get_datastore() self.notifier = hs.get_notifier() - @defer.inlineCallbacks - def on_PUT(self, request, user_id, room_id, tag): - requester = yield self.auth.get_user_by_req(request) + async def on_PUT(self, request, user_id, room_id, tag): + requester = await self.auth.get_user_by_req(request) if user_id != requester.user.to_string(): raise AuthError(403, "Cannot add tags for other users.") body = parse_json_object_from_request(request) - max_id = yield self.store.add_tag_to_room(user_id, room_id, tag, body) + max_id = await self.store.add_tag_to_room(user_id, room_id, tag, body) self.notifier.on_new_event("account_data_key", max_id, users=[user_id]) return 200, {} - @defer.inlineCallbacks - def on_DELETE(self, request, user_id, room_id, tag): - requester = yield self.auth.get_user_by_req(request) + async def on_DELETE(self, request, user_id, room_id, tag): + requester = await self.auth.get_user_by_req(request) if user_id != requester.user.to_string(): raise AuthError(403, "Cannot add tags for other users.") - max_id = yield self.store.remove_tag_from_room(user_id, room_id, tag) + max_id = await self.store.remove_tag_from_room(user_id, room_id, tag) self.notifier.on_new_event("account_data_key", max_id, users=[user_id]) diff --git a/synapse/rest/client/v2_alpha/thirdparty.py b/synapse/rest/client/v2_alpha/thirdparty.py index 2e8d672471..23709960ad 100644 --- a/synapse/rest/client/v2_alpha/thirdparty.py +++ b/synapse/rest/client/v2_alpha/thirdparty.py @@ -16,8 +16,6 @@ import logging -from twisted.internet import defer - from synapse.api.constants import ThirdPartyEntityKind from synapse.http.servlet import RestServlet @@ -35,11 +33,10 @@ class ThirdPartyProtocolsServlet(RestServlet): self.auth = hs.get_auth() self.appservice_handler = hs.get_application_service_handler() - @defer.inlineCallbacks - def on_GET(self, request): - yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request): + await self.auth.get_user_by_req(request, allow_guest=True) - protocols = yield self.appservice_handler.get_3pe_protocols() + protocols = await self.appservice_handler.get_3pe_protocols() return 200, protocols @@ -52,11 +49,10 @@ class ThirdPartyProtocolServlet(RestServlet): self.auth = hs.get_auth() self.appservice_handler = hs.get_application_service_handler() - @defer.inlineCallbacks - def on_GET(self, request, protocol): - yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, protocol): + await self.auth.get_user_by_req(request, allow_guest=True) - protocols = yield self.appservice_handler.get_3pe_protocols( + protocols = await self.appservice_handler.get_3pe_protocols( only_protocol=protocol ) if protocol in protocols: @@ -74,14 +70,13 @@ class ThirdPartyUserServlet(RestServlet): self.auth = hs.get_auth() self.appservice_handler = hs.get_application_service_handler() - @defer.inlineCallbacks - def on_GET(self, request, protocol): - yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, protocol): + await self.auth.get_user_by_req(request, allow_guest=True) fields = request.args fields.pop(b"access_token", None) - results = yield self.appservice_handler.query_3pe( + results = await self.appservice_handler.query_3pe( ThirdPartyEntityKind.USER, protocol, fields ) @@ -97,14 +92,13 @@ class ThirdPartyLocationServlet(RestServlet): self.auth = hs.get_auth() self.appservice_handler = hs.get_application_service_handler() - @defer.inlineCallbacks - def on_GET(self, request, protocol): - yield self.auth.get_user_by_req(request, allow_guest=True) + async def on_GET(self, request, protocol): + await self.auth.get_user_by_req(request, allow_guest=True) fields = request.args fields.pop(b"access_token", None) - results = yield self.appservice_handler.query_3pe( + results = await self.appservice_handler.query_3pe( ThirdPartyEntityKind.LOCATION, protocol, fields ) diff --git a/synapse/rest/client/v2_alpha/tokenrefresh.py b/synapse/rest/client/v2_alpha/tokenrefresh.py index 2da0f55811..83f3b6b70a 100644 --- a/synapse/rest/client/v2_alpha/tokenrefresh.py +++ b/synapse/rest/client/v2_alpha/tokenrefresh.py @@ -13,8 +13,6 @@ # See the License for the specific language governing permissions and # limitations under the License. -from twisted.internet import defer - from synapse.api.errors import AuthError from synapse.http.servlet import RestServlet @@ -32,8 +30,7 @@ class TokenRefreshRestServlet(RestServlet): def __init__(self, hs): super(TokenRefreshRestServlet, self).__init__() - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): raise AuthError(403, "tokenrefresh is no longer supported.") diff --git a/synapse/rest/client/v2_alpha/user_directory.py b/synapse/rest/client/v2_alpha/user_directory.py index 2863affbab..bef91a2d3e 100644 --- a/synapse/rest/client/v2_alpha/user_directory.py +++ b/synapse/rest/client/v2_alpha/user_directory.py @@ -15,8 +15,6 @@ import logging -from twisted.internet import defer - from synapse.api.errors import SynapseError from synapse.http.servlet import RestServlet, parse_json_object_from_request @@ -38,8 +36,7 @@ class UserDirectorySearchRestServlet(RestServlet): self.auth = hs.get_auth() self.user_directory_handler = hs.get_user_directory_handler() - @defer.inlineCallbacks - def on_POST(self, request): + async def on_POST(self, request): """Searches for users in directory Returns: @@ -56,7 +53,7 @@ class UserDirectorySearchRestServlet(RestServlet): ] } """ - requester = yield self.auth.get_user_by_req(request, allow_guest=False) + requester = await self.auth.get_user_by_req(request, allow_guest=False) user_id = requester.user.to_string() if not self.hs.config.user_directory_search_enabled: @@ -72,7 +69,7 @@ class UserDirectorySearchRestServlet(RestServlet): except Exception: raise SynapseError(400, "`search_term` is required field") - results = yield self.user_directory_handler.search_users( + results = await self.user_directory_handler.search_users( user_id, search_term, limit ) diff --git a/synapse/rest/client/versions.py b/synapse/rest/client/versions.py index bb30ce3f34..3d0fefb4df 100644 --- a/synapse/rest/client/versions.py +++ b/synapse/rest/client/versions.py @@ -1,5 +1,8 @@ # -*- coding: utf-8 -*- # Copyright 2016 OpenMarket Ltd +# Copyright 2017 Vector Creations Ltd +# Copyright 2018-2019 New Vector Ltd +# Copyright 2019 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -68,6 +71,8 @@ class VersionsRestServlet(RestServlet): # Implements support for label-based filtering as described in # MSC2326. "org.matrix.label_based_filtering": True, + # Implements support for cross signing as described in MSC1756 + "org.matrix.e2e_cross_signing": True, }, }, ) diff --git a/synapse/rest/media/v0/__init__.py b/synapse/rest/media/v0/__init__.py deleted file mode 100644 index e69de29bb2..0000000000 --- a/synapse/rest/media/v0/__init__.py +++ /dev/null diff --git a/synapse/rest/media/v0/content_repository.py b/synapse/rest/media/v0/content_repository.py deleted file mode 100644 index 86884c0ef4..0000000000 --- a/synapse/rest/media/v0/content_repository.py +++ /dev/null @@ -1,103 +0,0 @@ -# -*- coding: utf-8 -*- -# Copyright 2014-2016 OpenMarket Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -import base64 -import logging -import os -import re - -from canonicaljson import json - -from twisted.protocols.basic import FileSender -from twisted.web import resource, server - -from synapse.api.errors import Codes, cs_error -from synapse.http.server import finish_request, respond_with_json_bytes - -logger = logging.getLogger(__name__) - - -class ContentRepoResource(resource.Resource): - """Provides file uploading and downloading. - - Uploads are POSTed to wherever this Resource is linked to. This resource - returns a "content token" which can be used to GET this content again. The - token is typically a path, but it may not be. Tokens can expire, be - one-time uses, etc. - - In this case, the token is a path to the file and contains 3 interesting - sections: - - User ID base64d (for namespacing content to each user) - - random 24 char string - - Content type base64d (so we can return it when clients GET it) - - """ - - isLeaf = True - - def __init__(self, hs, directory): - resource.Resource.__init__(self) - self.hs = hs - self.directory = directory - - def render_GET(self, request): - # no auth here on purpose, to allow anyone to view, even across home - # servers. - - # TODO: A little crude here, we could do this better. - filename = request.path.decode("ascii").split("/")[-1] - # be paranoid - filename = re.sub("[^0-9A-z.-_]", "", filename) - - file_path = self.directory + "/" + filename - - logger.debug("Searching for %s", file_path) - - if os.path.isfile(file_path): - # filename has the content type - base64_contentype = filename.split(".")[1] - content_type = base64.urlsafe_b64decode(base64_contentype) - logger.info("Sending file %s", file_path) - f = open(file_path, "rb") - request.setHeader("Content-Type", content_type) - - # cache for at least a day. - # XXX: we might want to turn this off for data we don't want to - # recommend caching as it's sensitive or private - or at least - # select private. don't bother setting Expires as all our matrix - # clients are smart enough to be happy with Cache-Control (right?) - request.setHeader(b"Cache-Control", b"public,max-age=86400,s-maxage=86400") - - d = FileSender().beginFileTransfer(f, request) - - # after the file has been sent, clean up and finish the request - def cbFinished(ignored): - f.close() - finish_request(request) - - d.addCallback(cbFinished) - else: - respond_with_json_bytes( - request, - 404, - json.dumps(cs_error("Not found", code=Codes.NOT_FOUND)), - send_cors=True, - ) - - return server.NOT_DONE_YET - - def render_OPTIONS(self, request): - respond_with_json_bytes(request, 200, {}, send_cors=True) - return server.NOT_DONE_YET diff --git a/synapse/rest/media/v1/preview_url_resource.py b/synapse/rest/media/v1/preview_url_resource.py index 15c15a12f5..6b978be876 100644 --- a/synapse/rest/media/v1/preview_url_resource.py +++ b/synapse/rest/media/v1/preview_url_resource.py @@ -77,8 +77,8 @@ class PreviewUrlResource(DirectServeResource): treq_args={"browser_like_redirects": True}, ip_whitelist=hs.config.url_preview_ip_range_whitelist, ip_blacklist=hs.config.url_preview_ip_range_blacklist, - http_proxy=os.getenv("http_proxy"), - https_proxy=os.getenv("HTTPS_PROXY"), + http_proxy=os.getenvb(b"http_proxy"), + https_proxy=os.getenvb(b"HTTPS_PROXY"), ) self.media_repo = media_repo self.primary_base_path = media_repo.primary_base_path @@ -122,7 +122,7 @@ class PreviewUrlResource(DirectServeResource): pattern = entry[attrib] value = getattr(url_tuple, attrib) logger.debug( - "Matching attrib '%s' with value '%s' against" " pattern '%s'", + "Matching attrib '%s' with value '%s' against pattern '%s'", attrib, value, pattern, @@ -402,7 +402,7 @@ class PreviewUrlResource(DirectServeResource): logger.info("Running url preview cache expiry") - if not (yield self.store.has_completed_background_updates()): + if not (yield self.store.db.updates.has_completed_background_updates()): logger.info("Still running DB updates; skipping expiry") return diff --git a/synapse/rest/media/v1/thumbnailer.py b/synapse/rest/media/v1/thumbnailer.py index 8cf415e29d..c234ea7421 100644 --- a/synapse/rest/media/v1/thumbnailer.py +++ b/synapse/rest/media/v1/thumbnailer.py @@ -129,5 +129,8 @@ class Thumbnailer(object): def _encode_image(self, output_image, output_type): output_bytes_io = BytesIO() - output_image.save(output_bytes_io, self.FORMATS[output_type], quality=80) + fmt = self.FORMATS[output_type] + if fmt == "JPEG": + output_image = output_image.convert("RGB") + output_image.save(output_bytes_io, fmt, quality=80) return output_bytes_io diff --git a/synapse/server.py b/synapse/server.py index d0e6f850a4..4247675224 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -25,7 +25,6 @@ import abc import logging import os -from twisted.enterprise import adbapi from twisted.mail.smtp import sendmail from twisted.web.client import BrowserLikePolicyForHTTPS @@ -34,6 +33,7 @@ from synapse.api.filtering import Filtering from synapse.api.ratelimiting import Ratelimiter from synapse.appservice.api import ApplicationServiceApi from synapse.appservice.scheduler import ApplicationServiceScheduler +from synapse.config.homeserver import HomeServerConfig from synapse.crypto import context_factory from synapse.crypto.keyring import Keyring from synapse.events.builder import EventBuilderFactory @@ -132,7 +132,6 @@ class HomeServer(object): DEPENDENCIES = [ "http_client", - "db_pool", "federation_client", "federation_server", "handlers", @@ -209,16 +208,18 @@ class HomeServer(object): # instantiated during setup() for future return by get_datastore() DATASTORE_CLASS = abc.abstractproperty() - def __init__(self, hostname, reactor=None, **kwargs): + def __init__(self, hostname: str, config: HomeServerConfig, reactor=None, **kwargs): """ Args: hostname : The hostname for the server. + config: The full config for the homeserver. """ if not reactor: from twisted.internet import reactor self._reactor = reactor self.hostname = hostname + self.config = config self._building = {} self._listening_services = [] self.start_time = None @@ -237,12 +238,12 @@ class HomeServer(object): def setup(self): logger.info("Setting up.") - with self.get_db_conn() as conn: - datastore = self.DATASTORE_CLASS(conn, self) - self.datastores = DataStores(datastore, conn, self) - conn.commit() self.start_time = int(self.get_clock().time()) +<<<<<<< HEAD self.get_clock().looping_call(self.config.caches.resize_caches, 10000) +======= + self.datastores = DataStores(self.DATASTORE_CLASS, self) +>>>>>>> origin/develop logger.info("Finished setting up.") def setup_master(self): @@ -276,6 +277,9 @@ class HomeServer(object): def get_datastore(self): return self.datastores.main + def get_datastores(self): + return self.datastores + def get_config(self): return self.config @@ -319,8 +323,8 @@ class HomeServer(object): def build_proxied_http_client(self): return SimpleHttpClient( self, - http_proxy=os.getenv("http_proxy"), - https_proxy=os.getenv("HTTPS_PROXY"), + http_proxy=os.getenvb(b"http_proxy"), + https_proxy=os.getenvb(b"HTTPS_PROXY"), ) def build_room_creation_handler(self): @@ -425,31 +429,6 @@ class HomeServer(object): ) return MatrixFederationHttpClient(self, tls_client_options_factory) - def build_db_pool(self): - name = self.db_config["name"] - - return adbapi.ConnectionPool( - name, cp_reactor=self.get_reactor(), **self.db_config.get("args", {}) - ) - - def get_db_conn(self, run_new_connection=True): - """Makes a new connection to the database, skipping the db pool - - Returns: - Connection: a connection object implementing the PEP-249 spec - """ - # Any param beginning with cp_ is a parameter for adbapi, and should - # not be passed to the database engine. - db_params = { - k: v - for k, v in self.db_config.get("args", {}).items() - if not k.startswith("cp_") - } - db_conn = self.database_engine.module.connect(**db_params) - if run_new_connection: - self.database_engine.on_new_connection(db_conn) - return db_conn - def build_media_repository_resource(self): # build the media repo resource. This indirects through the HomeServer # to ensure that we only have a single instance of diff --git a/synapse/server_notices/consent_server_notices.py b/synapse/server_notices/consent_server_notices.py index 415e9c17d8..5736c56032 100644 --- a/synapse/server_notices/consent_server_notices.py +++ b/synapse/server_notices/consent_server_notices.py @@ -54,7 +54,7 @@ class ConsentServerNotices(object): ) if "body" not in self._server_notice_content: raise ConfigError( - "user_consent server_notice_consent must contain a 'body' " "key." + "user_consent server_notice_consent must contain a 'body' key." ) self._consent_uri_builder = ConsentURIBuilder(hs.config) diff --git a/synapse/server_notices/server_notices_manager.py b/synapse/server_notices/server_notices_manager.py index 2dac90578c..f7432c8d2f 100644 --- a/synapse/server_notices/server_notices_manager.py +++ b/synapse/server_notices/server_notices_manager.py @@ -105,7 +105,7 @@ class ServerNoticesManager(object): assert self._is_mine_id(user_id), "Cannot send server notices to remote users" - rooms = yield self._store.get_rooms_for_user_where_membership_is( + rooms = yield self._store.get_rooms_for_local_user_where_membership_is( user_id, [Membership.INVITE, Membership.JOIN] ) system_mxid = self._config.server_notices_mxid diff --git a/synapse/state/__init__.py b/synapse/state/__init__.py index e5630c7e1b..48324b2437 100644 --- a/synapse/state/__init__.py +++ b/synapse/state/__init__.py @@ -16,7 +16,7 @@ import logging from collections import namedtuple -from typing import Iterable, Optional +from typing import Dict, Iterable, List, Optional from six import iteritems, itervalues @@ -32,6 +32,8 @@ from synapse.events import EventBase from synapse.events.snapshot import EventContext from synapse.logging.utils import log_function from synapse.state import v1, v2 +from synapse.storage.data_stores.main.events_worker import EventRedactBehaviour +from synapse.types import StateMap from synapse.util.async_helpers import Linearizer from synapse.util.caches.expiringcache import ExpiringCache from synapse.util.metrics import Measure, measure_func @@ -414,6 +416,7 @@ class StateHandler(object): with Measure(self.clock, "state._resolve_events"): new_state = yield resolve_events_with_store( + event.room_id, room_version, state_set_ids, event_map=state_map, @@ -460,7 +463,7 @@ class StateResolutionHandler(object): not be called for a single state group Args: - room_id (str): room we are resolving for (used for logging) + room_id (str): room we are resolving for (used for logging and sanity checks) room_version (str): version of the room state_groups_ids (dict[int, dict[(str, str), str]]): map from state group id to the state in that state group @@ -516,6 +519,7 @@ class StateResolutionHandler(object): logger.info("Resolving conflicted state for %r", room_id) with Measure(self.clock, "state._resolve_events"): new_state = yield resolve_events_with_store( + room_id, room_version, list(itervalues(state_groups_ids)), event_map=event_map, @@ -587,36 +591,44 @@ def _make_state_cache_entry(new_state, state_groups_ids): ) -def resolve_events_with_store(room_version, state_sets, event_map, state_res_store): +def resolve_events_with_store( + room_id: str, + room_version: str, + state_sets: List[StateMap[str]], + event_map: Optional[Dict[str, EventBase]], + state_res_store: "StateResolutionStore", +): """ Args: - room_version(str): Version of the room + room_id: the room we are working in + + room_version: Version of the room - state_sets(list): List of dicts of (type, state_key) -> event_id, + state_sets: List of dicts of (type, state_key) -> event_id, which are the different state groups to resolve. - event_map(dict[str,FrozenEvent]|None): + event_map: a dict from event_id to event, for any events that we happen to have in flight (eg, those currently being persisted). This will be used as a starting point fof finding the state we need; any missing events will be requested via state_map_factory. - If None, all events will be fetched via state_map_factory. + If None, all events will be fetched via state_res_store. - state_res_store (StateResolutionStore) + state_res_store: a place to fetch events from - Returns + Returns: Deferred[dict[(str, str), str]]: a map from (type, state_key) to event_id. """ v = KNOWN_ROOM_VERSIONS[room_version] if v.state_res == StateResolutionVersions.V1: return v1.resolve_events_with_store( - state_sets, event_map, state_res_store.get_events + room_id, state_sets, event_map, state_res_store.get_events ) else: return v2.resolve_events_with_store( - room_version, state_sets, event_map, state_res_store + room_id, room_version, state_sets, event_map, state_res_store ) @@ -644,7 +656,7 @@ class StateResolutionStore(object): return self.store.get_events( event_ids, - check_redacted=False, + redact_behaviour=EventRedactBehaviour.AS_IS, get_prev_content=False, allow_rejected=allow_rejected, ) diff --git a/synapse/state/v1.py b/synapse/state/v1.py index a2f92d9ff9..d6c34ce3b7 100644 --- a/synapse/state/v1.py +++ b/synapse/state/v1.py @@ -15,6 +15,7 @@ import hashlib import logging +from typing import Callable, Dict, List, Optional from six import iteritems, iterkeys, itervalues @@ -24,6 +25,8 @@ from synapse import event_auth from synapse.api.constants import EventTypes from synapse.api.errors import AuthError from synapse.api.room_versions import RoomVersions +from synapse.events import EventBase +from synapse.types import StateMap logger = logging.getLogger(__name__) @@ -32,13 +35,20 @@ POWER_KEY = (EventTypes.PowerLevels, "") @defer.inlineCallbacks -def resolve_events_with_store(state_sets, event_map, state_map_factory): +def resolve_events_with_store( + room_id: str, + state_sets: List[StateMap[str]], + event_map: Optional[Dict[str, EventBase]], + state_map_factory: Callable, +): """ Args: - state_sets(list): List of dicts of (type, state_key) -> event_id, + room_id: the room we are working in + + state_sets: List of dicts of (type, state_key) -> event_id, which are the different state groups to resolve. - event_map(dict[str,FrozenEvent]|None): + event_map: a dict from event_id to event, for any events that we happen to have in flight (eg, those currently being persisted). This will be used as a starting point fof finding the state we need; any missing @@ -46,11 +56,11 @@ def resolve_events_with_store(state_sets, event_map, state_map_factory): If None, all events will be fetched via state_map_factory. - state_map_factory(func): will be called + state_map_factory: will be called with a list of event_ids that are needed, and should return with a Deferred of dict of event_id to event. - Returns + Returns: Deferred[dict[(str, str), str]]: a map from (type, state_key) to event_id. """ @@ -76,6 +86,14 @@ def resolve_events_with_store(state_sets, event_map, state_map_factory): if event_map is not None: state_map.update(event_map) + # everything in the state map should be in the right room + for event in state_map.values(): + if event.room_id != room_id: + raise Exception( + "Attempting to state-resolve for room %s with event %s which is in %s" + % (room_id, event.event_id, event.room_id,) + ) + # get the ids of the auth events which allow us to authenticate the # conflicted state, picking only from the unconflicting state. # @@ -95,6 +113,13 @@ def resolve_events_with_store(state_sets, event_map, state_map_factory): ) state_map_new = yield state_map_factory(new_needed_events) + for event in state_map_new.values(): + if event.room_id != room_id: + raise Exception( + "Attempting to state-resolve for room %s with event %s which is in %s" + % (room_id, event.event_id, event.room_id,) + ) + state_map.update(state_map_new) return _resolve_with_state( diff --git a/synapse/state/v2.py b/synapse/state/v2.py index b327c86f40..6216fdd204 100644 --- a/synapse/state/v2.py +++ b/synapse/state/v2.py @@ -16,29 +16,41 @@ import heapq import itertools import logging +from typing import Dict, List, Optional from six import iteritems, itervalues from twisted.internet import defer +import synapse.state from synapse import event_auth from synapse.api.constants import EventTypes from synapse.api.errors import AuthError +from synapse.events import EventBase +from synapse.types import StateMap logger = logging.getLogger(__name__) @defer.inlineCallbacks -def resolve_events_with_store(room_version, state_sets, event_map, state_res_store): +def resolve_events_with_store( + room_id: str, + room_version: str, + state_sets: List[StateMap[str]], + event_map: Optional[Dict[str, EventBase]], + state_res_store: "synapse.state.StateResolutionStore", +): """Resolves the state using the v2 state resolution algorithm Args: - room_version (str): The room version + room_id: the room we are working in + + room_version: The room version - state_sets(list): List of dicts of (type, state_key) -> event_id, + state_sets: List of dicts of (type, state_key) -> event_id, which are the different state groups to resolve. - event_map(dict[str,FrozenEvent]|None): + event_map: a dict from event_id to event, for any events that we happen to have in flight (eg, those currently being persisted). This will be used as a starting point fof finding the state we need; any missing @@ -46,9 +58,9 @@ def resolve_events_with_store(room_version, state_sets, event_map, state_res_sto If None, all events will be fetched via state_res_store. - state_res_store (StateResolutionStore) + state_res_store: - Returns + Returns: Deferred[dict[(str, str), str]]: a map from (type, state_key) to event_id. """ @@ -84,6 +96,14 @@ def resolve_events_with_store(room_version, state_sets, event_map, state_res_sto ) event_map.update(events) + # everything in the event map should be in the right room + for event in event_map.values(): + if event.room_id != room_id: + raise Exception( + "Attempting to state-resolve for room %s with event %s which is in %s" + % (room_id, event.event_id, event.room_id,) + ) + full_conflicted_set = set(eid for eid in full_conflicted_set if eid in event_map) logger.debug("%d full_conflicted_set entries", len(full_conflicted_set)) @@ -94,13 +114,14 @@ def resolve_events_with_store(room_version, state_sets, event_map, state_res_sto ) sorted_power_events = yield _reverse_topological_power_sort( - power_events, event_map, state_res_store, full_conflicted_set + room_id, power_events, event_map, state_res_store, full_conflicted_set ) logger.debug("sorted %d power events", len(sorted_power_events)) # Now sequentially auth each one resolved_state = yield _iterative_auth_checks( + room_id, room_version, sorted_power_events, unconflicted_state, @@ -121,13 +142,18 @@ def resolve_events_with_store(room_version, state_sets, event_map, state_res_sto pl = resolved_state.get((EventTypes.PowerLevels, ""), None) leftover_events = yield _mainline_sort( - leftover_events, pl, event_map, state_res_store + room_id, leftover_events, pl, event_map, state_res_store ) logger.debug("resolving remaining events") resolved_state = yield _iterative_auth_checks( - room_version, leftover_events, resolved_state, event_map, state_res_store + room_id, + room_version, + leftover_events, + resolved_state, + event_map, + state_res_store, ) logger.debug("resolved") @@ -141,11 +167,12 @@ def resolve_events_with_store(room_version, state_sets, event_map, state_res_sto @defer.inlineCallbacks -def _get_power_level_for_sender(event_id, event_map, state_res_store): +def _get_power_level_for_sender(room_id, event_id, event_map, state_res_store): """Return the power level of the sender of the given event according to their auth events. Args: + room_id (str) event_id (str) event_map (dict[str,FrozenEvent]) state_res_store (StateResolutionStore) @@ -153,20 +180,24 @@ def _get_power_level_for_sender(event_id, event_map, state_res_store): Returns: Deferred[int] """ - event = yield _get_event(event_id, event_map, state_res_store) + event = yield _get_event(room_id, event_id, event_map, state_res_store) pl = None for aid in event.auth_event_ids(): - aev = yield _get_event(aid, event_map, state_res_store) - if (aev.type, aev.state_key) == (EventTypes.PowerLevels, ""): + aev = yield _get_event( + room_id, aid, event_map, state_res_store, allow_none=True + ) + if aev and (aev.type, aev.state_key) == (EventTypes.PowerLevels, ""): pl = aev break if pl is None: # Couldn't find power level. Check if they're the creator of the room for aid in event.auth_event_ids(): - aev = yield _get_event(aid, event_map, state_res_store) - if (aev.type, aev.state_key) == (EventTypes.Create, ""): + aev = yield _get_event( + room_id, aid, event_map, state_res_store, allow_none=True + ) + if aev and (aev.type, aev.state_key) == (EventTypes.Create, ""): if aev.content.get("creator") == event.sender: return 100 break @@ -279,7 +310,7 @@ def _is_power_event(event): @defer.inlineCallbacks def _add_event_and_auth_chain_to_graph( - graph, event_id, event_map, state_res_store, auth_diff + graph, room_id, event_id, event_map, state_res_store, auth_diff ): """Helper function for _reverse_topological_power_sort that add the event and its auth chain (that is in the auth diff) to the graph @@ -287,6 +318,7 @@ def _add_event_and_auth_chain_to_graph( Args: graph (dict[str, set[str]]): A map from event ID to the events auth event IDs + room_id (str): the room we are working in event_id (str): Event to add to the graph event_map (dict[str,FrozenEvent]) state_res_store (StateResolutionStore) @@ -298,7 +330,7 @@ def _add_event_and_auth_chain_to_graph( eid = state.pop() graph.setdefault(eid, set()) - event = yield _get_event(eid, event_map, state_res_store) + event = yield _get_event(room_id, eid, event_map, state_res_store) for aid in event.auth_event_ids(): if aid in auth_diff: if aid not in graph: @@ -308,11 +340,14 @@ def _add_event_and_auth_chain_to_graph( @defer.inlineCallbacks -def _reverse_topological_power_sort(event_ids, event_map, state_res_store, auth_diff): +def _reverse_topological_power_sort( + room_id, event_ids, event_map, state_res_store, auth_diff +): """Returns a list of the event_ids sorted by reverse topological ordering, and then by power level and origin_server_ts Args: + room_id (str): the room we are working in event_ids (list[str]): The events to sort event_map (dict[str,FrozenEvent]) state_res_store (StateResolutionStore) @@ -325,12 +360,14 @@ def _reverse_topological_power_sort(event_ids, event_map, state_res_store, auth_ graph = {} for event_id in event_ids: yield _add_event_and_auth_chain_to_graph( - graph, event_id, event_map, state_res_store, auth_diff + graph, room_id, event_id, event_map, state_res_store, auth_diff ) event_to_pl = {} for event_id in graph: - pl = yield _get_power_level_for_sender(event_id, event_map, state_res_store) + pl = yield _get_power_level_for_sender( + room_id, event_id, event_map, state_res_store + ) event_to_pl[event_id] = pl def _get_power_order(event_id): @@ -348,20 +385,21 @@ def _reverse_topological_power_sort(event_ids, event_map, state_res_store, auth_ @defer.inlineCallbacks def _iterative_auth_checks( - room_version, event_ids, base_state, event_map, state_res_store + room_id, room_version, event_ids, base_state, event_map, state_res_store ): """Sequentially apply auth checks to each event in given list, updating the state as it goes along. Args: + room_id (str) room_version (str) event_ids (list[str]): Ordered list of events to apply auth checks to - base_state (dict[tuple[str, str], str]): The set of state to start with + base_state (StateMap[str]): The set of state to start with event_map (dict[str,FrozenEvent]) state_res_store (StateResolutionStore) Returns: - Deferred[dict[tuple[str, str], str]]: Returns the final updated state + Deferred[StateMap[str]]: Returns the final updated state """ resolved_state = base_state.copy() @@ -370,15 +408,22 @@ def _iterative_auth_checks( auth_events = {} for aid in event.auth_event_ids(): - ev = yield _get_event(aid, event_map, state_res_store) + ev = yield _get_event( + room_id, aid, event_map, state_res_store, allow_none=True + ) - if ev.rejected_reason is None: - auth_events[(ev.type, ev.state_key)] = ev + if not ev: + logger.warning( + "auth_event id %s for event %s is missing", aid, event_id + ) + else: + if ev.rejected_reason is None: + auth_events[(ev.type, ev.state_key)] = ev for key in event_auth.auth_types_for_event(event): if key in resolved_state: ev_id = resolved_state[key] - ev = yield _get_event(ev_id, event_map, state_res_store) + ev = yield _get_event(room_id, ev_id, event_map, state_res_store) if ev.rejected_reason is None: auth_events[key] = event_map[ev_id] @@ -400,11 +445,14 @@ def _iterative_auth_checks( @defer.inlineCallbacks -def _mainline_sort(event_ids, resolved_power_event_id, event_map, state_res_store): +def _mainline_sort( + room_id, event_ids, resolved_power_event_id, event_map, state_res_store +): """Returns a sorted list of event_ids sorted by mainline ordering based on the given event resolved_power_event_id Args: + room_id (str): room we're working in event_ids (list[str]): Events to sort resolved_power_event_id (str): The final resolved power level event ID event_map (dict[str,FrozenEvent]) @@ -417,12 +465,14 @@ def _mainline_sort(event_ids, resolved_power_event_id, event_map, state_res_stor pl = resolved_power_event_id while pl: mainline.append(pl) - pl_ev = yield _get_event(pl, event_map, state_res_store) + pl_ev = yield _get_event(room_id, pl, event_map, state_res_store) auth_events = pl_ev.auth_event_ids() pl = None for aid in auth_events: - ev = yield _get_event(aid, event_map, state_res_store) - if (ev.type, ev.state_key) == (EventTypes.PowerLevels, ""): + ev = yield _get_event( + room_id, aid, event_map, state_res_store, allow_none=True + ) + if ev and (ev.type, ev.state_key) == (EventTypes.PowerLevels, ""): pl = aid break @@ -457,6 +507,8 @@ def _get_mainline_depth_for_event(event, mainline_map, event_map, state_res_stor Deferred[int] """ + room_id = event.room_id + # We do an iterative search, replacing `event with the power level in its # auth events (if any) while event: @@ -468,8 +520,10 @@ def _get_mainline_depth_for_event(event, mainline_map, event_map, state_res_stor event = None for aid in auth_events: - aev = yield _get_event(aid, event_map, state_res_store) - if (aev.type, aev.state_key) == (EventTypes.PowerLevels, ""): + aev = yield _get_event( + room_id, aid, event_map, state_res_store, allow_none=True + ) + if aev and (aev.type, aev.state_key) == (EventTypes.PowerLevels, ""): event = aev break @@ -478,22 +532,37 @@ def _get_mainline_depth_for_event(event, mainline_map, event_map, state_res_stor @defer.inlineCallbacks -def _get_event(event_id, event_map, state_res_store): +def _get_event(room_id, event_id, event_map, state_res_store, allow_none=False): """Helper function to look up event in event_map, falling back to looking it up in the store Args: + room_id (str) event_id (str) event_map (dict[str,FrozenEvent]) state_res_store (StateResolutionStore) + allow_none (bool): if the event is not found, return None rather than raising + an exception Returns: - Deferred[FrozenEvent] + Deferred[Optional[FrozenEvent]] """ if event_id not in event_map: events = yield state_res_store.get_events([event_id], allow_rejected=True) event_map.update(events) - return event_map[event_id] + event = event_map.get(event_id) + + if event is None: + if allow_none: + return None + raise Exception("Unknown event %s" % (event_id,)) + + if event.room_id != room_id: + raise Exception( + "In state res for room %s, event %s is in %s" + % (room_id, event_id, event.room_id) + ) + return event def lexicographical_topological_sort(graph, key): diff --git a/synapse/storage/__init__.py b/synapse/storage/__init__.py index 0460fe8cc9..ec89f645d4 100644 --- a/synapse/storage/__init__.py +++ b/synapse/storage/__init__.py @@ -17,10 +17,10 @@ """ The storage layer is split up into multiple parts to allow Synapse to run against different configurations of databases (e.g. single or multiple -databases). The `data_stores` are classes that talk directly to a single -database and have associated schemas, background updates, etc. On top of those -there are (or will be) classes that provide high level interfaces that combine -calls to multiple `data_stores`. +databases). The `Database` class represents a single physical database. The +`data_stores` are classes that talk directly to a `Database` instance and have +associated schemas, background updates, etc. On top of those there are classes +that provide high level interfaces that combine calls to multiple `data_stores`. There are also schemas that get applied to every database, regardless of the data stores associated with them (e.g. the schema version tables), which are @@ -49,15 +49,3 @@ class Storage(object): self.persistence = EventsPersistenceStorage(hs, stores) self.purge_events = PurgeEventsStorage(hs, stores) self.state = StateGroupStorage(hs, stores) - - -def are_all_users_on_domain(txn, database_engine, domain): - sql = database_engine.convert_param_style( - "SELECT COUNT(*) FROM users WHERE name NOT LIKE ?" - ) - pat = "%:" + domain - txn.execute(sql, (pat,)) - num_not_matching = txn.fetchall()[0][0] - if num_not_matching == 0: - return True - return False diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py index ab596fa68d..3bb9381663 100644 --- a/synapse/storage/_base.py +++ b/synapse/storage/_base.py @@ -14,1400 +14,38 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -import itertools import logging import random -import sys -import threading -import time -from typing import Iterable, Tuple +from abc import ABCMeta -from six import PY2, iteritems, iterkeys, itervalues -from six.moves import builtins, intern, range +from six import PY2 +from six.moves import builtins from canonicaljson import json -from prometheus_client import Histogram -from twisted.internet import defer - -from synapse.api.errors import StoreError -from synapse.logging.context import LoggingContext, make_deferred_yieldable -from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.storage.engines import PostgresEngine, Sqlite3Engine +from synapse.storage.database import LoggingTransaction # noqa: F401 +from synapse.storage.database import make_in_list_sql_clause # noqa: F401 +from synapse.storage.database import Database from synapse.types import get_domain_from_id -from synapse.util import batch_iter -from synapse.util.caches.descriptors import Cache -from synapse.util.stringutils import exception_to_unicode - -# import a function which will return a monotonic time, in seconds -try: - # on python 3, use time.monotonic, since time.clock can go backwards - from time import monotonic as monotonic_time -except ImportError: - # ... but python 2 doesn't have it - from time import clock as monotonic_time logger = logging.getLogger(__name__) -try: - MAX_TXN_ID = sys.maxint - 1 -except AttributeError: - # python 3 does not have a maximum int value - MAX_TXN_ID = 2 ** 63 - 1 - -sql_logger = logging.getLogger("synapse.storage.SQL") -transaction_logger = logging.getLogger("synapse.storage.txn") -perf_logger = logging.getLogger("synapse.storage.TIME") - -sql_scheduling_timer = Histogram("synapse_storage_schedule_time", "sec") - -sql_query_timer = Histogram("synapse_storage_query_time", "sec", ["verb"]) -sql_txn_timer = Histogram("synapse_storage_transaction_time", "sec", ["desc"]) - - -# Unique indexes which have been added in background updates. Maps from table name -# to the name of the background update which added the unique index to that table. -# -# This is used by the upsert logic to figure out which tables are safe to do a proper -# UPSERT on: until the relevant background update has completed, we -# have to emulate an upsert by locking the table. -# -UNIQUE_INDEX_BACKGROUND_UPDATES = { - "user_ips": "user_ips_device_unique_index", - "device_lists_remote_extremeties": "device_lists_remote_extremeties_unique_idx", - "device_lists_remote_cache": "device_lists_remote_cache_unique_idx", - "event_search": "event_search_event_id_idx", -} - -# This is a special cache name we use to batch multiple invalidations of caches -# based on the current state when notifying workers over replication. -_CURRENT_STATE_CACHE_NAME = "cs_cache_fake" - -class LoggingTransaction(object): - """An object that almost-transparently proxies for the 'txn' object - passed to the constructor. Adds logging and metrics to the .execute() - method. +# some of our subclasses have abstract methods, so we use the ABCMeta metaclass. +class SQLBaseStore(metaclass=ABCMeta): + """Base class for data stores that holds helper functions. - Args: - txn: The database transcation object to wrap. - name (str): The name of this transactions for logging. - database_engine (Sqlite3Engine|PostgresEngine) - after_callbacks(list|None): A list that callbacks will be appended to - that have been added by `call_after` which should be run on - successful completion of the transaction. None indicates that no - callbacks should be allowed to be scheduled to run. - exception_callbacks(list|None): A list that callbacks will be appended - to that have been added by `call_on_exception` which should be run - if transaction ends with an error. None indicates that no callbacks - should be allowed to be scheduled to run. + Note that multiple instances of this class will exist as there will be one + per data store (and not one per physical database). """ - __slots__ = [ - "txn", - "name", - "database_engine", - "after_callbacks", - "exception_callbacks", - ] - - def __init__( - self, txn, name, database_engine, after_callbacks=None, exception_callbacks=None - ): - object.__setattr__(self, "txn", txn) - object.__setattr__(self, "name", name) - object.__setattr__(self, "database_engine", database_engine) - object.__setattr__(self, "after_callbacks", after_callbacks) - object.__setattr__(self, "exception_callbacks", exception_callbacks) - - def call_after(self, callback, *args, **kwargs): - """Call the given callback on the main twisted thread after the - transaction has finished. Used to invalidate the caches on the - correct thread. - """ - self.after_callbacks.append((callback, args, kwargs)) - - def call_on_exception(self, callback, *args, **kwargs): - self.exception_callbacks.append((callback, args, kwargs)) - - def __getattr__(self, name): - return getattr(self.txn, name) - - def __setattr__(self, name, value): - setattr(self.txn, name, value) - - def __iter__(self): - return self.txn.__iter__() - - def execute_batch(self, sql, args): - if isinstance(self.database_engine, PostgresEngine): - from psycopg2.extras import execute_batch - - self._do_execute(lambda *x: execute_batch(self.txn, *x), sql, args) - else: - for val in args: - self.execute(sql, val) - - def execute(self, sql, *args): - self._do_execute(self.txn.execute, sql, *args) - - def executemany(self, sql, *args): - self._do_execute(self.txn.executemany, sql, *args) - - def _make_sql_one_line(self, sql): - "Strip newlines out of SQL so that the loggers in the DB are on one line" - return " ".join(l.strip() for l in sql.splitlines() if l.strip()) - - def _do_execute(self, func, sql, *args): - sql = self._make_sql_one_line(sql) - - # TODO(paul): Maybe use 'info' and 'debug' for values? - sql_logger.debug("[SQL] {%s} %s", self.name, sql) - - sql = self.database_engine.convert_param_style(sql) - if args: - try: - sql_logger.debug("[SQL values] {%s} %r", self.name, args[0]) - except Exception: - # Don't let logging failures stop SQL from working - pass - - start = time.time() - - try: - return func(sql, *args) - except Exception as e: - logger.debug("[SQL FAIL] {%s} %s", self.name, e) - raise - finally: - secs = time.time() - start - sql_logger.debug("[SQL time] {%s} %f sec", self.name, secs) - sql_query_timer.labels(sql.split()[0]).observe(secs) - - -class PerformanceCounters(object): - def __init__(self): - self.current_counters = {} - self.previous_counters = {} - - def update(self, key, duration_secs): - count, cum_time = self.current_counters.get(key, (0, 0)) - count += 1 - cum_time += duration_secs - self.current_counters[key] = (count, cum_time) - - def interval(self, interval_duration_secs, limit=3): - counters = [] - for name, (count, cum_time) in iteritems(self.current_counters): - prev_count, prev_time = self.previous_counters.get(name, (0, 0)) - counters.append( - ( - (cum_time - prev_time) / interval_duration_secs, - count - prev_count, - name, - ) - ) - - self.previous_counters = dict(self.current_counters) - - counters.sort(reverse=True) - - top_n_counters = ", ".join( - "%s(%d): %.3f%%" % (name, count, 100 * ratio) - for ratio, count, name in counters[:limit] - ) - - return top_n_counters - - -class SQLBaseStore(object): - _TXN_ID = 0 - - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): self.hs = hs self._clock = hs.get_clock() - self._db_pool = hs.get_db_pool() - - self._previous_txn_total_time = 0 - self._current_txn_total_time = 0 - self._previous_loop_ts = 0 - - # TODO(paul): These can eventually be removed once the metrics code - # is running in mainline, and we have some nice monitoring frontends - # to watch it - self._txn_perf_counters = PerformanceCounters() - - self._get_event_cache = Cache( - "*getEvent*", keylen=3, max_entries=hs.config.event_cache_size - ) - - self._event_fetch_lock = threading.Condition() - self._event_fetch_list = [] - self._event_fetch_ongoing = 0 - - self._pending_ds = [] - - self.database_engine = hs.database_engine - - # A set of tables that are not safe to use native upserts in. - self._unsafe_to_upsert_tables = set(UNIQUE_INDEX_BACKGROUND_UPDATES.keys()) - - self._account_validity = self.hs.config.account_validity - - # We add the user_directory_search table to the blacklist on SQLite - # because the existing search table does not have an index, making it - # unsafe to use native upserts. - if isinstance(self.database_engine, Sqlite3Engine): - self._unsafe_to_upsert_tables.add("user_directory_search") - - if self.database_engine.can_native_upsert: - # Check ASAP (and then later, every 1s) to see if we have finished - # background updates of tables that aren't safe to update. - self._clock.call_later( - 0.0, - run_as_background_process, - "upsert_safety_check", - self._check_safe_to_upsert, - ) - + self.database_engine = database.engine + self.db = database self.rand = random.SystemRandom() - if self._account_validity.enabled: - self._clock.call_later( - 0.0, - run_as_background_process, - "account_validity_set_expiration_dates", - self._set_expiration_date_when_missing, - ) - - @defer.inlineCallbacks - def _check_safe_to_upsert(self): - """ - Is it safe to use native UPSERT? - - If there are background updates, we will need to wait, as they may be - the addition of indexes that set the UNIQUE constraint that we require. - - If the background updates have not completed, wait 15 sec and check again. - """ - updates = yield self._simple_select_list( - "background_updates", - keyvalues=None, - retcols=["update_name"], - desc="check_background_updates", - ) - updates = [x["update_name"] for x in updates] - - for table, update_name in UNIQUE_INDEX_BACKGROUND_UPDATES.items(): - if update_name not in updates: - logger.debug("Now safe to upsert in %s", table) - self._unsafe_to_upsert_tables.discard(table) - - # If there's any updates still running, reschedule to run. - if updates: - self._clock.call_later( - 15.0, - run_as_background_process, - "upsert_safety_check", - self._check_safe_to_upsert, - ) - - @defer.inlineCallbacks - def _set_expiration_date_when_missing(self): - """ - Retrieves the list of registered users that don't have an expiration date, and - adds an expiration date for each of them. - """ - - def select_users_with_no_expiration_date_txn(txn): - """Retrieves the list of registered users with no expiration date from the - database, filtering out deactivated users. - """ - sql = ( - "SELECT users.name FROM users" - " LEFT JOIN account_validity ON (users.name = account_validity.user_id)" - " WHERE account_validity.user_id is NULL AND users.deactivated = 0;" - ) - txn.execute(sql, []) - - res = self.cursor_to_dict(txn) - if res: - for user in res: - self.set_expiration_date_for_user_txn( - txn, user["name"], use_delta=True - ) - - yield self.runInteraction( - "get_users_with_no_expiration_date", - select_users_with_no_expiration_date_txn, - ) - - def set_expiration_date_for_user_txn(self, txn, user_id, use_delta=False): - """Sets an expiration date to the account with the given user ID. - - Args: - user_id (str): User ID to set an expiration date for. - use_delta (bool): If set to False, the expiration date for the user will be - now + validity period. If set to True, this expiration date will be a - random value in the [now + period - d ; now + period] range, d being a - delta equal to 10% of the validity period. - """ - now_ms = self._clock.time_msec() - expiration_ts = now_ms + self._account_validity.period - - if use_delta: - expiration_ts = self.rand.randrange( - expiration_ts - self._account_validity.startup_job_max_delta, - expiration_ts, - ) - - self._simple_upsert_txn( - txn, - "account_validity", - keyvalues={"user_id": user_id}, - values={"expiration_ts_ms": expiration_ts, "email_sent": False}, - ) - - def start_profiling(self): - self._previous_loop_ts = monotonic_time() - - def loop(): - curr = self._current_txn_total_time - prev = self._previous_txn_total_time - self._previous_txn_total_time = curr - - time_now = monotonic_time() - time_then = self._previous_loop_ts - self._previous_loop_ts = time_now - - duration = time_now - time_then - ratio = (curr - prev) / duration - - top_three_counters = self._txn_perf_counters.interval(duration, limit=3) - - perf_logger.info( - "Total database time: %.3f%% {%s}", ratio * 100, top_three_counters - ) - - self._clock.looping_call(loop, 10000) - - def _new_transaction( - self, conn, desc, after_callbacks, exception_callbacks, func, *args, **kwargs - ): - start = monotonic_time() - txn_id = self._TXN_ID - - # We don't really need these to be unique, so lets stop it from - # growing really large. - self._TXN_ID = (self._TXN_ID + 1) % (MAX_TXN_ID) - - name = "%s-%x" % (desc, txn_id) - - transaction_logger.debug("[TXN START] {%s}", name) - - try: - i = 0 - N = 5 - while True: - try: - txn = conn.cursor() - txn = LoggingTransaction( - txn, - name, - self.database_engine, - after_callbacks, - exception_callbacks, - ) - r = func(txn, *args, **kwargs) - conn.commit() - return r - except self.database_engine.module.OperationalError as e: - # This can happen if the database disappears mid - # transaction. - logger.warning( - "[TXN OPERROR] {%s} %s %d/%d", - name, - exception_to_unicode(e), - i, - N, - ) - if i < N: - i += 1 - try: - conn.rollback() - except self.database_engine.module.Error as e1: - logger.warning( - "[TXN EROLL] {%s} %s", name, exception_to_unicode(e1) - ) - continue - raise - except self.database_engine.module.DatabaseError as e: - if self.database_engine.is_deadlock(e): - logger.warning("[TXN DEADLOCK] {%s} %d/%d", name, i, N) - if i < N: - i += 1 - try: - conn.rollback() - except self.database_engine.module.Error as e1: - logger.warning( - "[TXN EROLL] {%s} %s", - name, - exception_to_unicode(e1), - ) - continue - raise - except Exception as e: - logger.debug("[TXN FAIL] {%s} %s", name, e) - raise - finally: - end = monotonic_time() - duration = end - start - - LoggingContext.current_context().add_database_transaction(duration) - - transaction_logger.debug("[TXN END] {%s} %f sec", name, duration) - - self._current_txn_total_time += duration - self._txn_perf_counters.update(desc, duration) - sql_txn_timer.labels(desc).observe(duration) - - @defer.inlineCallbacks - def runInteraction(self, desc, func, *args, **kwargs): - """Starts a transaction on the database and runs a given function - - Arguments: - desc (str): description of the transaction, for logging and metrics - func (func): callback function, which will be called with a - database transaction (twisted.enterprise.adbapi.Transaction) as - its first argument, followed by `args` and `kwargs`. - - args (list): positional args to pass to `func` - kwargs (dict): named args to pass to `func` - - Returns: - Deferred: The result of func - """ - after_callbacks = [] - exception_callbacks = [] - - if LoggingContext.current_context() == LoggingContext.sentinel: - logger.warning("Starting db txn '%s' from sentinel context", desc) - - try: - result = yield self.runWithConnection( - self._new_transaction, - desc, - after_callbacks, - exception_callbacks, - func, - *args, - **kwargs - ) - - for after_callback, after_args, after_kwargs in after_callbacks: - after_callback(*after_args, **after_kwargs) - except: # noqa: E722, as we reraise the exception this is fine. - for after_callback, after_args, after_kwargs in exception_callbacks: - after_callback(*after_args, **after_kwargs) - raise - - return result - - @defer.inlineCallbacks - def runWithConnection(self, func, *args, **kwargs): - """Wraps the .runWithConnection() method on the underlying db_pool. - - Arguments: - func (func): callback function, which will be called with a - database connection (twisted.enterprise.adbapi.Connection) as - its first argument, followed by `args` and `kwargs`. - args (list): positional args to pass to `func` - kwargs (dict): named args to pass to `func` - - Returns: - Deferred: The result of func - """ - parent_context = LoggingContext.current_context() - if parent_context == LoggingContext.sentinel: - logger.warning( - "Starting db connection from sentinel context: metrics will be lost" - ) - parent_context = None - - start_time = monotonic_time() - - def inner_func(conn, *args, **kwargs): - with LoggingContext("runWithConnection", parent_context) as context: - sched_duration_sec = monotonic_time() - start_time - sql_scheduling_timer.observe(sched_duration_sec) - context.add_database_scheduled(sched_duration_sec) - - if self.database_engine.is_connection_closed(conn): - logger.debug("Reconnecting closed database connection") - conn.reconnect() - - return func(conn, *args, **kwargs) - - result = yield make_deferred_yieldable( - self._db_pool.runWithConnection(inner_func, *args, **kwargs) - ) - - return result - - @staticmethod - def cursor_to_dict(cursor): - """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. - """ - col_headers = list(intern(str(column[0])) for column in cursor.description) - results = list(dict(zip(col_headers, row)) for row in cursor) - return results - - def _execute(self, desc, decoder, query, *args): - """Runs a single query for a result set. - - Args: - 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): - txn.execute(query, args) - if decoder: - return decoder(txn) - else: - return txn.fetchall() - - return self.runInteraction(desc, interaction) - - # "Simple" SQL API methods that operate on a single table with no JOINs, - # no complex WHERE clauses, just a dict of values for columns. - - @defer.inlineCallbacks - def _simple_insert(self, table, values, or_ignore=False, desc="_simple_insert"): - """Executes an INSERT query on the named table. - - Args: - table : string giving the table name - values : dict of new column names and values for them - or_ignore : bool stating whether an exception should be raised - when a conflicting row already exists. If True, False will be - returned by the function instead - desc : string giving a description of the transaction - - Returns: - bool: Whether the row was inserted or not. Only useful when - `or_ignore` is True - """ - try: - yield self.runInteraction(desc, self._simple_insert_txn, table, values) - except self.database_engine.module.IntegrityError: - # We have to do or_ignore flag at this layer, since we can't reuse - # a cursor after we receive an error from the db. - if not or_ignore: - raise - return False - return True - - @staticmethod - def _simple_insert_txn(txn, table, values): - keys, vals = zip(*values.items()) - - sql = "INSERT INTO %s (%s) VALUES(%s)" % ( - table, - ", ".join(k for k in keys), - ", ".join("?" for _ in keys), - ) - - txn.execute(sql, vals) - - def _simple_insert_many(self, table, values, desc): - return self.runInteraction(desc, self._simple_insert_many_txn, table, values) - - @staticmethod - def _simple_insert_many_txn(txn, table, values): - if not values: - return - - # This is a *slight* abomination to get a list of tuples of key names - # and a list of tuples of value names. - # - # i.e. [{"a": 1, "b": 2}, {"c": 3, "d": 4}] - # => [("a", "b",), ("c", "d",)] and [(1, 2,), (3, 4,)] - # - # The sort is to ensure that we don't rely on dictionary iteration - # order. - keys, vals = zip( - *[zip(*(sorted(i.items(), key=lambda kv: kv[0]))) for i in values if i] - ) - - for k in keys: - if k != keys[0]: - raise RuntimeError("All items must have the same keys") - - sql = "INSERT INTO %s (%s) VALUES(%s)" % ( - table, - ", ".join(k for k in keys[0]), - ", ".join("?" for _ in keys[0]), - ) - - txn.executemany(sql, vals) - - @defer.inlineCallbacks - def _simple_upsert( - self, - table, - keyvalues, - values, - insertion_values={}, - desc="_simple_upsert", - lock=True, - ): - """ - - `lock` should generally be set to True (the default), but can be set - to False if either of the following are true: - - * there is a UNIQUE INDEX on the key columns. In this case a conflict - will cause an IntegrityError in which case this function will retry - the update. - - * we somehow know that we are the only thread which will be updating - this table. - - Args: - table (str): The table to upsert into - keyvalues (dict): The unique key columns and their new values - values (dict): The nonunique columns and their new values - insertion_values (dict): additional key/values to use only when - inserting - lock (bool): True to lock the table when doing the upsert. - Returns: - Deferred(None or bool): Native upserts always return None. Emulated - upserts return True if a new entry was created, False if an existing - one was updated. - """ - attempts = 0 - while True: - try: - result = yield self.runInteraction( - desc, - self._simple_upsert_txn, - table, - keyvalues, - values, - insertion_values, - lock=lock, - ) - return result - except self.database_engine.module.IntegrityError as e: - attempts += 1 - if attempts >= 5: - # don't retry forever, because things other than races - # can cause IntegrityErrors - raise - - # presumably we raced with another transaction: let's retry. - logger.warning( - "IntegrityError when upserting into %s; retrying: %s", table, e - ) - - def _simple_upsert_txn( - self, txn, table, keyvalues, values, insertion_values={}, lock=True - ): - """ - Pick the UPSERT method which works best on the platform. Either the - native one (Pg9.5+, recent SQLites), or fall back to an emulated method. - - Args: - txn: The transaction to use. - table (str): The table to upsert into - keyvalues (dict): The unique key tables and their new values - values (dict): The nonunique columns and their new values - insertion_values (dict): additional key/values to use only when - inserting - lock (bool): True to lock the table when doing the upsert. - Returns: - None or bool: Native upserts always return None. Emulated - upserts return True if a new entry was created, False if an existing - one was updated. - """ - if ( - self.database_engine.can_native_upsert - and table not in self._unsafe_to_upsert_tables - ): - return self._simple_upsert_txn_native_upsert( - txn, table, keyvalues, values, insertion_values=insertion_values - ) - else: - return self._simple_upsert_txn_emulated( - txn, - table, - keyvalues, - values, - insertion_values=insertion_values, - lock=lock, - ) - - def _simple_upsert_txn_emulated( - self, txn, table, keyvalues, values, insertion_values={}, lock=True - ): - """ - Args: - table (str): The table to upsert into - keyvalues (dict): The unique key tables and their new values - values (dict): The nonunique columns and their new values - insertion_values (dict): additional key/values to use only when - inserting - lock (bool): True to lock the table when doing the upsert. - Returns: - bool: Return True if a new entry was created, False if an existing - one was updated. - """ - # We need to lock the table :(, unless we're *really* careful - if lock: - self.database_engine.lock_table(txn, table) - - def _getwhere(key): - # If the value we're passing in is None (aka NULL), we need to use - # IS, not =, as NULL = NULL equals NULL (False). - if keyvalues[key] is None: - return "%s IS ?" % (key,) - else: - return "%s = ?" % (key,) - - if not values: - # If `values` is empty, then all of the values we care about are in - # the unique key, so there is nothing to UPDATE. We can just do a - # SELECT instead to see if it exists. - sql = "SELECT 1 FROM %s WHERE %s" % ( - table, - " AND ".join(_getwhere(k) for k in keyvalues), - ) - sqlargs = list(keyvalues.values()) - txn.execute(sql, sqlargs) - if txn.fetchall(): - # We have an existing record. - return False - else: - # First try to update. - sql = "UPDATE %s SET %s WHERE %s" % ( - table, - ", ".join("%s = ?" % (k,) for k in values), - " AND ".join(_getwhere(k) for k in keyvalues), - ) - sqlargs = list(values.values()) + list(keyvalues.values()) - - txn.execute(sql, sqlargs) - if txn.rowcount > 0: - # successfully updated at least one row. - return False - - # We didn't find any existing rows, so insert a new one - allvalues = {} - allvalues.update(keyvalues) - allvalues.update(values) - allvalues.update(insertion_values) - - sql = "INSERT INTO %s (%s) VALUES (%s)" % ( - table, - ", ".join(k for k in allvalues), - ", ".join("?" for _ in allvalues), - ) - txn.execute(sql, list(allvalues.values())) - # successfully inserted - return True - - def _simple_upsert_txn_native_upsert( - self, txn, table, keyvalues, values, insertion_values={} - ): - """ - Use the native UPSERT functionality in recent PostgreSQL versions. - - Args: - table (str): The table to upsert into - keyvalues (dict): The unique key tables and their new values - values (dict): The nonunique columns and their new values - insertion_values (dict): additional key/values to use only when - inserting - Returns: - None - """ - allvalues = {} - allvalues.update(keyvalues) - allvalues.update(insertion_values) - - if not values: - latter = "NOTHING" - else: - allvalues.update(values) - latter = "UPDATE SET " + ", ".join(k + "=EXCLUDED." + k for k in values) - - sql = ("INSERT INTO %s (%s) VALUES (%s) " "ON CONFLICT (%s) DO %s") % ( - table, - ", ".join(k for k in allvalues), - ", ".join("?" for _ in allvalues), - ", ".join(k for k in keyvalues), - latter, - ) - txn.execute(sql, list(allvalues.values())) - - def _simple_upsert_many_txn( - self, txn, table, key_names, key_values, value_names, value_values - ): - """ - Upsert, many times. - - Args: - table (str): The table to upsert into - key_names (list[str]): The key column names. - key_values (list[list]): A list of each row's key column values. - value_names (list[str]): The value column names. If empty, no - values will be used, even if value_values is provided. - value_values (list[list]): A list of each row's value column values. - Returns: - None - """ - if ( - self.database_engine.can_native_upsert - and 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 - ) - else: - return self._simple_upsert_many_txn_emulated( - txn, table, key_names, key_values, value_names, value_values - ) - - def _simple_upsert_many_txn_emulated( - self, txn, table, key_names, key_values, value_names, value_values - ): - """ - Upsert, many times, but without native UPSERT support or batching. - - Args: - table (str): The table to upsert into - key_names (list[str]): The key column names. - key_values (list[list]): A list of each row's key column values. - value_names (list[str]): The value column names. If empty, no - values will be used, even if value_values is provided. - value_values (list[list]): A list of each row's value column values. - Returns: - None - """ - # 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))] - - for keyv, valv in zip(key_values, value_values): - _keys = {x: y for x, y in zip(key_names, keyv)} - _vals = {x: y for x, y in zip(value_names, valv)} - - self._simple_upsert_txn_emulated(txn, table, _keys, _vals) - - def _simple_upsert_many_txn_native_upsert( - self, txn, table, key_names, key_values, value_names, value_values - ): - """ - Upsert, many times, using batching where possible. - - Args: - table (str): The table to upsert into - key_names (list[str]): The key column names. - key_values (list[list]): A list of each row's key column values. - value_names (list[str]): The value column names. If empty, no - values will be used, even if value_values is provided. - value_values (list[list]): A list of each row's value column values. - Returns: - None - """ - allnames = [] - allnames.extend(key_names) - 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 - ) - - sql = "INSERT INTO %s (%s) VALUES (%s) ON CONFLICT (%s) DO %s" % ( - table, - ", ".join(k for k in allnames), - ", ".join("?" for _ in allnames), - ", ".join(key_names), - latter, - ) - - args = [] - - for x, y in zip(key_values, value_values): - args.append(tuple(x) + tuple(y)) - - return txn.execute_batch(sql, args) - - def _simple_select_one( - self, table, keyvalues, retcols, allow_none=False, desc="_simple_select_one" - ): - """Executes a SELECT query on the named table, which is expected to - return a single row, returning multiple columns from it. - - Args: - table : string giving the table name - keyvalues : dict of column names and values to select the row with - retcols : list of strings giving the names of the columns to return - - allow_none : If true, return None instead of failing if the SELECT - statement returns no rows - """ - return self.runInteraction( - desc, self._simple_select_one_txn, table, keyvalues, retcols, allow_none - ) - - def _simple_select_one_onecol( - self, - table, - keyvalues, - retcol, - allow_none=False, - desc="_simple_select_one_onecol", - ): - """Executes a SELECT query on the named table, which is expected to - return a single row, returning a single column from it. - - Args: - table : string giving the table name - keyvalues : dict of column names and values to select the row with - retcol : string giving the name of the column to return - """ - return self.runInteraction( - desc, - self._simple_select_one_onecol_txn, - table, - keyvalues, - retcol, - allow_none=allow_none, - ) - - @classmethod - def _simple_select_one_onecol_txn( - cls, txn, table, keyvalues, retcol, allow_none=False - ): - ret = cls._simple_select_onecol_txn( - txn, table=table, keyvalues=keyvalues, retcol=retcol - ) - - if ret: - return ret[0] - else: - if allow_none: - return None - else: - raise StoreError(404, "No row found") - - @staticmethod - def _simple_select_onecol_txn(txn, table, keyvalues, retcol): - sql = ("SELECT %(retcol)s FROM %(table)s") % {"retcol": retcol, "table": table} - - if keyvalues: - sql += " WHERE %s" % " AND ".join("%s = ?" % k for k in iterkeys(keyvalues)) - txn.execute(sql, list(keyvalues.values())) - else: - txn.execute(sql) - - return [r[0] for r in txn] - - def _simple_select_onecol( - self, table, keyvalues, retcol, desc="_simple_select_onecol" - ): - """Executes a SELECT query on the named table, which returns a list - comprising of the values of the named column from the selected rows. - - Args: - table (str): table name - keyvalues (dict|None): column names and values to select the rows with - retcol (str): column whos value we wish to retrieve. - - Returns: - Deferred: Results in a list - """ - return self.runInteraction( - desc, self._simple_select_onecol_txn, table, keyvalues, retcol - ) - - def _simple_select_list( - self, table, keyvalues, retcols, desc="_simple_select_list" - ): - """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 (str): the table name - keyvalues (dict[str, Any] | None): - column names and values to select the rows with, or None to not - apply a WHERE clause. - retcols (iterable[str]): the names of the columns to return - Returns: - defer.Deferred: resolves to list[dict[str, Any]] - """ - return self.runInteraction( - desc, self._simple_select_list_txn, table, keyvalues, retcols - ) - - @classmethod - def _simple_select_list_txn(cls, txn, table, keyvalues, retcols): - """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 (str): the table name - keyvalues (dict[str, T] | None): - column names and values to select the rows with, or None to not - apply a WHERE clause. - retcols (iterable[str]): the names of the columns to return - """ - if keyvalues: - sql = "SELECT %s FROM %s WHERE %s" % ( - ", ".join(retcols), - table, - " AND ".join("%s = ?" % (k,) for k in keyvalues), - ) - txn.execute(sql, list(keyvalues.values())) - else: - sql = "SELECT %s FROM %s" % (", ".join(retcols), table) - txn.execute(sql) - - return cls.cursor_to_dict(txn) - - @defer.inlineCallbacks - def _simple_select_many_batch( - self, - table, - column, - iterable, - retcols, - keyvalues={}, - desc="_simple_select_many_batch", - batch_size=100, - ): - """Executes a SELECT query on the named table, which may return zero or - more rows, returning the result as a list of dicts. - - Filters rows by if value of `column` is in `iterable`. - - Args: - table : string giving the table name - column : column name to test for inclusion against `iterable` - 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 - """ - results = [] - - if not iterable: - return results - - # iterables can not be sliced, so convert it to a list first - it_list = list(iterable) - - chunks = [ - it_list[i : i + batch_size] for i in range(0, len(it_list), batch_size) - ] - for chunk in chunks: - rows = yield self.runInteraction( - desc, - self._simple_select_many_txn, - table, - column, - chunk, - keyvalues, - retcols, - ) - - results.extend(rows) - - return results - - @classmethod - def _simple_select_many_txn(cls, txn, table, column, iterable, keyvalues, retcols): - """Executes a SELECT query on the named table, which may return zero or - more rows, returning the result as a list of dicts. - - Filters rows by if value of `column` is in `iterable`. - - Args: - txn : Transaction object - table : string giving the table name - column : column name to test for inclusion against `iterable` - 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 - """ - if not iterable: - return [] - - clause, values = make_in_list_sql_clause(txn.database_engine, column, iterable) - clauses = [clause] - - for key, value in iteritems(keyvalues): - clauses.append("%s = ?" % (key,)) - values.append(value) - - sql = "SELECT %s FROM %s WHERE %s" % ( - ", ".join(retcols), - table, - " AND ".join(clauses), - ) - - txn.execute(sql, values) - return cls.cursor_to_dict(txn) - - def _simple_update(self, table, keyvalues, updatevalues, desc): - return self.runInteraction( - desc, self._simple_update_txn, table, keyvalues, updatevalues - ) - - @staticmethod - def _simple_update_txn(txn, table, keyvalues, updatevalues): - if keyvalues: - where = "WHERE %s" % " AND ".join("%s = ?" % k for k in iterkeys(keyvalues)) - else: - where = "" - - update_sql = "UPDATE %s SET %s %s" % ( - table, - ", ".join("%s = ?" % (k,) for k in updatevalues), - where, - ) - - txn.execute(update_sql, list(updatevalues.values()) + list(keyvalues.values())) - - return txn.rowcount - - def _simple_update_one( - self, table, keyvalues, updatevalues, desc="_simple_update_one" - ): - """Executes an UPDATE query on the named table, setting new values for - columns in a row matching the key values. - - Args: - table : string giving the table name - keyvalues : dict of column names and values to select the row with - updatevalues : dict giving column names and values to update - retcols : optional list of column names to return - - If present, retcols gives a list of column names on which to perform - a SELECT statement *before* performing the UPDATE statement. The values - of these will be returned in a dict. - - These are performed within the same transaction, allowing an atomic - get-and-set. This can be used to implement compare-and-set by putting - the update column in the 'keyvalues' dict as well. - """ - return self.runInteraction( - desc, self._simple_update_one_txn, table, keyvalues, updatevalues - ) - - @classmethod - def _simple_update_one_txn(cls, txn, table, keyvalues, updatevalues): - rowcount = cls._simple_update_txn(txn, table, keyvalues, updatevalues) - - if rowcount == 0: - raise StoreError(404, "No row found (%s)" % (table,)) - if rowcount > 1: - raise StoreError(500, "More than one row matched (%s)" % (table,)) - - @staticmethod - def _simple_select_one_txn(txn, table, keyvalues, retcols, allow_none=False): - select_sql = "SELECT %s FROM %s WHERE %s" % ( - ", ".join(retcols), - table, - " AND ".join("%s = ?" % (k,) for k in keyvalues), - ) - - txn.execute(select_sql, list(keyvalues.values())) - row = txn.fetchone() - - if not row: - if allow_none: - return None - raise StoreError(404, "No row found (%s)" % (table,)) - if txn.rowcount > 1: - raise StoreError(500, "More than one row matched (%s)" % (table,)) - - return dict(zip(retcols, row)) - - def _simple_delete_one(self, table, keyvalues, desc="_simple_delete_one"): - """Executes a DELETE query on the named table, expecting to delete a - single row. - - Args: - table : string giving the table name - keyvalues : dict of column names and values to select the row with - """ - return self.runInteraction(desc, self._simple_delete_one_txn, table, keyvalues) - - @staticmethod - def _simple_delete_one_txn(txn, table, keyvalues): - """Executes a DELETE query on the named table, expecting to delete a - single row. - - Args: - table : string giving the table name - keyvalues : dict of column names and values to select the row with - """ - sql = "DELETE FROM %s WHERE %s" % ( - table, - " AND ".join("%s = ?" % (k,) for k in keyvalues), - ) - - txn.execute(sql, list(keyvalues.values())) - if txn.rowcount == 0: - raise StoreError(404, "No row found (%s)" % (table,)) - if txn.rowcount > 1: - raise StoreError(500, "More than one row matched (%s)" % (table,)) - - def _simple_delete(self, table, keyvalues, desc): - return self.runInteraction(desc, self._simple_delete_txn, table, keyvalues) - - @staticmethod - def _simple_delete_txn(txn, table, keyvalues): - sql = "DELETE FROM %s WHERE %s" % ( - table, - " AND ".join("%s = ?" % (k,) for k in keyvalues), - ) - - txn.execute(sql, list(keyvalues.values())) - return txn.rowcount - - def _simple_delete_many(self, table, column, iterable, keyvalues, desc): - return self.runInteraction( - desc, self._simple_delete_many_txn, table, column, iterable, keyvalues - ) - - @staticmethod - def _simple_delete_many_txn(txn, table, column, iterable, keyvalues): - """Executes a DELETE query on the named table. - - Filters rows by if value of `column` is in `iterable`. - - Args: - txn : Transaction object - table : string giving the table name - column : column name to test for inclusion against `iterable` - iterable : list - keyvalues : dict of column names and values to select the rows with - - Returns: - int: Number rows deleted - """ - if not iterable: - return 0 - - sql = "DELETE FROM %s" % table - - clause, values = make_in_list_sql_clause(txn.database_engine, column, iterable) - clauses = [clause] - - for key, value in iteritems(keyvalues): - clauses.append("%s = ?" % (key,)) - values.append(value) - - if clauses: - sql = "%s WHERE %s" % (sql, " AND ".join(clauses)) - txn.execute(sql, values) - - return txn.rowcount - - def _get_cache_dict( - self, db_conn, table, entity_column, stream_column, max_value, limit=100000 - ): - # Fetch a mapping of room_id -> max stream position for "recent" rooms. - # It doesn't really matter how many we get, the StreamChangeCache will - # do the right thing to ensure it respects the max size of cache. - sql = ( - "SELECT %(entity)s, MAX(%(stream)s) FROM %(table)s" - " WHERE %(stream)s > ? - %(limit)s" - " GROUP BY %(entity)s" - ) % { - "table": table, - "entity": entity_column, - "stream": stream_column, - "limit": limit, - } - - sql = self.database_engine.convert_param_style(sql) - - txn = db_conn.cursor() - txn.execute(sql, (int(max_value),)) - - cache = {row[0]: int(row[1]) for row in txn} - - txn.close() - - if cache: - min_val = min(itervalues(cache)) - else: - min_val = max_value - - return cache, min_val - - def _invalidate_cache_and_stream(self, txn, cache_func, keys): - """Invalidates the cache and adds it to the cache stream so slaves - will know to invalidate their caches. - - This should only be used to invalidate caches where slaves won't - otherwise know from other replication streams that the cache should - be invalidated. - """ - txn.call_after(cache_func.invalidate, keys) - self._send_invalidation_to_replication(txn, cache_func.__name__, keys) - - def _invalidate_state_caches_and_stream(self, txn, room_id, members_changed): - """Special case invalidation of caches based on current state. - - We special case this so that we can batch the cache invalidations into a - single replication poke. - - Args: - txn - room_id (str): Room where state changed - members_changed (iterable[str]): The user_ids of members that have changed - """ - txn.call_after(self._invalidate_state_caches, room_id, members_changed) - - if members_changed: - # We need to be careful that the size of the `members_changed` list - # isn't so large that it causes problems sending over replication, so we - # send them in chunks. - # Max line length is 16K, and max user ID length is 255, so 50 should - # be safe. - for chunk in batch_iter(members_changed, 50): - keys = itertools.chain([room_id], chunk) - self._send_invalidation_to_replication( - txn, _CURRENT_STATE_CACHE_NAME, keys - ) - else: - # if no members changed, we still need to invalidate the other caches. - self._send_invalidation_to_replication( - txn, _CURRENT_STATE_CACHE_NAME, [room_id] - ) - def _invalidate_state_caches(self, room_id, members_changed): """Invalidates caches that are based on the current state, but does not stream invalidations down replication. @@ -1441,226 +79,6 @@ class SQLBaseStore(object): # which is fine. pass - def _send_invalidation_to_replication(self, txn, cache_name, keys): - """Notifies replication that given cache has been invalidated. - - Note that this does *not* invalidate the cache locally. - - Args: - txn - cache_name (str) - keys (iterable[str]) - """ - - if isinstance(self.database_engine, PostgresEngine): - # get_next() returns a context manager which is designed to wrap - # the transaction. However, we want to only get an ID when we want - # to use it, here, so we need to call __enter__ manually, and have - # __exit__ called after the transaction finishes. - ctx = self._cache_id_gen.get_next() - stream_id = ctx.__enter__() - txn.call_on_exception(ctx.__exit__, None, None, None) - txn.call_after(ctx.__exit__, None, None, None) - txn.call_after(self.hs.get_notifier().on_new_replication_data) - - self._simple_insert_txn( - txn, - table="cache_invalidation_stream", - values={ - "stream_id": stream_id, - "cache_func": cache_name, - "keys": list(keys), - "invalidation_ts": self.clock.time_msec(), - }, - ) - - def get_all_updated_caches(self, last_id, current_id, limit): - if last_id == current_id: - return defer.succeed([]) - - def get_all_updated_caches_txn(txn): - # We purposefully don't bound by the current token, as we want to - # send across cache invalidations as quickly as possible. Cache - # invalidations are idempotent, so duplicates are fine. - sql = ( - "SELECT stream_id, cache_func, keys, invalidation_ts" - " FROM cache_invalidation_stream" - " WHERE stream_id > ? ORDER BY stream_id ASC LIMIT ?" - ) - txn.execute(sql, (last_id, limit)) - return txn.fetchall() - - return self.runInteraction("get_all_updated_caches", get_all_updated_caches_txn) - - def get_cache_stream_token(self): - if self._cache_id_gen: - return self._cache_id_gen.get_current_token() - else: - return 0 - - def _simple_select_list_paginate( - self, - table, - keyvalues, - orderby, - start, - limit, - retcols, - order_direction="ASC", - desc="_simple_select_list_paginate", - ): - """ - 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, - returning the result as a list of dicts. - - Args: - table (str): the table name - keyvalues (dict[str, T] | None): - column names and values to select the rows with, or None to not - apply a WHERE clause. - orderby (str): Column to order the results by. - start (int): Index to begin the query at. - limit (int): Number of results to return. - retcols (iterable[str]): the names of the columns to return - order_direction (str): Whether the results should be ordered "ASC" or "DESC". - Returns: - defer.Deferred: resolves to list[dict[str, Any]] - """ - return self.runInteraction( - desc, - self._simple_select_list_paginate_txn, - table, - keyvalues, - orderby, - start, - limit, - retcols, - order_direction=order_direction, - ) - - @classmethod - def _simple_select_list_paginate_txn( - cls, - txn, - table, - keyvalues, - orderby, - start, - limit, - retcols, - order_direction="ASC", - ): - """ - 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, - returning the result as a list of dicts. - - Args: - txn : Transaction object - table (str): the table name - keyvalues (dict[str, T] | None): - column names and values to select the rows with, or None to not - apply a WHERE clause. - orderby (str): Column to order the results by. - start (int): Index to begin the query at. - limit (int): Number of results to return. - retcols (iterable[str]): the names of the columns to return - order_direction (str): Whether the results should be ordered "ASC" or "DESC". - Returns: - defer.Deferred: resolves to list[dict[str, Any]] - """ - if order_direction not in ["ASC", "DESC"]: - raise ValueError("order_direction must be one of 'ASC' or 'DESC'.") - - if keyvalues: - where_clause = "WHERE " + " AND ".join("%s = ?" % (k,) for k in keyvalues) - else: - where_clause = "" - - sql = "SELECT %s FROM %s %s ORDER BY %s %s LIMIT ? OFFSET ?" % ( - ", ".join(retcols), - table, - where_clause, - orderby, - order_direction, - ) - txn.execute(sql, list(keyvalues.values()) + [limit, start]) - - return cls.cursor_to_dict(txn) - - def get_user_count_txn(self, txn): - """Get a total number of registered users in the users list. - - Args: - txn : Transaction object - Returns: - int : number of users - """ - sql_count = "SELECT COUNT(*) FROM users WHERE is_guest = 0;" - txn.execute(sql_count) - return txn.fetchone()[0] - - def _simple_search_list( - self, table, term, col, retcols, desc="_simple_search_list" - ): - """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 (str): the table name - term (str | None): - term for searching the table matched to a column. - col (str): column to query term should be matched to - retcols (iterable[str]): the names of the columns to return - Returns: - defer.Deferred: resolves to list[dict[str, Any]] or None - """ - - return self.runInteraction( - desc, self._simple_search_list_txn, table, term, col, retcols - ) - - @classmethod - def _simple_search_list_txn(cls, txn, table, term, col, retcols): - """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 (str): the table name - term (str | None): - term for searching the table matched to a column. - col (str): column to query term should be matched to - retcols (iterable[str]): the names of the columns to return - Returns: - defer.Deferred: resolves to list[dict[str, Any]] or None - """ - if term: - sql = "SELECT %s FROM %s WHERE %s LIKE ?" % (", ".join(retcols), table, col) - termvalues = ["%%" + term + "%%"] - txn.execute(sql, termvalues) - else: - return 0 - - return cls.cursor_to_dict(txn) - - @property - def database_engine_name(self): - return self.database_engine.module.__name__ - - def get_server_version(self): - """Returns a string describing the server version number""" - return self.database_engine.server_version - - -class _RollbackButIsFineException(Exception): - """ This exception is used to rollback a transaction without implying - something went wrong. - """ - - pass - def db_to_json(db_content): """ @@ -1689,30 +107,3 @@ def db_to_json(db_content): except Exception: logging.warning("Tried to decode '%r' as JSON and failed", db_content) raise - - -def make_in_list_sql_clause( - database_engine, column: str, iterable: Iterable -) -> Tuple[str, Iterable]: - """Returns an SQL clause that checks the given column is in the iterable. - - On SQLite this expands to `column IN (?, ?, ...)`, whereas on Postgres - it expands to `column = ANY(?)`. While both DBs support the `IN` form, - using the `ANY` form on postgres means that it views queries with - different length iterables as the same, helping the query stats. - - Args: - database_engine - column: Name of the column - iterable: The values to check the column against. - - Returns: - A tuple of SQL query and the args - """ - - if database_engine.supports_using_any_list: - # This should hopefully be faster, but also makes postgres query - # stats easier to understand. - return "%s = ANY(?)" % (column,), [list(iterable)] - else: - return "%s IN (%s)" % (column, ",".join("?" for _ in iterable)), list(iterable) diff --git a/synapse/storage/background_updates.py b/synapse/storage/background_updates.py index 37d469ffd7..bd547f35cf 100644 --- a/synapse/storage/background_updates.py +++ b/synapse/storage/background_updates.py @@ -14,6 +14,7 @@ # limitations under the License. import logging +from typing import Optional from canonicaljson import json @@ -22,7 +23,6 @@ from twisted.internet import defer from synapse.metrics.background_process_metrics import run_as_background_process from . import engines -from ._base import SQLBaseStore logger = logging.getLogger(__name__) @@ -74,7 +74,7 @@ class BackgroundUpdatePerformance(object): return float(self.total_item_count) / float(self.total_duration_ms) -class BackgroundUpdateStore(SQLBaseStore): +class BackgroundUpdater(object): """ Background updates are updates to the database that run in the background. Each update processes a batch of data at once. We attempt to limit the impact of each update by monitoring how long each batch takes to @@ -86,8 +86,10 @@ class BackgroundUpdateStore(SQLBaseStore): BACKGROUND_UPDATE_INTERVAL_MS = 1000 BACKGROUND_UPDATE_DURATION_MS = 100 - def __init__(self, db_conn, hs): - super(BackgroundUpdateStore, self).__init__(db_conn, hs) + def __init__(self, hs, database): + self._clock = hs.get_clock() + self.db = database + self._background_update_performance = {} self._background_update_queue = [] self._background_update_handlers = {} @@ -96,17 +98,14 @@ class BackgroundUpdateStore(SQLBaseStore): def start_doing_background_updates(self): run_as_background_process("background_updates", self.run_background_updates) - @defer.inlineCallbacks - def run_background_updates(self, sleep=True): + async def run_background_updates(self, sleep=True): logger.info("Starting background schema updates") while True: if sleep: - yield self.hs.get_clock().sleep( - self.BACKGROUND_UPDATE_INTERVAL_MS / 1000.0 - ) + await self._clock.sleep(self.BACKGROUND_UPDATE_INTERVAL_MS / 1000.0) try: - result = yield self.do_next_background_update( + result = await self.do_next_background_update( self.BACKGROUND_UPDATE_DURATION_MS ) except Exception: @@ -139,7 +138,7 @@ class BackgroundUpdateStore(SQLBaseStore): # otherwise, check if there are updates to be run. This is important, # as we may be running on a worker which doesn't perform the bg updates # itself, but still wants to wait for them to happen. - updates = yield self._simple_select_onecol( + updates = yield self.db.simple_select_onecol( "background_updates", keyvalues=None, retcol="1", @@ -161,7 +160,7 @@ class BackgroundUpdateStore(SQLBaseStore): if update_name in self._background_update_queue: return False - update_exists = await self._simple_select_one_onecol( + update_exists = await self.db.simple_select_one_onecol( "background_updates", keyvalues={"update_name": update_name}, retcol="1", @@ -171,20 +170,21 @@ class BackgroundUpdateStore(SQLBaseStore): return not update_exists - @defer.inlineCallbacks - def do_next_background_update(self, desired_duration_ms): + async def do_next_background_update( + self, desired_duration_ms: float + ) -> Optional[int]: """Does some amount of work on the next queued background update + Returns once some amount of work is done. + Args: desired_duration_ms(float): How long we want to spend updating. Returns: - A deferred that completes once some amount of work is done. - The deferred will have a value of None if there is currently - no more work to do. + None if there is no more work to do, otherwise an int """ if not self._background_update_queue: - updates = yield self._simple_select_list( + updates = await self.db.simple_select_list( "background_updates", keyvalues=None, retcols=("update_name", "depends_on"), @@ -202,11 +202,12 @@ class BackgroundUpdateStore(SQLBaseStore): update_name = self._background_update_queue.pop(0) self._background_update_queue.append(update_name) - res = yield self._do_background_update(update_name, desired_duration_ms) + res = await self._do_background_update(update_name, desired_duration_ms) return res - @defer.inlineCallbacks - def _do_background_update(self, update_name, desired_duration_ms): + async def _do_background_update( + self, update_name: str, desired_duration_ms: float + ) -> int: logger.info("Starting update batch on background update '%s'", update_name) update_handler = self._background_update_handlers[update_name] @@ -226,7 +227,7 @@ class BackgroundUpdateStore(SQLBaseStore): else: batch_size = self.DEFAULT_BACKGROUND_BATCH_SIZE - progress_json = yield self._simple_select_one_onecol( + progress_json = await self.db.simple_select_one_onecol( "background_updates", keyvalues={"update_name": update_name}, retcol="progress_json", @@ -235,7 +236,7 @@ class BackgroundUpdateStore(SQLBaseStore): progress = json.loads(progress_json) time_start = self._clock.time_msec() - items_updated = yield update_handler(progress, batch_size) + items_updated = await update_handler(progress, batch_size) time_stop = self._clock.time_msec() duration_ms = time_stop - time_start @@ -264,7 +265,9 @@ class BackgroundUpdateStore(SQLBaseStore): * A dict of the current progress * An integer count of the number of items to update in this batch. - The handler should return a deferred integer count of items updated. + The handler should return a deferred or coroutine which returns an integer count + of items updated. + The handler is responsible for updating the progress of the update. Args: @@ -380,7 +383,7 @@ class BackgroundUpdateStore(SQLBaseStore): logger.debug("[SQL] %s", sql) c.execute(sql) - if isinstance(self.database_engine, engines.PostgresEngine): + if isinstance(self.db.engine, engines.PostgresEngine): runner = create_index_psql elif psql_only: runner = None @@ -391,7 +394,7 @@ class BackgroundUpdateStore(SQLBaseStore): def updater(progress, batch_size): if runner is not None: logger.info("Adding index %s to %s", index_name, table) - yield self.runWithConnection(runner) + yield self.db.runWithConnection(runner) yield self._end_background_update(update_name) return 1 @@ -413,7 +416,7 @@ class BackgroundUpdateStore(SQLBaseStore): self._background_update_queue = [] progress_json = json.dumps(progress) - return self._simple_insert( + return self.db.simple_insert( "background_updates", {"update_name": update_name, "progress_json": progress_json}, ) @@ -429,10 +432,25 @@ class BackgroundUpdateStore(SQLBaseStore): self._background_update_queue = [ name for name in self._background_update_queue if name != update_name ] - return self._simple_delete_one( + return self.db.simple_delete_one( "background_updates", keyvalues={"update_name": update_name} ) + def _background_update_progress(self, update_name: str, progress: dict): + """Update the progress of a background update + + Args: + update_name: The name of the background update task + progress: The progress of the update. + """ + + return self.db.runInteraction( + "background_update_progress", + self._background_update_progress_txn, + update_name, + progress, + ) + def _background_update_progress_txn(self, txn, update_name, progress): """Update the progress of a background update @@ -444,7 +462,7 @@ class BackgroundUpdateStore(SQLBaseStore): progress_json = json.dumps(progress) - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, "background_updates", keyvalues={"update_name": update_name}, diff --git a/synapse/storage/data_stores/__init__.py b/synapse/storage/data_stores/__init__.py index cb184a98cc..e1d03429ca 100644 --- a/synapse/storage/data_stores/__init__.py +++ b/synapse/storage/data_stores/__init__.py @@ -13,14 +13,76 @@ # See the License for the specific language governing permissions and # limitations under the License. +import logging + +from synapse.storage.data_stores.state import StateGroupDataStore +from synapse.storage.database import Database, make_conn +from synapse.storage.engines import create_engine +from synapse.storage.prepare_database import prepare_database + +logger = logging.getLogger(__name__) + class DataStores(object): """The various data stores. These are low level interfaces to physical databases. + + Attributes: + main (DataStore) """ - def __init__(self, main_store, db_conn, hs): - # Note we pass in the main store here as workers use a different main + def __init__(self, main_store_class, hs): + # Note we pass in the main store class here as workers use a different main # store. - self.main = main_store + + self.databases = [] + self.main = None + self.state = None + + for database_config in hs.config.database.databases: + db_name = database_config.name + engine = create_engine(database_config.config) + + with make_conn(database_config, engine) as db_conn: + logger.info("Preparing database %r...", db_name) + + engine.check_database(db_conn) + prepare_database( + db_conn, engine, hs.config, data_stores=database_config.data_stores, + ) + + database = Database(hs, database_config, engine) + + if "main" in database_config.data_stores: + logger.info("Starting 'main' data store") + + # Sanity check we don't try and configure the main store on + # multiple databases. + if self.main: + raise Exception("'main' data store already configured") + + self.main = main_store_class(database, db_conn, hs) + + if "state" in database_config.data_stores: + logger.info("Starting 'state' data store") + + # Sanity check we don't try and configure the state store on + # multiple databases. + if self.state: + raise Exception("'state' data store already configured") + + self.state = StateGroupDataStore(database, db_conn, hs) + + db_conn.commit() + + self.databases.append(database) + + logger.info("Database %r prepared", db_name) + + # Sanity check that we have actually configured all the required stores. + if not self.main: + raise Exception("No 'main' data store configured") + + if not self.state: + raise Exception("No 'main' data store configured") diff --git a/synapse/storage/data_stores/main/__init__.py b/synapse/storage/data_stores/main/__init__.py index 10c940df1e..2700cca822 100644 --- a/synapse/storage/data_stores/main/__init__.py +++ b/synapse/storage/data_stores/main/__init__.py @@ -19,9 +19,8 @@ import calendar import logging import time -from twisted.internet import defer - from synapse.api.constants import PresenceState +from synapse.storage.database import Database from synapse.storage.engines import PostgresEngine from synapse.storage.util.id_generators import ( ChainedIdGenerator, @@ -32,6 +31,7 @@ from synapse.util.caches.stream_change_cache import StreamChangeCache from .account_data import AccountDataStore from .appservice import ApplicationServiceStore, ApplicationServiceTransactionStore +from .cache import CacheInvalidationStore from .client_ips import ClientIpStore from .deviceinbox import DeviceInboxStore from .devices import DeviceStore @@ -110,11 +110,22 @@ class DataStore( MonthlyActiveUsersStore, StatsStore, RelationsStore, + CacheInvalidationStore, ): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): self.hs = hs self._clock = hs.get_clock() - self.database_engine = hs.database_engine + self.database_engine = database.engine + + all_users_native = are_all_users_on_domain( + db_conn.cursor(), database.engine, hs.hostname + ) + if not all_users_native: + raise Exception( + "Found users in database not native to %s!\n" + "You cannot changed a synapse server_name after it's been configured" + % (hs.hostname,) + ) self._stream_id_gen = StreamIdGenerator( db_conn, @@ -169,9 +180,11 @@ class DataStore( else: self._cache_id_gen = None + super(DataStore, self).__init__(database, db_conn, hs) + self._presence_on_startup = self._get_active_presence(db_conn) - presence_cache_prefill, min_presence_val = self._get_cache_dict( + presence_cache_prefill, min_presence_val = self.db.get_cache_dict( db_conn, "presence_stream", entity_column="user_id", @@ -185,7 +198,7 @@ class DataStore( ) max_device_inbox_id = self._device_inbox_id_gen.get_current_token() - device_inbox_prefill, min_device_inbox_id = self._get_cache_dict( + device_inbox_prefill, min_device_inbox_id = self.db.get_cache_dict( db_conn, "device_inbox", entity_column="user_id", @@ -200,7 +213,7 @@ class DataStore( ) # The federation outbox and the local device inbox uses the same # stream_id generator. - device_outbox_prefill, min_device_outbox_id = self._get_cache_dict( + device_outbox_prefill, min_device_outbox_id = self.db.get_cache_dict( db_conn, "device_federation_outbox", entity_column="destination", @@ -226,7 +239,7 @@ class DataStore( ) events_max = self._stream_id_gen.get_current_token() - curr_state_delta_prefill, min_curr_state_delta_id = self._get_cache_dict( + curr_state_delta_prefill, min_curr_state_delta_id = self.db.get_cache_dict( db_conn, "current_state_delta_stream", entity_column="room_id", @@ -240,7 +253,7 @@ class DataStore( prefilled_cache=curr_state_delta_prefill, ) - _group_updates_prefill, min_group_updates_id = self._get_cache_dict( + _group_updates_prefill, min_group_updates_id = self.db.get_cache_dict( db_conn, "local_group_updates", entity_column="user_id", @@ -260,8 +273,6 @@ class DataStore( # Used in _generate_user_daily_visits to keep track of progress self._last_user_visit_update = self._get_start_of_day() - super(DataStore, self).__init__(db_conn, hs) - def take_presence_startup_info(self): active_on_startup = self._presence_on_startup self._presence_on_startup = None @@ -281,7 +292,7 @@ class DataStore( txn = db_conn.cursor() txn.execute(sql, (PresenceState.OFFLINE,)) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) txn.close() for row in rows: @@ -294,7 +305,7 @@ class DataStore( Counts the number of users who used this homeserver in the last 24 hours. """ yesterday = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24) - return self.runInteraction("count_daily_users", self._count_users, yesterday) + return self.db.runInteraction("count_daily_users", self._count_users, yesterday) def count_monthly_users(self): """ @@ -304,7 +315,7 @@ class DataStore( amongst other things, includes a 3 day grace period before a user counts. """ thirty_days_ago = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24 * 30) - return self.runInteraction( + return self.db.runInteraction( "count_monthly_users", self._count_users, thirty_days_ago ) @@ -404,7 +415,7 @@ class DataStore( return results - return self.runInteraction("count_r30_users", _count_r30_users) + return self.db.runInteraction("count_r30_users", _count_r30_users) def _get_start_of_day(self): """ @@ -469,50 +480,73 @@ class DataStore( # frequently self._last_user_visit_update = now - return self.runInteraction( + return self.db.runInteraction( "generate_user_daily_visits", _generate_user_daily_visits ) def get_users(self): - """Function to reterive a list of users in users table. + """Function to retrieve a list of users in users table. Args: Returns: defer.Deferred: resolves to list[dict[str, Any]] """ - return self._simple_select_list( + return self.db.simple_select_list( table="users", keyvalues={}, - retcols=["name", "password_hash", "is_guest", "admin", "user_type"], + retcols=[ + "name", + "password_hash", + "is_guest", + "admin", + "user_type", + "deactivated", + ], desc="get_users", ) - @defer.inlineCallbacks - def get_users_paginate(self, order, start, limit): - """Function to reterive a paginated list of users from - users list. This will return a json object, which contains - list of users and the total number of users in users table. + def get_users_paginate( + self, start, limit, name=None, guests=True, deactivated=False + ): + """Function to retrieve a paginated list of users from + users list. This will return a json list of users. Args: - order (str): column name to order the select by this column start (int): start number to begin the query from - limit (int): number of rows to reterive + limit (int): number of rows to retrieve + name (string): filter for user names + guests (bool): whether to in include guest users + deactivated (bool): whether to include deactivated users Returns: - defer.Deferred: resolves to json object {list[dict[str, Any]], count} + defer.Deferred: resolves to list[dict[str, Any]] """ - users = yield self.runInteraction( - "get_users_paginate", - self._simple_select_list_paginate_txn, + name_filter = {} + if name: + name_filter["name"] = "%" + name + "%" + + attr_filter = {} + if not guests: + attr_filter["is_guest"] = 0 + if not deactivated: + attr_filter["deactivated"] = 0 + + return self.db.simple_select_list_paginate( + desc="get_users_paginate", table="users", - keyvalues={"is_guest": False}, - orderby=order, + orderby="name", start=start, limit=limit, - retcols=["name", "password_hash", "is_guest", "admin", "user_type"], + filters=name_filter, + keyvalues=attr_filter, + retcols=[ + "name", + "password_hash", + "is_guest", + "admin", + "user_type", + "deactivated", + ], ) - count = yield self.runInteraction("get_users_paginate", self.get_user_count_txn) - retval = {"users": users, "total": count} - return retval def search_users(self, term): """Function to search users list for one or more users with @@ -524,10 +558,22 @@ class DataStore( Returns: defer.Deferred: resolves to list[dict[str, Any]] """ - return self._simple_search_list( + return self.db.simple_search_list( table="users", term=term, col="name", retcols=["name", "password_hash", "is_guest", "admin", "user_type"], desc="search_users", ) + + +def are_all_users_on_domain(txn, database_engine, domain): + sql = database_engine.convert_param_style( + "SELECT COUNT(*) FROM users WHERE name NOT LIKE ?" + ) + pat = "%:" + domain + txn.execute(sql, (pat,)) + num_not_matching = txn.fetchall()[0][0] + if num_not_matching == 0: + return True + return False diff --git a/synapse/storage/data_stores/main/account_data.py b/synapse/storage/data_stores/main/account_data.py index 6afbfc0d74..46b494b334 100644 --- a/synapse/storage/data_stores/main/account_data.py +++ b/synapse/storage/data_stores/main/account_data.py @@ -22,6 +22,7 @@ from canonicaljson import json from twisted.internet import defer from synapse.storage._base import SQLBaseStore +from synapse.storage.database import Database from synapse.storage.util.id_generators import StreamIdGenerator from synapse.util.caches.descriptors import cached, cachedInlineCallbacks from synapse.util.caches.stream_change_cache import StreamChangeCache @@ -38,13 +39,13 @@ class AccountDataWorkerStore(SQLBaseStore): # the abstract methods being implemented. __metaclass__ = abc.ABCMeta - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): account_max = self.get_max_account_data_stream_id() self._account_data_stream_cache = StreamChangeCache( "AccountDataAndTagsChangeCache", account_max ) - super(AccountDataWorkerStore, self).__init__(db_conn, hs) + super(AccountDataWorkerStore, self).__init__(database, db_conn, hs) @abc.abstractmethod def get_max_account_data_stream_id(self): @@ -67,7 +68,7 @@ class AccountDataWorkerStore(SQLBaseStore): """ def get_account_data_for_user_txn(txn): - rows = self._simple_select_list_txn( + rows = self.db.simple_select_list_txn( txn, "account_data", {"user_id": user_id}, @@ -78,7 +79,7 @@ class AccountDataWorkerStore(SQLBaseStore): row["account_data_type"]: json.loads(row["content"]) for row in rows } - rows = self._simple_select_list_txn( + rows = self.db.simple_select_list_txn( txn, "room_account_data", {"user_id": user_id}, @@ -92,7 +93,7 @@ class AccountDataWorkerStore(SQLBaseStore): return global_account_data, by_room - return self.runInteraction( + return self.db.runInteraction( "get_account_data_for_user", get_account_data_for_user_txn ) @@ -102,7 +103,7 @@ class AccountDataWorkerStore(SQLBaseStore): Returns: Deferred: A dict """ - result = yield self._simple_select_one_onecol( + result = yield self.db.simple_select_one_onecol( table="account_data", keyvalues={"user_id": user_id, "account_data_type": data_type}, retcol="content", @@ -127,7 +128,7 @@ class AccountDataWorkerStore(SQLBaseStore): """ def get_account_data_for_room_txn(txn): - rows = self._simple_select_list_txn( + rows = self.db.simple_select_list_txn( txn, "room_account_data", {"user_id": user_id, "room_id": room_id}, @@ -138,7 +139,7 @@ class AccountDataWorkerStore(SQLBaseStore): row["account_data_type"]: json.loads(row["content"]) for row in rows } - return self.runInteraction( + return self.db.runInteraction( "get_account_data_for_room", get_account_data_for_room_txn ) @@ -156,7 +157,7 @@ class AccountDataWorkerStore(SQLBaseStore): """ def get_account_data_for_room_and_type_txn(txn): - content_json = self._simple_select_one_onecol_txn( + content_json = self.db.simple_select_one_onecol_txn( txn, table="room_account_data", keyvalues={ @@ -170,7 +171,7 @@ class AccountDataWorkerStore(SQLBaseStore): return json.loads(content_json) if content_json else None - return self.runInteraction( + return self.db.runInteraction( "get_account_data_for_room_and_type", get_account_data_for_room_and_type_txn ) @@ -184,14 +185,14 @@ class AccountDataWorkerStore(SQLBaseStore): current_id(int): The position to fetch up to. Returns: A deferred pair of lists of tuples of stream_id int, user_id string, - room_id string, type string, and content string. + room_id string, and type string. """ if last_room_id == current_id and last_global_id == current_id: return defer.succeed(([], [])) def get_updated_account_data_txn(txn): sql = ( - "SELECT stream_id, user_id, account_data_type, content" + "SELECT stream_id, user_id, account_data_type" " FROM account_data WHERE ? < stream_id AND stream_id <= ?" " ORDER BY stream_id ASC LIMIT ?" ) @@ -199,7 +200,7 @@ class AccountDataWorkerStore(SQLBaseStore): global_results = txn.fetchall() sql = ( - "SELECT stream_id, user_id, room_id, account_data_type, content" + "SELECT stream_id, user_id, room_id, account_data_type" " FROM room_account_data WHERE ? < stream_id AND stream_id <= ?" " ORDER BY stream_id ASC LIMIT ?" ) @@ -207,7 +208,7 @@ class AccountDataWorkerStore(SQLBaseStore): room_results = txn.fetchall() return global_results, room_results - return self.runInteraction( + return self.db.runInteraction( "get_all_updated_account_data_txn", get_updated_account_data_txn ) @@ -250,9 +251,9 @@ class AccountDataWorkerStore(SQLBaseStore): user_id, int(stream_id) ) if not changed: - return {}, {} + return defer.succeed(({}, {})) - return self.runInteraction( + return self.db.runInteraction( "get_updated_account_data_for_user", get_updated_account_data_for_user_txn ) @@ -270,12 +271,12 @@ class AccountDataWorkerStore(SQLBaseStore): class AccountDataStore(AccountDataWorkerStore): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): self._account_data_id_gen = StreamIdGenerator( db_conn, "account_data_max_stream_id", "stream_id" ) - super(AccountDataStore, self).__init__(db_conn, hs) + super(AccountDataStore, self).__init__(database, db_conn, hs) def get_max_account_data_stream_id(self): """Get the current max stream id for the private user data stream @@ -300,9 +301,9 @@ class AccountDataStore(AccountDataWorkerStore): with self._account_data_id_gen.get_next() as next_id: # no need to lock here as room_account_data has a unique constraint - # on (user_id, room_id, account_data_type) so _simple_upsert will + # on (user_id, room_id, account_data_type) so simple_upsert will # retry if there is a conflict. - yield self._simple_upsert( + yield self.db.simple_upsert( desc="add_room_account_data", table="room_account_data", keyvalues={ @@ -346,9 +347,9 @@ class AccountDataStore(AccountDataWorkerStore): with self._account_data_id_gen.get_next() as next_id: # no need to lock here as account_data has a unique constraint on - # (user_id, account_data_type) so _simple_upsert will retry if + # (user_id, account_data_type) so simple_upsert will retry if # there is a conflict. - yield self._simple_upsert( + yield self.db.simple_upsert( desc="add_user_account_data", table="account_data", keyvalues={"user_id": user_id, "account_data_type": account_data_type}, @@ -388,4 +389,4 @@ class AccountDataStore(AccountDataWorkerStore): ) txn.execute(update_max_id_sql, (next_id, next_id)) - return self.runInteraction("update_account_data_max_stream_id", _update) + return self.db.runInteraction("update_account_data_max_stream_id", _update) diff --git a/synapse/storage/data_stores/main/appservice.py b/synapse/storage/data_stores/main/appservice.py index 81babf2029..b2f39649fd 100644 --- a/synapse/storage/data_stores/main/appservice.py +++ b/synapse/storage/data_stores/main/appservice.py @@ -24,6 +24,7 @@ from synapse.appservice import AppServiceTransaction from synapse.config.appservice import load_appservices from synapse.storage._base import SQLBaseStore from synapse.storage.data_stores.main.events_worker import EventsWorkerStore +from synapse.storage.database import Database logger = logging.getLogger(__name__) @@ -48,13 +49,13 @@ def _make_exclusive_regex(services_cache): class ApplicationServiceWorkerStore(SQLBaseStore): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): self.services_cache = load_appservices( hs.hostname, hs.config.app_service_config_files ) self.exclusive_user_regex = _make_exclusive_regex(self.services_cache) - super(ApplicationServiceWorkerStore, self).__init__(db_conn, hs) + super(ApplicationServiceWorkerStore, self).__init__(database, db_conn, hs) def get_app_services(self): return self.services_cache @@ -133,7 +134,7 @@ class ApplicationServiceTransactionWorkerStore( A Deferred which resolves to a list of ApplicationServices, which may be empty. """ - results = yield self._simple_select_list( + results = yield self.db.simple_select_list( "application_services_state", dict(state=state), ["as_id"] ) # NB: This assumes this class is linked with ApplicationServiceStore @@ -155,7 +156,7 @@ class ApplicationServiceTransactionWorkerStore( Returns: A Deferred which resolves to ApplicationServiceState. """ - result = yield self._simple_select_one( + result = yield self.db.simple_select_one( "application_services_state", dict(as_id=service.id), ["state"], @@ -175,7 +176,7 @@ class ApplicationServiceTransactionWorkerStore( Returns: A Deferred which resolves when the state was set successfully. """ - return self._simple_upsert( + return self.db.simple_upsert( "application_services_state", dict(as_id=service.id), dict(state=state) ) @@ -216,7 +217,7 @@ class ApplicationServiceTransactionWorkerStore( ) return AppServiceTransaction(service=service, id=new_txn_id, events=events) - return self.runInteraction("create_appservice_txn", _create_appservice_txn) + return self.db.runInteraction("create_appservice_txn", _create_appservice_txn) def complete_appservice_txn(self, txn_id, service): """Completes an application service transaction. @@ -249,7 +250,7 @@ class ApplicationServiceTransactionWorkerStore( ) # Set current txn_id for AS to 'txn_id' - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, "application_services_state", dict(as_id=service.id), @@ -257,11 +258,13 @@ class ApplicationServiceTransactionWorkerStore( ) # Delete txn - self._simple_delete_txn( + self.db.simple_delete_txn( txn, "application_services_txns", dict(txn_id=txn_id, as_id=service.id) ) - return self.runInteraction("complete_appservice_txn", _complete_appservice_txn) + return self.db.runInteraction( + "complete_appservice_txn", _complete_appservice_txn + ) @defer.inlineCallbacks def get_oldest_unsent_txn(self, service): @@ -283,7 +286,7 @@ class ApplicationServiceTransactionWorkerStore( " ORDER BY txn_id ASC LIMIT 1", (service.id,), ) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) if not rows: return None @@ -291,7 +294,7 @@ class ApplicationServiceTransactionWorkerStore( return entry - entry = yield self.runInteraction( + entry = yield self.db.runInteraction( "get_oldest_unsent_appservice_txn", _get_oldest_unsent_txn ) @@ -321,7 +324,7 @@ class ApplicationServiceTransactionWorkerStore( "UPDATE appservice_stream_position SET stream_ordering = ?", (pos,) ) - return self.runInteraction( + return self.db.runInteraction( "set_appservice_last_pos", set_appservice_last_pos_txn ) @@ -350,7 +353,7 @@ class ApplicationServiceTransactionWorkerStore( return upper_bound, [row[1] for row in rows] - upper_bound, event_ids = yield self.runInteraction( + upper_bound, event_ids = yield self.db.runInteraction( "get_new_events_for_appservice", get_new_events_for_appservice_txn ) diff --git a/synapse/storage/data_stores/main/cache.py b/synapse/storage/data_stores/main/cache.py new file mode 100644 index 0000000000..54ed8574c4 --- /dev/null +++ b/synapse/storage/data_stores/main/cache.py @@ -0,0 +1,133 @@ +# -*- coding: utf-8 -*- +# Copyright 2019 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +import itertools +import logging + +from twisted.internet import defer + +from synapse.storage._base import SQLBaseStore +from synapse.storage.engines import PostgresEngine +from synapse.util import batch_iter + +logger = logging.getLogger(__name__) + + +# This is a special cache name we use to batch multiple invalidations of caches +# based on the current state when notifying workers over replication. +CURRENT_STATE_CACHE_NAME = "cs_cache_fake" + + +class CacheInvalidationStore(SQLBaseStore): + def _invalidate_cache_and_stream(self, txn, cache_func, keys): + """Invalidates the cache and adds it to the cache stream so slaves + will know to invalidate their caches. + + This should only be used to invalidate caches where slaves won't + otherwise know from other replication streams that the cache should + be invalidated. + """ + txn.call_after(cache_func.invalidate, keys) + self._send_invalidation_to_replication(txn, cache_func.__name__, keys) + + def _invalidate_state_caches_and_stream(self, txn, room_id, members_changed): + """Special case invalidation of caches based on current state. + + We special case this so that we can batch the cache invalidations into a + single replication poke. + + Args: + txn + room_id (str): Room where state changed + members_changed (iterable[str]): The user_ids of members that have changed + """ + txn.call_after(self._invalidate_state_caches, room_id, members_changed) + + if members_changed: + # We need to be careful that the size of the `members_changed` list + # isn't so large that it causes problems sending over replication, so we + # send them in chunks. + # Max line length is 16K, and max user ID length is 255, so 50 should + # be safe. + for chunk in batch_iter(members_changed, 50): + keys = itertools.chain([room_id], chunk) + self._send_invalidation_to_replication( + txn, CURRENT_STATE_CACHE_NAME, keys + ) + else: + # if no members changed, we still need to invalidate the other caches. + self._send_invalidation_to_replication( + txn, CURRENT_STATE_CACHE_NAME, [room_id] + ) + + def _send_invalidation_to_replication(self, txn, cache_name, keys): + """Notifies replication that given cache has been invalidated. + + Note that this does *not* invalidate the cache locally. + + Args: + txn + cache_name (str) + keys (iterable[str]) + """ + + if isinstance(self.database_engine, PostgresEngine): + # get_next() returns a context manager which is designed to wrap + # the transaction. However, we want to only get an ID when we want + # to use it, here, so we need to call __enter__ manually, and have + # __exit__ called after the transaction finishes. + ctx = self._cache_id_gen.get_next() + stream_id = ctx.__enter__() + txn.call_on_exception(ctx.__exit__, None, None, None) + txn.call_after(ctx.__exit__, None, None, None) + txn.call_after(self.hs.get_notifier().on_new_replication_data) + + self.db.simple_insert_txn( + txn, + table="cache_invalidation_stream", + values={ + "stream_id": stream_id, + "cache_func": cache_name, + "keys": list(keys), + "invalidation_ts": self.clock.time_msec(), + }, + ) + + def get_all_updated_caches(self, last_id, current_id, limit): + if last_id == current_id: + return defer.succeed([]) + + def get_all_updated_caches_txn(txn): + # We purposefully don't bound by the current token, as we want to + # send across cache invalidations as quickly as possible. Cache + # invalidations are idempotent, so duplicates are fine. + sql = ( + "SELECT stream_id, cache_func, keys, invalidation_ts" + " FROM cache_invalidation_stream" + " WHERE stream_id > ? ORDER BY stream_id ASC LIMIT ?" + ) + txn.execute(sql, (last_id, limit)) + return txn.fetchall() + + return self.db.runInteraction( + "get_all_updated_caches", get_all_updated_caches_txn + ) + + def get_cache_stream_token(self): + if self._cache_id_gen: + return self._cache_id_gen.get_current_token() + else: + return 0 diff --git a/synapse/storage/data_stores/main/client_ips.py b/synapse/storage/data_stores/main/client_ips.py index 2f9ad008b5..ae188d0f49 100644 --- a/synapse/storage/data_stores/main/client_ips.py +++ b/synapse/storage/data_stores/main/client_ips.py @@ -20,8 +20,10 @@ from six import iteritems from twisted.internet import defer from synapse.metrics.background_process_metrics import wrap_as_background_process -from synapse.storage import background_updates -from synapse.storage._base import Cache +from synapse.storage._base import SQLBaseStore +from synapse.storage.database import Database +from synapse.util.caches import CACHE_SIZE_FACTOR +from synapse.util.caches.descriptors import Cache logger = logging.getLogger(__name__) @@ -31,41 +33,41 @@ logger = logging.getLogger(__name__) LAST_SEEN_GRANULARITY = 120 * 1000 -class ClientIpBackgroundUpdateStore(background_updates.BackgroundUpdateStore): - def __init__(self, db_conn, hs): - super(ClientIpBackgroundUpdateStore, self).__init__(db_conn, hs) +class ClientIpBackgroundUpdateStore(SQLBaseStore): + def __init__(self, database: Database, db_conn, hs): + super(ClientIpBackgroundUpdateStore, self).__init__(database, db_conn, hs) - self.register_background_index_update( + self.db.updates.register_background_index_update( "user_ips_device_index", index_name="user_ips_device_id", table="user_ips", columns=["user_id", "device_id", "last_seen"], ) - self.register_background_index_update( + self.db.updates.register_background_index_update( "user_ips_last_seen_index", index_name="user_ips_last_seen", table="user_ips", columns=["user_id", "last_seen"], ) - self.register_background_index_update( + self.db.updates.register_background_index_update( "user_ips_last_seen_only_index", index_name="user_ips_last_seen_only", table="user_ips", columns=["last_seen"], ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "user_ips_analyze", self._analyze_user_ip ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "user_ips_remove_dupes", self._remove_user_ip_dupes ) # Register a unique index - self.register_background_index_update( + self.db.updates.register_background_index_update( "user_ips_device_unique_index", index_name="user_ips_user_token_ip_unique_index", table="user_ips", @@ -74,12 +76,12 @@ class ClientIpBackgroundUpdateStore(background_updates.BackgroundUpdateStore): ) # Drop the old non-unique index - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "user_ips_drop_nonunique_index", self._remove_user_ip_nonunique ) # Update the last seen info in devices. - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "devices_last_seen", self._devices_last_seen_update ) @@ -90,8 +92,8 @@ class ClientIpBackgroundUpdateStore(background_updates.BackgroundUpdateStore): txn.execute("DROP INDEX IF EXISTS user_ips_user_ip") txn.close() - yield self.runWithConnection(f) - yield self._end_background_update("user_ips_drop_nonunique_index") + yield self.db.runWithConnection(f) + yield self.db.updates._end_background_update("user_ips_drop_nonunique_index") return 1 @defer.inlineCallbacks @@ -105,9 +107,9 @@ class ClientIpBackgroundUpdateStore(background_updates.BackgroundUpdateStore): def user_ips_analyze(txn): txn.execute("ANALYZE user_ips") - yield self.runInteraction("user_ips_analyze", user_ips_analyze) + yield self.db.runInteraction("user_ips_analyze", user_ips_analyze) - yield self._end_background_update("user_ips_analyze") + yield self.db.updates._end_background_update("user_ips_analyze") return 1 @@ -139,7 +141,7 @@ class ClientIpBackgroundUpdateStore(background_updates.BackgroundUpdateStore): return None # Get a last seen that has roughly `batch_size` since `begin_last_seen` - end_last_seen = yield self.runInteraction( + end_last_seen = yield self.db.runInteraction( "user_ips_dups_get_last_seen", get_last_seen ) @@ -270,14 +272,14 @@ class ClientIpBackgroundUpdateStore(background_updates.BackgroundUpdateStore): (user_id, access_token, ip, device_id, user_agent, last_seen), ) - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, "user_ips_remove_dupes", {"last_seen": end_last_seen} ) - yield self.runInteraction("user_ips_dups_remove", remove) + yield self.db.runInteraction("user_ips_dups_remove", remove) if last: - yield self._end_background_update("user_ips_remove_dupes") + yield self.db.updates._end_background_update("user_ips_remove_dupes") return batch_size @@ -343,7 +345,7 @@ class ClientIpBackgroundUpdateStore(background_updates.BackgroundUpdateStore): txn.execute_batch(sql, rows) _, _, _, user_id, device_id = rows[-1] - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, "devices_last_seen", {"last_user_id": user_id, "last_device_id": device_id}, @@ -351,24 +353,24 @@ class ClientIpBackgroundUpdateStore(background_updates.BackgroundUpdateStore): return len(rows) - updated = yield self.runInteraction( + updated = yield self.db.runInteraction( "_devices_last_seen_update", _devices_last_seen_update_txn ) if not updated: - yield self._end_background_update("devices_last_seen") + yield self.db.updates._end_background_update("devices_last_seen") return updated class ClientIpStore(ClientIpBackgroundUpdateStore): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): self.client_ip_last_seen = Cache( name="client_ip_last_seen", keylen=4, max_entries=50000 ) - super(ClientIpStore, self).__init__(db_conn, hs) + super(ClientIpStore, self).__init__(database, db_conn, hs) self.user_ips_max_age = hs.config.user_ips_max_age @@ -410,18 +412,18 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): def _update_client_ips_batch(self): # If the DB pool has already terminated, don't try updating - if not self.hs.get_db_pool().running: + if not self.db.is_running(): return to_update = self._batch_row_update self._batch_row_update = {} - return self.runInteraction( + return self.db.runInteraction( "_update_client_ips_batch", self._update_client_ips_batch_txn, to_update ) def _update_client_ips_batch_txn(self, txn, to_update): - if "user_ips" in self._unsafe_to_upsert_tables or ( + if "user_ips" in self.db._unsafe_to_upsert_tables or ( not self.database_engine.can_native_upsert ): self.database_engine.lock_table(txn, "user_ips") @@ -430,7 +432,7 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): (user_id, access_token, ip), (user_agent, device_id, last_seen) = entry try: - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="user_ips", keyvalues={ @@ -449,16 +451,18 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): # Technically an access token might not be associated with # a device so we need to check. if device_id: - self._simple_upsert_txn( + # this is always an update rather than an upsert: the row should + # already exist, and if it doesn't, that may be because it has been + # deleted, and we don't want to re-create it. + self.db.simple_update_txn( txn, table="devices", keyvalues={"user_id": user_id, "device_id": device_id}, - values={ + updatevalues={ "user_agent": user_agent, "last_seen": last_seen, "ip": ip, }, - lock=False, ) except Exception as e: # Failed to upsert, log and continue @@ -482,7 +486,7 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): if device_id is not None: keyvalues["device_id"] = device_id - res = yield self._simple_select_list( + res = yield self.db.simple_select_list( table="devices", keyvalues=keyvalues, retcols=("user_id", "ip", "user_agent", "device_id", "last_seen"), @@ -515,7 +519,7 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): user_agent, _, last_seen = self._batch_row_update[key] results[(access_token, ip)] = (user_agent, last_seen) - rows = yield self._simple_select_list( + rows = yield self.db.simple_select_list( table="user_ips", keyvalues={"user_id": user_id}, retcols=["access_token", "ip", "user_agent", "last_seen"], @@ -545,7 +549,9 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): # Nothing to do return - if not await self.has_completed_background_update("devices_last_seen"): + if not await self.db.updates.has_completed_background_update( + "devices_last_seen" + ): # Only start pruning if we have finished populating the devices # last seen info. return @@ -576,4 +582,4 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): def _prune_old_user_ips_txn(txn): txn.execute(sql, (timestamp,)) - await self.runInteraction("_prune_old_user_ips", _prune_old_user_ips_txn) + await self.db.runInteraction("_prune_old_user_ips", _prune_old_user_ips_txn) diff --git a/synapse/storage/data_stores/main/deviceinbox.py b/synapse/storage/data_stores/main/deviceinbox.py index 96cd0fb77a..0613b49f4a 100644 --- a/synapse/storage/data_stores/main/deviceinbox.py +++ b/synapse/storage/data_stores/main/deviceinbox.py @@ -21,7 +21,7 @@ from twisted.internet import defer from synapse.logging.opentracing import log_kv, set_tag, trace from synapse.storage._base import SQLBaseStore, make_in_list_sql_clause -from synapse.storage.background_updates import BackgroundUpdateStore +from synapse.storage.database import Database from synapse.util.caches.expiringcache import ExpiringCache logger = logging.getLogger(__name__) @@ -69,7 +69,7 @@ class DeviceInboxWorkerStore(SQLBaseStore): stream_pos = current_stream_id return messages, stream_pos - return self.runInteraction( + return self.db.runInteraction( "get_new_messages_for_device", get_new_messages_for_device_txn ) @@ -109,7 +109,7 @@ class DeviceInboxWorkerStore(SQLBaseStore): txn.execute(sql, (user_id, device_id, up_to_stream_id)) return txn.rowcount - count = yield self.runInteraction( + count = yield self.db.runInteraction( "delete_messages_for_device", delete_messages_for_device_txn ) @@ -178,7 +178,7 @@ class DeviceInboxWorkerStore(SQLBaseStore): stream_pos = current_stream_id return messages, stream_pos - return self.runInteraction( + return self.db.runInteraction( "get_new_device_msgs_for_remote", get_new_messages_for_remote_destination_txn, ) @@ -203,25 +203,25 @@ class DeviceInboxWorkerStore(SQLBaseStore): ) txn.execute(sql, (destination, up_to_stream_id)) - return self.runInteraction( + return self.db.runInteraction( "delete_device_msgs_for_remote", delete_messages_for_remote_destination_txn ) -class DeviceInboxBackgroundUpdateStore(BackgroundUpdateStore): +class DeviceInboxBackgroundUpdateStore(SQLBaseStore): DEVICE_INBOX_STREAM_ID = "device_inbox_stream_drop" - def __init__(self, db_conn, hs): - super(DeviceInboxBackgroundUpdateStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(DeviceInboxBackgroundUpdateStore, self).__init__(database, db_conn, hs) - self.register_background_index_update( + self.db.updates.register_background_index_update( "device_inbox_stream_index", index_name="device_inbox_stream_id_user_id", table="device_inbox", columns=["stream_id", "user_id"], ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( self.DEVICE_INBOX_STREAM_ID, self._background_drop_index_device_inbox ) @@ -232,9 +232,9 @@ class DeviceInboxBackgroundUpdateStore(BackgroundUpdateStore): txn.execute("DROP INDEX IF EXISTS device_inbox_stream_id") txn.close() - yield self.runWithConnection(reindex_txn) + yield self.db.runWithConnection(reindex_txn) - yield self._end_background_update(self.DEVICE_INBOX_STREAM_ID) + yield self.db.updates._end_background_update(self.DEVICE_INBOX_STREAM_ID) return 1 @@ -242,8 +242,8 @@ class DeviceInboxBackgroundUpdateStore(BackgroundUpdateStore): class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore): DEVICE_INBOX_STREAM_ID = "device_inbox_stream_drop" - def __init__(self, db_conn, hs): - super(DeviceInboxStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(DeviceInboxStore, self).__init__(database, db_conn, hs) # Map of (user_id, device_id) to the last stream_id that has been # deleted up to. This is so that we can no op deletions. @@ -294,7 +294,7 @@ class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore) with self._device_inbox_id_gen.get_next() as stream_id: now_ms = self.clock.time_msec() - yield self.runInteraction( + yield self.db.runInteraction( "add_messages_to_device_inbox", add_messages_txn, now_ms, stream_id ) for user_id in local_messages_by_user_then_device.keys(): @@ -314,7 +314,7 @@ class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore) # Check if we've already inserted a matching message_id for that # origin. This can happen if the origin doesn't receive our # acknowledgement from the first time we received the message. - already_inserted = self._simple_select_one_txn( + already_inserted = self.db.simple_select_one_txn( txn, table="device_federation_inbox", keyvalues={"origin": origin, "message_id": message_id}, @@ -326,7 +326,7 @@ class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore) # Add an entry for this message_id so that we know we've processed # it. - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="device_federation_inbox", values={ @@ -344,7 +344,7 @@ class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore) with self._device_inbox_id_gen.get_next() as stream_id: now_ms = self.clock.time_msec() - yield self.runInteraction( + yield self.db.runInteraction( "add_messages_from_remote_to_device_inbox", add_messages_txn, now_ms, @@ -358,21 +358,8 @@ class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore) def _add_messages_to_local_device_inbox_txn( self, txn, stream_id, messages_by_user_then_device ): - # Compatible method of performing an upsert - sql = "SELECT stream_id FROM device_max_stream_id" - - txn.execute(sql) - rows = txn.fetchone() - if rows: - db_stream_id = rows[0] - if db_stream_id < stream_id: - # Insert the new stream_id - sql = "UPDATE device_max_stream_id SET stream_id = ?" - else: - # No rows, perform an insert - sql = "INSERT INTO device_max_stream_id (stream_id) VALUES (?)" - - txn.execute(sql, (stream_id,)) + sql = "UPDATE device_max_stream_id" " SET stream_id = ?" " WHERE stream_id < ?" + txn.execute(sql, (stream_id, stream_id)) local_by_user_then_device = {} for user_id, messages_by_device in messages_by_user_then_device.items(): @@ -380,7 +367,7 @@ class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore) devices = list(messages_by_device.keys()) if len(devices) == 1 and devices[0] == "*": # Handle wildcard device_ids. - sql = "SELECT device_id FROM devices" " WHERE user_id = ?" + sql = "SELECT device_id FROM devices WHERE user_id = ?" txn.execute(sql, (user_id,)) message_json = json.dumps(messages_by_device["*"]) for row in txn: @@ -465,6 +452,6 @@ class DeviceInboxStore(DeviceInboxWorkerStore, DeviceInboxBackgroundUpdateStore) return rows - return self.runInteraction( + return self.db.runInteraction( "get_all_new_device_messages", get_all_new_device_messages_txn ) diff --git a/synapse/storage/data_stores/main/devices.py b/synapse/storage/data_stores/main/devices.py index 71f62036c0..9a828231c4 100644 --- a/synapse/storage/data_stores/main/devices.py +++ b/synapse/storage/data_stores/main/devices.py @@ -30,16 +30,16 @@ from synapse.logging.opentracing import ( whitelisted_homeserver, ) from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.storage._base import ( - Cache, - SQLBaseStore, - db_to_json, - make_in_list_sql_clause, -) -from synapse.storage.background_updates import BackgroundUpdateStore +from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause +from synapse.storage.database import Database from synapse.types import get_verify_key_from_cross_signing_key from synapse.util import batch_iter -from synapse.util.caches.descriptors import cached, cachedInlineCallbacks, cachedList +from synapse.util.caches.descriptors import ( + Cache, + cached, + cachedInlineCallbacks, + cachedList, +) logger = logging.getLogger(__name__) @@ -61,7 +61,7 @@ class DeviceWorkerStore(SQLBaseStore): Raises: StoreError: if the device is not found """ - return self._simple_select_one( + return self.db.simple_select_one( table="devices", keyvalues={"user_id": user_id, "device_id": device_id, "hidden": False}, retcols=("user_id", "device_id", "display_name"), @@ -80,7 +80,7 @@ class DeviceWorkerStore(SQLBaseStore): containing "device_id", "user_id" and "display_name" for each device. """ - devices = yield self._simple_select_list( + devices = yield self.db.simple_select_list( table="devices", keyvalues={"user_id": user_id, "hidden": False}, retcols=("user_id", "device_id", "display_name"), @@ -122,7 +122,7 @@ class DeviceWorkerStore(SQLBaseStore): # consider the device update to be too large, and simply skip the # stream_id; the rationale being that such a large device list update # is likely an error. - updates = yield self.runInteraction( + updates = yield self.db.runInteraction( "get_device_updates_by_remote", self._get_device_updates_by_remote_txn, destination, @@ -283,7 +283,7 @@ class DeviceWorkerStore(SQLBaseStore): """ devices = ( - yield self.runInteraction( + yield self.db.runInteraction( "_get_e2e_device_keys_txn", self._get_e2e_device_keys_txn, query_map.keys(), @@ -340,12 +340,12 @@ class DeviceWorkerStore(SQLBaseStore): rows = txn.fetchall() return rows[0][0] - return self.runInteraction("get_last_device_update_for_remote_user", f) + return self.db.runInteraction("get_last_device_update_for_remote_user", f) def mark_as_sent_devices_by_remote(self, destination, stream_id): """Mark that updates have successfully been sent to the destination. """ - return self.runInteraction( + return self.db.runInteraction( "mark_as_sent_devices_by_remote", self._mark_as_sent_devices_by_remote_txn, destination, @@ -399,7 +399,7 @@ class DeviceWorkerStore(SQLBaseStore): """ with self._device_list_id_gen.get_next() as stream_id: - yield self.runInteraction( + yield self.db.runInteraction( "add_user_sig_change_to_streams", self._add_user_signature_change_txn, from_user_id, @@ -414,7 +414,7 @@ class DeviceWorkerStore(SQLBaseStore): from_user_id, stream_id, ) - self._simple_insert_txn( + self.db.simple_insert_txn( txn, "user_signature_stream", values={ @@ -466,7 +466,7 @@ class DeviceWorkerStore(SQLBaseStore): @cachedInlineCallbacks(num_args=2, tree=True) def _get_cached_user_device(self, user_id, device_id): - content = yield self._simple_select_one_onecol( + content = yield self.db.simple_select_one_onecol( table="device_lists_remote_cache", keyvalues={"user_id": user_id, "device_id": device_id}, retcol="content", @@ -476,7 +476,7 @@ class DeviceWorkerStore(SQLBaseStore): @cachedInlineCallbacks() def _get_cached_devices_for_user(self, user_id): - devices = yield self._simple_select_list( + devices = yield self.db.simple_select_list( table="device_lists_remote_cache", keyvalues={"user_id": user_id}, retcols=("device_id", "content"), @@ -492,7 +492,7 @@ class DeviceWorkerStore(SQLBaseStore): Returns: (stream_id, devices) """ - return self.runInteraction( + return self.db.runInteraction( "get_devices_with_keys_by_user", self._get_devices_with_keys_by_user_txn, user_id, @@ -565,7 +565,7 @@ class DeviceWorkerStore(SQLBaseStore): return changes - return self.runInteraction( + return self.db.runInteraction( "get_users_whose_devices_changed", _get_users_whose_devices_changed_txn ) @@ -584,7 +584,7 @@ class DeviceWorkerStore(SQLBaseStore): SELECT DISTINCT user_ids FROM user_signature_stream WHERE from_user_id = ? AND stream_id > ? """ - rows = yield self._execute( + rows = yield self.db.execute( "get_users_whose_signatures_changed", None, sql, user_id, from_key ) return set(user for row in rows for user in json.loads(row[0])) @@ -605,7 +605,7 @@ class DeviceWorkerStore(SQLBaseStore): WHERE ? < stream_id AND stream_id <= ? GROUP BY user_id, destination """ - return self._execute( + return self.db.execute( "get_all_device_list_changes_for_remotes", None, sql, from_key, to_key ) @@ -614,7 +614,7 @@ class DeviceWorkerStore(SQLBaseStore): """Get the last stream_id we got for a user. May be None if we haven't got any information for them. """ - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="device_lists_remote_extremeties", keyvalues={"user_id": user_id}, retcol="stream_id", @@ -628,7 +628,7 @@ class DeviceWorkerStore(SQLBaseStore): inlineCallbacks=True, ) def get_device_list_last_stream_id_for_remotes(self, user_ids): - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="device_lists_remote_extremeties", column="user_id", iterable=user_ids, @@ -642,11 +642,11 @@ class DeviceWorkerStore(SQLBaseStore): return results -class DeviceBackgroundUpdateStore(BackgroundUpdateStore): - def __init__(self, db_conn, hs): - super(DeviceBackgroundUpdateStore, self).__init__(db_conn, hs) +class DeviceBackgroundUpdateStore(SQLBaseStore): + def __init__(self, database: Database, db_conn, hs): + super(DeviceBackgroundUpdateStore, self).__init__(database, db_conn, hs) - self.register_background_index_update( + self.db.updates.register_background_index_update( "device_lists_stream_idx", index_name="device_lists_stream_user_id", table="device_lists_stream", @@ -654,7 +654,7 @@ class DeviceBackgroundUpdateStore(BackgroundUpdateStore): ) # create a unique index on device_lists_remote_cache - self.register_background_index_update( + self.db.updates.register_background_index_update( "device_lists_remote_cache_unique_idx", index_name="device_lists_remote_cache_unique_id", table="device_lists_remote_cache", @@ -663,7 +663,7 @@ class DeviceBackgroundUpdateStore(BackgroundUpdateStore): ) # And one on device_lists_remote_extremeties - self.register_background_index_update( + self.db.updates.register_background_index_update( "device_lists_remote_extremeties_unique_idx", index_name="device_lists_remote_extremeties_unique_idx", table="device_lists_remote_extremeties", @@ -672,7 +672,7 @@ class DeviceBackgroundUpdateStore(BackgroundUpdateStore): ) # once they complete, we can remove the old non-unique indexes. - self.register_background_update_handler( + self.db.updates.register_background_update_handler( DROP_DEVICE_LIST_STREAMS_NON_UNIQUE_INDEXES, self._drop_device_list_streams_non_unique_indexes, ) @@ -685,14 +685,16 @@ class DeviceBackgroundUpdateStore(BackgroundUpdateStore): txn.execute("DROP INDEX IF EXISTS device_lists_remote_extremeties_id") txn.close() - yield self.runWithConnection(f) - yield self._end_background_update(DROP_DEVICE_LIST_STREAMS_NON_UNIQUE_INDEXES) + yield self.db.runWithConnection(f) + yield self.db.updates._end_background_update( + DROP_DEVICE_LIST_STREAMS_NON_UNIQUE_INDEXES + ) return 1 class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): - def __init__(self, db_conn, hs): - super(DeviceStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(DeviceStore, self).__init__(database, db_conn, hs) # Map of (user_id, device_id) -> bool. If there is an entry that implies # the device exists. @@ -722,7 +724,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): return False try: - inserted = yield self._simple_insert( + inserted = yield self.db.simple_insert( "devices", values={ "user_id": user_id, @@ -736,7 +738,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): if not inserted: # if the device already exists, check if it's a real device, or # if the device ID is reserved by something else - hidden = yield self._simple_select_one_onecol( + hidden = yield self.db.simple_select_one_onecol( "devices", keyvalues={"user_id": user_id, "device_id": device_id}, retcol="hidden", @@ -771,7 +773,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): Returns: defer.Deferred """ - yield self._simple_delete_one( + yield self.db.simple_delete_one( table="devices", keyvalues={"user_id": user_id, "device_id": device_id, "hidden": False}, desc="delete_device", @@ -789,7 +791,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): Returns: defer.Deferred """ - yield self._simple_delete_many( + yield self.db.simple_delete_many( table="devices", column="device_id", iterable=device_ids, @@ -818,7 +820,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): updates["display_name"] = new_display_name if not updates: return defer.succeed(None) - return self._simple_update_one( + return self.db.simple_update_one( table="devices", keyvalues={"user_id": user_id, "device_id": device_id, "hidden": False}, updatevalues=updates, @@ -829,7 +831,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): def mark_remote_user_device_list_as_unsubscribed(self, user_id): """Mark that we no longer track device lists for remote user. """ - yield self._simple_delete( + yield self.db.simple_delete( table="device_lists_remote_extremeties", keyvalues={"user_id": user_id}, desc="mark_remote_user_device_list_as_unsubscribed", @@ -853,7 +855,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): Returns: Deferred[None] """ - return self.runInteraction( + return self.db.runInteraction( "update_remote_device_list_cache_entry", self._update_remote_device_list_cache_entry_txn, user_id, @@ -866,7 +868,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): self, txn, user_id, device_id, content, stream_id ): if content.get("deleted"): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="device_lists_remote_cache", keyvalues={"user_id": user_id, "device_id": device_id}, @@ -874,7 +876,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): txn.call_after(self.device_id_exists_cache.invalidate, (user_id, device_id)) else: - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="device_lists_remote_cache", keyvalues={"user_id": user_id, "device_id": device_id}, @@ -890,7 +892,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): self.get_device_list_last_stream_id_for_remote.invalidate, (user_id,) ) - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="device_lists_remote_extremeties", keyvalues={"user_id": user_id}, @@ -914,7 +916,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): Returns: Deferred[None] """ - return self.runInteraction( + return self.db.runInteraction( "update_remote_device_list_cache", self._update_remote_device_list_cache_txn, user_id, @@ -923,11 +925,11 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): ) def _update_remote_device_list_cache_txn(self, txn, user_id, devices, stream_id): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="device_lists_remote_cache", keyvalues={"user_id": user_id} ) - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="device_lists_remote_cache", values=[ @@ -946,7 +948,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): self.get_device_list_last_stream_id_for_remote.invalidate, (user_id,) ) - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="device_lists_remote_extremeties", keyvalues={"user_id": user_id}, @@ -962,7 +964,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): (if any) should be poked. """ with self._device_list_id_gen.get_next() as stream_id: - yield self.runInteraction( + yield self.db.runInteraction( "add_device_change_to_streams", self._add_device_change_txn, user_id, @@ -995,7 +997,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): [(user_id, device_id, stream_id) for device_id in device_ids], ) - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="device_lists_stream", values=[ @@ -1006,7 +1008,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): context = get_active_span_text_map() - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="device_lists_outbound_pokes", values=[ @@ -1069,7 +1071,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): return run_as_background_process( "prune_old_outbound_device_pokes", - self.runInteraction, + self.db.runInteraction, "_prune_old_outbound_device_pokes", _prune_txn, ) diff --git a/synapse/storage/data_stores/main/directory.py b/synapse/storage/data_stores/main/directory.py index 297966d9f4..c9e7de7d12 100644 --- a/synapse/storage/data_stores/main/directory.py +++ b/synapse/storage/data_stores/main/directory.py @@ -36,7 +36,7 @@ class DirectoryWorkerStore(SQLBaseStore): Deferred: results in namedtuple with keys "room_id" and "servers" or None if no association can be found """ - room_id = yield self._simple_select_one_onecol( + room_id = yield self.db.simple_select_one_onecol( "room_aliases", {"room_alias": room_alias.to_string()}, "room_id", @@ -47,7 +47,7 @@ class DirectoryWorkerStore(SQLBaseStore): if not room_id: return None - servers = yield self._simple_select_onecol( + servers = yield self.db.simple_select_onecol( "room_alias_servers", {"room_alias": room_alias.to_string()}, "server", @@ -60,7 +60,7 @@ class DirectoryWorkerStore(SQLBaseStore): return RoomAliasMapping(room_id, room_alias.to_string(), servers) def get_room_alias_creator(self, room_alias): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="room_aliases", keyvalues={"room_alias": room_alias}, retcol="creator", @@ -69,7 +69,7 @@ class DirectoryWorkerStore(SQLBaseStore): @cached(max_entries=5000) def get_aliases_for_room(self, room_id): - return self._simple_select_onecol( + return self.db.simple_select_onecol( "room_aliases", {"room_id": room_id}, "room_alias", @@ -93,7 +93,7 @@ class DirectoryStore(DirectoryWorkerStore): """ def alias_txn(txn): - self._simple_insert_txn( + self.db.simple_insert_txn( txn, "room_aliases", { @@ -103,7 +103,7 @@ class DirectoryStore(DirectoryWorkerStore): }, ) - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="room_alias_servers", values=[ @@ -117,7 +117,9 @@ class DirectoryStore(DirectoryWorkerStore): ) try: - ret = yield self.runInteraction("create_room_alias_association", alias_txn) + ret = yield self.db.runInteraction( + "create_room_alias_association", alias_txn + ) except self.database_engine.module.IntegrityError: raise SynapseError( 409, "Room alias %s already exists" % room_alias.to_string() @@ -126,7 +128,7 @@ class DirectoryStore(DirectoryWorkerStore): @defer.inlineCallbacks def delete_room_alias(self, room_alias): - room_id = yield self.runInteraction( + room_id = yield self.db.runInteraction( "delete_room_alias", self._delete_room_alias_txn, room_alias ) @@ -168,6 +170,6 @@ class DirectoryStore(DirectoryWorkerStore): txn, self.get_aliases_for_room, (new_room_id,) ) - return self.runInteraction( + return self.db.runInteraction( "_update_aliases_for_room_txn", _update_aliases_for_room_txn ) diff --git a/synapse/storage/data_stores/main/e2e_room_keys.py b/synapse/storage/data_stores/main/e2e_room_keys.py index 1cbbae5b63..84594cf0a9 100644 --- a/synapse/storage/data_stores/main/e2e_room_keys.py +++ b/synapse/storage/data_stores/main/e2e_room_keys.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2017 New Vector Ltd +# Copyright 2019 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. @@ -24,49 +25,8 @@ from synapse.storage._base import SQLBaseStore class EndToEndRoomKeyStore(SQLBaseStore): @defer.inlineCallbacks - def get_e2e_room_key(self, user_id, version, room_id, session_id): - """Get the encrypted E2E room key for a given session from a given - backup version of room_keys. We only store the 'best' room key for a given - session at a given time, as determined by the handler. - - Args: - user_id(str): the user whose backup we're querying - version(str): the version ID of the backup for the set of keys we're querying - room_id(str): the ID of the room whose keys we're querying. - This is a bit redundant as it's implied by the session_id, but - we include for consistency with the rest of the API. - session_id(str): the session whose room_key we're querying. - - Returns: - A deferred dict giving the session_data and message metadata for - this room key. - """ - - row = yield self._simple_select_one( - table="e2e_room_keys", - keyvalues={ - "user_id": user_id, - "version": version, - "room_id": room_id, - "session_id": session_id, - }, - retcols=( - "first_message_index", - "forwarded_count", - "is_verified", - "session_data", - ), - desc="get_e2e_room_key", - ) - - row["session_data"] = json.loads(row["session_data"]) - - return row - - @defer.inlineCallbacks - def set_e2e_room_key(self, user_id, version, room_id, session_id, room_key): - """Replaces or inserts the encrypted E2E room key for a given session in - a given backup + def update_e2e_room_key(self, user_id, version, room_id, session_id, room_key): + """Replaces the encrypted E2E room key for a given session in a given backup Args: user_id(str): the user whose backup we're setting @@ -78,7 +38,7 @@ class EndToEndRoomKeyStore(SQLBaseStore): StoreError """ - yield self._simple_upsert( + yield self.db.simple_update_one( table="e2e_room_keys", keyvalues={ "user_id": user_id, @@ -86,21 +46,51 @@ class EndToEndRoomKeyStore(SQLBaseStore): "room_id": room_id, "session_id": session_id, }, - values={ + updatevalues={ "first_message_index": room_key["first_message_index"], "forwarded_count": room_key["forwarded_count"], "is_verified": room_key["is_verified"], "session_data": json.dumps(room_key["session_data"]), }, - lock=False, + desc="update_e2e_room_key", ) - log_kv( - { - "message": "Set room key", - "room_id": room_id, - "session_id": session_id, - "room_key": room_key, - } + + @defer.inlineCallbacks + def add_e2e_room_keys(self, user_id, version, room_keys): + """Bulk add room keys to a given backup. + + Args: + user_id (str): the user whose backup we're adding to + version (str): the version ID of the backup for the set of keys we're adding to + room_keys (iterable[(str, str, dict)]): the keys to add, in the form + (roomID, sessionID, keyData) + """ + + values = [] + for (room_id, session_id, room_key) in room_keys: + values.append( + { + "user_id": user_id, + "version": version, + "room_id": room_id, + "session_id": session_id, + "first_message_index": room_key["first_message_index"], + "forwarded_count": room_key["forwarded_count"], + "is_verified": room_key["is_verified"], + "session_data": json.dumps(room_key["session_data"]), + } + ) + log_kv( + { + "message": "Set room key", + "room_id": room_id, + "session_id": session_id, + "room_key": room_key, + } + ) + + yield self.db.simple_insert_many( + table="e2e_room_keys", values=values, desc="add_e2e_room_keys" ) @trace @@ -110,11 +100,11 @@ class EndToEndRoomKeyStore(SQLBaseStore): room, or a given session. Args: - user_id(str): the user whose backup we're querying - version(str): the version ID of the backup for the set of keys we're querying - room_id(str): Optional. the ID of the room whose keys we're querying, if any. + user_id (str): the user whose backup we're querying + version (str): the version ID of the backup for the set of keys we're querying + room_id (str): Optional. the ID of the room whose keys we're querying, if any. If not specified, we return the keys for all the rooms in the backup. - session_id(str): Optional. the session whose room_key we're querying, if any. + session_id (str): Optional. the session whose room_key we're querying, if any. If specified, we also require the room_id to be specified. If not specified, we return all the keys in this version of the backup (or for the specified room) @@ -135,7 +125,7 @@ class EndToEndRoomKeyStore(SQLBaseStore): if session_id: keyvalues["session_id"] = session_id - rows = yield self._simple_select_list( + rows = yield self.db.simple_select_list( table="e2e_room_keys", keyvalues=keyvalues, retcols=( @@ -162,6 +152,95 @@ class EndToEndRoomKeyStore(SQLBaseStore): return sessions + def get_e2e_room_keys_multi(self, user_id, version, room_keys): + """Get multiple room keys at a time. The difference between this function and + get_e2e_room_keys is that this function can be used to retrieve + multiple specific keys at a time, whereas get_e2e_room_keys is used for + getting all the keys in a backup version, all the keys for a room, or a + specific key. + + Args: + user_id (str): the user whose backup we're querying + version (str): the version ID of the backup we're querying about + room_keys (dict[str, dict[str, iterable[str]]]): a map from + room ID -> {"session": [session ids]} indicating the session IDs + that we want to query + + Returns: + Deferred[dict[str, dict[str, dict]]]: a map of room IDs to session IDs to room key + """ + + return self.db.runInteraction( + "get_e2e_room_keys_multi", + self._get_e2e_room_keys_multi_txn, + user_id, + version, + room_keys, + ) + + @staticmethod + def _get_e2e_room_keys_multi_txn(txn, user_id, version, room_keys): + if not room_keys: + return {} + + where_clauses = [] + params = [user_id, version] + for room_id, room in room_keys.items(): + sessions = list(room["sessions"]) + if not sessions: + continue + params.append(room_id) + params.extend(sessions) + where_clauses.append( + "(room_id = ? AND session_id IN (%s))" + % (",".join(["?" for _ in sessions]),) + ) + + # check if we're actually querying something + if not where_clauses: + return {} + + sql = """ + SELECT room_id, session_id, first_message_index, forwarded_count, + is_verified, session_data + FROM e2e_room_keys + WHERE user_id = ? AND version = ? AND (%s) + """ % ( + " OR ".join(where_clauses) + ) + + txn.execute(sql, params) + + ret = {} + + for row in txn: + room_id = row[0] + session_id = row[1] + ret.setdefault(room_id, {}) + ret[room_id][session_id] = { + "first_message_index": row[2], + "forwarded_count": row[3], + "is_verified": row[4], + "session_data": json.loads(row[5]), + } + + return ret + + def count_e2e_room_keys(self, user_id, version): + """Get the number of keys in a backup version. + + Args: + user_id (str): the user whose backup we're querying + version (str): the version ID of the backup we're querying about + """ + + return self.db.simple_select_one_onecol( + table="e2e_room_keys", + keyvalues={"user_id": user_id, "version": version}, + retcol="COUNT(*)", + desc="count_e2e_room_keys", + ) + @trace @defer.inlineCallbacks def delete_e2e_room_keys(self, user_id, version, room_id=None, session_id=None): @@ -188,7 +267,7 @@ class EndToEndRoomKeyStore(SQLBaseStore): if session_id: keyvalues["session_id"] = session_id - yield self._simple_delete( + yield self.db.simple_delete( table="e2e_room_keys", keyvalues=keyvalues, desc="delete_e2e_room_keys" ) @@ -219,6 +298,7 @@ class EndToEndRoomKeyStore(SQLBaseStore): version(str) algorithm(str) auth_data(object): opaque dict supplied by the client + etag(int): tag of the keys in the backup """ def _get_e2e_room_keys_version_info_txn(txn): @@ -232,17 +312,19 @@ class EndToEndRoomKeyStore(SQLBaseStore): # it isn't there. raise StoreError(404, "No row found") - result = self._simple_select_one_txn( + result = self.db.simple_select_one_txn( txn, table="e2e_room_keys_versions", keyvalues={"user_id": user_id, "version": this_version, "deleted": 0}, - retcols=("version", "algorithm", "auth_data"), + retcols=("version", "algorithm", "auth_data", "etag"), ) result["auth_data"] = json.loads(result["auth_data"]) result["version"] = str(result["version"]) + if result["etag"] is None: + result["etag"] = 0 return result - return self.runInteraction( + return self.db.runInteraction( "get_e2e_room_keys_version_info", _get_e2e_room_keys_version_info_txn ) @@ -270,7 +352,7 @@ class EndToEndRoomKeyStore(SQLBaseStore): new_version = str(int(current_version) + 1) - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="e2e_room_keys_versions", values={ @@ -283,26 +365,38 @@ class EndToEndRoomKeyStore(SQLBaseStore): return new_version - return self.runInteraction( + return self.db.runInteraction( "create_e2e_room_keys_version_txn", _create_e2e_room_keys_version_txn ) @trace - def update_e2e_room_keys_version(self, user_id, version, info): + def update_e2e_room_keys_version( + self, user_id, version, info=None, version_etag=None + ): """Update a given backup version Args: user_id(str): the user whose backup version we're updating version(str): the version ID of the backup version we're updating - info(dict): the new backup version info to store + info (dict): the new backup version info to store. If None, then + the backup version info is not updated + version_etag (Optional[int]): etag of the keys in the backup. If + None, then the etag is not updated """ + updatevalues = {} - return self._simple_update( - table="e2e_room_keys_versions", - keyvalues={"user_id": user_id, "version": version}, - updatevalues={"auth_data": json.dumps(info["auth_data"])}, - desc="update_e2e_room_keys_version", - ) + if info is not None and "auth_data" in info: + updatevalues["auth_data"] = json.dumps(info["auth_data"]) + if version_etag is not None: + updatevalues["etag"] = version_etag + + if updatevalues: + return self.db.simple_update( + table="e2e_room_keys_versions", + keyvalues={"user_id": user_id, "version": version}, + updatevalues=updatevalues, + desc="update_e2e_room_keys_version", + ) @trace def delete_e2e_room_keys_version(self, user_id, version=None): @@ -326,19 +420,19 @@ class EndToEndRoomKeyStore(SQLBaseStore): else: this_version = version - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="e2e_room_keys", keyvalues={"user_id": user_id, "version": this_version}, ) - return self._simple_update_one_txn( + return self.db.simple_update_one_txn( txn, table="e2e_room_keys_versions", keyvalues={"user_id": user_id, "version": this_version}, updatevalues={"deleted": 1}, ) - return self.runInteraction( + return self.db.runInteraction( "delete_e2e_room_keys_version", _delete_e2e_room_keys_version_txn ) diff --git a/synapse/storage/data_stores/main/end_to_end_keys.py b/synapse/storage/data_stores/main/end_to_end_keys.py index 073412a78d..e551606f9d 100644 --- a/synapse/storage/data_stores/main/end_to_end_keys.py +++ b/synapse/storage/data_stores/main/end_to_end_keys.py @@ -14,15 +14,18 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +from typing import Dict, List + from six import iteritems from canonicaljson import encode_canonical_json, json +from twisted.enterprise.adbapi import Connection from twisted.internet import defer from synapse.logging.opentracing import log_kv, set_tag, trace from synapse.storage._base import SQLBaseStore, db_to_json -from synapse.util.caches.descriptors import cached +from synapse.util.caches.descriptors import cached, cachedList class EndToEndKeyWorkerStore(SQLBaseStore): @@ -48,7 +51,7 @@ class EndToEndKeyWorkerStore(SQLBaseStore): if not query_list: return {} - results = yield self.runInteraction( + results = yield self.db.runInteraction( "get_e2e_device_keys", self._get_e2e_device_keys_txn, query_list, @@ -125,7 +128,7 @@ class EndToEndKeyWorkerStore(SQLBaseStore): ) txn.execute(sql, query_params) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) result = {} for row in rows: @@ -138,20 +141,35 @@ class EndToEndKeyWorkerStore(SQLBaseStore): result.setdefault(user_id, {})[device_id] = None # get signatures on the device - signature_sql = ( - "SELECT * " " FROM e2e_cross_signing_signatures " " WHERE %s" - ) % (" OR ".join("(" + q + ")" for q in signature_query_clauses)) + signature_sql = ("SELECT * FROM e2e_cross_signing_signatures WHERE %s") % ( + " OR ".join("(" + q + ")" for q in signature_query_clauses) + ) txn.execute(signature_sql, signature_query_params) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) + # add each cross-signing signature to the correct device in the result dict. for row in rows: + signing_user_id = row["user_id"] + signing_key_id = row["key_id"] target_user_id = row["target_user_id"] target_device_id = row["target_device_id"] - if target_user_id in result and target_device_id in result[target_user_id]: - result[target_user_id][target_device_id].setdefault( - "signatures", {} - ).setdefault(row["user_id"], {})[row["key_id"]] = row["signature"] + signature = row["signature"] + + target_user_result = result.get(target_user_id) + if not target_user_result: + continue + + target_device_result = target_user_result.get(target_device_id) + if not target_device_result: + # note that target_device_result will be None for deleted devices. + continue + + target_device_signatures = target_device_result.setdefault("signatures", {}) + signing_user_signatures = target_device_signatures.setdefault( + signing_user_id, {} + ) + signing_user_signatures[signing_key_id] = signature log_kv(result) return result @@ -171,7 +189,7 @@ class EndToEndKeyWorkerStore(SQLBaseStore): key_id) to json string for key """ - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="e2e_one_time_keys_json", column="key_id", iterable=key_ids, @@ -204,7 +222,7 @@ class EndToEndKeyWorkerStore(SQLBaseStore): # a unique constraint. If there is a race of two calls to # `add_e2e_one_time_keys` then they'll conflict and we will only # insert one set. - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="e2e_one_time_keys_json", values=[ @@ -223,7 +241,7 @@ class EndToEndKeyWorkerStore(SQLBaseStore): txn, self.count_e2e_one_time_keys, (user_id, device_id) ) - yield self.runInteraction( + yield self.db.runInteraction( "add_e2e_one_time_keys_insert", _add_e2e_one_time_keys ) @@ -246,7 +264,9 @@ class EndToEndKeyWorkerStore(SQLBaseStore): result[algorithm] = key_count return result - return self.runInteraction("count_e2e_one_time_keys", _count_e2e_one_time_keys) + return self.db.runInteraction( + "count_e2e_one_time_keys", _count_e2e_one_time_keys + ) def _get_e2e_cross_signing_key_txn(self, txn, user_id, key_type, from_user_id=None): """Returns a user's cross-signing key. @@ -254,7 +274,7 @@ class EndToEndKeyWorkerStore(SQLBaseStore): Args: txn (twisted.enterprise.adbapi.Connection): db connection user_id (str): the user whose key is being requested - key_type (str): the type of key that is being set: either 'master' + key_type (str): the type of key that is being requested: either 'master' for a master key, 'self_signing' for a self-signing key, or 'user_signing' for a user-signing key from_user_id (str): if specified, signatures made by this user on @@ -299,15 +319,17 @@ class EndToEndKeyWorkerStore(SQLBaseStore): """Returns a user's cross-signing key. Args: - user_id (str): the user whose self-signing key is being requested - key_type (str): the type of cross-signing key to get + user_id (str): the user whose key is being requested + key_type (str): the type of key that is being requested: either 'master' + for a master key, 'self_signing' for a self-signing key, or + 'user_signing' for a user-signing key from_user_id (str): if specified, signatures made by this user on the self-signing key will be included in the result Returns: dict of the key data or None if not found """ - return self.runInteraction( + return self.db.runInteraction( "get_e2e_cross_signing_key", self._get_e2e_cross_signing_key_txn, user_id, @@ -315,6 +337,206 @@ class EndToEndKeyWorkerStore(SQLBaseStore): from_user_id, ) + @cached(num_args=1) + def _get_bare_e2e_cross_signing_keys(self, user_id): + """Dummy function. Only used to make a cache for + _get_bare_e2e_cross_signing_keys_bulk. + """ + raise NotImplementedError() + + @cachedList( + cached_method_name="_get_bare_e2e_cross_signing_keys", + list_name="user_ids", + num_args=1, + ) + def _get_bare_e2e_cross_signing_keys_bulk( + self, user_ids: List[str] + ) -> Dict[str, Dict[str, dict]]: + """Returns the cross-signing keys for a set of users. The output of this + function should be passed to _get_e2e_cross_signing_signatures_txn if + the signatures for the calling user need to be fetched. + + Args: + user_ids (list[str]): the users whose keys are being requested + + Returns: + dict[str, dict[str, dict]]: mapping from user ID to key type to key + data. If a user's cross-signing keys were not found, either + their user ID will not be in the dict, or their user ID will map + to None. + + """ + return self.db.runInteraction( + "get_bare_e2e_cross_signing_keys_bulk", + self._get_bare_e2e_cross_signing_keys_bulk_txn, + user_ids, + ) + + def _get_bare_e2e_cross_signing_keys_bulk_txn( + self, txn: Connection, user_ids: List[str], + ) -> Dict[str, Dict[str, dict]]: + """Returns the cross-signing keys for a set of users. The output of this + function should be passed to _get_e2e_cross_signing_signatures_txn if + the signatures for the calling user need to be fetched. + + Args: + txn (twisted.enterprise.adbapi.Connection): db connection + user_ids (list[str]): the users whose keys are being requested + + Returns: + dict[str, dict[str, dict]]: mapping from user ID to key type to key + data. If a user's cross-signing keys were not found, their user + ID will not be in the dict. + + """ + result = {} + + batch_size = 100 + chunks = [ + user_ids[i : i + batch_size] for i in range(0, len(user_ids), batch_size) + ] + for user_chunk in chunks: + sql = """ + SELECT k.user_id, k.keytype, k.keydata, k.stream_id + FROM e2e_cross_signing_keys k + INNER JOIN (SELECT user_id, keytype, MAX(stream_id) AS stream_id + FROM e2e_cross_signing_keys + GROUP BY user_id, keytype) s + USING (user_id, stream_id, keytype) + WHERE k.user_id IN (%s) + """ % ( + ",".join("?" for u in user_chunk), + ) + query_params = [] + query_params.extend(user_chunk) + + txn.execute(sql, query_params) + rows = self.db.cursor_to_dict(txn) + + for row in rows: + user_id = row["user_id"] + key_type = row["keytype"] + key = json.loads(row["keydata"]) + user_info = result.setdefault(user_id, {}) + user_info[key_type] = key + + return result + + def _get_e2e_cross_signing_signatures_txn( + self, txn: Connection, keys: Dict[str, Dict[str, dict]], from_user_id: str, + ) -> Dict[str, Dict[str, dict]]: + """Returns the cross-signing signatures made by a user on a set of keys. + + Args: + txn (twisted.enterprise.adbapi.Connection): db connection + keys (dict[str, dict[str, dict]]): a map of user ID to key type to + key data. This dict will be modified to add signatures. + from_user_id (str): fetch the signatures made by this user + + Returns: + dict[str, dict[str, dict]]: mapping from user ID to key type to key + data. The return value will be the same as the keys argument, + with the modifications included. + """ + + # find out what cross-signing keys (a.k.a. devices) we need to get + # signatures for. This is a map of (user_id, device_id) to key type + # (device_id is the key's public part). + devices = {} + + for user_id, user_info in keys.items(): + if user_info is None: + continue + for key_type, key in user_info.items(): + device_id = None + for k in key["keys"].values(): + device_id = k + devices[(user_id, device_id)] = key_type + + device_list = list(devices) + + # split into batches + batch_size = 100 + chunks = [ + device_list[i : i + batch_size] + for i in range(0, len(device_list), batch_size) + ] + for user_chunk in chunks: + sql = """ + SELECT target_user_id, target_device_id, key_id, signature + FROM e2e_cross_signing_signatures + WHERE user_id = ? + AND (%s) + """ % ( + " OR ".join( + "(target_user_id = ? AND target_device_id = ?)" for d in devices + ) + ) + query_params = [from_user_id] + for item in devices: + # item is a (user_id, device_id) tuple + query_params.extend(item) + + txn.execute(sql, query_params) + rows = self.db.cursor_to_dict(txn) + + # and add the signatures to the appropriate keys + for row in rows: + key_id = row["key_id"] + target_user_id = row["target_user_id"] + target_device_id = row["target_device_id"] + 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 + # need to recursively copy the dicts that will be modified. + user_info = keys[target_user_id] = keys[target_user_id].copy() + target_user_key = user_info[key_type] = user_info[key_type].copy() + if "signatures" in target_user_key: + signatures = target_user_key["signatures"] = target_user_key[ + "signatures" + ].copy() + if from_user_id in signatures: + user_sigs = signatures[from_user_id] = signatures[from_user_id] + user_sigs[key_id] = row["signature"] + else: + signatures[from_user_id] = {key_id: row["signature"]} + else: + target_user_key["signatures"] = { + from_user_id: {key_id: row["signature"]} + } + + return keys + + @defer.inlineCallbacks + def get_e2e_cross_signing_keys_bulk( + self, user_ids: List[str], from_user_id: str = None + ) -> defer.Deferred: + """Returns the cross-signing keys for a set of users. + + Args: + user_ids (list[str]): the users whose keys are being requested + from_user_id (str): if specified, signatures made by this user on + the self-signing keys will be included in the result + + Returns: + Deferred[dict[str, dict[str, dict]]]: map of user ID to key type to + key data. If a user's cross-signing keys were not found, either + their user ID will not be in the dict, or their user ID will map + to None. + """ + + result = yield self._get_bare_e2e_cross_signing_keys_bulk(user_ids) + + if from_user_id: + result = yield self.db.runInteraction( + "get_e2e_cross_signing_signatures", + self._get_e2e_cross_signing_signatures_txn, + result, + from_user_id, + ) + + return result + def get_all_user_signature_changes_for_remotes(self, from_key, to_key): """Return a list of changes from the user signature stream to notify remotes. Note that the user signature stream represents when a user signs their @@ -335,7 +557,7 @@ class EndToEndKeyWorkerStore(SQLBaseStore): WHERE ? < stream_id AND stream_id <= ? GROUP BY user_id """ - return self._execute( + return self.db.execute( "get_all_user_signature_changes_for_remotes", None, sql, from_key, to_key ) @@ -352,7 +574,7 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): set_tag("time_now", time_now) set_tag("device_keys", device_keys) - old_key_json = self._simple_select_one_onecol_txn( + old_key_json = self.db.simple_select_one_onecol_txn( txn, table="e2e_device_keys_json", keyvalues={"user_id": user_id, "device_id": device_id}, @@ -368,7 +590,7 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): log_kv({"Message": "Device key already stored."}) return False - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="e2e_device_keys_json", keyvalues={"user_id": user_id, "device_id": device_id}, @@ -377,7 +599,7 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): log_kv({"message": "Device keys stored."}) return True - return self.runInteraction("set_e2e_device_keys", _set_e2e_device_keys_txn) + return self.db.runInteraction("set_e2e_device_keys", _set_e2e_device_keys_txn) def claim_e2e_one_time_keys(self, query_list): """Take a list of one time keys out of the database""" @@ -416,7 +638,9 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): ) return result - return self.runInteraction("claim_e2e_one_time_keys", _claim_e2e_one_time_keys) + return self.db.runInteraction( + "claim_e2e_one_time_keys", _claim_e2e_one_time_keys + ) def delete_e2e_keys_by_device(self, user_id, device_id): def delete_e2e_keys_by_device_txn(txn): @@ -427,12 +651,12 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): "user_id": user_id, } ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="e2e_device_keys_json", keyvalues={"user_id": user_id, "device_id": device_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="e2e_one_time_keys_json", keyvalues={"user_id": user_id, "device_id": device_id}, @@ -441,7 +665,7 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): txn, self.count_e2e_one_time_keys, (user_id, device_id) ) - return self.runInteraction( + return self.db.runInteraction( "delete_e2e_keys_by_device", delete_e2e_keys_by_device_txn ) @@ -477,7 +701,7 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): # The "keys" property must only have one entry, which will be the public # key, so we just grab the first value in there pubkey = next(iter(key["keys"].values())) - self._simple_insert_txn( + self.db.simple_insert_txn( txn, "devices", values={ @@ -490,7 +714,7 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): # and finally, store the key itself with self._cross_signing_id_gen.get_next() as stream_id: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, "e2e_cross_signing_keys", values={ @@ -501,6 +725,10 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): }, ) + self._invalidate_cache_and_stream( + txn, self._get_bare_e2e_cross_signing_keys, (user_id,) + ) + def set_e2e_cross_signing_key(self, user_id, key_type, key): """Set a user's cross-signing key. @@ -509,7 +737,7 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): key_type (str): the type of cross-signing key to set key (dict): the key data """ - return self.runInteraction( + return self.db.runInteraction( "add_e2e_cross_signing_key", self._set_e2e_cross_signing_key_txn, user_id, @@ -524,7 +752,7 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): user_id (str): the user who made the signatures signatures (iterable[SignatureListItem]): signatures to add """ - return self._simple_insert_many( + return self.db.simple_insert_many( "e2e_cross_signing_signatures", [ { diff --git a/synapse/storage/data_stores/main/event_federation.py b/synapse/storage/data_stores/main/event_federation.py index 90bef0cd2c..60c67457b4 100644 --- a/synapse/storage/data_stores/main/event_federation.py +++ b/synapse/storage/data_stores/main/event_federation.py @@ -14,13 +14,10 @@ # limitations under the License. import itertools import logging -import random from six.moves import range from six.moves.queue import Empty, PriorityQueue -from unpaddedbase64 import encode_base64 - from twisted.internet import defer from synapse.api.errors import StoreError @@ -28,6 +25,7 @@ from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage._base import SQLBaseStore, make_in_list_sql_clause from synapse.storage.data_stores.main.events_worker import EventsWorkerStore from synapse.storage.data_stores.main.signatures import SignatureWorkerStore +from synapse.storage.database import Database from synapse.util.caches.descriptors import cached logger = logging.getLogger(__name__) @@ -58,7 +56,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas Returns: list of event_ids """ - return self.runInteraction( + return self.db.runInteraction( "get_auth_chain_ids", self._get_auth_chain_ids_txn, event_ids, include_given ) @@ -90,12 +88,12 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas return list(results) def get_oldest_events_in_room(self, room_id): - return self.runInteraction( + return self.db.runInteraction( "get_oldest_events_in_room", self._get_oldest_events_in_room_txn, room_id ) def get_oldest_events_with_depth_in_room(self, room_id): - return self.runInteraction( + return self.db.runInteraction( "get_oldest_events_with_depth_in_room", self.get_oldest_events_with_depth_in_room_txn, room_id, @@ -126,7 +124,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas Returns Deferred[int] """ - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="events", column="event_id", iterable=event_ids, @@ -140,15 +138,14 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas return max(row["depth"] for row in rows) def _get_oldest_events_in_room_txn(self, txn, room_id): - return self._simple_select_onecol_txn( + return self.db.simple_select_onecol_txn( txn, table="event_backward_extremities", keyvalues={"room_id": room_id}, retcol="event_id", ) - @defer.inlineCallbacks - def get_prev_events_for_room(self, room_id): + def get_prev_events_for_room(self, room_id: str): """ Gets a subset of the current forward extremities in the given room. @@ -159,40 +156,29 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas room_id (str): room_id Returns: - Deferred[list[(str, dict[str, str], int)]] - for each event, a tuple of (event_id, hashes, depth) - where *hashes* is a map from algorithm to hash. + Deferred[List[str]]: the event ids of the forward extremites + """ - res = yield self.get_latest_event_ids_and_hashes_in_room(room_id) - if len(res) > 10: - # Sort by reverse depth, so we point to the most recent. - res.sort(key=lambda a: -a[2]) - # we use half of the limit for the actual most recent events, and - # the other half to randomly point to some of the older events, to - # make sure that we don't completely ignore the older events. - res = res[0:5] + random.sample(res[5:], 5) + return self.db.runInteraction( + "get_prev_events_for_room", self._get_prev_events_for_room_txn, room_id + ) - return res + def _get_prev_events_for_room_txn(self, txn, room_id: str): + # we just use the 10 newest events. Older events will become + # prev_events of future events. - def get_latest_event_ids_and_hashes_in_room(self, room_id): + sql = """ + SELECT e.event_id FROM event_forward_extremities AS f + INNER JOIN events AS e USING (event_id) + WHERE f.room_id = ? + ORDER BY e.depth DESC + LIMIT 10 """ - Gets the current forward extremities in the given room - Args: - room_id (str): room_id + txn.execute(sql, (room_id,)) - Returns: - Deferred[list[(str, dict[str, str], int)]] - for each event, a tuple of (event_id, hashes, depth) - where *hashes* is a map from algorithm to hash. - """ - - return self.runInteraction( - "get_latest_event_ids_and_hashes_in_room", - self._get_latest_event_ids_and_hashes_in_room, - room_id, - ) + return [row[0] for row in txn] def get_rooms_with_many_extremities(self, min_count, limit, room_id_filter): """Get the top rooms with at least N extremities. @@ -229,49 +215,28 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas txn.execute(sql, query_args) return [room_id for room_id, in txn] - return self.runInteraction( + return self.db.runInteraction( "get_rooms_with_many_extremities", _get_rooms_with_many_extremities_txn ) @cached(max_entries=5000, iterable=True) def get_latest_event_ids_in_room(self, room_id): - return self._simple_select_onecol( + return self.db.simple_select_onecol( table="event_forward_extremities", keyvalues={"room_id": room_id}, retcol="event_id", desc="get_latest_event_ids_in_room", ) - def _get_latest_event_ids_and_hashes_in_room(self, txn, room_id): - sql = ( - "SELECT e.event_id, e.depth FROM events as e " - "INNER JOIN event_forward_extremities as f " - "ON e.event_id = f.event_id " - "AND e.room_id = f.room_id " - "WHERE f.room_id = ?" - ) - - txn.execute(sql, (room_id,)) - - results = [] - for event_id, depth in txn.fetchall(): - hashes = self._get_event_reference_hashes_txn(txn, event_id) - prev_hashes = { - k: encode_base64(v) for k, v in hashes.items() if k == "sha256" - } - results.append((event_id, prev_hashes, depth)) - - return results - def get_min_depth(self, room_id): """ For hte given room, get the minimum depth we have seen for it. """ - return self.runInteraction( + return self.db.runInteraction( "get_min_depth", self._get_min_depth_interaction, room_id ) def _get_min_depth_interaction(self, txn, room_id): - min_depth = self._simple_select_one_onecol_txn( + min_depth = self.db.simple_select_one_onecol_txn( txn, table="room_depth", keyvalues={"room_id": room_id}, @@ -337,7 +302,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas txn.execute(sql, (stream_ordering, room_id)) return [event_id for event_id, in txn] - return self.runInteraction( + return self.db.runInteraction( "get_forward_extremeties_for_room", get_forward_extremeties_for_room_txn ) @@ -352,7 +317,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas limit (int) """ return ( - self.runInteraction( + self.db.runInteraction( "get_backfill_events", self._get_backfill_events, room_id, @@ -383,7 +348,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas queue = PriorityQueue() for event_id in event_list: - depth = self._simple_select_one_onecol_txn( + depth = self.db.simple_select_one_onecol_txn( txn, table="events", keyvalues={"event_id": event_id, "room_id": room_id}, @@ -415,7 +380,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas @defer.inlineCallbacks def get_missing_events(self, room_id, earliest_events, latest_events, limit): - ids = yield self.runInteraction( + ids = yield self.db.runInteraction( "get_missing_events", self._get_missing_events, room_id, @@ -468,7 +433,7 @@ class EventFederationWorkerStore(EventsWorkerStore, SignatureWorkerStore, SQLBas Returns: Deferred[list[str]] """ - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="event_edges", column="prev_event_id", iterable=event_ids, @@ -491,10 +456,10 @@ class EventFederationStore(EventFederationWorkerStore): EVENT_AUTH_STATE_ONLY = "event_auth_state_only" - def __init__(self, db_conn, hs): - super(EventFederationStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(EventFederationStore, self).__init__(database, db_conn, hs) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( self.EVENT_AUTH_STATE_ONLY, self._background_delete_non_state_event_auth ) @@ -505,10 +470,10 @@ class EventFederationStore(EventFederationWorkerStore): def _update_min_depth_for_room_txn(self, txn, room_id, depth): min_depth = self._get_min_depth_interaction(txn, room_id) - if min_depth and depth >= min_depth: + if min_depth is not None and depth >= min_depth: return - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="room_depth", keyvalues={"room_id": room_id}, @@ -520,7 +485,7 @@ class EventFederationStore(EventFederationWorkerStore): For the given event, update the event edges table and forward and backward extremities tables. """ - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="event_edges", values=[ @@ -604,13 +569,13 @@ class EventFederationStore(EventFederationWorkerStore): return run_as_background_process( "delete_old_forward_extrem_cache", - self.runInteraction, + self.db.runInteraction, "_delete_old_forward_extrem_cache", _delete_old_forward_extrem_cache_txn, ) def clean_room_for_join(self, room_id): - return self.runInteraction( + return self.db.runInteraction( "clean_room_for_join", self._clean_room_for_join_txn, room_id ) @@ -654,17 +619,17 @@ class EventFederationStore(EventFederationWorkerStore): "max_stream_id_exclusive": min_stream_id, } - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, self.EVENT_AUTH_STATE_ONLY, new_progress ) return min_stream_id >= target_min_stream_id - result = yield self.runInteraction( + result = yield self.db.runInteraction( self.EVENT_AUTH_STATE_ONLY, delete_event_auth ) if not result: - yield self._end_background_update(self.EVENT_AUTH_STATE_ONLY) + yield self.db.updates._end_background_update(self.EVENT_AUTH_STATE_ONLY) return batch_size diff --git a/synapse/storage/data_stores/main/event_push_actions.py b/synapse/storage/data_stores/main/event_push_actions.py index 04ce21ac66..9988a6d3fc 100644 --- a/synapse/storage/data_stores/main/event_push_actions.py +++ b/synapse/storage/data_stores/main/event_push_actions.py @@ -24,6 +24,7 @@ from twisted.internet import defer from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage._base import LoggingTransaction, SQLBaseStore +from synapse.storage.database import Database from synapse.util.caches.descriptors import cachedInlineCallbacks logger = logging.getLogger(__name__) @@ -68,8 +69,8 @@ def _deserialize_action(actions, is_highlight): class EventPushActionsWorkerStore(SQLBaseStore): - def __init__(self, db_conn, hs): - super(EventPushActionsWorkerStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(EventPushActionsWorkerStore, self).__init__(database, db_conn, hs) # These get correctly set by _find_stream_orderings_for_times_txn self.stream_ordering_month_ago = None @@ -93,7 +94,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): def get_unread_event_push_actions_by_room_for_user( self, room_id, user_id, last_read_event_id ): - ret = yield self.runInteraction( + ret = yield self.db.runInteraction( "get_unread_event_push_actions_by_room", self._get_unread_counts_by_receipt_txn, room_id, @@ -177,7 +178,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): txn.execute(sql, (min_stream_ordering, max_stream_ordering)) return [r[0] for r in txn] - ret = yield self.runInteraction("get_push_action_users_in_range", f) + ret = yield self.db.runInteraction("get_push_action_users_in_range", f) return ret @defer.inlineCallbacks @@ -229,7 +230,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): txn.execute(sql, args) return txn.fetchall() - after_read_receipt = yield self.runInteraction( + after_read_receipt = yield self.db.runInteraction( "get_unread_push_actions_for_user_in_range_http_arr", get_after_receipt ) @@ -257,7 +258,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): txn.execute(sql, args) return txn.fetchall() - no_read_receipt = yield self.runInteraction( + no_read_receipt = yield self.db.runInteraction( "get_unread_push_actions_for_user_in_range_http_nrr", get_no_receipt ) @@ -329,7 +330,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): txn.execute(sql, args) return txn.fetchall() - after_read_receipt = yield self.runInteraction( + after_read_receipt = yield self.db.runInteraction( "get_unread_push_actions_for_user_in_range_email_arr", get_after_receipt ) @@ -357,7 +358,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): txn.execute(sql, args) return txn.fetchall() - no_read_receipt = yield self.runInteraction( + no_read_receipt = yield self.db.runInteraction( "get_unread_push_actions_for_user_in_range_email_nrr", get_no_receipt ) @@ -407,7 +408,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): txn.execute(sql, (user_id, min_stream_ordering)) return bool(txn.fetchone()) - return self.runInteraction( + return self.db.runInteraction( "get_if_maybe_push_in_range_for_user", _get_if_maybe_push_in_range_for_user_txn, ) @@ -441,7 +442,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): ) def _add_push_actions_to_staging_txn(txn): - # We don't use _simple_insert_many here to avoid the overhead + # We don't use simple_insert_many here to avoid the overhead # of generating lists of dicts. sql = """ @@ -458,7 +459,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): ), ) - return self.runInteraction( + return self.db.runInteraction( "add_push_actions_to_staging", _add_push_actions_to_staging_txn ) @@ -472,7 +473,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): """ try: - res = yield self._simple_delete( + res = yield self.db.simple_delete( table="event_push_actions_staging", keyvalues={"event_id": event_id}, desc="remove_push_actions_from_staging", @@ -489,7 +490,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): def _find_stream_orderings_for_times(self): return run_as_background_process( "event_push_action_stream_orderings", - self.runInteraction, + self.db.runInteraction, "_find_stream_orderings_for_times", self._find_stream_orderings_for_times_txn, ) @@ -525,7 +526,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): Deferred[int]: stream ordering of the first event received on/after the timestamp """ - return self.runInteraction( + return self.db.runInteraction( "_find_first_stream_ordering_after_ts_txn", self._find_first_stream_ordering_after_ts_txn, ts, @@ -611,17 +612,17 @@ class EventPushActionsWorkerStore(SQLBaseStore): class EventPushActionsStore(EventPushActionsWorkerStore): EPA_HIGHLIGHT_INDEX = "epa_highlight_index" - def __init__(self, db_conn, hs): - super(EventPushActionsStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(EventPushActionsStore, self).__init__(database, db_conn, hs) - self.register_background_index_update( + self.db.updates.register_background_index_update( self.EPA_HIGHLIGHT_INDEX, index_name="event_push_actions_u_highlight", table="event_push_actions", columns=["user_id", "stream_ordering"], ) - self.register_background_index_update( + self.db.updates.register_background_index_update( "event_push_actions_highlights_index", index_name="event_push_actions_highlights_index", table="event_push_actions", @@ -677,7 +678,7 @@ class EventPushActionsStore(EventPushActionsWorkerStore): ) for event, _ in events_and_contexts: - user_ids = self._simple_select_onecol_txn( + user_ids = self.db.simple_select_onecol_txn( txn, table="event_push_actions_staging", keyvalues={"event_id": event.event_id}, @@ -727,9 +728,9 @@ class EventPushActionsStore(EventPushActionsWorkerStore): " LIMIT ?" % (before_clause,) ) txn.execute(sql, args) - return self.cursor_to_dict(txn) + return self.db.cursor_to_dict(txn) - push_actions = yield self.runInteraction("get_push_actions_for_user", f) + push_actions = yield self.db.runInteraction("get_push_actions_for_user", f) for pa in push_actions: pa["actions"] = _deserialize_action(pa["actions"], pa["highlight"]) return push_actions @@ -748,7 +749,7 @@ class EventPushActionsStore(EventPushActionsWorkerStore): txn.execute(sql, (stream_ordering,)) return txn.fetchone() - result = yield self.runInteraction("get_time_of_last_push_action_before", f) + result = yield self.db.runInteraction("get_time_of_last_push_action_before", f) return result[0] if result else None @defer.inlineCallbacks @@ -757,7 +758,9 @@ class EventPushActionsStore(EventPushActionsWorkerStore): txn.execute("SELECT MAX(stream_ordering) FROM event_push_actions") return txn.fetchone() - result = yield self.runInteraction("get_latest_push_action_stream_ordering", f) + result = yield self.db.runInteraction( + "get_latest_push_action_stream_ordering", f + ) return result[0] or 0 def _remove_push_actions_for_event_id_txn(self, txn, room_id, event_id): @@ -830,7 +833,7 @@ class EventPushActionsStore(EventPushActionsWorkerStore): while True: logger.info("Rotating notifications") - caught_up = yield self.runInteraction( + caught_up = yield self.db.runInteraction( "_rotate_notifs", self._rotate_notifs_txn ) if caught_up: @@ -844,7 +847,7 @@ class EventPushActionsStore(EventPushActionsWorkerStore): the archiving process has caught up or not. """ - old_rotate_stream_ordering = self._simple_select_one_onecol_txn( + old_rotate_stream_ordering = self.db.simple_select_one_onecol_txn( txn, table="event_push_summary_stream_ordering", keyvalues={}, @@ -880,7 +883,7 @@ class EventPushActionsStore(EventPushActionsWorkerStore): return caught_up def _rotate_notifs_before_txn(self, txn, rotate_to_stream_ordering): - old_rotate_stream_ordering = self._simple_select_one_onecol_txn( + old_rotate_stream_ordering = self.db.simple_select_one_onecol_txn( txn, table="event_push_summary_stream_ordering", keyvalues={}, @@ -912,7 +915,7 @@ class EventPushActionsStore(EventPushActionsWorkerStore): # If the `old.user_id` above is NULL then we know there isn't already an # entry in the table, so we simply insert it. Otherwise we update the # existing table. - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="event_push_summary", values=[ diff --git a/synapse/storage/data_stores/main/events.py b/synapse/storage/data_stores/main/events.py index 878f7568a6..e9fe63037b 100644 --- a/synapse/storage/data_stores/main/events.py +++ b/synapse/storage/data_stores/main/events.py @@ -38,10 +38,10 @@ from synapse.logging.utils import log_function from synapse.metrics import BucketCollector from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage._base import make_in_list_sql_clause -from synapse.storage.background_updates import BackgroundUpdateStore from synapse.storage.data_stores.main.event_federation import EventFederationStore from synapse.storage.data_stores.main.events_worker import EventsWorkerStore from synapse.storage.data_stores.main.state import StateGroupWorkerStore +from synapse.storage.database import Database from synapse.types import RoomStreamToken, get_domain_from_id from synapse.util import batch_iter from synapse.util.caches.descriptors import cached, cachedInlineCallbacks @@ -94,13 +94,10 @@ def _retry_on_integrity_error(func): # inherits from EventFederationStore so that we can call _update_backward_extremities # and _handle_mult_prev_events (though arguably those could both be moved in here) class EventsStore( - StateGroupWorkerStore, - EventFederationStore, - EventsWorkerStore, - BackgroundUpdateStore, + StateGroupWorkerStore, EventFederationStore, EventsWorkerStore, ): - def __init__(self, db_conn, hs): - super(EventsStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(EventsStore, self).__init__(database, db_conn, hs) # Collect metrics on the number of forward extremities that exist. # Counter of number of extremities to count @@ -130,6 +127,9 @@ class EventsStore( if self.hs.config.redaction_retention_period is not None: hs.get_clock().looping_call(_censor_redactions, 5 * 60 * 1000) + self._ephemeral_messages_enabled = hs.config.enable_ephemeral_messages + self.is_mine_id = hs.is_mine_id + @defer.inlineCallbacks def _read_forward_extremities(self): def fetch(txn): @@ -141,7 +141,7 @@ class EventsStore( ) return txn.fetchall() - res = yield self.runInteraction("read_forward_extremities", fetch) + res = yield self.db.runInteraction("read_forward_extremities", fetch) self._current_forward_extremities_amount = c_counter(list(x[0] for x in res)) @_retry_on_integrity_error @@ -206,7 +206,7 @@ class EventsStore( for (event, context), stream in zip(events_and_contexts, stream_orderings): event.internal_metadata.stream_ordering = stream - yield self.runInteraction( + yield self.db.runInteraction( "persist_events", self._persist_events_txn, events_and_contexts=events_and_contexts, @@ -279,7 +279,7 @@ class EventsStore( results.extend(r[0] for r in txn if not json.loads(r[1]).get("soft_failed")) for chunk in batch_iter(event_ids, 100): - yield self.runInteraction( + yield self.db.runInteraction( "_get_events_which_are_prevs", _get_events_which_are_prevs_txn, chunk ) @@ -343,7 +343,7 @@ class EventsStore( existing_prevs.add(prev_event_id) for chunk in batch_iter(event_ids, 100): - yield self.runInteraction( + yield self.db.runInteraction( "_get_prevs_before_rejected", _get_prevs_before_rejected_txn, chunk ) @@ -430,7 +430,7 @@ class EventsStore( # event's auth chain, but its easier for now just to store them (and # it doesn't take much storage compared to storing the entire event # anyway). - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="event_auth", values=[ @@ -548,6 +548,34 @@ class EventsStore( ], ) + # 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.executemany( + "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.executemany( + """INSERT INTO local_current_membership + (room_id, user_id, event_id, membership) + VALUES (?, ?, ?, (SELECT membership FROM room_memberships WHERE event_id = ?)) + """, + [ + (room_id, key[1], 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._curr_state_delta_stream_cache.entity_has_changed, room_id, @@ -578,12 +606,12 @@ class EventsStore( self, txn, new_forward_extremities, max_stream_order ): for room_id, new_extrem in iteritems(new_forward_extremities): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="event_forward_extremities", keyvalues={"room_id": room_id} ) txn.call_after(self.get_latest_event_ids_in_room.invalidate, (room_id,)) - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="event_forward_extremities", values=[ @@ -596,7 +624,7 @@ class EventsStore( # new stream_ordering to new forward extremeties in the room. # This allows us to later efficiently look up the forward extremeties # for a room before a given stream_ordering - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="stream_ordering_to_exterm", values=[ @@ -713,16 +741,14 @@ class EventsStore( metadata_json = encode_json(event.internal_metadata.get_dict()) - sql = ( - "UPDATE event_json SET internal_metadata = ?" " WHERE event_id = ?" - ) + sql = "UPDATE event_json SET internal_metadata = ? WHERE event_id = ?" txn.execute(sql, (metadata_json, event.event_id)) # Add an entry to the ex_outlier_stream table to replicate the # change in outlier status to our workers. stream_order = event.internal_metadata.stream_ordering state_group_id = context.state_group - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="ex_outlier_stream", values={ @@ -732,7 +758,7 @@ class EventsStore( }, ) - sql = "UPDATE events SET outlier = ?" " WHERE event_id = ?" + sql = "UPDATE events SET outlier = ? WHERE event_id = ?" txn.execute(sql, (False, event.event_id)) # Update the event_backward_extremities table now that this @@ -794,7 +820,7 @@ class EventsStore( d.pop("redacted_because", None) return d - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="event_json", values=[ @@ -811,7 +837,7 @@ class EventsStore( ], ) - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="events", values=[ @@ -841,7 +867,7 @@ class EventsStore( # If we're persisting an unredacted event we go and ensure # that we mark any redactions that reference this event as # requiring censoring. - self._simple_update_txn( + self.db.simple_update_txn( txn, table="redactions", keyvalues={"redacts": event.event_id}, @@ -929,6 +955,9 @@ class EventsStore( elif event.type == EventTypes.Redaction: # Insert into the redactions table. self._store_redaction(txn, event) + elif event.type == EventTypes.Retention: + # Update the room_retention table. + self._store_retention_policy_for_room_txn(txn, event) self._handle_event_relations(txn, event) @@ -939,6 +968,12 @@ class EventsStore( txn, event.event_id, labels, event.room_id, event.depth ) + if self._ephemeral_messages_enabled: + # If there's an expiry timestamp on the event, store it. + expiry_ts = event.content.get(EventContentFields.SELF_DESTRUCT_AFTER) + if isinstance(expiry_ts, int) and not event.is_state(): + self._insert_event_expiry_txn(txn, event.event_id, expiry_ts) + # Insert into the room_memberships table. self._store_room_members_txn( txn, @@ -974,7 +1009,7 @@ class EventsStore( state_values.append(vals) - self._simple_insert_many_txn(txn, table="state_events", values=state_values) + self.db.simple_insert_many_txn(txn, table="state_events", values=state_values) # Prefill the event cache self._add_to_cache(txn, events_and_contexts) @@ -1005,7 +1040,7 @@ class EventsStore( ) txn.execute(sql + clause, args) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) for row in rows: event = ev_map[row["event_id"]] if not row["rejects"] and not row["redacts"]: @@ -1023,7 +1058,7 @@ class EventsStore( # invalidate the cache for the redacted event txn.call_after(self._invalidate_get_event_cache, event.redacts) - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="redactions", values={ @@ -1033,20 +1068,25 @@ class EventsStore( }, ) - @defer.inlineCallbacks - def _censor_redactions(self): + async def _censor_redactions(self): """Censors all redactions older than the configured period that haven't been censored yet. By censor we mean update the event_json table with the redacted event. - - Returns: - Deferred """ if self.hs.config.redaction_retention_period is None: return + if not ( + await self.db.updates.has_completed_background_update( + "redactions_have_censored_ts_idx" + ) + ): + # We don't want to run this until the appropriate index has been + # created. + return + before_ts = self._clock.time_msec() - self.hs.config.redaction_retention_period # We fetch all redactions that: @@ -1068,15 +1108,15 @@ class EventsStore( LIMIT ? """ - rows = yield self._execute( + rows = await self.db.execute( "_censor_redactions_fetch", None, sql, before_ts, 100 ) updates = [] for redaction_id, event_id in rows: - redaction_event = yield self.get_event(redaction_id, allow_none=True) - original_event = yield self.get_event( + redaction_event = await self.get_event(redaction_id, allow_none=True) + original_event = await self.get_event( event_id, allow_rejected=True, allow_none=True ) @@ -1100,21 +1140,32 @@ class EventsStore( def _update_censor_txn(txn): for redaction_id, event_id, pruned_json in updates: if pruned_json: - self._simple_update_one_txn( - txn, - table="event_json", - keyvalues={"event_id": event_id}, - updatevalues={"json": pruned_json}, - ) + self._censor_event_txn(txn, event_id, pruned_json) - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, table="redactions", keyvalues={"event_id": redaction_id}, updatevalues={"have_censored": True}, ) - yield self.runInteraction("_update_censor_txn", _update_censor_txn) + await self.db.runInteraction("_update_censor_txn", _update_censor_txn) + + def _censor_event_txn(self, txn, event_id, pruned_json): + """Censor an event by replacing its JSON in the event_json table with the + provided pruned JSON. + + Args: + txn (LoggingTransaction): The database transaction. + event_id (str): The ID of the event to censor. + pruned_json (str): The pruned JSON + """ + self.db.simple_update_one_txn( + txn, + table="event_json", + keyvalues={"event_id": event_id}, + updatevalues={"json": pruned_json}, + ) @defer.inlineCallbacks def count_daily_messages(self): @@ -1135,7 +1186,7 @@ class EventsStore( (count,) = txn.fetchone() return count - ret = yield self.runInteraction("count_messages", _count_messages) + ret = yield self.db.runInteraction("count_messages", _count_messages) return ret @defer.inlineCallbacks @@ -1156,7 +1207,7 @@ class EventsStore( (count,) = txn.fetchone() return count - ret = yield self.runInteraction("count_daily_sent_messages", _count_messages) + ret = yield self.db.runInteraction("count_daily_sent_messages", _count_messages) return ret @defer.inlineCallbacks @@ -1171,7 +1222,7 @@ class EventsStore( (count,) = txn.fetchone() return count - ret = yield self.runInteraction("count_daily_active_rooms", _count) + ret = yield self.db.runInteraction("count_daily_active_rooms", _count) return ret def get_current_backfill_token(self): @@ -1223,7 +1274,7 @@ class EventsStore( return new_event_updates - return self.runInteraction( + return self.db.runInteraction( "get_all_new_forward_event_rows", get_all_new_forward_event_rows ) @@ -1268,7 +1319,7 @@ class EventsStore( return new_event_updates - return self.runInteraction( + return self.db.runInteraction( "get_all_new_backfill_event_rows", get_all_new_backfill_event_rows ) @@ -1361,7 +1412,7 @@ class EventsStore( backward_ex_outliers, ) - return self.runInteraction("get_all_new_events", get_all_new_events_txn) + return self.db.runInteraction("get_all_new_events", get_all_new_events_txn) def purge_history(self, room_id, token, delete_local_events): """Deletes room history before a certain point @@ -1381,7 +1432,7 @@ class EventsStore( deleted events. """ - return self.runInteraction( + return self.db.runInteraction( "purge_history", self._purge_history_txn, room_id, @@ -1479,7 +1530,7 @@ class EventsStore( # We do joins against events_to_purge for e.g. calculating state # groups to purge, etc., so lets make an index. - txn.execute("CREATE INDEX events_to_purge_id" " ON events_to_purge(event_id)") + txn.execute("CREATE INDEX events_to_purge_id ON events_to_purge(event_id)") txn.execute("SELECT event_id, should_delete FROM events_to_purge") event_rows = txn.fetchall() @@ -1629,7 +1680,7 @@ class EventsStore( Deferred[List[int]]: The list of state groups to delete. """ - return self.runInteraction("purge_room", self._purge_room_txn, room_id) + return self.db.runInteraction("purge_room", self._purge_room_txn, room_id) def _purge_room_txn(self, txn, room_id): # First we fetch all the state groups that should be deleted, before @@ -1702,6 +1753,7 @@ class EventsStore( "local_invites", "room_account_data", "room_tags", + "local_current_membership", ): logger.info("[purge] removing %s from %s", room_id, table) txn.execute("DELETE FROM %s WHERE room_id=?" % (table,), (room_id,)) @@ -1735,163 +1787,6 @@ class EventsStore( return state_groups - def purge_unreferenced_state_groups( - self, room_id: str, state_groups_to_delete - ) -> defer.Deferred: - """Deletes no longer referenced state groups and de-deltas any state - groups that reference them. - - Args: - room_id: The room the state groups belong to (must all be in the - same room). - state_groups_to_delete (Collection[int]): Set of all state groups - to delete. - """ - - return self.runInteraction( - "purge_unreferenced_state_groups", - self._purge_unreferenced_state_groups, - room_id, - state_groups_to_delete, - ) - - def _purge_unreferenced_state_groups(self, txn, room_id, state_groups_to_delete): - logger.info( - "[purge] found %i state groups to delete", len(state_groups_to_delete) - ) - - rows = self._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 = set( - row["state_group"] - for row in rows - if row["state_group"] not in state_groups_to_delete - ) - - logger.info( - "[purge] de-delta-ing %i remaining state groups", - len(remaining_state_groups), - ) - - # Now we turn the state groups that reference to-be-deleted state - # groups to non delta versions. - for sg in remaining_state_groups: - logger.info("[purge] de-delta-ing remaining state group %s", sg) - curr_state = self._get_state_groups_from_groups_txn(txn, [sg]) - curr_state = curr_state[sg] - - self._simple_delete_txn( - txn, table="state_groups_state", keyvalues={"state_group": sg} - ) - - self._simple_delete_txn( - txn, table="state_group_edges", keyvalues={"state_group": sg} - ) - - self._simple_insert_many_txn( - txn, - table="state_groups_state", - values=[ - { - "state_group": sg, - "room_id": room_id, - "type": key[0], - "state_key": key[1], - "event_id": state_id, - } - for key, state_id in iteritems(curr_state) - ], - ) - - logger.info("[purge] removing redundant state groups") - txn.executemany( - "DELETE FROM state_groups_state WHERE state_group = ?", - ((sg,) for sg in state_groups_to_delete), - ) - txn.executemany( - "DELETE FROM state_groups WHERE id = ?", - ((sg,) for sg in state_groups_to_delete), - ) - - @defer.inlineCallbacks - def get_previous_state_groups(self, state_groups): - """Fetch the previous groups of the given state groups. - - Args: - state_groups (Iterable[int]) - - Returns: - Deferred[dict[int, int]]: mapping from state group to previous - state group. - """ - - rows = yield self._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", - ) - - return {row["state_group"]: row["prev_state_group"] for row in rows} - - def purge_room_state(self, room_id, state_groups_to_delete): - """Deletes all record of a room from state tables - - Args: - room_id (str): - state_groups_to_delete (list[int]): State groups to delete - """ - - return self.runInteraction( - "purge_room_state", - self._purge_room_state_txn, - room_id, - state_groups_to_delete, - ) - - def _purge_room_state_txn(self, txn, room_id, state_groups_to_delete): - # first we have to delete the state groups states - logger.info("[purge] removing %s from state_groups_state", room_id) - - self._simple_delete_many_txn( - txn, - table="state_groups_state", - column="state_group", - iterable=state_groups_to_delete, - keyvalues={}, - ) - - # ... and the state group edges - logger.info("[purge] removing %s from state_group_edges", room_id) - - self._simple_delete_many_txn( - txn, - table="state_group_edges", - column="state_group", - iterable=state_groups_to_delete, - keyvalues={}, - ) - - # ... and the state groups - logger.info("[purge] removing %s from state_groups", room_id) - - self._simple_delete_many_txn( - txn, - table="state_groups", - column="id", - iterable=state_groups_to_delete, - keyvalues={}, - ) - async def is_event_after(self, event_id1, event_id2): """Returns True if event_id1 is after event_id2 in the stream """ @@ -1901,7 +1796,7 @@ class EventsStore( @cachedInlineCallbacks(max_entries=5000) def _get_event_ordering(self, event_id): - res = yield self._simple_select_one( + res = yield self.db.simple_select_one( table="events", retcols=["topological_ordering", "stream_ordering"], keyvalues={"event_id": event_id}, @@ -1924,7 +1819,7 @@ class EventsStore( txn.execute(sql, (from_token, to_token, limit)) return txn.fetchall() - return self.runInteraction( + return self.db.runInteraction( "get_all_updated_current_state_deltas", get_all_updated_current_state_deltas_txn, ) @@ -1942,7 +1837,7 @@ class EventsStore( room_id (str): The ID of the room the event was sent to. topological_ordering (int): The position of the event in the room's topology. """ - return self._simple_insert_many_txn( + return self.db.simple_insert_many_txn( txn=txn, table="event_labels", values=[ @@ -1956,6 +1851,101 @@ class EventsStore( ], ) + def _insert_event_expiry_txn(self, txn, event_id, expiry_ts): + """Save the expiry timestamp associated with a given event ID. + + Args: + txn (LoggingTransaction): The database transaction to use. + event_id (str): The event ID the expiry timestamp is associated with. + expiry_ts (int): The timestamp at which to expire (delete) the event. + """ + return self.db.simple_insert_txn( + txn=txn, + table="event_expiry", + values={"event_id": event_id, "expiry_ts": expiry_ts}, + ) + + @defer.inlineCallbacks + def expire_event(self, event_id): + """Retrieve and expire an event that has expired, and delete its associated + expiry timestamp. If the event can't be retrieved, delete its associated + timestamp so we don't try to expire it again in the future. + + Args: + event_id (str): The ID of the event to delete. + """ + # Try to retrieve the event's content from the database or the event cache. + event = yield self.get_event(event_id) + + def delete_expired_event_txn(txn): + # Delete the expiry timestamp associated with this event from the database. + self._delete_event_expiry_txn(txn, event_id) + + if not event: + # If we can't find the event, log a warning and delete the expiry date + # from the database so that we don't try to expire it again in the + # future. + logger.warning( + "Can't expire event %s because we don't have it.", event_id + ) + return + + # Prune the event's dict then convert it to JSON. + pruned_json = encode_json(prune_event_dict(event.get_dict())) + + # Update the event_json table to replace the event's JSON with the pruned + # JSON. + self._censor_event_txn(txn, event.event_id, pruned_json) + + # We need to invalidate the event cache entry for this event because we + # changed its content in the database. We can't call + # self._invalidate_cache_and_stream because self.get_event_cache isn't of the + # right type. + txn.call_after(self._get_event_cache.invalidate, (event.event_id,)) + # Send that invalidation to replication so that other workers also invalidate + # the event cache. + self._send_invalidation_to_replication( + txn, "_get_event_cache", (event.event_id,) + ) + + yield self.db.runInteraction("delete_expired_event", delete_expired_event_txn) + + def _delete_event_expiry_txn(self, txn, event_id): + """Delete the expiry timestamp associated with an event ID without deleting the + actual event. + + Args: + txn (LoggingTransaction): The transaction to use to perform the deletion. + event_id (str): The event ID to delete the associated expiry timestamp of. + """ + return self.db.simple_delete_txn( + txn=txn, table="event_expiry", keyvalues={"event_id": event_id} + ) + + def get_next_event_to_expire(self): + """Retrieve the entry with the lowest expiry timestamp in the event_expiry + table, or None if there's no more event to expire. + + Returns: Deferred[Optional[Tuple[str, int]]] + A tuple containing the event ID as its first element and an expiry timestamp + as its second one, if there's at least one row in the event_expiry table. + None otherwise. + """ + + def get_next_event_to_expire_txn(txn): + txn.execute( + """ + SELECT event_id, expiry_ts FROM event_expiry + ORDER BY expiry_ts ASC LIMIT 1 + """ + ) + + return txn.fetchone() + + return self.db.runInteraction( + desc="get_next_event_to_expire", func=get_next_event_to_expire_txn + ) + AllNewEventsResult = namedtuple( "AllNewEventsResult", diff --git a/synapse/storage/data_stores/main/events_bg_updates.py b/synapse/storage/data_stores/main/events_bg_updates.py index 0ed59ef48e..5177b71016 100644 --- a/synapse/storage/data_stores/main/events_bg_updates.py +++ b/synapse/storage/data_stores/main/events_bg_updates.py @@ -22,30 +22,30 @@ from canonicaljson import json from twisted.internet import defer from synapse.api.constants import EventContentFields -from synapse.storage._base import make_in_list_sql_clause -from synapse.storage.background_updates import BackgroundUpdateStore +from synapse.storage._base import SQLBaseStore, make_in_list_sql_clause +from synapse.storage.database import Database logger = logging.getLogger(__name__) -class EventsBackgroundUpdatesStore(BackgroundUpdateStore): +class EventsBackgroundUpdatesStore(SQLBaseStore): EVENT_ORIGIN_SERVER_TS_NAME = "event_origin_server_ts" EVENT_FIELDS_SENDER_URL_UPDATE_NAME = "event_fields_sender_url" DELETE_SOFT_FAILED_EXTREMITIES = "delete_soft_failed_extremities" - def __init__(self, db_conn, hs): - super(EventsBackgroundUpdatesStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(EventsBackgroundUpdatesStore, self).__init__(database, db_conn, hs) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( self.EVENT_ORIGIN_SERVER_TS_NAME, self._background_reindex_origin_server_ts ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( self.EVENT_FIELDS_SENDER_URL_UPDATE_NAME, self._background_reindex_fields_sender, ) - self.register_background_index_update( + self.db.updates.register_background_index_update( "event_contains_url_index", index_name="event_contains_url_index", table="events", @@ -56,7 +56,7 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): # an event_id index on event_search is useful for the purge_history # api. Plus it means we get to enforce some integrity with a UNIQUE # clause - self.register_background_index_update( + self.db.updates.register_background_index_update( "event_search_event_id_idx", index_name="event_search_event_id_idx", table="event_search", @@ -65,16 +65,16 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): psql_only=True, ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( self.DELETE_SOFT_FAILED_EXTREMITIES, self._cleanup_extremities_bg_update ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "redactions_received_ts", self._redactions_received_ts ) # This index gets deleted in `event_fix_redactions_bytes` update - self.register_background_index_update( + self.db.updates.register_background_index_update( "event_fix_redactions_bytes_create_index", index_name="redactions_censored_redacts", table="redactions", @@ -82,14 +82,22 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): where_clause="have_censored", ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "event_fix_redactions_bytes", self._event_fix_redactions_bytes ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "event_store_labels", self._event_store_labels ) + self.db.updates.register_background_index_update( + "redactions_have_censored_ts_idx", + index_name="redactions_have_censored_ts", + table="redactions", + columns=["received_ts"], + where_clause="NOT have_censored", + ) + @defer.inlineCallbacks def _background_reindex_fields_sender(self, progress, batch_size): target_min_stream_id = progress["target_min_stream_id_inclusive"] @@ -145,18 +153,20 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): "rows_inserted": rows_inserted + len(rows), } - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, self.EVENT_FIELDS_SENDER_URL_UPDATE_NAME, progress ) return len(rows) - result = yield self.runInteraction( + result = yield self.db.runInteraction( self.EVENT_FIELDS_SENDER_URL_UPDATE_NAME, reindex_txn ) if not result: - yield self._end_background_update(self.EVENT_FIELDS_SENDER_URL_UPDATE_NAME) + yield self.db.updates._end_background_update( + self.EVENT_FIELDS_SENDER_URL_UPDATE_NAME + ) return result @@ -189,7 +199,7 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): chunks = [event_ids[i : i + 100] for i in range(0, len(event_ids), 100)] for chunk in chunks: - ev_rows = self._simple_select_many_txn( + ev_rows = self.db.simple_select_many_txn( txn, table="event_json", column="event_id", @@ -222,18 +232,20 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): "rows_inserted": rows_inserted + len(rows_to_update), } - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, self.EVENT_ORIGIN_SERVER_TS_NAME, progress ) return len(rows_to_update) - result = yield self.runInteraction( + result = yield self.db.runInteraction( self.EVENT_ORIGIN_SERVER_TS_NAME, reindex_search_txn ) if not result: - yield self._end_background_update(self.EVENT_ORIGIN_SERVER_TS_NAME) + yield self.db.updates._end_background_update( + self.EVENT_ORIGIN_SERVER_TS_NAME + ) return result @@ -366,7 +378,7 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): to_delete.intersection_update(original_set) - deleted = self._simple_delete_many_txn( + deleted = self.db.simple_delete_many_txn( txn=txn, table="event_forward_extremities", column="event_id", @@ -382,7 +394,7 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): if deleted: # We now need to invalidate the caches of these rooms - rows = self._simple_select_many_txn( + rows = self.db.simple_select_many_txn( txn, table="events", column="event_id", @@ -396,7 +408,7 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): self.get_latest_event_ids_in_room.invalidate, (room_id,) ) - self._simple_delete_many_txn( + self.db.simple_delete_many_txn( txn=txn, table="_extremities_to_check", column="event_id", @@ -406,17 +418,19 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): return len(original_set) - num_handled = yield self.runInteraction( + num_handled = yield self.db.runInteraction( "_cleanup_extremities_bg_update", _cleanup_extremities_bg_update_txn ) if not num_handled: - yield self._end_background_update(self.DELETE_SOFT_FAILED_EXTREMITIES) + yield self.db.updates._end_background_update( + self.DELETE_SOFT_FAILED_EXTREMITIES + ) def _drop_table_txn(txn): txn.execute("DROP TABLE _extremities_to_check") - yield self.runInteraction( + yield self.db.runInteraction( "_cleanup_extremities_bg_update_drop_table", _drop_table_txn ) @@ -464,18 +478,18 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): txn.execute(sql, (self._clock.time_msec(), last_event_id, upper_event_id)) - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, "redactions_received_ts", {"last_event_id": upper_event_id} ) return len(rows) - count = yield self.runInteraction( + count = yield self.db.runInteraction( "_redactions_received_ts", _redactions_received_ts_txn ) if not count: - yield self._end_background_update("redactions_received_ts") + yield self.db.updates._end_background_update("redactions_received_ts") return count @@ -501,11 +515,11 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): txn.execute("DROP INDEX redactions_censored_redacts") - yield self.runInteraction( + yield self.db.runInteraction( "_event_fix_redactions_bytes", _event_fix_redactions_bytes_txn ) - yield self._end_background_update("event_fix_redactions_bytes") + yield self.db.updates._end_background_update("event_fix_redactions_bytes") return 1 @@ -530,39 +544,46 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): nbrows = 0 last_row_event_id = "" for (event_id, event_json_raw) in results: - event_json = json.loads(event_json_raw) - - self._simple_insert_many_txn( - txn=txn, - table="event_labels", - values=[ - { - "event_id": event_id, - "label": label, - "room_id": event_json["room_id"], - "topological_ordering": event_json["depth"], - } - for label in event_json["content"].get( - EventContentFields.LABELS, [] - ) - if isinstance(label, str) - ], - ) + try: + event_json = json.loads(event_json_raw) + + self.db.simple_insert_many_txn( + txn=txn, + table="event_labels", + values=[ + { + "event_id": event_id, + "label": label, + "room_id": event_json["room_id"], + "topological_ordering": event_json["depth"], + } + for label in event_json["content"].get( + EventContentFields.LABELS, [] + ) + if isinstance(label, str) + ], + ) + except Exception as e: + logger.warning( + "Unable to load event %s (no labels will be imported): %s", + event_id, + e, + ) nbrows += 1 last_row_event_id = event_id - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, "event_store_labels", {"last_event_id": last_row_event_id} ) return nbrows - num_rows = yield self.runInteraction( + num_rows = yield self.db.runInteraction( desc="event_store_labels", func=_event_store_labels_txn ) if not num_rows: - yield self._end_background_update("event_store_labels") + yield self.db.updates._end_background_update("event_store_labels") return num_rows diff --git a/synapse/storage/data_stores/main/events_worker.py b/synapse/storage/data_stores/main/events_worker.py index 4c4b76bd93..0cce5232f5 100644 --- a/synapse/storage/data_stores/main/events_worker.py +++ b/synapse/storage/data_stores/main/events_worker.py @@ -17,9 +17,12 @@ from __future__ import division import itertools import logging +import threading from collections import namedtuple +from typing import List, Optional from canonicaljson import json +from constantly import NamedConstant, Names from twisted.internet import defer @@ -32,8 +35,10 @@ from synapse.events.utils import prune_event from synapse.logging.context import LoggingContext, PreserveLoggingContext from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage._base import SQLBaseStore, make_in_list_sql_clause +from synapse.storage.database import Database from synapse.types import get_domain_from_id from synapse.util import batch_iter +from synapse.util.caches.descriptors import Cache from synapse.util.metrics import Measure logger = logging.getLogger(__name__) @@ -52,7 +57,28 @@ EVENT_QUEUE_TIMEOUT_S = 0.1 # Timeout when waiting for requests for events _EventCacheEntry = namedtuple("_EventCacheEntry", ("event", "redacted_event")) +class EventRedactBehaviour(Names): + """ + What to do when retrieving a redacted event from the database. + """ + + AS_IS = NamedConstant() + REDACT = NamedConstant() + BLOCK = NamedConstant() + + class EventsWorkerStore(SQLBaseStore): + def __init__(self, database: Database, db_conn, hs): + super(EventsWorkerStore, self).__init__(database, db_conn, hs) + + self._get_event_cache = Cache( + "*getEvent*", keylen=3, max_entries=hs.config.event_cache_size + ) + + self._event_fetch_lock = threading.Condition() + self._event_fetch_list = [] + self._event_fetch_ongoing = 0 + def get_received_ts(self, event_id): """Get received_ts (when it was persisted) for the event. @@ -65,7 +91,7 @@ class EventsWorkerStore(SQLBaseStore): Deferred[int|None]: Timestamp in milliseconds, or None for events that were persisted before received_ts was implemented. """ - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="events", keyvalues={"event_id": event_id}, retcol="received_ts", @@ -104,32 +130,41 @@ class EventsWorkerStore(SQLBaseStore): return ts - return self.runInteraction( + return self.db.runInteraction( "get_approximate_received_ts", _get_approximate_received_ts_txn ) @defer.inlineCallbacks def get_event( self, - event_id, - check_redacted=True, - get_prev_content=False, - allow_rejected=False, - allow_none=False, - check_room_id=None, + event_id: str, + redact_behaviour: EventRedactBehaviour = EventRedactBehaviour.REDACT, + get_prev_content: bool = False, + allow_rejected: bool = False, + allow_none: bool = False, + check_room_id: Optional[str] = None, ): """Get an event from the database by event_id. Args: - event_id (str): The event_id of the event to fetch - check_redacted (bool): If True, check if event has been redacted - and redact it. - get_prev_content (bool): If True and event is a state event, + event_id: The event_id of the event to fetch + + redact_behaviour: Determine what to do with a redacted event. Possible values: + * AS_IS - Return the full event body with no redacted content + * REDACT - Return the event but with a redacted body + * DISALLOW - Do not return redacted events (behave as per allow_none + if the event is redacted) + + get_prev_content: If True and event is a state event, include the previous states content in the unsigned field. - allow_rejected (bool): If True return rejected events. - allow_none (bool): If True, return None if no event found, if + + allow_rejected: If True, return rejected events. Otherwise, + behave as per allow_none. + + allow_none: If True, return None if no event found, if False throw a NotFoundError - check_room_id (str|None): if not None, check the room of the found event. + + check_room_id: if not None, check the room of the found event. If there is a mismatch, behave as per allow_none. Returns: @@ -140,7 +175,7 @@ class EventsWorkerStore(SQLBaseStore): events = yield self.get_events_as_list( [event_id], - check_redacted=check_redacted, + redact_behaviour=redact_behaviour, get_prev_content=get_prev_content, allow_rejected=allow_rejected, ) @@ -159,27 +194,34 @@ class EventsWorkerStore(SQLBaseStore): @defer.inlineCallbacks def get_events( self, - event_ids, - check_redacted=True, - get_prev_content=False, - allow_rejected=False, + event_ids: List[str], + redact_behaviour: EventRedactBehaviour = EventRedactBehaviour.REDACT, + get_prev_content: bool = False, + allow_rejected: bool = False, ): """Get events from the database Args: - event_ids (list): The event_ids of the events to fetch - check_redacted (bool): If True, check if event has been redacted - and redact it. - get_prev_content (bool): If True and event is a state event, + event_ids: The event_ids of the events to fetch + + redact_behaviour: Determine what to do with a redacted event. Possible + values: + * AS_IS - Return the full event body with no redacted content + * REDACT - Return the event but with a redacted body + * DISALLOW - Do not return redacted events (omit them from the response) + + get_prev_content: If True and event is a state event, include the previous states content in the unsigned field. - allow_rejected (bool): If True return rejected events. + + allow_rejected: If True, return rejected events. Otherwise, + omits rejeted events from the response. Returns: Deferred : Dict from event_id to event. """ events = yield self.get_events_as_list( event_ids, - check_redacted=check_redacted, + redact_behaviour=redact_behaviour, get_prev_content=get_prev_content, allow_rejected=allow_rejected, ) @@ -189,21 +231,29 @@ class EventsWorkerStore(SQLBaseStore): @defer.inlineCallbacks def get_events_as_list( self, - event_ids, - check_redacted=True, - get_prev_content=False, - allow_rejected=False, + event_ids: List[str], + redact_behaviour: EventRedactBehaviour = EventRedactBehaviour.REDACT, + get_prev_content: bool = False, + allow_rejected: bool = False, ): """Get events from the database and return in a list in the same order as given by `event_ids` arg. + Unknown events will be omitted from the response. + Args: - event_ids (list): The event_ids of the events to fetch - check_redacted (bool): If True, check if event has been redacted - and redact it. - get_prev_content (bool): If True and event is a state event, + event_ids: The event_ids of the events to fetch + + redact_behaviour: Determine what to do with a redacted event. Possible values: + * AS_IS - Return the full event body with no redacted content + * REDACT - Return the event but with a redacted body + * DISALLOW - Do not return redacted events (omit them from the response) + + get_prev_content: If True and event is a state event, include the previous states content in the unsigned field. - allow_rejected (bool): If True return rejected events. + + allow_rejected: If True, return rejected events. Otherwise, + omits rejected events from the response. Returns: Deferred[list[EventBase]]: List of events fetched from the database. The @@ -305,10 +355,14 @@ class EventsWorkerStore(SQLBaseStore): # Update the cache to save doing the checks again. entry.event.internal_metadata.recheck_redaction = False - if check_redacted and entry.redacted_event: - event = entry.redacted_event - else: - event = entry.event + event = entry.event + + if entry.redacted_event: + if redact_behaviour == EventRedactBehaviour.BLOCK: + # Skip this event + continue + elif redact_behaviour == EventRedactBehaviour.REDACT: + event = entry.redacted_event events.append(event) @@ -332,9 +386,14 @@ class EventsWorkerStore(SQLBaseStore): If events are pulled from the database, they will be cached for future lookups. + Unknown events are omitted from the response. + Args: + event_ids (Iterable[str]): The event_ids of the events to fetch - allow_rejected (bool): Whether to include rejected events + + allow_rejected (bool): Whether to include rejected events. If False, + rejected events are omitted from the response. Returns: Deferred[Dict[str, _EventCacheEntry]]: @@ -439,7 +498,7 @@ class EventsWorkerStore(SQLBaseStore): event_id for events, _ in event_list for event_id in events ) - row_dict = self._new_transaction( + row_dict = self.db.new_transaction( conn, "do_fetch", [], [], self._fetch_event_rows, events_to_fetch ) @@ -469,9 +528,13 @@ class EventsWorkerStore(SQLBaseStore): Returned events will be added to the cache for future lookups. + Unknown events are omitted from the response. + Args: event_ids (Iterable[str]): The event_ids of the events to fetch - allow_rejected (bool): Whether to include rejected events + + allow_rejected (bool): Whether to include rejected events. If False, + rejected events are omitted from the response. Returns: Deferred[Dict[str, _EventCacheEntry]]: @@ -571,7 +634,7 @@ class EventsWorkerStore(SQLBaseStore): if should_start: run_as_background_process( - "fetch_events", self.runWithConnection, self._do_fetch + "fetch_events", self.db.runWithConnection, self._do_fetch ) logger.debug("Loading %d events: %s", len(events), events) @@ -732,7 +795,7 @@ class EventsWorkerStore(SQLBaseStore): """Given a list of event ids, check if we have already processed and stored them as non outliers. """ - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="events", retcols=("event_id",), column="event_id", @@ -767,42 +830,10 @@ class EventsWorkerStore(SQLBaseStore): # break the input up into chunks of 100 input_iterator = iter(event_ids) for chunk in iter(lambda: list(itertools.islice(input_iterator, 100)), []): - yield self.runInteraction("have_seen_events", have_seen_events_txn, chunk) - return results - - def get_seen_events_with_rejections(self, event_ids): - """Given a list of event ids, check if we rejected them. - - Args: - event_ids (list[str]) - - Returns: - Deferred[dict[str, str|None): - Has an entry for each event id we already have seen. Maps to - the rejected reason string if we rejected the event, else maps - to None. - """ - if not event_ids: - return defer.succeed({}) - - def f(txn): - sql = ( - "SELECT e.event_id, reason FROM events as e " - "LEFT JOIN rejections as r ON e.event_id = r.event_id " - "WHERE e.event_id = ?" + yield self.db.runInteraction( + "have_seen_events", have_seen_events_txn, chunk ) - - res = {} - for event_id in event_ids: - txn.execute(sql, (event_id,)) - row = txn.fetchone() - if row: - _, rejected = row - res[event_id] = rejected - - return res - - return self.runInteraction("get_seen_events_with_rejections", f) + return results def _get_total_state_event_counts_txn(self, txn, room_id): """ @@ -828,7 +859,7 @@ class EventsWorkerStore(SQLBaseStore): Returns: Deferred[int] """ - return self.runInteraction( + return self.db.runInteraction( "get_total_state_event_counts", self._get_total_state_event_counts_txn, room_id, @@ -853,7 +884,7 @@ class EventsWorkerStore(SQLBaseStore): Returns: Deferred[int] """ - return self.runInteraction( + return self.db.runInteraction( "get_current_state_event_counts", self._get_current_state_event_counts_txn, room_id, diff --git a/synapse/storage/data_stores/main/filtering.py b/synapse/storage/data_stores/main/filtering.py index a2a2a67927..342d6622a4 100644 --- a/synapse/storage/data_stores/main/filtering.py +++ b/synapse/storage/data_stores/main/filtering.py @@ -30,7 +30,7 @@ class FilteringStore(SQLBaseStore): except ValueError: raise SynapseError(400, "Invalid filter ID", Codes.INVALID_PARAM) - def_json = yield self._simple_select_one_onecol( + def_json = yield self.db.simple_select_one_onecol( table="user_filters", keyvalues={"user_id": user_localpart, "filter_id": filter_id}, retcol="filter_json", @@ -55,7 +55,7 @@ class FilteringStore(SQLBaseStore): if filter_id_response is not None: return filter_id_response[0] - sql = "SELECT MAX(filter_id) FROM user_filters " "WHERE user_id = ?" + sql = "SELECT MAX(filter_id) FROM user_filters WHERE user_id = ?" txn.execute(sql, (user_localpart,)) max_id = txn.fetchone()[0] if max_id is None: @@ -71,4 +71,4 @@ class FilteringStore(SQLBaseStore): return filter_id - return self.runInteraction("add_user_filter", _do_txn) + return self.db.runInteraction("add_user_filter", _do_txn) diff --git a/synapse/storage/data_stores/main/group_server.py b/synapse/storage/data_stores/main/group_server.py index 5ded539af8..6acd45e9f3 100644 --- a/synapse/storage/data_stores/main/group_server.py +++ b/synapse/storage/data_stores/main/group_server.py @@ -35,7 +35,7 @@ class GroupServerStore(SQLBaseStore): * "invite" * "open" """ - return self._simple_update_one( + return self.db.simple_update_one( table="groups", keyvalues={"group_id": group_id}, updatevalues={"join_policy": join_policy}, @@ -43,7 +43,7 @@ class GroupServerStore(SQLBaseStore): ) def get_group(self, group_id): - return self._simple_select_one( + return self.db.simple_select_one( table="groups", keyvalues={"group_id": group_id}, retcols=( @@ -65,7 +65,7 @@ class GroupServerStore(SQLBaseStore): if not include_private: keyvalues["is_public"] = True - return self._simple_select_list( + return self.db.simple_select_list( table="group_users", keyvalues=keyvalues, retcols=("user_id", "is_public", "is_admin"), @@ -75,7 +75,7 @@ class GroupServerStore(SQLBaseStore): def get_invited_users_in_group(self, group_id): # TODO: Pagination - return self._simple_select_onecol( + return self.db.simple_select_onecol( table="group_invites", keyvalues={"group_id": group_id}, retcol="user_id", @@ -89,7 +89,7 @@ class GroupServerStore(SQLBaseStore): if not include_private: keyvalues["is_public"] = True - return self._simple_select_list( + return self.db.simple_select_list( table="group_rooms", keyvalues=keyvalues, retcols=("room_id", "is_public"), @@ -153,10 +153,12 @@ class GroupServerStore(SQLBaseStore): return rooms, categories - return self.runInteraction("get_rooms_for_summary", _get_rooms_for_summary_txn) + return self.db.runInteraction( + "get_rooms_for_summary", _get_rooms_for_summary_txn + ) def add_room_to_summary(self, group_id, room_id, category_id, order, is_public): - return self.runInteraction( + return self.db.runInteraction( "add_room_to_summary", self._add_room_to_summary_txn, group_id, @@ -180,7 +182,7 @@ class GroupServerStore(SQLBaseStore): an order of 1 will put the room first. Otherwise, the room gets added to the end. """ - room_in_group = self._simple_select_one_onecol_txn( + room_in_group = self.db.simple_select_one_onecol_txn( txn, table="group_rooms", keyvalues={"group_id": group_id, "room_id": room_id}, @@ -193,7 +195,7 @@ class GroupServerStore(SQLBaseStore): if category_id is None: category_id = _DEFAULT_CATEGORY_ID else: - cat_exists = self._simple_select_one_onecol_txn( + cat_exists = self.db.simple_select_one_onecol_txn( txn, table="group_room_categories", keyvalues={"group_id": group_id, "category_id": category_id}, @@ -204,7 +206,7 @@ class GroupServerStore(SQLBaseStore): raise SynapseError(400, "Category doesn't exist") # TODO: Check category is part of summary already - cat_exists = self._simple_select_one_onecol_txn( + cat_exists = self.db.simple_select_one_onecol_txn( txn, table="group_summary_room_categories", keyvalues={"group_id": group_id, "category_id": category_id}, @@ -224,7 +226,7 @@ class GroupServerStore(SQLBaseStore): (group_id, category_id, group_id, category_id), ) - existing = self._simple_select_one_txn( + existing = self.db.simple_select_one_txn( txn, table="group_summary_rooms", keyvalues={ @@ -257,7 +259,7 @@ class GroupServerStore(SQLBaseStore): to_update["room_order"] = order if is_public is not None: to_update["is_public"] = is_public - self._simple_update_txn( + self.db.simple_update_txn( txn, table="group_summary_rooms", keyvalues={ @@ -271,7 +273,7 @@ class GroupServerStore(SQLBaseStore): if is_public is None: is_public = True - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="group_summary_rooms", values={ @@ -287,7 +289,7 @@ class GroupServerStore(SQLBaseStore): if category_id is None: category_id = _DEFAULT_CATEGORY_ID - return self._simple_delete( + return self.db.simple_delete( table="group_summary_rooms", keyvalues={ "group_id": group_id, @@ -299,7 +301,7 @@ class GroupServerStore(SQLBaseStore): @defer.inlineCallbacks def get_group_categories(self, group_id): - rows = yield self._simple_select_list( + rows = yield self.db.simple_select_list( table="group_room_categories", keyvalues={"group_id": group_id}, retcols=("category_id", "is_public", "profile"), @@ -316,7 +318,7 @@ class GroupServerStore(SQLBaseStore): @defer.inlineCallbacks def get_group_category(self, group_id, category_id): - category = yield self._simple_select_one( + category = yield self.db.simple_select_one( table="group_room_categories", keyvalues={"group_id": group_id, "category_id": category_id}, retcols=("is_public", "profile"), @@ -343,7 +345,7 @@ class GroupServerStore(SQLBaseStore): else: update_values["is_public"] = is_public - return self._simple_upsert( + return self.db.simple_upsert( table="group_room_categories", keyvalues={"group_id": group_id, "category_id": category_id}, values=update_values, @@ -352,7 +354,7 @@ class GroupServerStore(SQLBaseStore): ) def remove_group_category(self, group_id, category_id): - return self._simple_delete( + return self.db.simple_delete( table="group_room_categories", keyvalues={"group_id": group_id, "category_id": category_id}, desc="remove_group_category", @@ -360,7 +362,7 @@ class GroupServerStore(SQLBaseStore): @defer.inlineCallbacks def get_group_roles(self, group_id): - rows = yield self._simple_select_list( + rows = yield self.db.simple_select_list( table="group_roles", keyvalues={"group_id": group_id}, retcols=("role_id", "is_public", "profile"), @@ -377,7 +379,7 @@ class GroupServerStore(SQLBaseStore): @defer.inlineCallbacks def get_group_role(self, group_id, role_id): - role = yield self._simple_select_one( + role = yield self.db.simple_select_one( table="group_roles", keyvalues={"group_id": group_id, "role_id": role_id}, retcols=("is_public", "profile"), @@ -404,7 +406,7 @@ class GroupServerStore(SQLBaseStore): else: update_values["is_public"] = is_public - return self._simple_upsert( + return self.db.simple_upsert( table="group_roles", keyvalues={"group_id": group_id, "role_id": role_id}, values=update_values, @@ -413,14 +415,14 @@ class GroupServerStore(SQLBaseStore): ) def remove_group_role(self, group_id, role_id): - return self._simple_delete( + return self.db.simple_delete( table="group_roles", keyvalues={"group_id": group_id, "role_id": role_id}, desc="remove_group_role", ) def add_user_to_summary(self, group_id, user_id, role_id, order, is_public): - return self.runInteraction( + return self.db.runInteraction( "add_user_to_summary", self._add_user_to_summary_txn, group_id, @@ -444,7 +446,7 @@ class GroupServerStore(SQLBaseStore): an order of 1 will put the user first. Otherwise, the user gets added to the end. """ - user_in_group = self._simple_select_one_onecol_txn( + user_in_group = self.db.simple_select_one_onecol_txn( txn, table="group_users", keyvalues={"group_id": group_id, "user_id": user_id}, @@ -457,7 +459,7 @@ class GroupServerStore(SQLBaseStore): if role_id is None: role_id = _DEFAULT_ROLE_ID else: - role_exists = self._simple_select_one_onecol_txn( + role_exists = self.db.simple_select_one_onecol_txn( txn, table="group_roles", keyvalues={"group_id": group_id, "role_id": role_id}, @@ -468,7 +470,7 @@ class GroupServerStore(SQLBaseStore): raise SynapseError(400, "Role doesn't exist") # TODO: Check role is part of the summary already - role_exists = self._simple_select_one_onecol_txn( + role_exists = self.db.simple_select_one_onecol_txn( txn, table="group_summary_roles", keyvalues={"group_id": group_id, "role_id": role_id}, @@ -488,7 +490,7 @@ class GroupServerStore(SQLBaseStore): (group_id, role_id, group_id, role_id), ) - existing = self._simple_select_one_txn( + existing = self.db.simple_select_one_txn( txn, table="group_summary_users", keyvalues={"group_id": group_id, "user_id": user_id, "role_id": role_id}, @@ -517,7 +519,7 @@ class GroupServerStore(SQLBaseStore): to_update["user_order"] = order if is_public is not None: to_update["is_public"] = is_public - self._simple_update_txn( + self.db.simple_update_txn( txn, table="group_summary_users", keyvalues={ @@ -531,7 +533,7 @@ class GroupServerStore(SQLBaseStore): if is_public is None: is_public = True - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="group_summary_users", values={ @@ -547,7 +549,7 @@ class GroupServerStore(SQLBaseStore): if role_id is None: role_id = _DEFAULT_ROLE_ID - return self._simple_delete( + return self.db.simple_delete( table="group_summary_users", keyvalues={"group_id": group_id, "role_id": role_id, "user_id": user_id}, desc="remove_user_from_summary", @@ -561,7 +563,7 @@ class GroupServerStore(SQLBaseStore): Deferred[list[str]]: A twisted.Deferred containing a list of group ids containing this room """ - return self._simple_select_onecol( + return self.db.simple_select_onecol( table="group_rooms", keyvalues={"room_id": room_id}, retcol="group_id", @@ -625,12 +627,12 @@ class GroupServerStore(SQLBaseStore): return users, roles - return self.runInteraction( + return self.db.runInteraction( "get_users_for_summary_by_role", _get_users_for_summary_txn ) def is_user_in_group(self, user_id, group_id): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="group_users", keyvalues={"group_id": group_id, "user_id": user_id}, retcol="user_id", @@ -639,7 +641,7 @@ class GroupServerStore(SQLBaseStore): ).addCallback(lambda r: bool(r)) def is_user_admin_in_group(self, group_id, user_id): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="group_users", keyvalues={"group_id": group_id, "user_id": user_id}, retcol="is_admin", @@ -650,7 +652,7 @@ class GroupServerStore(SQLBaseStore): def add_group_invite(self, group_id, user_id): """Record that the group server has invited a user """ - return self._simple_insert( + return self.db.simple_insert( table="group_invites", values={"group_id": group_id, "user_id": user_id}, desc="add_group_invite", @@ -659,7 +661,7 @@ class GroupServerStore(SQLBaseStore): def is_user_invited_to_local_group(self, group_id, user_id): """Has the group server invited a user? """ - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="group_invites", keyvalues={"group_id": group_id, "user_id": user_id}, retcol="user_id", @@ -682,7 +684,7 @@ class GroupServerStore(SQLBaseStore): """ def _get_users_membership_in_group_txn(txn): - row = self._simple_select_one_txn( + row = self.db.simple_select_one_txn( txn, table="group_users", keyvalues={"group_id": group_id, "user_id": user_id}, @@ -697,7 +699,7 @@ class GroupServerStore(SQLBaseStore): "is_privileged": row["is_admin"], } - row = self._simple_select_one_onecol_txn( + row = self.db.simple_select_one_onecol_txn( txn, table="group_invites", keyvalues={"group_id": group_id, "user_id": user_id}, @@ -710,7 +712,7 @@ class GroupServerStore(SQLBaseStore): return {} - return self.runInteraction( + return self.db.runInteraction( "get_users_membership_info_in_group", _get_users_membership_in_group_txn ) @@ -738,7 +740,7 @@ class GroupServerStore(SQLBaseStore): """ def _add_user_to_group_txn(txn): - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="group_users", values={ @@ -749,14 +751,14 @@ class GroupServerStore(SQLBaseStore): }, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_invites", keyvalues={"group_id": group_id, "user_id": user_id}, ) if local_attestation: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="group_attestations_renewals", values={ @@ -766,7 +768,7 @@ class GroupServerStore(SQLBaseStore): }, ) if remote_attestation: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="group_attestations_remote", values={ @@ -777,49 +779,49 @@ class GroupServerStore(SQLBaseStore): }, ) - return self.runInteraction("add_user_to_group", _add_user_to_group_txn) + return self.db.runInteraction("add_user_to_group", _add_user_to_group_txn) def remove_user_from_group(self, group_id, user_id): def _remove_user_from_group_txn(txn): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_users", keyvalues={"group_id": group_id, "user_id": user_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_invites", keyvalues={"group_id": group_id, "user_id": user_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_attestations_renewals", keyvalues={"group_id": group_id, "user_id": user_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_attestations_remote", keyvalues={"group_id": group_id, "user_id": user_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_summary_users", keyvalues={"group_id": group_id, "user_id": user_id}, ) - return self.runInteraction( + return self.db.runInteraction( "remove_user_from_group", _remove_user_from_group_txn ) def add_room_to_group(self, group_id, room_id, is_public): - return self._simple_insert( + return self.db.simple_insert( table="group_rooms", values={"group_id": group_id, "room_id": room_id, "is_public": is_public}, desc="add_room_to_group", ) def update_room_in_group_visibility(self, group_id, room_id, is_public): - return self._simple_update( + return self.db.simple_update( table="group_rooms", keyvalues={"group_id": group_id, "room_id": room_id}, updatevalues={"is_public": is_public}, @@ -828,26 +830,26 @@ class GroupServerStore(SQLBaseStore): def remove_room_from_group(self, group_id, room_id): def _remove_room_from_group_txn(txn): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_rooms", keyvalues={"group_id": group_id, "room_id": room_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_summary_rooms", keyvalues={"group_id": group_id, "room_id": room_id}, ) - return self.runInteraction( + return self.db.runInteraction( "remove_room_from_group", _remove_room_from_group_txn ) def get_publicised_groups_for_user(self, user_id): """Get all groups a user is publicising """ - return self._simple_select_onecol( + return self.db.simple_select_onecol( table="local_group_membership", keyvalues={"user_id": user_id, "membership": "join", "is_publicised": True}, retcol="group_id", @@ -857,7 +859,7 @@ class GroupServerStore(SQLBaseStore): def update_group_publicity(self, group_id, user_id, publicise): """Update whether the user is publicising their membership of the group """ - return self._simple_update_one( + return self.db.simple_update_one( table="local_group_membership", keyvalues={"group_id": group_id, "user_id": user_id}, updatevalues={"is_publicised": publicise}, @@ -893,12 +895,12 @@ class GroupServerStore(SQLBaseStore): def _register_user_group_membership_txn(txn, next_id): # TODO: Upsert? - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="local_group_membership", keyvalues={"group_id": group_id, "user_id": user_id}, ) - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="local_group_membership", values={ @@ -911,7 +913,7 @@ class GroupServerStore(SQLBaseStore): }, ) - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="local_group_updates", values={ @@ -930,7 +932,7 @@ class GroupServerStore(SQLBaseStore): if membership == "join": if local_attestation: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="group_attestations_renewals", values={ @@ -940,7 +942,7 @@ class GroupServerStore(SQLBaseStore): }, ) if remote_attestation: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="group_attestations_remote", values={ @@ -951,12 +953,12 @@ class GroupServerStore(SQLBaseStore): }, ) else: - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_attestations_renewals", keyvalues={"group_id": group_id, "user_id": user_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="group_attestations_remote", keyvalues={"group_id": group_id, "user_id": user_id}, @@ -965,7 +967,7 @@ class GroupServerStore(SQLBaseStore): return next_id with self._group_updates_id_gen.get_next() as next_id: - res = yield self.runInteraction( + res = yield self.db.runInteraction( "register_user_group_membership", _register_user_group_membership_txn, next_id, @@ -976,7 +978,7 @@ class GroupServerStore(SQLBaseStore): def create_group( self, group_id, user_id, name, avatar_url, short_description, long_description ): - yield self._simple_insert( + yield self.db.simple_insert( table="groups", values={ "group_id": group_id, @@ -991,7 +993,7 @@ class GroupServerStore(SQLBaseStore): @defer.inlineCallbacks def update_group_profile(self, group_id, profile): - yield self._simple_update_one( + yield self.db.simple_update_one( table="groups", keyvalues={"group_id": group_id}, updatevalues=profile, @@ -1008,16 +1010,16 @@ class GroupServerStore(SQLBaseStore): WHERE valid_until_ms <= ? """ txn.execute(sql, (valid_until_ms,)) - return self.cursor_to_dict(txn) + return self.db.cursor_to_dict(txn) - return self.runInteraction( + return self.db.runInteraction( "get_attestations_need_renewals", _get_attestations_need_renewals_txn ) def update_attestation_renewal(self, group_id, user_id, attestation): """Update an attestation that we have renewed """ - return self._simple_update_one( + return self.db.simple_update_one( table="group_attestations_renewals", keyvalues={"group_id": group_id, "user_id": user_id}, updatevalues={"valid_until_ms": attestation["valid_until_ms"]}, @@ -1027,7 +1029,7 @@ class GroupServerStore(SQLBaseStore): def update_remote_attestion(self, group_id, user_id, attestation): """Update an attestation that a remote has renewed """ - return self._simple_update_one( + return self.db.simple_update_one( table="group_attestations_remote", keyvalues={"group_id": group_id, "user_id": user_id}, updatevalues={ @@ -1046,7 +1048,7 @@ class GroupServerStore(SQLBaseStore): group_id (str) user_id (str) """ - return self._simple_delete( + return self.db.simple_delete( table="group_attestations_renewals", keyvalues={"group_id": group_id, "user_id": user_id}, desc="remove_attestation_renewal", @@ -1057,7 +1059,7 @@ class GroupServerStore(SQLBaseStore): """Get the attestation that proves the remote agrees that the user is in the group. """ - row = yield self._simple_select_one( + row = yield self.db.simple_select_one( table="group_attestations_remote", keyvalues={"group_id": group_id, "user_id": user_id}, retcols=("valid_until_ms", "attestation_json"), @@ -1072,7 +1074,7 @@ class GroupServerStore(SQLBaseStore): return None def get_joined_groups(self, user_id): - return self._simple_select_onecol( + return self.db.simple_select_onecol( table="local_group_membership", keyvalues={"user_id": user_id, "membership": "join"}, retcol="group_id", @@ -1099,7 +1101,7 @@ class GroupServerStore(SQLBaseStore): for row in txn ] - return self.runInteraction( + return self.db.runInteraction( "get_all_groups_for_user", _get_all_groups_for_user_txn ) @@ -1109,7 +1111,7 @@ class GroupServerStore(SQLBaseStore): user_id, from_token ) if not has_changed: - return [] + return defer.succeed([]) def _get_groups_changes_for_user_txn(txn): sql = """ @@ -1129,7 +1131,7 @@ class GroupServerStore(SQLBaseStore): for group_id, membership, gtype, content_json in txn ] - return self.runInteraction( + return self.db.runInteraction( "get_groups_changes_for_user", _get_groups_changes_for_user_txn ) @@ -1139,7 +1141,7 @@ class GroupServerStore(SQLBaseStore): from_token ) if not has_changed: - return [] + return defer.succeed([]) def _get_all_groups_changes_txn(txn): sql = """ @@ -1154,7 +1156,7 @@ class GroupServerStore(SQLBaseStore): for stream_id, group_id, user_id, gtype, content_json in txn ] - return self.runInteraction( + return self.db.runInteraction( "get_all_groups_changes", _get_all_groups_changes_txn ) @@ -1188,8 +1190,8 @@ class GroupServerStore(SQLBaseStore): ] for table in tables: - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table=table, keyvalues={"group_id": group_id} ) - return self.runInteraction("delete_group", _delete_group_txn) + return self.db.runInteraction("delete_group", _delete_group_txn) diff --git a/synapse/storage/data_stores/main/keys.py b/synapse/storage/data_stores/main/keys.py index ebc7db3ed6..6b12f5a75f 100644 --- a/synapse/storage/data_stores/main/keys.py +++ b/synapse/storage/data_stores/main/keys.py @@ -92,7 +92,7 @@ class KeyStore(SQLBaseStore): _get_keys(txn, batch) return keys - return self.runInteraction("get_server_verify_keys", _txn) + return self.db.runInteraction("get_server_verify_keys", _txn) def store_server_verify_keys(self, from_server, ts_added_ms, verify_keys): """Stores NACL verification keys for remote servers. @@ -127,9 +127,9 @@ class KeyStore(SQLBaseStore): f((i,)) return res - return self.runInteraction( + return self.db.runInteraction( "store_server_verify_keys", - self._simple_upsert_many_txn, + self.db.simple_upsert_many_txn, table="server_signature_keys", key_names=("server_name", "key_id"), key_values=key_values, @@ -157,7 +157,7 @@ class KeyStore(SQLBaseStore): ts_valid_until_ms (int): The time when this json stops being valid. key_json (bytes): The encoded JSON. """ - return self._simple_upsert( + return self.db.simple_upsert( table="server_keys_json", keyvalues={ "server_name": server_name, @@ -196,7 +196,7 @@ class KeyStore(SQLBaseStore): keyvalues["key_id"] = key_id if from_server is not None: keyvalues["from_server"] = from_server - rows = self._simple_select_list_txn( + rows = self.db.simple_select_list_txn( txn, "server_keys_json", keyvalues=keyvalues, @@ -211,4 +211,4 @@ class KeyStore(SQLBaseStore): results[(server_name, key_id, from_server)] = rows return results - return self.runInteraction("get_server_keys_json", _get_server_keys_json_txn) + return self.db.runInteraction("get_server_keys_json", _get_server_keys_json_txn) diff --git a/synapse/storage/data_stores/main/media_repository.py b/synapse/storage/data_stores/main/media_repository.py index 84b5f3ad5e..80ca36dedf 100644 --- a/synapse/storage/data_stores/main/media_repository.py +++ b/synapse/storage/data_stores/main/media_repository.py @@ -12,14 +12,17 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -from synapse.storage.background_updates import BackgroundUpdateStore +from synapse.storage._base import SQLBaseStore +from synapse.storage.database import Database -class MediaRepositoryBackgroundUpdateStore(BackgroundUpdateStore): - def __init__(self, db_conn, hs): - super(MediaRepositoryBackgroundUpdateStore, self).__init__(db_conn, hs) +class MediaRepositoryBackgroundUpdateStore(SQLBaseStore): + def __init__(self, database: Database, db_conn, hs): + super(MediaRepositoryBackgroundUpdateStore, self).__init__( + database, db_conn, hs + ) - self.register_background_index_update( + self.db.updates.register_background_index_update( update_name="local_media_repository_url_idx", index_name="local_media_repository_url_idx", table="local_media_repository", @@ -31,15 +34,15 @@ class MediaRepositoryBackgroundUpdateStore(BackgroundUpdateStore): class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): """Persistence for attachments and avatars""" - def __init__(self, db_conn, hs): - super(MediaRepositoryStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(MediaRepositoryStore, self).__init__(database, db_conn, hs) def get_local_media(self, media_id): """Get the metadata for a local piece of media Returns: None if the media_id doesn't exist. """ - return self._simple_select_one( + return self.db.simple_select_one( "local_media_repository", {"media_id": media_id}, ( @@ -64,7 +67,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): user_id, url_cache=None, ): - return self._simple_insert( + return self.db.simple_insert( "local_media_repository", { "media_id": media_id, @@ -124,12 +127,12 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): ) ) - return self.runInteraction("get_url_cache", get_url_cache_txn) + return self.db.runInteraction("get_url_cache", get_url_cache_txn) def store_url_cache( self, url, response_code, etag, expires_ts, og, media_id, download_ts ): - return self._simple_insert( + return self.db.simple_insert( "local_media_repository_url_cache", { "url": url, @@ -144,7 +147,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): ) def get_local_media_thumbnails(self, media_id): - return self._simple_select_list( + return self.db.simple_select_list( "local_media_repository_thumbnails", {"media_id": media_id}, ( @@ -166,7 +169,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): thumbnail_method, thumbnail_length, ): - return self._simple_insert( + return self.db.simple_insert( "local_media_repository_thumbnails", { "media_id": media_id, @@ -180,7 +183,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): ) def get_cached_remote_media(self, origin, media_id): - return self._simple_select_one( + return self.db.simple_select_one( "remote_media_cache", {"media_origin": origin, "media_id": media_id}, ( @@ -205,7 +208,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): upload_name, filesystem_id, ): - return self._simple_insert( + return self.db.simple_insert( "remote_media_cache", { "media_origin": origin, @@ -250,10 +253,12 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): txn.executemany(sql, ((time_ms, media_id) for media_id in local_media)) - return self.runInteraction("update_cached_last_access_time", update_cache_txn) + return self.db.runInteraction( + "update_cached_last_access_time", update_cache_txn + ) def get_remote_media_thumbnails(self, origin, media_id): - return self._simple_select_list( + return self.db.simple_select_list( "remote_media_cache_thumbnails", {"media_origin": origin, "media_id": media_id}, ( @@ -278,7 +283,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): thumbnail_method, thumbnail_length, ): - return self._simple_insert( + return self.db.simple_insert( "remote_media_cache_thumbnails", { "media_origin": origin, @@ -300,24 +305,24 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): " WHERE last_access_ts < ?" ) - return self._execute( - "get_remote_media_before", self.cursor_to_dict, sql, before_ts + return self.db.execute( + "get_remote_media_before", self.db.cursor_to_dict, sql, before_ts ) def delete_remote_media(self, media_origin, media_id): def delete_remote_media_txn(txn): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, "remote_media_cache", keyvalues={"media_origin": media_origin, "media_id": media_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, "remote_media_cache_thumbnails", keyvalues={"media_origin": media_origin, "media_id": media_id}, ) - return self.runInteraction("delete_remote_media", delete_remote_media_txn) + return self.db.runInteraction("delete_remote_media", delete_remote_media_txn) def get_expired_url_cache(self, now_ts): sql = ( @@ -331,18 +336,20 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): txn.execute(sql, (now_ts,)) return [row[0] for row in txn] - return self.runInteraction("get_expired_url_cache", _get_expired_url_cache_txn) + return self.db.runInteraction( + "get_expired_url_cache", _get_expired_url_cache_txn + ) def delete_url_cache(self, media_ids): if len(media_ids) == 0: return - sql = "DELETE FROM local_media_repository_url_cache" " WHERE media_id = ?" + sql = "DELETE FROM local_media_repository_url_cache WHERE media_id = ?" def _delete_url_cache_txn(txn): txn.executemany(sql, [(media_id,) for media_id in media_ids]) - return self.runInteraction("delete_url_cache", _delete_url_cache_txn) + return self.db.runInteraction("delete_url_cache", _delete_url_cache_txn) def get_url_cache_media_before(self, before_ts): sql = ( @@ -356,7 +363,7 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): txn.execute(sql, (before_ts,)) return [row[0] for row in txn] - return self.runInteraction( + return self.db.runInteraction( "get_url_cache_media_before", _get_url_cache_media_before_txn ) @@ -365,14 +372,14 @@ class MediaRepositoryStore(MediaRepositoryBackgroundUpdateStore): return def _delete_url_cache_media_txn(txn): - sql = "DELETE FROM local_media_repository" " WHERE media_id = ?" + sql = "DELETE FROM local_media_repository WHERE media_id = ?" txn.executemany(sql, [(media_id,) for media_id in media_ids]) - sql = "DELETE FROM local_media_repository_thumbnails" " WHERE media_id = ?" + sql = "DELETE FROM local_media_repository_thumbnails WHERE media_id = ?" txn.executemany(sql, [(media_id,) for media_id in media_ids]) - return self.runInteraction( + return self.db.runInteraction( "delete_url_cache_media", _delete_url_cache_media_txn ) diff --git a/synapse/storage/data_stores/main/monthly_active_users.py b/synapse/storage/data_stores/main/monthly_active_users.py index b41c3d317a..27158534cb 100644 --- a/synapse/storage/data_stores/main/monthly_active_users.py +++ b/synapse/storage/data_stores/main/monthly_active_users.py @@ -17,6 +17,7 @@ import logging from twisted.internet import defer from synapse.storage._base import SQLBaseStore +from synapse.storage.database import Database from synapse.util.caches.descriptors import cached logger = logging.getLogger(__name__) @@ -27,13 +28,13 @@ LAST_SEEN_GRANULARITY = 60 * 60 * 1000 class MonthlyActiveUsersStore(SQLBaseStore): - def __init__(self, dbconn, hs): - super(MonthlyActiveUsersStore, self).__init__(None, hs) + def __init__(self, database: Database, db_conn, hs): + super(MonthlyActiveUsersStore, self).__init__(database, db_conn, hs) self._clock = hs.get_clock() self.hs = hs # Do not add more reserved users than the total allowable number - self._new_transaction( - dbconn, + self.db.new_transaction( + db_conn, "initialise_mau_threepids", [], [], @@ -146,7 +147,7 @@ class MonthlyActiveUsersStore(SQLBaseStore): txn.execute(sql, query_args) reserved_users = yield self.get_registered_reserved_users() - yield self.runInteraction( + yield self.db.runInteraction( "reap_monthly_active_users", _reap_users, reserved_users ) # It seems poor to invalidate the whole cache, Postgres supports @@ -174,7 +175,7 @@ class MonthlyActiveUsersStore(SQLBaseStore): (count,) = txn.fetchone() return count - return self.runInteraction("count_users", _count_users) + return self.db.runInteraction("count_users", _count_users) @defer.inlineCallbacks def get_registered_reserved_users(self): @@ -217,7 +218,7 @@ class MonthlyActiveUsersStore(SQLBaseStore): if is_support: return - yield self.runInteraction( + yield self.db.runInteraction( "upsert_monthly_active_user", self.upsert_monthly_active_user_txn, user_id ) @@ -261,7 +262,7 @@ class MonthlyActiveUsersStore(SQLBaseStore): # never be a big table and alternative approaches (batching multiple # upserts into a single txn) introduced a lot of extra complexity. # See https://github.com/matrix-org/synapse/issues/3854 for more - is_insert = self._simple_upsert_txn( + is_insert = self.db.simple_upsert_txn( txn, table="monthly_active_users", keyvalues={"user_id": user_id}, @@ -281,7 +282,7 @@ class MonthlyActiveUsersStore(SQLBaseStore): """ - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="monthly_active_users", keyvalues={"user_id": user_id}, retcol="timestamp", diff --git a/synapse/storage/data_stores/main/openid.py b/synapse/storage/data_stores/main/openid.py index 79b40044d9..cc21437e92 100644 --- a/synapse/storage/data_stores/main/openid.py +++ b/synapse/storage/data_stores/main/openid.py @@ -3,7 +3,7 @@ from synapse.storage._base import SQLBaseStore class OpenIdStore(SQLBaseStore): def insert_open_id_token(self, token, ts_valid_until_ms, user_id): - return self._simple_insert( + return self.db.simple_insert( table="open_id_tokens", values={ "token": token, @@ -28,4 +28,6 @@ class OpenIdStore(SQLBaseStore): else: return rows[0][0] - return self.runInteraction("get_user_id_for_token", get_user_id_for_token_txn) + return self.db.runInteraction( + "get_user_id_for_token", get_user_id_for_token_txn + ) diff --git a/synapse/storage/data_stores/main/presence.py b/synapse/storage/data_stores/main/presence.py index 523ed6575e..a2c83e0867 100644 --- a/synapse/storage/data_stores/main/presence.py +++ b/synapse/storage/data_stores/main/presence.py @@ -29,7 +29,7 @@ class PresenceStore(SQLBaseStore): ) with stream_ordering_manager as stream_orderings: - yield self.runInteraction( + yield self.db.runInteraction( "update_presence", self._update_presence_txn, stream_orderings, @@ -46,7 +46,7 @@ class PresenceStore(SQLBaseStore): txn.call_after(self._get_presence_for_user.invalidate, (state.user_id,)) # Actually insert new rows - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="presence_stream", values=[ @@ -88,7 +88,7 @@ class PresenceStore(SQLBaseStore): txn.execute(sql, (last_id, current_id)) return txn.fetchall() - return self.runInteraction( + return self.db.runInteraction( "get_all_presence_updates", get_all_presence_updates_txn ) @@ -103,7 +103,7 @@ class PresenceStore(SQLBaseStore): inlineCallbacks=True, ) def get_presence_for_users(self, user_ids): - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="presence_stream", column="user_id", iterable=user_ids, @@ -129,7 +129,7 @@ class PresenceStore(SQLBaseStore): return self._presence_id_gen.get_current_token() def allow_presence_visible(self, observed_localpart, observer_userid): - return self._simple_insert( + return self.db.simple_insert( table="presence_allow_inbound", values={ "observed_user_id": observed_localpart, @@ -140,7 +140,7 @@ class PresenceStore(SQLBaseStore): ) def disallow_presence_visible(self, observed_localpart, observer_userid): - return self._simple_delete_one( + return self.db.simple_delete_one( table="presence_allow_inbound", keyvalues={ "observed_user_id": observed_localpart, diff --git a/synapse/storage/data_stores/main/profile.py b/synapse/storage/data_stores/main/profile.py index e4e8a1c1d6..2b52cf9c1a 100644 --- a/synapse/storage/data_stores/main/profile.py +++ b/synapse/storage/data_stores/main/profile.py @@ -24,7 +24,7 @@ class ProfileWorkerStore(SQLBaseStore): @defer.inlineCallbacks def get_profileinfo(self, user_localpart): try: - profile = yield self._simple_select_one( + profile = yield self.db.simple_select_one( table="profiles", keyvalues={"user_id": user_localpart}, retcols=("displayname", "avatar_url"), @@ -42,7 +42,7 @@ class ProfileWorkerStore(SQLBaseStore): ) def get_profile_displayname(self, user_localpart): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="profiles", keyvalues={"user_id": user_localpart}, retcol="displayname", @@ -50,7 +50,7 @@ class ProfileWorkerStore(SQLBaseStore): ) def get_profile_avatar_url(self, user_localpart): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="profiles", keyvalues={"user_id": user_localpart}, retcol="avatar_url", @@ -58,7 +58,7 @@ class ProfileWorkerStore(SQLBaseStore): ) def get_from_remote_profile_cache(self, user_id): - return self._simple_select_one( + return self.db.simple_select_one( table="remote_profile_cache", keyvalues={"user_id": user_id}, retcols=("displayname", "avatar_url"), @@ -67,12 +67,12 @@ class ProfileWorkerStore(SQLBaseStore): ) def create_profile(self, user_localpart): - return self._simple_insert( + return self.db.simple_insert( table="profiles", values={"user_id": user_localpart}, desc="create_profile" ) def set_profile_displayname(self, user_localpart, new_displayname): - return self._simple_update_one( + return self.db.simple_update_one( table="profiles", keyvalues={"user_id": user_localpart}, updatevalues={"displayname": new_displayname}, @@ -80,7 +80,7 @@ class ProfileWorkerStore(SQLBaseStore): ) def set_profile_avatar_url(self, user_localpart, new_avatar_url): - return self._simple_update_one( + return self.db.simple_update_one( table="profiles", keyvalues={"user_id": user_localpart}, updatevalues={"avatar_url": new_avatar_url}, @@ -95,7 +95,7 @@ class ProfileStore(ProfileWorkerStore): This should only be called when `is_subscribed_remote_profile_for_user` would return true for the user. """ - return self._simple_upsert( + return self.db.simple_upsert( table="remote_profile_cache", keyvalues={"user_id": user_id}, values={ @@ -107,7 +107,7 @@ class ProfileStore(ProfileWorkerStore): ) def update_remote_profile_cache(self, user_id, displayname, avatar_url): - return self._simple_update( + return self.db.simple_update( table="remote_profile_cache", keyvalues={"user_id": user_id}, values={ @@ -125,7 +125,7 @@ class ProfileStore(ProfileWorkerStore): """ subscribed = yield self.is_subscribed_remote_profile_for_user(user_id) if not subscribed: - yield self._simple_delete( + yield self.db.simple_delete( table="remote_profile_cache", keyvalues={"user_id": user_id}, desc="delete_remote_profile_cache", @@ -144,9 +144,9 @@ class ProfileStore(ProfileWorkerStore): txn.execute(sql, (last_checked,)) - return self.cursor_to_dict(txn) + return self.db.cursor_to_dict(txn) - return self.runInteraction( + return self.db.runInteraction( "get_remote_profile_cache_entries_that_expire", _get_remote_profile_cache_entries_that_expire_txn, ) @@ -155,7 +155,7 @@ class ProfileStore(ProfileWorkerStore): def is_subscribed_remote_profile_for_user(self, user_id): """Check whether we are interested in a remote user's profile. """ - res = yield self._simple_select_one_onecol( + res = yield self.db.simple_select_one_onecol( table="group_users", keyvalues={"user_id": user_id}, retcol="user_id", @@ -166,7 +166,7 @@ class ProfileStore(ProfileWorkerStore): if res: return True - res = yield self._simple_select_one_onecol( + res = yield self.db.simple_select_one_onecol( table="group_invites", keyvalues={"user_id": user_id}, retcol="user_id", diff --git a/synapse/storage/data_stores/main/push_rule.py b/synapse/storage/data_stores/main/push_rule.py index b520062d84..e2673ae073 100644 --- a/synapse/storage/data_stores/main/push_rule.py +++ b/synapse/storage/data_stores/main/push_rule.py @@ -27,6 +27,7 @@ from synapse.storage.data_stores.main.appservice import ApplicationServiceWorker from synapse.storage.data_stores.main.pusher import PusherWorkerStore from synapse.storage.data_stores.main.receipts import ReceiptsWorkerStore from synapse.storage.data_stores.main.roommember import RoomMemberWorkerStore +from synapse.storage.database import Database from synapse.storage.push_rule import InconsistentRuleException, RuleNotFoundException from synapse.util.caches.descriptors import cachedInlineCallbacks, cachedList from synapse.util.caches.stream_change_cache import StreamChangeCache @@ -72,10 +73,10 @@ class PushRulesWorkerStore( # the abstract methods being implemented. __metaclass__ = abc.ABCMeta - def __init__(self, db_conn, hs): - super(PushRulesWorkerStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(PushRulesWorkerStore, self).__init__(database, db_conn, hs) - push_rules_prefill, push_rules_id = self._get_cache_dict( + push_rules_prefill, push_rules_id = self.db.get_cache_dict( db_conn, "push_rules_stream", entity_column="user_id", @@ -100,7 +101,7 @@ class PushRulesWorkerStore( @cachedInlineCallbacks(max_entries=5000) def get_push_rules_for_user(self, user_id): - rows = yield self._simple_select_list( + rows = yield self.db.simple_select_list( table="push_rules", keyvalues={"user_name": user_id}, retcols=( @@ -124,7 +125,7 @@ class PushRulesWorkerStore( @cachedInlineCallbacks(max_entries=5000) def get_push_rules_enabled_for_user(self, user_id): - results = yield self._simple_select_list( + results = yield self.db.simple_select_list( table="push_rules_enable", keyvalues={"user_name": user_id}, retcols=("user_name", "rule_id", "enabled"), @@ -146,7 +147,7 @@ class PushRulesWorkerStore( (count,) = txn.fetchone() return bool(count) - return self.runInteraction( + return self.db.runInteraction( "have_push_rules_changed", have_push_rules_changed_txn ) @@ -162,7 +163,7 @@ class PushRulesWorkerStore( results = {user_id: [] for user_id in user_ids} - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="push_rules", column="user_name", iterable=user_ids, @@ -243,7 +244,7 @@ class PushRulesWorkerStore( # To do this we set the state_group to a new object as object() != object() state_group = object() - current_state_ids = yield context.get_current_state_ids(self) + current_state_ids = yield context.get_current_state_ids() result = yield self._bulk_get_push_rules_for_room( event.room_id, state_group, current_state_ids, event=event ) @@ -320,7 +321,7 @@ class PushRulesWorkerStore( results = {user_id: {} for user_id in user_ids} - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="push_rules_enable", column="user_name", iterable=user_ids, @@ -350,7 +351,7 @@ class PushRuleStore(PushRulesWorkerStore): with self._push_rules_stream_id_gen.get_next() as ids: stream_id, event_stream_ordering = ids if before or after: - yield self.runInteraction( + yield self.db.runInteraction( "_add_push_rule_relative_txn", self._add_push_rule_relative_txn, stream_id, @@ -364,7 +365,7 @@ class PushRuleStore(PushRulesWorkerStore): after, ) else: - yield self.runInteraction( + yield self.db.runInteraction( "_add_push_rule_highest_priority_txn", self._add_push_rule_highest_priority_txn, stream_id, @@ -395,7 +396,7 @@ class PushRuleStore(PushRulesWorkerStore): relative_to_rule = before or after - res = self._simple_select_one_txn( + res = self.db.simple_select_one_txn( txn, table="push_rules", keyvalues={"user_name": user_id, "rule_id": relative_to_rule}, @@ -499,7 +500,7 @@ class PushRuleStore(PushRulesWorkerStore): actions_json, update_stream=True, ): - """Specialised version of _simple_upsert_txn that picks a push_rule_id + """Specialised version of simple_upsert_txn that picks a push_rule_id using the _push_rule_id_gen if it needs to insert the rule. It assumes that the "push_rules" table is locked""" @@ -518,7 +519,7 @@ class PushRuleStore(PushRulesWorkerStore): # We didn't update a row with the given rule_id so insert one push_rule_id = self._push_rule_id_gen.get_next() - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="push_rules", values={ @@ -561,7 +562,7 @@ class PushRuleStore(PushRulesWorkerStore): """ def delete_push_rule_txn(txn, stream_id, event_stream_ordering): - self._simple_delete_one_txn( + self.db.simple_delete_one_txn( txn, "push_rules", {"user_name": user_id, "rule_id": rule_id} ) @@ -571,7 +572,7 @@ class PushRuleStore(PushRulesWorkerStore): with self._push_rules_stream_id_gen.get_next() as ids: stream_id, event_stream_ordering = ids - yield self.runInteraction( + yield self.db.runInteraction( "delete_push_rule", delete_push_rule_txn, stream_id, @@ -582,7 +583,7 @@ class PushRuleStore(PushRulesWorkerStore): def set_push_rule_enabled(self, user_id, rule_id, enabled): with self._push_rules_stream_id_gen.get_next() as ids: stream_id, event_stream_ordering = ids - yield self.runInteraction( + yield self.db.runInteraction( "_set_push_rule_enabled_txn", self._set_push_rule_enabled_txn, stream_id, @@ -596,7 +597,7 @@ class PushRuleStore(PushRulesWorkerStore): self, txn, stream_id, event_stream_ordering, user_id, rule_id, enabled ): new_id = self._push_rules_enable_id_gen.get_next() - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, "push_rules_enable", {"user_name": user_id, "rule_id": rule_id}, @@ -636,7 +637,7 @@ class PushRuleStore(PushRulesWorkerStore): update_stream=False, ) else: - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, "push_rules", {"user_name": user_id, "rule_id": rule_id}, @@ -655,7 +656,7 @@ class PushRuleStore(PushRulesWorkerStore): with self._push_rules_stream_id_gen.get_next() as ids: stream_id, event_stream_ordering = ids - yield self.runInteraction( + yield self.db.runInteraction( "set_push_rule_actions", set_push_rule_actions_txn, stream_id, @@ -675,7 +676,7 @@ class PushRuleStore(PushRulesWorkerStore): if data is not None: values.update(data) - self._simple_insert_txn(txn, "push_rules_stream", values=values) + self.db.simple_insert_txn(txn, "push_rules_stream", values=values) txn.call_after(self.get_push_rules_for_user.invalidate, (user_id,)) txn.call_after(self.get_push_rules_enabled_for_user.invalidate, (user_id,)) @@ -699,7 +700,7 @@ class PushRuleStore(PushRulesWorkerStore): txn.execute(sql, (last_id, current_id, limit)) return txn.fetchall() - return self.runInteraction( + return self.db.runInteraction( "get_all_push_rule_updates", get_all_push_rule_updates_txn ) diff --git a/synapse/storage/data_stores/main/pusher.py b/synapse/storage/data_stores/main/pusher.py index d76861cdc0..6b03233262 100644 --- a/synapse/storage/data_stores/main/pusher.py +++ b/synapse/storage/data_stores/main/pusher.py @@ -15,8 +15,7 @@ # limitations under the License. import logging - -import six +from typing import Iterable, Iterator from canonicaljson import encode_canonical_json, json @@ -27,21 +26,16 @@ from synapse.util.caches.descriptors import cachedInlineCallbacks, cachedList logger = logging.getLogger(__name__) -if six.PY2: - db_binary_type = six.moves.builtins.buffer -else: - db_binary_type = memoryview - class PusherWorkerStore(SQLBaseStore): - def _decode_pushers_rows(self, rows): + def _decode_pushers_rows(self, rows: Iterable[dict]) -> Iterator[dict]: + """JSON-decode the data in the rows returned from the `pushers` table + + Drops any rows whose data cannot be decoded + """ for r in rows: dataJson = r["data"] - r["data"] = None try: - if isinstance(dataJson, db_binary_type): - dataJson = str(dataJson).decode("UTF8") - r["data"] = json.loads(dataJson) except Exception as e: logger.warning( @@ -50,16 +44,13 @@ class PusherWorkerStore(SQLBaseStore): dataJson, e.args[0], ) - pass - - if isinstance(r["pushkey"], db_binary_type): - r["pushkey"] = str(r["pushkey"]).decode("UTF8") + continue - return rows + yield r @defer.inlineCallbacks def user_has_pusher(self, user_id): - ret = yield self._simple_select_one_onecol( + ret = yield self.db.simple_select_one_onecol( "pushers", {"user_name": user_id}, "id", allow_none=True ) return ret is not None @@ -72,7 +63,7 @@ class PusherWorkerStore(SQLBaseStore): @defer.inlineCallbacks def get_pushers_by(self, keyvalues): - ret = yield self._simple_select_list( + ret = yield self.db.simple_select_list( "pushers", keyvalues, [ @@ -100,11 +91,11 @@ class PusherWorkerStore(SQLBaseStore): def get_all_pushers(self): def get_pushers(txn): txn.execute("SELECT * FROM pushers") - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) return self._decode_pushers_rows(rows) - rows = yield self.runInteraction("get_all_pushers", get_pushers) + rows = yield self.db.runInteraction("get_all_pushers", get_pushers) return rows def get_all_updated_pushers(self, last_id, current_id, limit): @@ -134,7 +125,7 @@ class PusherWorkerStore(SQLBaseStore): return updated, deleted - return self.runInteraction( + return self.db.runInteraction( "get_all_updated_pushers", get_all_updated_pushers_txn ) @@ -177,7 +168,7 @@ class PusherWorkerStore(SQLBaseStore): return results - return self.runInteraction( + return self.db.runInteraction( "get_all_updated_pushers_rows", get_all_updated_pushers_rows_txn ) @@ -193,7 +184,7 @@ class PusherWorkerStore(SQLBaseStore): inlineCallbacks=True, ) def get_if_users_have_pushers(self, user_ids): - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="pushers", column="user_name", iterable=user_ids, @@ -229,8 +220,8 @@ class PusherStore(PusherWorkerStore): ): with self._pushers_id_gen.get_next() as stream_id: # no need to lock because `pushers` has a unique key on - # (app_id, pushkey, user_name) so _simple_upsert will retry - yield self._simple_upsert( + # (app_id, pushkey, user_name) so simple_upsert will retry + yield self.db.simple_upsert( table="pushers", keyvalues={"app_id": app_id, "pushkey": pushkey, "user_name": user_id}, values={ @@ -255,7 +246,7 @@ class PusherStore(PusherWorkerStore): if user_has_pusher is not True: # invalidate, since we the user might not have had a pusher before - yield self.runInteraction( + yield self.db.runInteraction( "add_pusher", self._invalidate_cache_and_stream, self.get_if_user_has_pusher, @@ -269,7 +260,7 @@ class PusherStore(PusherWorkerStore): txn, self.get_if_user_has_pusher, (user_id,) ) - self._simple_delete_one_txn( + self.db.simple_delete_one_txn( txn, "pushers", {"app_id": app_id, "pushkey": pushkey, "user_name": user_id}, @@ -278,7 +269,7 @@ class PusherStore(PusherWorkerStore): # it's possible for us to end up with duplicate rows for # (app_id, pushkey, user_id) at different stream_ids, but that # doesn't really matter. - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="deleted_pushers", values={ @@ -290,13 +281,13 @@ class PusherStore(PusherWorkerStore): ) with self._pushers_id_gen.get_next() as stream_id: - yield self.runInteraction("delete_pusher", delete_pusher_txn, stream_id) + yield self.db.runInteraction("delete_pusher", delete_pusher_txn, stream_id) @defer.inlineCallbacks def update_pusher_last_stream_ordering( self, app_id, pushkey, user_id, last_stream_ordering ): - yield self._simple_update_one( + yield self.db.simple_update_one( "pushers", {"app_id": app_id, "pushkey": pushkey, "user_name": user_id}, {"last_stream_ordering": last_stream_ordering}, @@ -319,7 +310,7 @@ class PusherStore(PusherWorkerStore): Returns: Deferred[bool]: True if the pusher still exists; False if it has been deleted. """ - updated = yield self._simple_update( + updated = yield self.db.simple_update( table="pushers", keyvalues={"app_id": app_id, "pushkey": pushkey, "user_name": user_id}, updatevalues={ @@ -333,7 +324,7 @@ class PusherStore(PusherWorkerStore): @defer.inlineCallbacks def update_pusher_failing_since(self, app_id, pushkey, user_id, failing_since): - yield self._simple_update( + yield self.db.simple_update( table="pushers", keyvalues={"app_id": app_id, "pushkey": pushkey, "user_name": user_id}, updatevalues={"failing_since": failing_since}, @@ -342,7 +333,7 @@ class PusherStore(PusherWorkerStore): @defer.inlineCallbacks def get_throttle_params_by_room(self, pusher_id): - res = yield self._simple_select_list( + res = yield self.db.simple_select_list( "pusher_throttle", {"pusher": pusher_id}, ["room_id", "last_sent_ts", "throttle_ms"], @@ -361,8 +352,8 @@ class PusherStore(PusherWorkerStore): @defer.inlineCallbacks def set_throttle_params(self, pusher_id, room_id, params): # no need to lock because `pusher_throttle` has a primary key on - # (pusher, room_id) so _simple_upsert will retry - yield self._simple_upsert( + # (pusher, room_id) so simple_upsert will retry + yield self.db.simple_upsert( "pusher_throttle", {"pusher": pusher_id, "room_id": room_id}, params, diff --git a/synapse/storage/data_stores/main/receipts.py b/synapse/storage/data_stores/main/receipts.py index 0c24430f28..96e54d145e 100644 --- a/synapse/storage/data_stores/main/receipts.py +++ b/synapse/storage/data_stores/main/receipts.py @@ -22,6 +22,7 @@ from canonicaljson import json from twisted.internet import defer from synapse.storage._base import SQLBaseStore, make_in_list_sql_clause +from synapse.storage.database import Database from synapse.storage.util.id_generators import StreamIdGenerator from synapse.util.caches.descriptors import cached, cachedInlineCallbacks, cachedList from synapse.util.caches.stream_change_cache import StreamChangeCache @@ -38,8 +39,8 @@ class ReceiptsWorkerStore(SQLBaseStore): # the abstract methods being implemented. __metaclass__ = abc.ABCMeta - def __init__(self, db_conn, hs): - super(ReceiptsWorkerStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(ReceiptsWorkerStore, self).__init__(database, db_conn, hs) self._receipts_stream_cache = StreamChangeCache( "ReceiptsRoomChangeCache", self.get_max_receipt_stream_id() @@ -61,7 +62,7 @@ class ReceiptsWorkerStore(SQLBaseStore): @cached(num_args=2) def get_receipts_for_room(self, room_id, receipt_type): - return self._simple_select_list( + return self.db.simple_select_list( table="receipts_linearized", keyvalues={"room_id": room_id, "receipt_type": receipt_type}, retcols=("user_id", "event_id"), @@ -70,7 +71,7 @@ class ReceiptsWorkerStore(SQLBaseStore): @cached(num_args=3) def get_last_receipt_event_id_for_user(self, user_id, room_id, receipt_type): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="receipts_linearized", keyvalues={ "room_id": room_id, @@ -84,7 +85,7 @@ class ReceiptsWorkerStore(SQLBaseStore): @cachedInlineCallbacks(num_args=2) def get_receipts_for_user(self, user_id, receipt_type): - rows = yield self._simple_select_list( + rows = yield self.db.simple_select_list( table="receipts_linearized", keyvalues={"user_id": user_id, "receipt_type": receipt_type}, retcols=("room_id", "event_id"), @@ -108,7 +109,7 @@ class ReceiptsWorkerStore(SQLBaseStore): txn.execute(sql, (user_id,)) return txn.fetchall() - rows = yield self.runInteraction("get_receipts_for_user_with_orderings", f) + rows = yield self.db.runInteraction("get_receipts_for_user_with_orderings", f) return { row[0]: { "event_id": row[1], @@ -187,11 +188,11 @@ class ReceiptsWorkerStore(SQLBaseStore): txn.execute(sql, (room_id, to_key)) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) return rows - rows = yield self.runInteraction("get_linearized_receipts_for_room", f) + rows = yield self.db.runInteraction("get_linearized_receipts_for_room", f) if not rows: return [] @@ -237,9 +238,11 @@ class ReceiptsWorkerStore(SQLBaseStore): txn.execute(sql + clause, [to_key] + list(args)) - return self.cursor_to_dict(txn) + return self.db.cursor_to_dict(txn) - txn_results = yield self.runInteraction("_get_linearized_receipts_for_rooms", f) + txn_results = yield self.db.runInteraction( + "_get_linearized_receipts_for_rooms", f + ) results = {} for row in txn_results: @@ -280,9 +283,9 @@ class ReceiptsWorkerStore(SQLBaseStore): args.append(limit) txn.execute(sql, args) - return (r[0:5] + (json.loads(r[5]),) for r in txn) + return list(r[0:5] + (json.loads(r[5]),) for r in txn) - return self.runInteraction( + return self.db.runInteraction( "get_all_updated_receipts", get_all_updated_receipts_txn ) @@ -313,14 +316,14 @@ class ReceiptsWorkerStore(SQLBaseStore): class ReceiptsStore(ReceiptsWorkerStore): - def __init__(self, db_conn, hs): + def __init__(self, database: Database, db_conn, hs): # We instantiate this first as the ReceiptsWorkerStore constructor # needs to be able to call get_max_receipt_stream_id self._receipts_id_gen = StreamIdGenerator( db_conn, "receipts_linearized", "stream_id" ) - super(ReceiptsStore, self).__init__(db_conn, hs) + super(ReceiptsStore, self).__init__(database, db_conn, hs) def get_max_receipt_stream_id(self): return self._receipts_id_gen.get_current_token() @@ -335,7 +338,7 @@ class ReceiptsStore(ReceiptsWorkerStore): otherwise, the rx timestamp of the event that the RR corresponds to (or 0 if the event is unknown) """ - res = self._simple_select_one_txn( + res = self.db.simple_select_one_txn( txn, table="events", retcols=["stream_ordering", "received_ts"], @@ -388,7 +391,7 @@ class ReceiptsStore(ReceiptsWorkerStore): (user_id, room_id, receipt_type), ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="receipts_linearized", keyvalues={ @@ -398,7 +401,7 @@ class ReceiptsStore(ReceiptsWorkerStore): }, ) - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="receipts_linearized", values={ @@ -453,13 +456,13 @@ class ReceiptsStore(ReceiptsWorkerStore): else: raise RuntimeError("Unrecognized event_ids: %r" % (event_ids,)) - linearized_event_id = yield self.runInteraction( + linearized_event_id = yield self.db.runInteraction( "insert_receipt_conv", graph_to_linear ) stream_id_manager = self._receipts_id_gen.get_next() with stream_id_manager as stream_id: - event_ts = yield self.runInteraction( + event_ts = yield self.db.runInteraction( "insert_linearized_receipt", self.insert_linearized_receipt_txn, room_id, @@ -488,7 +491,7 @@ class ReceiptsStore(ReceiptsWorkerStore): return stream_id, max_persisted_id def insert_graph_receipt(self, room_id, receipt_type, user_id, event_ids, data): - return self.runInteraction( + return self.db.runInteraction( "insert_graph_receipt", self.insert_graph_receipt_txn, room_id, @@ -514,7 +517,7 @@ class ReceiptsStore(ReceiptsWorkerStore): self._get_linearized_receipts_for_room.invalidate_many, (room_id,) ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="receipts_graph", keyvalues={ @@ -523,7 +526,7 @@ class ReceiptsStore(ReceiptsWorkerStore): "user_id": user_id, }, ) - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="receipts_graph", values={ diff --git a/synapse/storage/data_stores/main/registration.py b/synapse/storage/data_stores/main/registration.py index ee1b2b2bbf..cb4b2b39a0 100644 --- a/synapse/storage/data_stores/main/registration.py +++ b/synapse/storage/data_stores/main/registration.py @@ -19,7 +19,6 @@ import logging import re from six import iterkeys -from six.moves import range from twisted.internet import defer from twisted.internet.defer import Deferred @@ -27,8 +26,8 @@ from twisted.internet.defer import Deferred from synapse.api.constants import UserTypes from synapse.api.errors import Codes, StoreError, SynapseError, ThreepidValidationError from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.storage import background_updates from synapse.storage._base import SQLBaseStore +from synapse.storage.database import Database from synapse.types import UserID from synapse.util.caches.descriptors import cached, cachedInlineCallbacks @@ -38,26 +37,28 @@ logger = logging.getLogger(__name__) class RegistrationWorkerStore(SQLBaseStore): - def __init__(self, db_conn, hs): - super(RegistrationWorkerStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(RegistrationWorkerStore, self).__init__(database, db_conn, hs) self.config = hs.config self.clock = hs.get_clock() @cached() def get_user_by_id(self, user_id): - return self._simple_select_one( + return self.db.simple_select_one( table="users", keyvalues={"name": user_id}, retcols=[ "name", "password_hash", "is_guest", + "admin", "consent_version", "consent_server_notice_sent", "appservice_id", "creation_ts", "user_type", + "deactivated", ], allow_none=True, desc="get_user_by_id", @@ -95,7 +96,7 @@ class RegistrationWorkerStore(SQLBaseStore): including the keys `name`, `is_guest`, `device_id`, `token_id`, `valid_until_ms`. """ - return self.runInteraction( + return self.db.runInteraction( "get_user_by_access_token", self._query_for_auth, token ) @@ -110,7 +111,7 @@ class RegistrationWorkerStore(SQLBaseStore): otherwise int representation of the timestamp (as a number of milliseconds since epoch). """ - res = yield self._simple_select_one_onecol( + res = yield self.db.simple_select_one_onecol( table="account_validity", keyvalues={"user_id": user_id}, retcol="expiration_ts_ms", @@ -138,7 +139,7 @@ class RegistrationWorkerStore(SQLBaseStore): """ def set_account_validity_for_user_txn(txn): - self._simple_update_txn( + self.db.simple_update_txn( txn=txn, table="account_validity", keyvalues={"user_id": user_id}, @@ -152,7 +153,7 @@ class RegistrationWorkerStore(SQLBaseStore): txn, self.get_expiration_ts_for_user, (user_id,) ) - yield self.runInteraction( + yield self.db.runInteraction( "set_account_validity_for_user", set_account_validity_for_user_txn ) @@ -168,7 +169,7 @@ class RegistrationWorkerStore(SQLBaseStore): Raises: StoreError: The provided token is already set for another user. """ - yield self._simple_update_one( + yield self.db.simple_update_one( table="account_validity", keyvalues={"user_id": user_id}, updatevalues={"renewal_token": renewal_token}, @@ -185,7 +186,7 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: defer.Deferred[str]: The ID of the user to which the token belongs. """ - res = yield self._simple_select_one_onecol( + res = yield self.db.simple_select_one_onecol( table="account_validity", keyvalues={"renewal_token": renewal_token}, retcol="user_id", @@ -204,7 +205,7 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: defer.Deferred[str]: The renewal token associated with this user ID. """ - res = yield self._simple_select_one_onecol( + res = yield self.db.simple_select_one_onecol( table="account_validity", keyvalues={"user_id": user_id}, retcol="renewal_token", @@ -230,9 +231,9 @@ class RegistrationWorkerStore(SQLBaseStore): ) values = [False, now_ms, renew_at] txn.execute(sql, values) - return self.cursor_to_dict(txn) + return self.db.cursor_to_dict(txn) - res = yield self.runInteraction( + res = yield self.db.runInteraction( "get_users_expiring_soon", select_users_txn, self.clock.time_msec(), @@ -251,7 +252,7 @@ class RegistrationWorkerStore(SQLBaseStore): email_sent (bool): Flag which indicates whether a renewal email has been sent to this user. """ - yield self._simple_update_one( + yield self.db.simple_update_one( table="account_validity", keyvalues={"user_id": user_id}, updatevalues={"email_sent": email_sent}, @@ -266,7 +267,7 @@ class RegistrationWorkerStore(SQLBaseStore): Args: user_id (str): ID of the user to remove from the account validity table. """ - yield self._simple_delete_one( + yield self.db.simple_delete_one( table="account_validity", keyvalues={"user_id": user_id}, desc="delete_account_validity_for_user", @@ -282,7 +283,7 @@ class RegistrationWorkerStore(SQLBaseStore): Returns (bool): true iff the user is a server admin, false otherwise. """ - res = yield self._simple_select_one_onecol( + res = yield self.db.simple_select_one_onecol( table="users", keyvalues={"name": user.to_string()}, retcol="admin", @@ -300,7 +301,7 @@ class RegistrationWorkerStore(SQLBaseStore): admin (bool): true iff the user is to be a server admin, false otherwise. """ - return self._simple_update_one( + return self.db.simple_update_one( table="users", keyvalues={"name": user.to_string()}, updatevalues={"admin": 1 if admin else 0}, @@ -317,7 +318,7 @@ class RegistrationWorkerStore(SQLBaseStore): ) txn.execute(sql, (token,)) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) if rows: return rows[0] @@ -333,7 +334,9 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: Deferred[bool]: True if user 'user_type' is null or empty string """ - res = yield self.runInteraction("is_real_user", self.is_real_user_txn, user_id) + res = yield self.db.runInteraction( + "is_real_user", self.is_real_user_txn, user_id + ) return res @cachedInlineCallbacks() @@ -346,13 +349,13 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: Deferred[bool]: True if user is of type UserTypes.SUPPORT """ - res = yield self.runInteraction( + res = yield self.db.runInteraction( "is_support_user", self.is_support_user_txn, user_id ) return res def is_real_user_txn(self, txn, user_id): - res = self._simple_select_one_onecol_txn( + res = self.db.simple_select_one_onecol_txn( txn=txn, table="users", keyvalues={"name": user_id}, @@ -362,7 +365,7 @@ class RegistrationWorkerStore(SQLBaseStore): return res is None def is_support_user_txn(self, txn, user_id): - res = self._simple_select_one_onecol_txn( + res = self.db.simple_select_one_onecol_txn( txn=txn, table="users", keyvalues={"name": user_id}, @@ -377,13 +380,11 @@ class RegistrationWorkerStore(SQLBaseStore): """ def f(txn): - sql = ( - "SELECT name, password_hash FROM users" " WHERE lower(name) = lower(?)" - ) + sql = "SELECT name, password_hash FROM users WHERE lower(name) = lower(?)" txn.execute(sql, (user_id,)) return dict(txn) - return self.runInteraction("get_users_by_id_case_insensitive", f) + return self.db.runInteraction("get_users_by_id_case_insensitive", f) async def get_user_by_external_id( self, auth_provider: str, external_id: str @@ -397,7 +398,7 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: str|None: the mxid of the user, or None if they are not known """ - return await self._simple_select_one_onecol( + return await self.db.simple_select_one_onecol( table="user_external_ids", keyvalues={"auth_provider": auth_provider, "external_id": external_id}, retcol="user_id", @@ -411,12 +412,12 @@ class RegistrationWorkerStore(SQLBaseStore): def _count_users(txn): txn.execute("SELECT COUNT(*) AS users FROM users") - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) if rows: return rows[0]["users"] return 0 - ret = yield self.runInteraction("count_users", _count_users) + ret = yield self.db.runInteraction("count_users", _count_users) return ret def count_daily_user_type(self): @@ -448,7 +449,7 @@ class RegistrationWorkerStore(SQLBaseStore): results[row[0]] = row[1] return results - return self.runInteraction("count_daily_user_type", _count_daily_user_type) + return self.db.runInteraction("count_daily_user_type", _count_daily_user_type) @defer.inlineCallbacks def count_nonbridged_users(self): @@ -462,7 +463,7 @@ class RegistrationWorkerStore(SQLBaseStore): (count,) = txn.fetchone() return count - ret = yield self.runInteraction("count_users", _count_users) + ret = yield self.db.runInteraction("count_users", _count_users) return ret @defer.inlineCallbacks @@ -471,12 +472,12 @@ class RegistrationWorkerStore(SQLBaseStore): def _count_users(txn): txn.execute("SELECT COUNT(*) AS users FROM users where user_type is null") - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) if rows: return rows[0]["users"] return 0 - ret = yield self.runInteraction("count_real_users", _count_users) + ret = yield self.db.runInteraction("count_real_users", _count_users) return ret @defer.inlineCallbacks @@ -484,12 +485,8 @@ class RegistrationWorkerStore(SQLBaseStore): """ Gets the localpart of the next generated user ID. - Generated user IDs are integers, and we aim for them to be as small as - we can. Unfortunately, it's possible some of them are already taken by - existing users, and there may be gaps in the already taken range. This - function returns the start of the first allocatable gap. This is to - avoid the case of ID 1000 being pre-allocated and starting at 1001 while - 0-999 are available. + Generated user IDs are integers, so we find the largest integer user ID + already taken and return that plus one. """ def _find_next_generated_user_id(txn): @@ -499,19 +496,18 @@ class RegistrationWorkerStore(SQLBaseStore): regex = re.compile(r"^@(\d+):") - found = set() + max_found = 0 for (user_id,) in txn: match = regex.search(user_id) if match: - found.add(int(match.group(1))) - for i in range(len(found) + 1): - if i not in found: - return i + max_found = max(int(match.group(1)), max_found) + + return max_found + 1 return ( ( - yield self.runInteraction( + yield self.db.runInteraction( "find_next_generated_user_id", _find_next_generated_user_id ) ) @@ -528,7 +524,7 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: Deferred[str|None]: user id or None if no user id/threepid mapping exists """ - user_id = yield self.runInteraction( + user_id = yield self.db.runInteraction( "get_user_id_by_threepid", self.get_user_id_by_threepid_txn, medium, address ) return user_id @@ -544,7 +540,7 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: str|None: user id or None if no user id/threepid mapping exists """ - ret = self._simple_select_one_txn( + ret = self.db.simple_select_one_txn( txn, "user_threepids", {"medium": medium, "address": address}, @@ -557,7 +553,7 @@ class RegistrationWorkerStore(SQLBaseStore): @defer.inlineCallbacks def user_add_threepid(self, user_id, medium, address, validated_at, added_at): - yield self._simple_upsert( + yield self.db.simple_upsert( "user_threepids", {"medium": medium, "address": address}, {"user_id": user_id, "validated_at": validated_at, "added_at": added_at}, @@ -565,7 +561,7 @@ class RegistrationWorkerStore(SQLBaseStore): @defer.inlineCallbacks def user_get_threepids(self, user_id): - ret = yield self._simple_select_list( + ret = yield self.db.simple_select_list( "user_threepids", {"user_id": user_id}, ["medium", "address", "validated_at", "added_at"], @@ -574,9 +570,22 @@ class RegistrationWorkerStore(SQLBaseStore): return ret def user_delete_threepid(self, user_id, medium, address): - return self._simple_delete( + return self.db.simple_delete( "user_threepids", keyvalues={"user_id": user_id, "medium": medium, "address": address}, + desc="user_delete_threepid", + ) + + def user_delete_threepids(self, user_id: str): + """Delete all threepid this user has bound + + Args: + user_id: The user id to delete all threepids of + + """ + return self.db.simple_delete( + "user_threepids", + keyvalues={"user_id": user_id}, desc="user_delete_threepids", ) @@ -596,7 +605,7 @@ class RegistrationWorkerStore(SQLBaseStore): """ # We need to use an upsert, in case they user had already bound the # threepid - return self._simple_upsert( + return self.db.simple_upsert( table="user_threepid_id_server", keyvalues={ "user_id": user_id, @@ -622,7 +631,7 @@ class RegistrationWorkerStore(SQLBaseStore): medium (str): The medium of the threepid (e.g "email") address (str): The address of the threepid (e.g "bob@example.com") """ - return self._simple_select_list( + return self.db.simple_select_list( table="user_threepid_id_server", keyvalues={"user_id": user_id}, retcols=["medium", "address"], @@ -643,7 +652,7 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: Deferred """ - return self._simple_delete( + return self.db.simple_delete( table="user_threepid_id_server", keyvalues={ "user_id": user_id, @@ -666,7 +675,7 @@ class RegistrationWorkerStore(SQLBaseStore): Returns: Deferred[list[str]]: Resolves to a list of identity servers """ - return self._simple_select_onecol( + return self.db.simple_select_onecol( table="user_threepid_id_server", keyvalues={"user_id": user_id, "medium": medium, "address": address}, retcol="id_server", @@ -684,7 +693,7 @@ class RegistrationWorkerStore(SQLBaseStore): defer.Deferred(bool): The requested value. """ - res = yield self._simple_select_one_onecol( + res = yield self.db.simple_select_one_onecol( table="users", keyvalues={"name": user_id}, retcol="deactivated", @@ -751,13 +760,13 @@ class RegistrationWorkerStore(SQLBaseStore): sql += " LIMIT 1" txn.execute(sql, list(keyvalues.values())) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) if not rows: return None return rows[0] - return self.runInteraction( + return self.db.runInteraction( "get_threepid_validation_session", get_threepid_validation_session_txn ) @@ -771,39 +780,37 @@ class RegistrationWorkerStore(SQLBaseStore): """ def delete_threepid_session_txn(txn): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="threepid_validation_token", keyvalues={"session_id": session_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="threepid_validation_session", keyvalues={"session_id": session_id}, ) - return self.runInteraction( + return self.db.runInteraction( "delete_threepid_session", delete_threepid_session_txn ) -class RegistrationBackgroundUpdateStore( - RegistrationWorkerStore, background_updates.BackgroundUpdateStore -): - def __init__(self, db_conn, hs): - super(RegistrationBackgroundUpdateStore, self).__init__(db_conn, hs) +class RegistrationBackgroundUpdateStore(RegistrationWorkerStore): + def __init__(self, database: Database, db_conn, hs): + super(RegistrationBackgroundUpdateStore, self).__init__(database, db_conn, hs) self.clock = hs.get_clock() self.config = hs.config - self.register_background_index_update( + self.db.updates.register_background_index_update( "access_tokens_device_index", index_name="access_tokens_device_id", table="access_tokens", columns=["user_id", "device_id"], ) - self.register_background_index_update( + self.db.updates.register_background_index_update( "users_creation_ts", index_name="users_creation_ts", table="users", @@ -813,13 +820,13 @@ class RegistrationBackgroundUpdateStore( # we no longer use refresh tokens, but it's possible that some people # might have a background update queued to build this index. Just # clear the background update. - self.register_noop_background_update("refresh_tokens_device_index") + self.db.updates.register_noop_background_update("refresh_tokens_device_index") - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "user_threepids_grandfather", self._bg_user_threepids_grandfather ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "users_set_deactivated_flag", self._background_update_set_deactivated_flag ) @@ -852,7 +859,7 @@ class RegistrationBackgroundUpdateStore( (last_user, batch_size), ) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) if not rows: return True, 0 @@ -866,7 +873,7 @@ class RegistrationBackgroundUpdateStore( logger.info("Marked %d rows as deactivated", rows_processed_nb) - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, "users_set_deactivated_flag", {"user_id": rows[-1]["name"]} ) @@ -875,12 +882,12 @@ class RegistrationBackgroundUpdateStore( else: return False, len(rows) - end, nb_processed = yield self.runInteraction( + end, nb_processed = yield self.db.runInteraction( "users_set_deactivated_flag", _background_update_set_deactivated_flag_txn ) if end: - yield self._end_background_update("users_set_deactivated_flag") + yield self.db.updates._end_background_update("users_set_deactivated_flag") return nb_processed @@ -906,21 +913,29 @@ class RegistrationBackgroundUpdateStore( txn.executemany(sql, [(id_server,) for id_server in id_servers]) if id_servers: - yield self.runInteraction( + yield self.db.runInteraction( "_bg_user_threepids_grandfather", _bg_user_threepids_grandfather_txn ) - yield self._end_background_update("user_threepids_grandfather") + yield self.db.updates._end_background_update("user_threepids_grandfather") return 1 class RegistrationStore(RegistrationBackgroundUpdateStore): - def __init__(self, db_conn, hs): - super(RegistrationStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(RegistrationStore, self).__init__(database, db_conn, hs) self._account_validity = hs.config.account_validity + if self._account_validity.enabled: + self._clock.call_later( + 0.0, + run_as_background_process, + "account_validity_set_expiration_dates", + self._set_expiration_date_when_missing, + ) + # Create a background job for culling expired 3PID validity tokens def start_cull(): # run as a background process to make sure that the database transactions @@ -948,7 +963,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): """ next_id = self._access_tokens_id_gen.get_next() - yield self._simple_insert( + yield self.db.simple_insert( "access_tokens", { "id": next_id, @@ -990,7 +1005,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): Raises: StoreError if the user_id could not be registered. """ - return self.runInteraction( + return self.db.runInteraction( "register_user", self._register_user, user_id, @@ -1024,7 +1039,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): # Ensure that the guest user actually exists # ``allow_none=False`` makes this raise an exception # if the row isn't in the database. - self._simple_select_one_txn( + self.db.simple_select_one_txn( txn, "users", keyvalues={"name": user_id, "is_guest": 1}, @@ -1032,7 +1047,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): allow_none=False, ) - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, "users", keyvalues={"name": user_id, "is_guest": 1}, @@ -1046,7 +1061,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): }, ) else: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, "users", values={ @@ -1101,7 +1116,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): external_id: id on that system user_id: complete mxid that it is mapped to """ - return self._simple_insert( + return self.db.simple_insert( table="user_external_ids", values={ "auth_provider": auth_provider, @@ -1119,12 +1134,14 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): """ def user_set_password_hash_txn(txn): - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, "users", {"name": user_id}, {"password_hash": password_hash} ) self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,)) - return self.runInteraction("user_set_password_hash", user_set_password_hash_txn) + return self.db.runInteraction( + "user_set_password_hash", user_set_password_hash_txn + ) def user_set_consent_version(self, user_id, consent_version): """Updates the user table to record privacy policy consent @@ -1139,7 +1156,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): """ def f(txn): - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, table="users", keyvalues={"name": user_id}, @@ -1147,7 +1164,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): ) self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,)) - return self.runInteraction("user_set_consent_version", f) + return self.db.runInteraction("user_set_consent_version", f) def user_set_consent_server_notice_sent(self, user_id, consent_version): """Updates the user table to record that we have sent the user a server @@ -1163,7 +1180,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): """ def f(txn): - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, table="users", keyvalues={"name": user_id}, @@ -1171,7 +1188,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): ) self._invalidate_cache_and_stream(txn, self.get_user_by_id, (user_id,)) - return self.runInteraction("user_set_consent_server_notice_sent", f) + return self.db.runInteraction("user_set_consent_server_notice_sent", f) def user_delete_access_tokens(self, user_id, except_token_id=None, device_id=None): """ @@ -1217,11 +1234,11 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): return tokens_and_devices - return self.runInteraction("user_delete_access_tokens", f) + return self.db.runInteraction("user_delete_access_tokens", f) def delete_access_token(self, access_token): def f(txn): - self._simple_delete_one_txn( + self.db.simple_delete_one_txn( txn, table="access_tokens", keyvalues={"token": access_token} ) @@ -1229,11 +1246,11 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): txn, self.get_user_by_access_token, (access_token,) ) - return self.runInteraction("delete_access_token", f) + return self.db.runInteraction("delete_access_token", f) @cachedInlineCallbacks() def is_guest(self, user_id): - res = yield self._simple_select_one_onecol( + res = yield self.db.simple_select_one_onecol( table="users", keyvalues={"name": user_id}, retcol="is_guest", @@ -1248,7 +1265,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): Adds a user to the table of users who need to be parted from all the rooms they're in """ - return self._simple_insert( + return self.db.simple_insert( "users_pending_deactivation", values={"user_id": user_id}, desc="add_user_pending_deactivation", @@ -1261,7 +1278,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): """ # XXX: This should be simple_delete_one but we failed to put a unique index on # the table, so somehow duplicate entries have ended up in it. - return self._simple_delete( + return self.db.simple_delete( "users_pending_deactivation", keyvalues={"user_id": user_id}, desc="del_user_pending_deactivation", @@ -1272,7 +1289,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): Gets one user from the table of users waiting to be parted from all the rooms they're in. """ - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( "users_pending_deactivation", keyvalues={}, retcol="user_id", @@ -1302,7 +1319,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): # Insert everything into a transaction in order to run atomically def validate_threepid_session_txn(txn): - row = self._simple_select_one_txn( + row = self.db.simple_select_one_txn( txn, table="threepid_validation_session", keyvalues={"session_id": session_id}, @@ -1320,7 +1337,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): 400, "This client_secret does not match the provided session_id" ) - row = self._simple_select_one_txn( + row = self.db.simple_select_one_txn( txn, table="threepid_validation_token", keyvalues={"session_id": session_id, "token": token}, @@ -1345,7 +1362,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): ) # Looks good. Validate the session - self._simple_update_txn( + self.db.simple_update_txn( txn, table="threepid_validation_session", keyvalues={"session_id": session_id}, @@ -1355,7 +1372,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): return next_link # Return next_link if it exists - return self.runInteraction( + return self.db.runInteraction( "validate_threepid_session_txn", validate_threepid_session_txn ) @@ -1388,7 +1405,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): if validated_at: insertion_values["validated_at"] = validated_at - return self._simple_upsert( + return self.db.simple_upsert( table="threepid_validation_session", keyvalues={"session_id": session_id}, values={"last_send_attempt": send_attempt}, @@ -1426,7 +1443,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): def start_or_continue_validation_session_txn(txn): # Create or update a validation session - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="threepid_validation_session", keyvalues={"session_id": session_id}, @@ -1439,7 +1456,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): ) # Create a new validation token with this session ID - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="threepid_validation_token", values={ @@ -1450,7 +1467,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): }, ) - return self.runInteraction( + return self.db.runInteraction( "start_or_continue_validation_session", start_or_continue_validation_session_txn, ) @@ -1465,7 +1482,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): """ return txn.execute(sql, (ts,)) - return self.runInteraction( + return self.db.runInteraction( "cull_expired_threepid_validation_tokens", cull_expired_threepid_validation_tokens_txn, self.clock.time_msec(), @@ -1480,7 +1497,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): deactivated (bool): The value to set for `deactivated`. """ - yield self.runInteraction( + yield self.db.runInteraction( "set_user_deactivated_status", self.set_user_deactivated_status_txn, user_id, @@ -1488,7 +1505,7 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): ) def set_user_deactivated_status_txn(self, txn, user_id, deactivated): - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn=txn, table="users", keyvalues={"name": user_id}, @@ -1497,3 +1514,59 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): self._invalidate_cache_and_stream( txn, self.get_user_deactivated_status, (user_id,) ) + + @defer.inlineCallbacks + def _set_expiration_date_when_missing(self): + """ + Retrieves the list of registered users that don't have an expiration date, and + adds an expiration date for each of them. + """ + + def select_users_with_no_expiration_date_txn(txn): + """Retrieves the list of registered users with no expiration date from the + database, filtering out deactivated users. + """ + sql = ( + "SELECT users.name FROM users" + " LEFT JOIN account_validity ON (users.name = account_validity.user_id)" + " WHERE account_validity.user_id is NULL AND users.deactivated = 0;" + ) + txn.execute(sql, []) + + res = self.db.cursor_to_dict(txn) + if res: + for user in res: + self.set_expiration_date_for_user_txn( + txn, user["name"], use_delta=True + ) + + yield self.db.runInteraction( + "get_users_with_no_expiration_date", + select_users_with_no_expiration_date_txn, + ) + + def set_expiration_date_for_user_txn(self, txn, user_id, use_delta=False): + """Sets an expiration date to the account with the given user ID. + + Args: + user_id (str): User ID to set an expiration date for. + use_delta (bool): If set to False, the expiration date for the user will be + now + validity period. If set to True, this expiration date will be a + random value in the [now + period - d ; now + period] range, d being a + delta equal to 10% of the validity period. + """ + now_ms = self._clock.time_msec() + expiration_ts = now_ms + self._account_validity.period + + if use_delta: + expiration_ts = self.rand.randrange( + expiration_ts - self._account_validity.startup_job_max_delta, + expiration_ts, + ) + + self.db.simple_upsert_txn( + txn, + "account_validity", + keyvalues={"user_id": user_id}, + values={"expiration_ts_ms": expiration_ts, "email_sent": False}, + ) diff --git a/synapse/storage/data_stores/main/rejections.py b/synapse/storage/data_stores/main/rejections.py index 7d5de0ea2e..1c07c7a425 100644 --- a/synapse/storage/data_stores/main/rejections.py +++ b/synapse/storage/data_stores/main/rejections.py @@ -22,7 +22,7 @@ logger = logging.getLogger(__name__) class RejectionsStore(SQLBaseStore): def _store_rejections_txn(self, txn, event_id, reason): - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="rejections", values={ @@ -33,7 +33,7 @@ class RejectionsStore(SQLBaseStore): ) def get_rejection_reason(self, event_id): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="rejections", retcol="reason", keyvalues={"event_id": event_id}, diff --git a/synapse/storage/data_stores/main/relations.py b/synapse/storage/data_stores/main/relations.py index 858f65582b..046c2b4845 100644 --- a/synapse/storage/data_stores/main/relations.py +++ b/synapse/storage/data_stores/main/relations.py @@ -129,7 +129,7 @@ class RelationsWorkerStore(SQLBaseStore): chunk=list(events[:limit]), next_batch=next_batch, prev_batch=from_token ) - return self.runInteraction( + return self.db.runInteraction( "get_recent_references_for_event", _get_recent_references_for_event_txn ) @@ -223,7 +223,7 @@ class RelationsWorkerStore(SQLBaseStore): chunk=list(events[:limit]), next_batch=next_batch, prev_batch=from_token ) - return self.runInteraction( + return self.db.runInteraction( "get_aggregation_groups_for_event", _get_aggregation_groups_for_event_txn ) @@ -268,7 +268,7 @@ class RelationsWorkerStore(SQLBaseStore): if row: return row[0] - edit_id = yield self.runInteraction( + edit_id = yield self.db.runInteraction( "get_applicable_edit", _get_applicable_edit_txn ) @@ -318,7 +318,7 @@ class RelationsWorkerStore(SQLBaseStore): return bool(txn.fetchone()) - return self.runInteraction( + return self.db.runInteraction( "get_if_user_has_annotated_event", _get_if_user_has_annotated_event ) @@ -352,7 +352,7 @@ class RelationsStore(RelationsWorkerStore): aggregation_key = relation.get("key") - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="event_relations", values={ @@ -380,6 +380,6 @@ class RelationsStore(RelationsWorkerStore): redacted_event_id (str): The event that was redacted. """ - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="event_relations", keyvalues={"event_id": redacted_event_id} ) diff --git a/synapse/storage/data_stores/main/room.py b/synapse/storage/data_stores/main/room.py index 67bb1b6f60..49bab62be3 100644 --- a/synapse/storage/data_stores/main/room.py +++ b/synapse/storage/data_stores/main/room.py @@ -17,15 +17,20 @@ import collections import logging import re -from typing import Optional, Tuple +from abc import abstractmethod +from typing import List, Optional, Tuple + +from six import integer_types from canonicaljson import json from twisted.internet import defer +from synapse.api.constants import EventTypes from synapse.api.errors import StoreError from synapse.storage._base import SQLBaseStore from synapse.storage.data_stores.main.search import SearchStore +from synapse.storage.database import Database from synapse.types import ThirdPartyInstanceID from synapse.util.caches.descriptors import cached, cachedInlineCallbacks @@ -42,6 +47,11 @@ RatelimitOverride = collections.namedtuple( class RoomWorkerStore(SQLBaseStore): + def __init__(self, database: Database, db_conn, hs): + super(RoomWorkerStore, self).__init__(database, db_conn, hs) + + self.config = hs.config + def get_room(self, room_id): """Retrieve a room. @@ -50,7 +60,7 @@ class RoomWorkerStore(SQLBaseStore): Returns: A dict containing the room information, or None if the room is unknown. """ - return self._simple_select_one( + return self.db.simple_select_one( table="rooms", keyvalues={"room_id": room_id}, retcols=("room_id", "is_public", "creator"), @@ -59,7 +69,7 @@ class RoomWorkerStore(SQLBaseStore): ) def get_public_room_ids(self): - return self._simple_select_onecol( + return self.db.simple_select_onecol( table="rooms", keyvalues={"is_public": True}, retcol="room_id", @@ -116,7 +126,7 @@ class RoomWorkerStore(SQLBaseStore): txn.execute(sql, query_args) return txn.fetchone()[0] - return self.runInteraction("count_public_rooms", _count_public_rooms_txn) + return self.db.runInteraction("count_public_rooms", _count_public_rooms_txn) @defer.inlineCallbacks def get_largest_public_rooms( @@ -249,21 +259,21 @@ class RoomWorkerStore(SQLBaseStore): def _get_largest_public_rooms_txn(txn): txn.execute(sql, query_args) - results = self.cursor_to_dict(txn) + results = self.db.cursor_to_dict(txn) if not forwards: results.reverse() return results - ret_val = yield self.runInteraction( + ret_val = yield self.db.runInteraction( "get_largest_public_rooms", _get_largest_public_rooms_txn ) defer.returnValue(ret_val) @cached(max_entries=10000) def is_room_blocked(self, room_id): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="blocked_rooms", keyvalues={"room_id": room_id}, retcol="1", @@ -284,7 +294,7 @@ class RoomWorkerStore(SQLBaseStore): of RatelimitOverride are None or 0 then ratelimitng has been disabled for that user entirely. """ - row = yield self._simple_select_one( + row = yield self.db.simple_select_one( table="ratelimit_override", keyvalues={"user_id": user_id}, retcols=("messages_per_second", "burst_count"), @@ -300,8 +310,453 @@ class RoomWorkerStore(SQLBaseStore): else: return None + @cachedInlineCallbacks() + def get_retention_policy_for_room(self, room_id): + """Get the retention policy for a given room. + + If no retention policy has been found for this room, returns a policy defined + by the configured default policy (which has None as both the 'min_lifetime' and + the 'max_lifetime' if no default policy has been defined in the server's + configuration). + + Args: + room_id (str): The ID of the room to get the retention policy of. + + Returns: + dict[int, int]: "min_lifetime" and "max_lifetime" for this room. + """ + + def get_retention_policy_for_room_txn(txn): + txn.execute( + """ + SELECT min_lifetime, max_lifetime FROM room_retention + INNER JOIN current_state_events USING (event_id, room_id) + WHERE room_id = ?; + """, + (room_id,), + ) + + return self.db.cursor_to_dict(txn) + + ret = yield self.db.runInteraction( + "get_retention_policy_for_room", get_retention_policy_for_room_txn, + ) + + # If we don't know this room ID, ret will be None, in this case return the default + # policy. + if not ret: + defer.returnValue( + { + "min_lifetime": self.config.retention_default_min_lifetime, + "max_lifetime": self.config.retention_default_max_lifetime, + } + ) + + row = ret[0] + + # If one of the room's policy's attributes isn't defined, use the matching + # attribute from the default policy. + # The default values will be None if no default policy has been defined, or if one + # of the attributes is missing from the default policy. + if row["min_lifetime"] is None: + row["min_lifetime"] = self.config.retention_default_min_lifetime + + if row["max_lifetime"] is None: + row["max_lifetime"] = self.config.retention_default_max_lifetime + + defer.returnValue(row) + + def get_media_mxcs_in_room(self, room_id): + """Retrieves all the local and remote media MXC URIs in a given room + + Args: + room_id (str) + + Returns: + The local and remote media as a lists of tuples where the key is + the hostname and the value is the media ID. + """ + + def _get_media_mxcs_in_room_txn(txn): + local_mxcs, remote_mxcs = self._get_media_mxcs_in_room_txn(txn, room_id) + local_media_mxcs = [] + remote_media_mxcs = [] + + # Convert the IDs to MXC URIs + for media_id in local_mxcs: + local_media_mxcs.append("mxc://%s/%s" % (self.hs.hostname, media_id)) + for hostname, media_id in remote_mxcs: + remote_media_mxcs.append("mxc://%s/%s" % (hostname, media_id)) + + return local_media_mxcs, remote_media_mxcs + + return self.db.runInteraction( + "get_media_ids_in_room", _get_media_mxcs_in_room_txn + ) + + def quarantine_media_ids_in_room(self, room_id, quarantined_by): + """For a room loops through all events with media and quarantines + the associated media + """ + + logger.info("Quarantining media in room: %s", room_id) + + def _quarantine_media_in_room_txn(txn): + local_mxcs, remote_mxcs = self._get_media_mxcs_in_room_txn(txn, room_id) + total_media_quarantined = 0 + + # Now update all the tables to set the quarantined_by flag + + txn.executemany( + """ + UPDATE local_media_repository + SET quarantined_by = ? + WHERE media_id = ? + """, + ((quarantined_by, media_id) for media_id in local_mxcs), + ) + + txn.executemany( + """ + UPDATE remote_media_cache + SET quarantined_by = ? + WHERE media_origin = ? AND media_id = ? + """, + ( + (quarantined_by, origin, media_id) + for origin, media_id in remote_mxcs + ), + ) + + total_media_quarantined += len(local_mxcs) + total_media_quarantined += len(remote_mxcs) + + return total_media_quarantined + + return self.db.runInteraction( + "quarantine_media_in_room", _quarantine_media_in_room_txn + ) + + def _get_media_mxcs_in_room_txn(self, txn, room_id): + """Retrieves all the local and remote media MXC URIs in a given room + + Args: + txn (cursor) + room_id (str) + + Returns: + The local and remote media as a lists of tuples where the key is + the hostname and the value is the media ID. + """ + mxc_re = re.compile("^mxc://([^/]+)/([^/#?]+)") + + sql = """ + SELECT stream_ordering, json FROM events + JOIN event_json USING (room_id, event_id) + WHERE room_id = ? + %(where_clause)s + AND contains_url = ? AND outlier = ? + ORDER BY stream_ordering DESC + LIMIT ? + """ + txn.execute(sql % {"where_clause": ""}, (room_id, True, False, 100)) + + local_media_mxcs = [] + remote_media_mxcs = [] + + while True: + next_token = None + for stream_ordering, content_json in txn: + next_token = stream_ordering + event_json = json.loads(content_json) + content = event_json["content"] + content_url = content.get("url") + thumbnail_url = content.get("info", {}).get("thumbnail_url") + + for url in (content_url, thumbnail_url): + if not url: + continue + matches = mxc_re.match(url) + if matches: + hostname = matches.group(1) + media_id = matches.group(2) + if hostname == self.hs.hostname: + local_media_mxcs.append(media_id) + else: + remote_media_mxcs.append((hostname, media_id)) + + if next_token is None: + # We've gone through the whole room, so we're finished. + break + + txn.execute( + sql % {"where_clause": "AND stream_ordering < ?"}, + (room_id, next_token, True, False, 100), + ) + + return local_media_mxcs, remote_media_mxcs + + def quarantine_media_by_id( + self, server_name: str, media_id: str, quarantined_by: str, + ): + """quarantines a single local or remote media id + + Args: + server_name: The name of the server that holds this media + media_id: The ID of the media to be quarantined + quarantined_by: The user ID that initiated the quarantine request + """ + logger.info("Quarantining media: %s/%s", server_name, media_id) + is_local = server_name == self.config.server_name + + def _quarantine_media_by_id_txn(txn): + local_mxcs = [media_id] if is_local else [] + remote_mxcs = [(server_name, media_id)] if not is_local else [] + + return self._quarantine_media_txn( + txn, local_mxcs, remote_mxcs, quarantined_by + ) + + return self.db.runInteraction( + "quarantine_media_by_user", _quarantine_media_by_id_txn + ) + + def quarantine_media_ids_by_user(self, user_id: str, quarantined_by: str): + """quarantines all local media associated with a single user + + Args: + user_id: The ID of the user to quarantine media of + quarantined_by: The ID of the user who made the quarantine request + """ + + def _quarantine_media_by_user_txn(txn): + local_media_ids = self._get_media_ids_by_user_txn(txn, user_id) + return self._quarantine_media_txn(txn, local_media_ids, [], quarantined_by) + + return self.db.runInteraction( + "quarantine_media_by_user", _quarantine_media_by_user_txn + ) + + def _get_media_ids_by_user_txn(self, txn, user_id: str, filter_quarantined=True): + """Retrieves local media IDs by a given user + + Args: + txn (cursor) + user_id: The ID of the user to retrieve media IDs of + + Returns: + The local and remote media as a lists of tuples where the key is + the hostname and the value is the media ID. + """ + # Local media + sql = """ + SELECT media_id + FROM local_media_repository + WHERE user_id = ? + """ + if filter_quarantined: + sql += "AND quarantined_by IS NULL" + txn.execute(sql, (user_id,)) + + local_media_ids = [row[0] for row in txn] + + # TODO: Figure out all remote media a user has referenced in a message + + return local_media_ids + + def _quarantine_media_txn( + self, + txn, + local_mxcs: List[str], + remote_mxcs: List[Tuple[str, str]], + quarantined_by: str, + ) -> int: + """Quarantine local and remote media items + + Args: + txn (cursor) + local_mxcs: A list of local mxc URLs + remote_mxcs: A list of (remote server, media id) tuples representing + remote mxc URLs + quarantined_by: The ID of the user who initiated the quarantine request + Returns: + The total number of media items quarantined + """ + total_media_quarantined = 0 + + # Update all the tables to set the quarantined_by flag + txn.executemany( + """ + UPDATE local_media_repository + SET quarantined_by = ? + WHERE media_id = ? + """, + ((quarantined_by, media_id) for media_id in local_mxcs), + ) + + txn.executemany( + """ + UPDATE remote_media_cache + SET quarantined_by = ? + WHERE media_origin = ? AND media_id = ? + """, + ((quarantined_by, origin, media_id) for origin, media_id in remote_mxcs), + ) + + total_media_quarantined += len(local_mxcs) + total_media_quarantined += len(remote_mxcs) + + return total_media_quarantined + + +class RoomBackgroundUpdateStore(SQLBaseStore): + REMOVE_TOMESTONED_ROOMS_BG_UPDATE = "remove_tombstoned_rooms_from_directory" + + def __init__(self, database: Database, db_conn, hs): + super(RoomBackgroundUpdateStore, self).__init__(database, db_conn, hs) + + self.config = hs.config + + self.db.updates.register_background_update_handler( + "insert_room_retention", self._background_insert_retention, + ) + + self.db.updates.register_background_update_handler( + self.REMOVE_TOMESTONED_ROOMS_BG_UPDATE, + self._remove_tombstoned_rooms_from_directory, + ) + + @defer.inlineCallbacks + def _background_insert_retention(self, progress, batch_size): + """Retrieves a list of all rooms within a range and inserts an entry for each of + them into the room_retention table. + NULLs the property's columns if missing from the retention event in the room's + state (or NULLs all of them if there's no retention event in the room's state), + so that we fall back to the server's retention policy. + """ + + last_room = progress.get("room_id", "") + + def _background_insert_retention_txn(txn): + txn.execute( + """ + SELECT state.room_id, state.event_id, events.json + FROM current_state_events as state + LEFT JOIN event_json AS events ON (state.event_id = events.event_id) + WHERE state.room_id > ? AND state.type = '%s' + ORDER BY state.room_id ASC + LIMIT ?; + """ + % EventTypes.Retention, + (last_room, batch_size), + ) + + rows = self.db.cursor_to_dict(txn) + + if not rows: + return True + + for row in rows: + if not row["json"]: + retention_policy = {} + else: + ev = json.loads(row["json"]) + retention_policy = json.dumps(ev["content"]) + + self.db.simple_insert_txn( + txn=txn, + table="room_retention", + values={ + "room_id": row["room_id"], + "event_id": row["event_id"], + "min_lifetime": retention_policy.get("min_lifetime"), + "max_lifetime": retention_policy.get("max_lifetime"), + }, + ) + + logger.info("Inserted %d rows into room_retention", len(rows)) + + self.db.updates._background_update_progress_txn( + txn, "insert_room_retention", {"room_id": rows[-1]["room_id"]} + ) + + if batch_size > len(rows): + return True + else: + return False + + end = yield self.db.runInteraction( + "insert_room_retention", _background_insert_retention_txn, + ) + + if end: + yield self.db.updates._end_background_update("insert_room_retention") + + defer.returnValue(batch_size) + + async def _remove_tombstoned_rooms_from_directory( + self, progress, batch_size + ) -> int: + """Removes any rooms with tombstone events from the room directory + + Nowadays this is handled by the room upgrade handler, but we may have some + that got left behind + """ + + last_room = progress.get("room_id", "") + + def _get_rooms(txn): + txn.execute( + """ + SELECT room_id + FROM rooms r + INNER JOIN current_state_events cse USING (room_id) + WHERE room_id > ? AND r.is_public + AND cse.type = '%s' AND cse.state_key = '' + ORDER BY room_id ASC + LIMIT ?; + """ + % EventTypes.Tombstone, + (last_room, batch_size), + ) + + return [row[0] for row in txn] + + rooms = await self.db.runInteraction( + "get_tombstoned_directory_rooms", _get_rooms + ) + + if not rooms: + await self.db.updates._end_background_update( + self.REMOVE_TOMESTONED_ROOMS_BG_UPDATE + ) + return 0 + + for room_id in rooms: + logger.info("Removing tombstoned room %s from the directory", room_id) + await self.set_room_is_public(room_id, False) + + await self.db.updates._background_update_progress( + self.REMOVE_TOMESTONED_ROOMS_BG_UPDATE, {"room_id": rooms[-1]} + ) + + return len(rooms) + + @abstractmethod + def set_room_is_public(self, room_id, is_public): + # this will need to be implemented if a background update is performed with + # existing (tombstoned, public) rooms in the database. + # + # It's overridden by RoomStore for the synapse master. + raise NotImplementedError() + + +class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore, SearchStore): + def __init__(self, database: Database, db_conn, hs): + super(RoomStore, self).__init__(database, db_conn, hs) + + self.config = hs.config -class RoomStore(RoomWorkerStore, SearchStore): @defer.inlineCallbacks def store_room(self, room_id, room_creator_user_id, is_public): """Stores a room. @@ -317,7 +772,7 @@ class RoomStore(RoomWorkerStore, SearchStore): try: def store_room_txn(txn, next_id): - self._simple_insert_txn( + self.db.simple_insert_txn( txn, "rooms", { @@ -327,7 +782,7 @@ class RoomStore(RoomWorkerStore, SearchStore): }, ) if is_public: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="public_room_list_stream", values={ @@ -338,7 +793,7 @@ class RoomStore(RoomWorkerStore, SearchStore): ) with self._public_room_id_gen.get_next() as next_id: - yield self.runInteraction("store_room_txn", store_room_txn, next_id) + yield self.db.runInteraction("store_room_txn", store_room_txn, next_id) except Exception as e: logger.error("store_room with room_id=%s failed: %s", room_id, e) raise StoreError(500, "Problem creating room.") @@ -346,14 +801,14 @@ class RoomStore(RoomWorkerStore, SearchStore): @defer.inlineCallbacks def set_room_is_public(self, room_id, is_public): def set_room_is_public_txn(txn, next_id): - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, table="rooms", keyvalues={"room_id": room_id}, updatevalues={"is_public": is_public}, ) - entries = self._simple_select_list_txn( + entries = self.db.simple_select_list_txn( txn, table="public_room_list_stream", keyvalues={ @@ -371,7 +826,7 @@ class RoomStore(RoomWorkerStore, SearchStore): add_to_stream = bool(entries[-1]["visibility"]) != is_public if add_to_stream: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="public_room_list_stream", values={ @@ -384,7 +839,7 @@ class RoomStore(RoomWorkerStore, SearchStore): ) with self._public_room_id_gen.get_next() as next_id: - yield self.runInteraction( + yield self.db.runInteraction( "set_room_is_public", set_room_is_public_txn, next_id ) self.hs.get_notifier().on_new_replication_data() @@ -411,7 +866,7 @@ class RoomStore(RoomWorkerStore, SearchStore): def set_room_is_public_appservice_txn(txn, next_id): if is_public: try: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="appservice_room_list", values={ @@ -424,7 +879,7 @@ class RoomStore(RoomWorkerStore, SearchStore): # We've already inserted, nothing to do. return else: - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="appservice_room_list", keyvalues={ @@ -434,7 +889,7 @@ class RoomStore(RoomWorkerStore, SearchStore): }, ) - entries = self._simple_select_list_txn( + entries = self.db.simple_select_list_txn( txn, table="public_room_list_stream", keyvalues={ @@ -452,7 +907,7 @@ class RoomStore(RoomWorkerStore, SearchStore): add_to_stream = bool(entries[-1]["visibility"]) != is_public if add_to_stream: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="public_room_list_stream", values={ @@ -465,7 +920,7 @@ class RoomStore(RoomWorkerStore, SearchStore): ) with self._public_room_id_gen.get_next() as next_id: - yield self.runInteraction( + yield self.db.runInteraction( "set_room_is_public_appservice", set_room_is_public_appservice_txn, next_id, @@ -482,7 +937,7 @@ class RoomStore(RoomWorkerStore, SearchStore): row = txn.fetchone() return row[0] or 0 - return self.runInteraction("get_rooms", f) + return self.db.runInteraction("get_rooms", f) def _store_room_topic_txn(self, txn, event): if hasattr(event, "content") and "topic" in event.content: @@ -502,11 +957,40 @@ class RoomStore(RoomWorkerStore, SearchStore): txn, event, "content.body", event.content["body"] ) + def _store_retention_policy_for_room_txn(self, txn, event): + if hasattr(event, "content") and ( + "min_lifetime" in event.content or "max_lifetime" in event.content + ): + if ( + "min_lifetime" in event.content + and not isinstance(event.content.get("min_lifetime"), integer_types) + ) or ( + "max_lifetime" in event.content + and not isinstance(event.content.get("max_lifetime"), integer_types) + ): + # Ignore the event if one of the value isn't an integer. + return + + self.db.simple_insert_txn( + txn=txn, + table="room_retention", + values={ + "room_id": event.room_id, + "event_id": event.event_id, + "min_lifetime": event.content.get("min_lifetime"), + "max_lifetime": event.content.get("max_lifetime"), + }, + ) + + self._invalidate_cache_and_stream( + txn, self.get_retention_policy_for_room, (event.room_id,) + ) + def add_event_report( self, room_id, event_id, user_id, reason, content, received_ts ): next_id = self._event_reports_id_gen.get_next() - return self._simple_insert( + return self.db.simple_insert( table="event_reports", values={ "id": next_id, @@ -539,7 +1023,9 @@ class RoomStore(RoomWorkerStore, SearchStore): if prev_id == current_id: return defer.succeed([]) - return self.runInteraction("get_all_new_public_rooms", get_all_new_public_rooms) + return self.db.runInteraction( + "get_all_new_public_rooms", get_all_new_public_rooms + ) @defer.inlineCallbacks def block_room(self, room_id, user_id): @@ -552,134 +1038,102 @@ class RoomStore(RoomWorkerStore, SearchStore): Returns: Deferred """ - yield self._simple_upsert( + yield self.db.simple_upsert( table="blocked_rooms", keyvalues={"room_id": room_id}, values={}, insertion_values={"user_id": user_id}, desc="block_room", ) - yield self.runInteraction( + yield self.db.runInteraction( "block_room_invalidation", self._invalidate_cache_and_stream, self.is_room_blocked, (room_id,), ) - def get_media_mxcs_in_room(self, room_id): - """Retrieves all the local and remote media MXC URIs in a given room + @defer.inlineCallbacks + def get_rooms_for_retention_period_in_range( + self, min_ms, max_ms, include_null=False + ): + """Retrieves all of the rooms within the given retention range. + + Optionally includes the rooms which don't have a retention policy. Args: - room_id (str) + min_ms (int|None): Duration in milliseconds that define the lower limit of + the range to handle (exclusive). If None, doesn't set a lower limit. + max_ms (int|None): Duration in milliseconds that define the upper limit of + the range to handle (inclusive). If None, doesn't set an upper limit. + include_null (bool): Whether to include rooms which retention policy is NULL + in the returned set. Returns: - The local and remote media as a lists of tuples where the key is - the hostname and the value is the media ID. + dict[str, dict]: The rooms within this range, along with their retention + policy. The key is "room_id", and maps to a dict describing the retention + policy associated with this room ID. The keys for this nested dict are + "min_lifetime" (int|None), and "max_lifetime" (int|None). """ - def _get_media_mxcs_in_room_txn(txn): - local_mxcs, remote_mxcs = self._get_media_mxcs_in_room_txn(txn, room_id) - local_media_mxcs = [] - remote_media_mxcs = [] + def get_rooms_for_retention_period_in_range_txn(txn): + range_conditions = [] + args = [] - # Convert the IDs to MXC URIs - for media_id in local_mxcs: - local_media_mxcs.append("mxc://%s/%s" % (self.hs.hostname, media_id)) - for hostname, media_id in remote_mxcs: - remote_media_mxcs.append("mxc://%s/%s" % (hostname, media_id)) - - return local_media_mxcs, remote_media_mxcs - - return self.runInteraction("get_media_ids_in_room", _get_media_mxcs_in_room_txn) - - def quarantine_media_ids_in_room(self, room_id, quarantined_by): - """For a room loops through all events with media and quarantines - the associated media - """ - - def _quarantine_media_in_room_txn(txn): - local_mxcs, remote_mxcs = self._get_media_mxcs_in_room_txn(txn, room_id) - total_media_quarantined = 0 + if min_ms is not None: + range_conditions.append("max_lifetime > ?") + args.append(min_ms) - # Now update all the tables to set the quarantined_by flag + if max_ms is not None: + range_conditions.append("max_lifetime <= ?") + args.append(max_ms) - txn.executemany( + # Do a first query which will retrieve the rooms that have a retention policy + # in their current state. + sql = """ + SELECT room_id, min_lifetime, max_lifetime FROM room_retention + INNER JOIN current_state_events USING (event_id, room_id) """ - UPDATE local_media_repository - SET quarantined_by = ? - WHERE media_id = ? - """, - ((quarantined_by, media_id) for media_id in local_mxcs), - ) - txn.executemany( - """ - UPDATE remote_media_cache - SET quarantined_by = ? - WHERE media_origin = ? AND media_id = ? - """, - ( - (quarantined_by, origin, media_id) - for origin, media_id in remote_mxcs - ), - ) + if len(range_conditions): + sql += " WHERE (" + " AND ".join(range_conditions) + ")" - total_media_quarantined += len(local_mxcs) - total_media_quarantined += len(remote_mxcs) + if include_null: + sql += " OR max_lifetime IS NULL" - return total_media_quarantined + txn.execute(sql, args) - return self.runInteraction( - "quarantine_media_in_room", _quarantine_media_in_room_txn - ) + rows = self.db.cursor_to_dict(txn) + rooms_dict = {} - def _get_media_mxcs_in_room_txn(self, txn, room_id): - """Retrieves all the local and remote media MXC URIs in a given room + for row in rows: + rooms_dict[row["room_id"]] = { + "min_lifetime": row["min_lifetime"], + "max_lifetime": row["max_lifetime"], + } - Args: - txn (cursor) - room_id (str) + if include_null: + # If required, do a second query that retrieves all of the rooms we know + # of so we can handle rooms with no retention policy. + sql = "SELECT DISTINCT room_id FROM current_state_events" - Returns: - The local and remote media as a lists of tuples where the key is - the hostname and the value is the media ID. - """ - mxc_re = re.compile("^mxc://([^/]+)/([^/#?]+)") + txn.execute(sql) - next_token = self.get_current_events_token() + 1 - local_media_mxcs = [] - remote_media_mxcs = [] + rows = self.db.cursor_to_dict(txn) - while next_token: - sql = """ - SELECT stream_ordering, json FROM events - JOIN event_json USING (room_id, event_id) - WHERE room_id = ? - AND stream_ordering < ? - AND contains_url = ? AND outlier = ? - ORDER BY stream_ordering DESC - LIMIT ? - """ - txn.execute(sql, (room_id, next_token, True, False, 100)) + # 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"]] = { + "min_lifetime": None, + "max_lifetime": None, + } - next_token = None - for stream_ordering, content_json in txn: - next_token = stream_ordering - event_json = json.loads(content_json) - content = event_json["content"] - content_url = content.get("url") - thumbnail_url = content.get("info", {}).get("thumbnail_url") + return rooms_dict - for url in (content_url, thumbnail_url): - if not url: - continue - matches = mxc_re.match(url) - if matches: - hostname = matches.group(1) - media_id = matches.group(2) - if hostname == self.hs.hostname: - local_media_mxcs.append(media_id) - else: - remote_media_mxcs.append((hostname, media_id)) + rooms = yield self.db.runInteraction( + "get_rooms_for_retention_period_in_range", + get_rooms_for_retention_period_in_range_txn, + ) - return local_media_mxcs, remote_media_mxcs + defer.returnValue(rooms) diff --git a/synapse/storage/data_stores/main/roommember.py b/synapse/storage/data_stores/main/roommember.py index 2af24a20b7..9acef7c950 100644 --- a/synapse/storage/data_stores/main/roommember.py +++ b/synapse/storage/data_stores/main/roommember.py @@ -15,6 +15,7 @@ # limitations under the License. import logging +from typing import Iterable, List from six import iteritems, itervalues @@ -25,9 +26,13 @@ from twisted.internet import defer from synapse.api.constants import EventTypes, Membership from synapse.metrics import LaterGauge from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.storage._base import LoggingTransaction, make_in_list_sql_clause -from synapse.storage.background_updates import BackgroundUpdateStore +from synapse.storage._base import ( + LoggingTransaction, + SQLBaseStore, + make_in_list_sql_clause, +) from synapse.storage.data_stores.main.events_worker import EventsWorkerStore +from synapse.storage.database import Database from synapse.storage.engines import Sqlite3Engine from synapse.storage.roommember import ( GetRoomsForUserWithStreamOrdering, @@ -50,8 +55,8 @@ _CURRENT_STATE_MEMBERSHIP_UPDATE_NAME = "current_state_events_membership" class RoomMemberWorkerStore(EventsWorkerStore): - def __init__(self, db_conn, hs): - super(RoomMemberWorkerStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(RoomMemberWorkerStore, self).__init__(database, db_conn, hs) # Is the current_state_events.membership up to date? Or is the # background update still running? @@ -115,7 +120,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): txn.execute(query) return list(txn)[0][0] - count = yield self.runInteraction("get_known_servers", _transact) + count = yield self.db.runInteraction("get_known_servers", _transact) # We always know about ourselves, even if we have nothing in # room_memberships (for example, the server is new). @@ -127,7 +132,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): membership column is up to date """ - pending_update = self._simple_select_one_txn( + pending_update = self.db.simple_select_one_txn( txn, table="background_updates", keyvalues={"update_name": _CURRENT_STATE_MEMBERSHIP_UPDATE_NAME}, @@ -143,7 +148,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): 15.0, run_as_background_process, "_check_safe_current_state_events_membership_updated", - self.runInteraction, + self.db.runInteraction, "_check_safe_current_state_events_membership_updated", self._check_safe_current_state_events_membership_updated_txn, ) @@ -160,7 +165,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): @cached(max_entries=100000, iterable=True) def get_users_in_room(self, room_id): - return self.runInteraction( + return self.db.runInteraction( "get_users_in_room", self.get_users_in_room_txn, room_id ) @@ -268,7 +273,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): return res - return self.runInteraction("get_room_summary", _get_room_summary_txn) + return self.db.runInteraction("get_room_summary", _get_room_summary_txn) def _get_user_counts_in_room_txn(self, txn, room_id): """ @@ -292,19 +297,22 @@ class RoomMemberWorkerStore(EventsWorkerStore): return {row[0]: row[1] for row in txn} @cached() - def get_invited_rooms_for_user(self, user_id): - """ Get all the rooms the user is invited to + def get_invited_rooms_for_local_user(self, user_id): + """ Get all the rooms the *local* user is invited to + Args: user_id (str): The user ID. Returns: A deferred list of RoomsForUser. """ - return self.get_rooms_for_user_where_membership_is(user_id, [Membership.INVITE]) + return self.get_rooms_for_local_user_where_membership_is( + user_id, [Membership.INVITE] + ) @defer.inlineCallbacks - def get_invite_for_user_in_room(self, user_id, room_id): - """Gets the invite for the given user and room + def get_invite_for_local_user_in_room(self, user_id, room_id): + """Gets the invite for the given *local* user and room Args: user_id (str) @@ -314,15 +322,15 @@ class RoomMemberWorkerStore(EventsWorkerStore): Deferred: Resolves to either a RoomsForUser or None if no invite was found. """ - invites = yield self.get_invited_rooms_for_user(user_id) + invites = yield self.get_invited_rooms_for_local_user(user_id) for invite in invites: if invite.room_id == room_id: return invite return None @defer.inlineCallbacks - def get_rooms_for_user_where_membership_is(self, user_id, membership_list): - """ Get all the rooms for this user where the membership for this user + def get_rooms_for_local_user_where_membership_is(self, user_id, membership_list): + """ Get all the rooms for this *local* user where the membership for this user matches one in the membership list. Filters out forgotten rooms. @@ -338,9 +346,9 @@ class RoomMemberWorkerStore(EventsWorkerStore): if not membership_list: return defer.succeed(None) - rooms = yield self.runInteraction( - "get_rooms_for_user_where_membership_is", - self._get_rooms_for_user_where_membership_is_txn, + rooms = yield self.db.runInteraction( + "get_rooms_for_local_user_where_membership_is", + self._get_rooms_for_local_user_where_membership_is_txn, user_id, membership_list, ) @@ -349,76 +357,42 @@ class RoomMemberWorkerStore(EventsWorkerStore): forgotten_rooms = yield self.get_forgotten_rooms_for_user(user_id) return [room for room in rooms if room.room_id not in forgotten_rooms] - def _get_rooms_for_user_where_membership_is_txn( + def _get_rooms_for_local_user_where_membership_is_txn( self, txn, user_id, membership_list ): + # Paranoia check. + if not self.hs.is_mine_id(user_id): + raise Exception( + "Cannot call 'get_rooms_for_local_user_where_membership_is' on non-local user %r" + % (user_id,), + ) - do_invite = Membership.INVITE in membership_list - membership_list = [m for m in membership_list if m != Membership.INVITE] - - results = [] - if membership_list: - if self._current_state_events_membership_up_to_date: - clause, args = make_in_list_sql_clause( - self.database_engine, "c.membership", membership_list - ) - sql = """ - SELECT room_id, e.sender, c.membership, event_id, e.stream_ordering - FROM current_state_events AS c - INNER JOIN events AS e USING (room_id, event_id) - WHERE - c.type = 'm.room.member' - AND state_key = ? - AND %s - """ % ( - clause, - ) - else: - clause, args = make_in_list_sql_clause( - self.database_engine, "m.membership", membership_list - ) - sql = """ - SELECT room_id, e.sender, m.membership, event_id, e.stream_ordering - FROM current_state_events AS c - INNER JOIN room_memberships AS m USING (room_id, event_id) - INNER JOIN events AS e USING (room_id, event_id) - WHERE - c.type = 'm.room.member' - AND state_key = ? - AND %s - """ % ( - clause, - ) - - txn.execute(sql, (user_id, *args)) - results = [RoomsForUser(**r) for r in self.cursor_to_dict(txn)] + clause, args = make_in_list_sql_clause( + self.database_engine, "c.membership", membership_list + ) - if do_invite: - sql = ( - "SELECT i.room_id, inviter, i.event_id, e.stream_ordering" - " FROM local_invites as i" - " INNER JOIN events as e USING (event_id)" - " WHERE invitee = ? AND locally_rejected is NULL" - " AND replaced_by is NULL" - ) + sql = """ + SELECT room_id, e.sender, c.membership, event_id, e.stream_ordering + FROM local_current_membership AS c + INNER JOIN events AS e USING (room_id, event_id) + WHERE + user_id = ? + AND %s + """ % ( + clause, + ) - txn.execute(sql, (user_id,)) - results.extend( - RoomsForUser( - room_id=r["room_id"], - sender=r["inviter"], - event_id=r["event_id"], - stream_ordering=r["stream_ordering"], - membership=Membership.INVITE, - ) - for r in self.cursor_to_dict(txn) - ) + txn.execute(sql, (user_id, *args)) + results = [RoomsForUser(**r) for r in self.db.cursor_to_dict(txn)] return results - @cachedInlineCallbacks(max_entries=500000, iterable=True) + @cached(max_entries=500000, iterable=True) def get_rooms_for_user_with_stream_ordering(self, user_id): - """Returns a set of room_ids the user is currently joined to + """Returns a set of room_ids the user is currently joined to. + + If a remote user only returns rooms this server is currently + participating in. Args: user_id (str) @@ -428,17 +402,49 @@ class RoomMemberWorkerStore(EventsWorkerStore): the rooms the user is in currently, along with the stream ordering of the most recent join for that user and room. """ - rooms = yield self.get_rooms_for_user_where_membership_is( - user_id, membership_list=[Membership.JOIN] - ) - return frozenset( - GetRoomsForUserWithStreamOrdering(r.room_id, r.stream_ordering) - for r in rooms + return self.db.runInteraction( + "get_rooms_for_user_with_stream_ordering", + self._get_rooms_for_user_with_stream_ordering_txn, + user_id, ) + def _get_rooms_for_user_with_stream_ordering_txn(self, txn, user_id): + # We use `current_state_events` here and not `local_current_membership` + # as a) this gets called with remote users and b) this only gets called + # for rooms the server is participating in. + if self._current_state_events_membership_up_to_date: + sql = """ + SELECT room_id, e.stream_ordering + FROM current_state_events AS c + INNER JOIN events AS e USING (room_id, event_id) + WHERE + c.type = 'm.room.member' + AND state_key = ? + AND c.membership = ? + """ + else: + sql = """ + SELECT room_id, e.stream_ordering + FROM current_state_events AS c + INNER JOIN room_memberships AS m USING (room_id, event_id) + INNER JOIN events AS e USING (room_id, event_id) + WHERE + c.type = 'm.room.member' + AND state_key = ? + AND m.membership = ? + """ + + txn.execute(sql, (user_id, Membership.JOIN)) + results = frozenset(GetRoomsForUserWithStreamOrdering(*row) for row in txn) + + return results + @defer.inlineCallbacks def get_rooms_for_user(self, user_id, on_invalidate=None): - """Returns a set of room_ids the user is currently joined to + """Returns a set of room_ids the user is currently joined to. + + If a remote user only returns rooms this server is currently + participating in. """ rooms = yield self.get_rooms_for_user_with_stream_ordering( user_id, on_invalidate=on_invalidate @@ -472,7 +478,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): # To do this we set the state_group to a new object as object() != object() state_group = object() - current_state_ids = yield context.get_current_state_ids(self) + current_state_ids = yield context.get_current_state_ids() result = yield self._get_joined_users_from_context( event.room_id, state_group, current_state_ids, event=event, context=context ) @@ -602,7 +608,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): to `user_id` and ProfileInfo (or None if not join event). """ - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="room_memberships", column="event_id", iterable=event_ids, @@ -642,7 +648,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): # the returned user actually has the correct domain. like_clause = "%:" + host - rows = yield self._execute("is_host_joined", None, sql, room_id, like_clause) + rows = yield self.db.execute("is_host_joined", None, sql, room_id, like_clause) if not rows: return False @@ -682,7 +688,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): # the returned user actually has the correct domain. like_clause = "%:" + host - rows = yield self._execute("was_host_joined", None, sql, room_id, like_clause) + rows = yield self.db.execute("was_host_joined", None, sql, room_id, like_clause) if not rows: return False @@ -752,7 +758,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): rows = txn.fetchall() return rows[0][0] - count = yield self.runInteraction("did_forget_membership", f) + count = yield self.db.runInteraction("did_forget_membership", f) return count == 0 @cached() @@ -789,7 +795,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): txn.execute(sql, (user_id,)) return set(row[0] for row in txn if row[1] == 0) - return self.runInteraction( + return self.db.runInteraction( "get_forgotten_rooms_for_user", _get_forgotten_rooms_for_user_txn ) @@ -804,7 +810,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): Deferred[set[str]]: Set of room IDs. """ - room_ids = yield self._simple_select_onecol( + room_ids = yield self.db.simple_select_onecol( table="room_memberships", keyvalues={"membership": Membership.JOIN, "user_id": user_id}, retcol="room_id", @@ -813,18 +819,34 @@ class RoomMemberWorkerStore(EventsWorkerStore): return set(room_ids) + def get_membership_from_event_ids( + self, member_event_ids: Iterable[str] + ) -> List[dict]: + """Get user_id and membership of a set of event IDs. + """ + + return self.db.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", + ) + -class RoomMemberBackgroundUpdateStore(BackgroundUpdateStore): - def __init__(self, db_conn, hs): - super(RoomMemberBackgroundUpdateStore, self).__init__(db_conn, hs) - self.register_background_update_handler( +class RoomMemberBackgroundUpdateStore(SQLBaseStore): + def __init__(self, database: Database, db_conn, hs): + super(RoomMemberBackgroundUpdateStore, self).__init__(database, db_conn, hs) + self.db.updates.register_background_update_handler( _MEMBERSHIP_PROFILE_UPDATE_NAME, self._background_add_membership_profile ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( _CURRENT_STATE_MEMBERSHIP_UPDATE_NAME, self._background_current_state_membership, ) - self.register_background_index_update( + self.db.updates.register_background_index_update( "room_membership_forgotten_idx", index_name="room_memberships_user_room_forgotten", table="room_memberships", @@ -857,7 +879,7 @@ class RoomMemberBackgroundUpdateStore(BackgroundUpdateStore): txn.execute(sql, (target_min_stream_id, max_stream_id, batch_size)) - rows = self.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) if not rows: return 0 @@ -892,18 +914,20 @@ class RoomMemberBackgroundUpdateStore(BackgroundUpdateStore): "max_stream_id_exclusive": min_stream_id, } - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, _MEMBERSHIP_PROFILE_UPDATE_NAME, progress ) return len(rows) - result = yield self.runInteraction( + result = yield self.db.runInteraction( _MEMBERSHIP_PROFILE_UPDATE_NAME, add_membership_profile_txn ) if not result: - yield self._end_background_update(_MEMBERSHIP_PROFILE_UPDATE_NAME) + yield self.db.updates._end_background_update( + _MEMBERSHIP_PROFILE_UPDATE_NAME + ) return result @@ -942,7 +966,7 @@ class RoomMemberBackgroundUpdateStore(BackgroundUpdateStore): last_processed_room = next_room - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, _CURRENT_STATE_MEMBERSHIP_UPDATE_NAME, {"last_processed_room": last_processed_room}, @@ -954,26 +978,28 @@ class RoomMemberBackgroundUpdateStore(BackgroundUpdateStore): # string, which will compare before all room IDs correctly. last_processed_room = progress.get("last_processed_room", "") - row_count, finished = yield self.runInteraction( + row_count, finished = yield self.db.runInteraction( "_background_current_state_membership_update", _background_current_state_membership_txn, last_processed_room, ) if finished: - yield self._end_background_update(_CURRENT_STATE_MEMBERSHIP_UPDATE_NAME) + yield self.db.updates._end_background_update( + _CURRENT_STATE_MEMBERSHIP_UPDATE_NAME + ) return row_count class RoomMemberStore(RoomMemberWorkerStore, RoomMemberBackgroundUpdateStore): - def __init__(self, db_conn, hs): - super(RoomMemberStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(RoomMemberStore, self).__init__(database, db_conn, hs) def _store_room_members_txn(self, txn, events, backfilled): """Store a room member in the database. """ - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="room_memberships", values=[ @@ -997,7 +1023,7 @@ class RoomMemberStore(RoomMemberWorkerStore, RoomMemberBackgroundUpdateStore): event.internal_metadata.stream_ordering, ) txn.call_after( - self.get_invited_rooms_for_user.invalidate, (event.state_key,) + self.get_invited_rooms_for_local_user.invalidate, (event.state_key,) ) # We update the local_invites table only if the event is "current", @@ -1011,7 +1037,7 @@ class RoomMemberStore(RoomMemberWorkerStore, RoomMemberBackgroundUpdateStore): is_mine = self.hs.is_mine_id(event.state_key) if is_new_state and is_mine: if event.membership == Membership.INVITE: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="local_invites", values={ @@ -1039,6 +1065,27 @@ class RoomMemberStore(RoomMemberWorkerStore, RoomMemberBackgroundUpdateStore): ), ) + # We also update the `local_current_membership` table with + # latest invite info. This will usually get updated by the + # `current_state_events` handling, unless its an outlier. + if event.internal_metadata.is_outlier(): + # This should only happen for out of band memberships, so + # we add a paranoia check. + assert event.internal_metadata.is_out_of_band_membership() + + self.db.simple_upsert_txn( + txn, + table="local_current_membership", + keyvalues={ + "room_id": event.room_id, + "user_id": event.state_key, + }, + values={ + "event_id": event.event_id, + "membership": event.membership, + }, + ) + @defer.inlineCallbacks def locally_reject_invite(self, user_id, room_id): sql = ( @@ -1050,8 +1097,17 @@ class RoomMemberStore(RoomMemberWorkerStore, RoomMemberBackgroundUpdateStore): def f(txn, stream_ordering): txn.execute(sql, (stream_ordering, True, room_id, user_id)) + # We also clear this entry from `local_current_membership`. + # Ideally we'd point to a leave event, but we don't have one, so + # nevermind. + self.db.simple_delete_txn( + txn, + table="local_current_membership", + keyvalues={"room_id": room_id, "user_id": user_id}, + ) + with self._stream_id_gen.get_next() as stream_ordering: - yield self.runInteraction("locally_reject_invite", f, stream_ordering) + yield self.db.runInteraction("locally_reject_invite", f, stream_ordering) def forget(self, user_id, room_id): """Indicate that user_id wishes to discard history for room_id.""" @@ -1074,7 +1130,7 @@ class RoomMemberStore(RoomMemberWorkerStore, RoomMemberBackgroundUpdateStore): txn, self.get_forgotten_rooms_for_user, (user_id,) ) - return self.runInteraction("forget_membership", f) + return self.db.runInteraction("forget_membership", f) class _JoinedHostsCache(object): diff --git a/synapse/storage/data_stores/main/schema/delta/32/remove_indices.sql b/synapse/storage/data_stores/main/schema/delta/32/remove_indices.sql index 4219cdd06a..2de50d408c 100644 --- a/synapse/storage/data_stores/main/schema/delta/32/remove_indices.sql +++ b/synapse/storage/data_stores/main/schema/delta/32/remove_indices.sql @@ -20,7 +20,6 @@ DROP INDEX IF EXISTS events_room_id; -- Prefix of events_room_stream DROP INDEX IF EXISTS events_order; -- Prefix of events_order_topo_stream_room DROP INDEX IF EXISTS events_topological_ordering; -- Prefix of events_order_topo_stream_room DROP INDEX IF EXISTS events_stream_ordering; -- Duplicate of PRIMARY KEY -DROP INDEX IF EXISTS state_groups_id; -- Duplicate of PRIMARY KEY DROP INDEX IF EXISTS event_to_state_groups_id; -- Duplicate of PRIMARY KEY DROP INDEX IF EXISTS event_push_actions_room_id_event_id_user_id_profile_tag; -- Duplicate of UNIQUE CONSTRAINT diff --git a/synapse/storage/data_stores/main/schema/delta/56/device_stream_id_insert.sql b/synapse/storage/data_stores/main/schema/delta/56/device_stream_id_insert.sql new file mode 100644 index 0000000000..c2f557fde9 --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/device_stream_id_insert.sql @@ -0,0 +1,20 @@ +/* Copyright 2019 The Matrix.org Foundation C.I.C. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +-- This line already existed in deltas/35/device_stream_id but was not included in the +-- 54 full schema SQL. Add some SQL here to insert the missing row if it does not exist +INSERT INTO device_max_stream_id (stream_id) SELECT 0 WHERE NOT EXISTS ( + SELECT * from device_max_stream_id +); \ No newline at end of file diff --git a/synapse/storage/data_stores/main/schema/delta/56/event_expiry.sql b/synapse/storage/data_stores/main/schema/delta/56/event_expiry.sql new file mode 100644 index 0000000000..81a36a8b1d --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/event_expiry.sql @@ -0,0 +1,21 @@ +/* Copyright 2019 The Matrix.org Foundation C.I.C. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +CREATE TABLE IF NOT EXISTS event_expiry ( + event_id TEXT PRIMARY KEY, + expiry_ts BIGINT NOT NULL +); + +CREATE INDEX event_expiry_expiry_ts_idx ON event_expiry(expiry_ts); diff --git a/synapse/storage/data_stores/main/schema/delta/56/nuke_empty_communities_from_db.sql b/synapse/storage/data_stores/main/schema/delta/56/nuke_empty_communities_from_db.sql new file mode 100644 index 0000000000..4f24c1405d --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/nuke_empty_communities_from_db.sql @@ -0,0 +1,29 @@ +/* Copyright 2019 Werner Sembach + * + * 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. + */ + +-- Groups/communities now get deleted when the last member leaves. This is a one time cleanup to remove old groups/communities that were already empty before that change was made. +DELETE FROM group_attestations_remote WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_attestations_renewals WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_invites WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_roles WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_room_categories WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_rooms WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_summary_roles WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_summary_room_categories WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_summary_rooms WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM group_summary_users WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM local_group_membership WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM local_group_updates WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); +DELETE FROM groups WHERE group_id IN (SELECT group_id FROM groups WHERE NOT EXISTS (SELECT group_id FROM group_users WHERE group_id = groups.group_id)); diff --git a/synapse/storage/data_stores/main/schema/delta/56/redaction_censor.sql b/synapse/storage/data_stores/main/schema/delta/56/redaction_censor.sql index fe51b02309..ea95db0ed7 100644 --- a/synapse/storage/data_stores/main/schema/delta/56/redaction_censor.sql +++ b/synapse/storage/data_stores/main/schema/delta/56/redaction_censor.sql @@ -14,4 +14,3 @@ */ ALTER TABLE redactions ADD COLUMN have_censored BOOL NOT NULL DEFAULT false; -CREATE INDEX redactions_have_censored ON redactions(event_id) WHERE not have_censored; diff --git a/synapse/storage/data_stores/main/schema/delta/56/redaction_censor2.sql b/synapse/storage/data_stores/main/schema/delta/56/redaction_censor2.sql index 77a5eca499..49ce35d794 100644 --- a/synapse/storage/data_stores/main/schema/delta/56/redaction_censor2.sql +++ b/synapse/storage/data_stores/main/schema/delta/56/redaction_censor2.sql @@ -14,7 +14,9 @@ */ ALTER TABLE redactions ADD COLUMN received_ts BIGINT; -CREATE INDEX redactions_have_censored_ts ON redactions(received_ts) WHERE not have_censored; INSERT INTO background_updates (update_name, progress_json) VALUES ('redactions_received_ts', '{}'); + +INSERT INTO background_updates (update_name, progress_json) VALUES + ('redactions_have_censored_ts_idx', '{}'); diff --git a/synapse/storage/data_stores/main/schema/delta/56/redaction_censor4.sql b/synapse/storage/data_stores/main/schema/delta/56/redaction_censor4.sql new file mode 100644 index 0000000000..b7550f6f4e --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/redaction_censor4.sql @@ -0,0 +1,16 @@ +/* Copyright 2019 The Matrix.org Foundation C.I.C. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +DROP INDEX IF EXISTS redactions_have_censored; diff --git a/synapse/storage/data_stores/main/schema/delta/56/remove_tombstoned_rooms_from_directory.sql b/synapse/storage/data_stores/main/schema/delta/56/remove_tombstoned_rooms_from_directory.sql new file mode 100644 index 0000000000..aeb17813d3 --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/remove_tombstoned_rooms_from_directory.sql @@ -0,0 +1,18 @@ +/* Copyright 2020 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. + */ + +-- Now that #6232 is a thing, we can remove old rooms from the directory. +INSERT INTO background_updates (update_name, progress_json) VALUES + ('remove_tombstoned_rooms_from_directory', '{}'); diff --git a/synapse/storage/data_stores/main/schema/delta/56/room_key_etag.sql b/synapse/storage/data_stores/main/schema/delta/56/room_key_etag.sql new file mode 100644 index 0000000000..7d70dd071e --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/room_key_etag.sql @@ -0,0 +1,17 @@ +/* Copyright 2019 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. + */ + +-- store the current etag of backup version +ALTER TABLE e2e_room_keys_versions ADD COLUMN etag BIGINT; diff --git a/synapse/storage/data_stores/main/schema/delta/56/room_retention.sql b/synapse/storage/data_stores/main/schema/delta/56/room_retention.sql new file mode 100644 index 0000000000..ee6cdf7a14 --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/room_retention.sql @@ -0,0 +1,33 @@ +/* Copyright 2019 New Vector Ltd + * + * 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. + */ + +-- Tracks the retention policy of a room. +-- A NULL max_lifetime or min_lifetime means that the matching property is not defined in +-- the room's retention policy state event. +-- If a room doesn't have a retention policy state event in its state, both max_lifetime +-- and min_lifetime are NULL. +CREATE TABLE IF NOT EXISTS room_retention( + room_id TEXT, + event_id TEXT, + min_lifetime BIGINT, + max_lifetime BIGINT, + + PRIMARY KEY(room_id, event_id) +); + +CREATE INDEX room_retention_max_lifetime_idx on room_retention(max_lifetime); + +INSERT INTO background_updates (update_name, progress_json) VALUES + ('insert_room_retention', '{}'); diff --git a/synapse/storage/data_stores/main/schema/delta/56/signing_keys.sql b/synapse/storage/data_stores/main/schema/delta/56/signing_keys.sql index 27a96123e3..5c5fffcafb 100644 --- a/synapse/storage/data_stores/main/schema/delta/56/signing_keys.sql +++ b/synapse/storage/data_stores/main/schema/delta/56/signing_keys.sql @@ -40,7 +40,8 @@ CREATE TABLE IF NOT EXISTS e2e_cross_signing_signatures ( signature TEXT NOT NULL ); -CREATE UNIQUE INDEX e2e_cross_signing_signatures_idx ON e2e_cross_signing_signatures(user_id, target_user_id, target_device_id); +-- replaced by the index created in signing_keys_nonunique_signatures.sql +-- CREATE UNIQUE INDEX e2e_cross_signing_signatures_idx ON e2e_cross_signing_signatures(user_id, target_user_id, target_device_id); -- stream of user signature updates CREATE TABLE IF NOT EXISTS user_signature_stream ( diff --git a/synapse/storage/data_stores/main/schema/delta/56/signing_keys_nonunique_signatures.sql b/synapse/storage/data_stores/main/schema/delta/56/signing_keys_nonunique_signatures.sql new file mode 100644 index 0000000000..0aa90ebf0c --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/signing_keys_nonunique_signatures.sql @@ -0,0 +1,22 @@ +/* Copyright 2019 The Matrix.org Foundation C.I.C. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/* The cross-signing signatures index should not be a unique index, because a + * user may upload multiple signatures for the same target user. The previous + * index was unique, so delete it if it's there and create a new non-unique + * index. */ + +DROP INDEX IF EXISTS e2e_cross_signing_signatures_idx; CREATE INDEX IF NOT +EXISTS e2e_cross_signing_signatures2_idx ON e2e_cross_signing_signatures(user_id, target_user_id, target_device_id); diff --git a/synapse/storage/data_stores/main/schema/delta/57/local_current_membership.py b/synapse/storage/data_stores/main/schema/delta/57/local_current_membership.py new file mode 100644 index 0000000000..601c236c4a --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/57/local_current_membership.py @@ -0,0 +1,97 @@ +# -*- coding: utf-8 -*- +# Copyright 2020 New Vector Ltd +# +# 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. + + +# We create a new table called `local_current_membership` that stores the latest +# membership state of local users in rooms, which helps track leaves/bans/etc +# even if the server has left the room (and so has deleted the room from +# `current_state_events`). This will also include outstanding invites for local +# users for rooms the server isn't in. +# +# If the server isn't and hasn't been in the room then it will only include +# outsstanding invites, and not e.g. pre-emptive bans of local users. +# +# If the server later rejoins a room `local_current_membership` can simply be +# replaced with the new current state of the room (which results in the +# equivalent behaviour as if the server had remained in the room). + + +def run_upgrade(cur, database_engine, config, *args, **kwargs): + # We need to do the insert in `run_upgrade` section as we don't have access + # to `config` in `run_create`. + + # This upgrade may take a bit of time for large servers (e.g. one minute for + # matrix.org) but means we avoid a lots of book keeping required to do it as + # a background update. + + # We check if the `current_state_events.membership` is up to date by + # checking if the relevant background update has finished. If it has + # finished we can avoid doing a join against `room_memberships`, which + # speesd things up. + cur.execute( + """SELECT 1 FROM background_updates + WHERE update_name = 'current_state_events_membership' + """ + ) + current_state_membership_up_to_date = not bool(cur.fetchone()) + + # Cheekily drop and recreate indices, as that is faster. + cur.execute("DROP INDEX local_current_membership_idx") + cur.execute("DROP INDEX local_current_membership_room_idx") + + if current_state_membership_up_to_date: + sql = """ + INSERT INTO local_current_membership (room_id, user_id, event_id, membership) + SELECT c.room_id, state_key AS user_id, event_id, c.membership + FROM current_state_events AS c + WHERE type = 'm.room.member' AND c.membership IS NOT NULL AND state_key like '%' || ? + """ + else: + # We can't rely on the membership column, so we need to join against + # `room_memberships`. + sql = """ + INSERT INTO local_current_membership (room_id, user_id, event_id, membership) + SELECT c.room_id, state_key AS user_id, event_id, r.membership + FROM current_state_events AS c + INNER JOIN room_memberships AS r USING (event_id) + WHERE type = 'm.room.member' and state_key like '%' || ? + """ + cur.execute(sql, (config.server_name,)) + + cur.execute( + "CREATE UNIQUE INDEX local_current_membership_idx ON local_current_membership(user_id, room_id)" + ) + cur.execute( + "CREATE INDEX local_current_membership_room_idx ON local_current_membership(room_id)" + ) + + +def run_create(cur, database_engine, *args, **kwargs): + cur.execute( + """ + CREATE TABLE local_current_membership ( + room_id TEXT NOT NULL, + user_id TEXT NOT NULL, + event_id TEXT NOT NULL, + membership TEXT NOT NULL + )""" + ) + + cur.execute( + "CREATE UNIQUE INDEX local_current_membership_idx ON local_current_membership(user_id, room_id)" + ) + cur.execute( + "CREATE INDEX local_current_membership_room_idx ON local_current_membership(room_id)" + ) diff --git a/synapse/storage/data_stores/main/schema/full_schemas/54/full.sql.postgres b/synapse/storage/data_stores/main/schema/full_schemas/54/full.sql.postgres index 4ad2929f32..889a9a0ce4 100644 --- a/synapse/storage/data_stores/main/schema/full_schemas/54/full.sql.postgres +++ b/synapse/storage/data_stores/main/schema/full_schemas/54/full.sql.postgres @@ -975,40 +975,6 @@ CREATE TABLE state_events ( -CREATE TABLE state_group_edges ( - state_group bigint NOT NULL, - prev_state_group bigint NOT NULL -); - - - -CREATE SEQUENCE state_group_id_seq - START WITH 1 - INCREMENT BY 1 - NO MINVALUE - NO MAXVALUE - CACHE 1; - - - -CREATE TABLE state_groups ( - id bigint NOT NULL, - room_id text NOT NULL, - event_id text NOT NULL -); - - - -CREATE TABLE state_groups_state ( - state_group bigint NOT NULL, - room_id text NOT NULL, - type text NOT NULL, - state_key text NOT NULL, - event_id text NOT NULL -); - - - CREATE TABLE stats_stream_pos ( lock character(1) DEFAULT 'X'::bpchar NOT NULL, stream_id bigint, @@ -1482,12 +1448,6 @@ ALTER TABLE ONLY state_events ADD CONSTRAINT state_events_event_id_key UNIQUE (event_id); - -ALTER TABLE ONLY state_groups - ADD CONSTRAINT state_groups_pkey PRIMARY KEY (id); - - - ALTER TABLE ONLY stats_stream_pos ADD CONSTRAINT stats_stream_pos_lock_key UNIQUE (lock); @@ -1928,18 +1888,6 @@ CREATE UNIQUE INDEX room_stats_room_ts ON room_stats USING btree (room_id, ts); -CREATE INDEX state_group_edges_idx ON state_group_edges USING btree (state_group); - - - -CREATE INDEX state_group_edges_prev_idx ON state_group_edges USING btree (prev_state_group); - - - -CREATE INDEX state_groups_state_type_idx ON state_groups_state USING btree (state_group, type, state_key); - - - CREATE INDEX stream_ordering_to_exterm_idx ON stream_ordering_to_exterm USING btree (stream_ordering); diff --git a/synapse/storage/data_stores/main/schema/full_schemas/54/full.sql.sqlite b/synapse/storage/data_stores/main/schema/full_schemas/54/full.sql.sqlite index bad33291e7..a0411ede7e 100644 --- a/synapse/storage/data_stores/main/schema/full_schemas/54/full.sql.sqlite +++ b/synapse/storage/data_stores/main/schema/full_schemas/54/full.sql.sqlite @@ -42,8 +42,6 @@ CREATE INDEX ev_edges_id ON event_edges(event_id); CREATE INDEX ev_edges_prev_id ON event_edges(prev_event_id); CREATE TABLE room_depth( room_id TEXT NOT NULL, min_depth INTEGER NOT NULL, UNIQUE (room_id) ); CREATE INDEX room_depth_room ON room_depth(room_id); -CREATE TABLE state_groups( id BIGINT PRIMARY KEY, room_id TEXT NOT NULL, event_id TEXT NOT NULL ); -CREATE TABLE state_groups_state( state_group BIGINT NOT NULL, room_id TEXT NOT NULL, type TEXT NOT NULL, state_key TEXT NOT NULL, event_id TEXT NOT NULL ); CREATE TABLE event_to_state_groups( event_id TEXT NOT NULL, state_group BIGINT NOT NULL, UNIQUE (event_id) ); CREATE TABLE local_media_repository ( media_id TEXT, media_type TEXT, media_length INTEGER, created_ts BIGINT, upload_name TEXT, user_id TEXT, quarantined_by TEXT, url_cache TEXT, last_access_ts BIGINT, UNIQUE (media_id) ); CREATE TABLE local_media_repository_thumbnails ( media_id TEXT, thumbnail_width INTEGER, thumbnail_height INTEGER, thumbnail_type TEXT, thumbnail_method TEXT, thumbnail_length INTEGER, UNIQUE ( media_id, thumbnail_width, thumbnail_height, thumbnail_type ) ); @@ -120,9 +118,6 @@ CREATE TABLE device_max_stream_id ( stream_id BIGINT NOT NULL ); CREATE TABLE public_room_list_stream ( stream_id BIGINT NOT NULL, room_id TEXT NOT NULL, visibility BOOLEAN NOT NULL , appservice_id TEXT, network_id TEXT); CREATE INDEX public_room_list_stream_idx on public_room_list_stream( stream_id ); CREATE INDEX public_room_list_stream_rm_idx on public_room_list_stream( room_id, stream_id ); -CREATE TABLE state_group_edges( state_group BIGINT NOT NULL, prev_state_group BIGINT NOT NULL ); -CREATE INDEX state_group_edges_idx ON state_group_edges(state_group); -CREATE INDEX state_group_edges_prev_idx ON state_group_edges(prev_state_group); CREATE TABLE stream_ordering_to_exterm ( stream_ordering BIGINT NOT NULL, room_id TEXT NOT NULL, event_id TEXT NOT NULL ); CREATE INDEX stream_ordering_to_exterm_idx on stream_ordering_to_exterm( stream_ordering ); CREATE INDEX stream_ordering_to_exterm_rm_idx on stream_ordering_to_exterm( room_id, stream_ordering ); @@ -254,6 +249,5 @@ CREATE INDEX user_ips_last_seen_only ON user_ips (last_seen); CREATE INDEX users_creation_ts ON users (creation_ts); CREATE INDEX event_to_state_groups_sg_index ON event_to_state_groups (state_group); CREATE UNIQUE INDEX device_lists_remote_cache_unique_id ON device_lists_remote_cache (user_id, device_id); -CREATE INDEX state_groups_state_type_idx ON state_groups_state(state_group, type, state_key); CREATE UNIQUE INDEX device_lists_remote_extremeties_unique_idx ON device_lists_remote_extremeties (user_id); CREATE UNIQUE INDEX user_ips_user_token_ip_unique_index ON user_ips (user_id, access_token, ip); diff --git a/synapse/storage/data_stores/main/schema/full_schemas/54/stream_positions.sql b/synapse/storage/data_stores/main/schema/full_schemas/54/stream_positions.sql index c265fd20e2..91d21b2921 100644 --- a/synapse/storage/data_stores/main/schema/full_schemas/54/stream_positions.sql +++ b/synapse/storage/data_stores/main/schema/full_schemas/54/stream_positions.sql @@ -5,3 +5,4 @@ INSERT INTO federation_stream_position (type, stream_id) SELECT 'events', coales INSERT INTO user_directory_stream_pos (stream_id) VALUES (0); INSERT INTO stats_stream_pos (stream_id) VALUES (0); INSERT INTO event_push_summary_stream_ordering (stream_ordering) VALUES (0); +-- device_max_stream_id is handled separately in 56/device_stream_id_insert.sql \ No newline at end of file diff --git a/synapse/storage/data_stores/main/schema/full_schemas/README.md b/synapse/storage/data_stores/main/schema/full_schemas/README.md new file mode 100644 index 0000000000..bbd3f18604 --- /dev/null +++ b/synapse/storage/data_stores/main/schema/full_schemas/README.md @@ -0,0 +1,13 @@ +# Building full schema dumps + +These schemas need to be made from a database that has had all background updates run. + +To do so, use `scripts-dev/make_full_schema.sh`. This will produce +`full.sql.postgres ` and `full.sql.sqlite` files. + +Ensure postgres is installed and your user has the ability to run bash commands +such as `createdb`. + +``` +./scripts-dev/make_full_schema.sh -p postgres_username -o output_dir/ +``` diff --git a/synapse/storage/data_stores/main/schema/full_schemas/README.txt b/synapse/storage/data_stores/main/schema/full_schemas/README.txt deleted file mode 100644 index d3f6401344..0000000000 --- a/synapse/storage/data_stores/main/schema/full_schemas/README.txt +++ /dev/null @@ -1,19 +0,0 @@ -Building full schema dumps -========================== - -These schemas need to be made from a database that has had all background updates run. - -Postgres --------- - -$ pg_dump --format=plain --schema-only --no-tablespaces --no-acl --no-owner $DATABASE_NAME| sed -e '/^--/d' -e 's/public\.//g' -e '/^SET /d' -e '/^SELECT /d' > full.sql.postgres - -SQLite ------- - -$ sqlite3 $DATABASE_FILE ".schema" > full.sql.sqlite - -After ------ - -Delete the CREATE statements for "sqlite_stat1", "schema_version", "applied_schema_deltas", and "applied_module_schemas". \ No newline at end of file diff --git a/synapse/storage/data_stores/main/search.py b/synapse/storage/data_stores/main/search.py index d1d7c6863d..47ebb8a214 100644 --- a/synapse/storage/data_stores/main/search.py +++ b/synapse/storage/data_stores/main/search.py @@ -24,8 +24,9 @@ from canonicaljson import json from twisted.internet import defer from synapse.api.errors import SynapseError -from synapse.storage._base import make_in_list_sql_clause -from synapse.storage.background_updates import BackgroundUpdateStore +from synapse.storage._base import SQLBaseStore, make_in_list_sql_clause +from synapse.storage.data_stores.main.events_worker import EventRedactBehaviour +from synapse.storage.database import Database from synapse.storage.engines import PostgresEngine, Sqlite3Engine logger = logging.getLogger(__name__) @@ -36,23 +37,23 @@ SearchEntry = namedtuple( ) -class SearchBackgroundUpdateStore(BackgroundUpdateStore): +class SearchBackgroundUpdateStore(SQLBaseStore): EVENT_SEARCH_UPDATE_NAME = "event_search" EVENT_SEARCH_ORDER_UPDATE_NAME = "event_search_order" EVENT_SEARCH_USE_GIST_POSTGRES_NAME = "event_search_postgres_gist" EVENT_SEARCH_USE_GIN_POSTGRES_NAME = "event_search_postgres_gin" - def __init__(self, db_conn, hs): - super(SearchBackgroundUpdateStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SearchBackgroundUpdateStore, self).__init__(database, db_conn, hs) if not hs.config.enable_search: return - self.register_background_update_handler( + self.db.updates.register_background_update_handler( self.EVENT_SEARCH_UPDATE_NAME, self._background_reindex_search ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( self.EVENT_SEARCH_ORDER_UPDATE_NAME, self._background_reindex_search_order ) @@ -61,9 +62,11 @@ class SearchBackgroundUpdateStore(BackgroundUpdateStore): # a GIN index. However, it's possible that some people might still have # the background update queued, so we register a handler to clear the # background update. - self.register_noop_background_update(self.EVENT_SEARCH_USE_GIST_POSTGRES_NAME) + self.db.updates.register_noop_background_update( + self.EVENT_SEARCH_USE_GIST_POSTGRES_NAME + ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( self.EVENT_SEARCH_USE_GIN_POSTGRES_NAME, self._background_reindex_gin_search ) @@ -93,7 +96,7 @@ class SearchBackgroundUpdateStore(BackgroundUpdateStore): # 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.cursor_to_dict(txn) + rows = self.db.cursor_to_dict(txn) if not rows: return 0 @@ -153,18 +156,18 @@ class SearchBackgroundUpdateStore(BackgroundUpdateStore): "rows_inserted": rows_inserted + len(event_search_rows), } - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, self.EVENT_SEARCH_UPDATE_NAME, progress ) return len(event_search_rows) - result = yield self.runInteraction( + result = yield self.db.runInteraction( self.EVENT_SEARCH_UPDATE_NAME, reindex_search_txn ) if not result: - yield self._end_background_update(self.EVENT_SEARCH_UPDATE_NAME) + yield self.db.updates._end_background_update(self.EVENT_SEARCH_UPDATE_NAME) return result @@ -206,9 +209,11 @@ class SearchBackgroundUpdateStore(BackgroundUpdateStore): conn.set_session(autocommit=False) if isinstance(self.database_engine, PostgresEngine): - yield self.runWithConnection(create_index) + yield self.db.runWithConnection(create_index) - yield self._end_background_update(self.EVENT_SEARCH_USE_GIN_POSTGRES_NAME) + yield self.db.updates._end_background_update( + self.EVENT_SEARCH_USE_GIN_POSTGRES_NAME + ) return 1 @defer.inlineCallbacks @@ -237,14 +242,14 @@ class SearchBackgroundUpdateStore(BackgroundUpdateStore): ) conn.set_session(autocommit=False) - yield self.runWithConnection(create_index) + yield self.db.runWithConnection(create_index) pg = dict(progress) pg["have_added_indexes"] = True - yield self.runInteraction( + yield self.db.runInteraction( self.EVENT_SEARCH_ORDER_UPDATE_NAME, - self._background_update_progress_txn, + self.db.updates._background_update_progress_txn, self.EVENT_SEARCH_ORDER_UPDATE_NAME, pg, ) @@ -274,18 +279,20 @@ class SearchBackgroundUpdateStore(BackgroundUpdateStore): "have_added_indexes": True, } - self._background_update_progress_txn( + self.db.updates._background_update_progress_txn( txn, self.EVENT_SEARCH_ORDER_UPDATE_NAME, progress ) return len(rows), True - num_rows, finished = yield self.runInteraction( + num_rows, finished = yield self.db.runInteraction( self.EVENT_SEARCH_ORDER_UPDATE_NAME, reindex_search_txn ) if not finished: - yield self._end_background_update(self.EVENT_SEARCH_ORDER_UPDATE_NAME) + yield self.db.updates._end_background_update( + self.EVENT_SEARCH_ORDER_UPDATE_NAME + ) return num_rows @@ -337,8 +344,8 @@ class SearchBackgroundUpdateStore(BackgroundUpdateStore): class SearchStore(SearchBackgroundUpdateStore): - def __init__(self, db_conn, hs): - super(SearchStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(SearchStore, self).__init__(database, db_conn, hs) def store_event_search_txn(self, txn, event, key, value): """Add event to the search table @@ -378,7 +385,7 @@ class SearchStore(SearchBackgroundUpdateStore): """ clauses = [] - search_query = search_query = _parse_query(self.database_engine, search_term) + search_query = _parse_query(self.database_engine, search_term) args = [] @@ -441,11 +448,18 @@ class SearchStore(SearchBackgroundUpdateStore): # entire table from the database. sql += " ORDER BY rank DESC LIMIT 500" - results = yield self._execute("search_msgs", self.cursor_to_dict, sql, *args) + results = yield self.db.execute( + "search_msgs", self.db.cursor_to_dict, sql, *args + ) results = list(filter(lambda row: row["room_id"] in room_ids, results)) - events = yield self.get_events_as_list([r["event_id"] for r in results]) + # We set redact_behaviour to BLOCK here to prevent redacted events being returned in + # search results (which is a data leak) + events = yield self.get_events_as_list( + [r["event_id"] for r in results], + redact_behaviour=EventRedactBehaviour.BLOCK, + ) event_map = {ev.event_id: ev for ev in events} @@ -455,8 +469,8 @@ class SearchStore(SearchBackgroundUpdateStore): count_sql += " GROUP BY room_id" - count_results = yield self._execute( - "search_rooms_count", self.cursor_to_dict, count_sql, *count_args + count_results = yield self.db.execute( + "search_rooms_count", self.db.cursor_to_dict, count_sql, *count_args ) count = sum(row["count"] for row in count_results if row["room_id"] in room_ids) @@ -487,7 +501,7 @@ class SearchStore(SearchBackgroundUpdateStore): """ clauses = [] - search_query = search_query = _parse_query(self.database_engine, search_term) + search_query = _parse_query(self.database_engine, search_term) args = [] @@ -586,11 +600,18 @@ class SearchStore(SearchBackgroundUpdateStore): args.append(limit) - results = yield self._execute("search_rooms", self.cursor_to_dict, sql, *args) + results = yield self.db.execute( + "search_rooms", self.db.cursor_to_dict, sql, *args + ) results = list(filter(lambda row: row["room_id"] in room_ids, results)) - events = yield self.get_events_as_list([r["event_id"] for r in results]) + # We set redact_behaviour to BLOCK here to prevent redacted events being returned in + # search results (which is a data leak) + events = yield self.get_events_as_list( + [r["event_id"] for r in results], + redact_behaviour=EventRedactBehaviour.BLOCK, + ) event_map = {ev.event_id: ev for ev in events} @@ -600,8 +621,8 @@ class SearchStore(SearchBackgroundUpdateStore): count_sql += " GROUP BY room_id" - count_results = yield self._execute( - "search_rooms_count", self.cursor_to_dict, count_sql, *count_args + count_results = yield self.db.execute( + "search_rooms_count", self.db.cursor_to_dict, count_sql, *count_args ) count = sum(row["count"] for row in count_results if row["room_id"] in room_ids) @@ -686,7 +707,7 @@ class SearchStore(SearchBackgroundUpdateStore): return highlight_words - return self.runInteraction("_find_highlights", f) + return self.db.runInteraction("_find_highlights", f) def _to_postgres_options(options_dict): diff --git a/synapse/storage/data_stores/main/signatures.py b/synapse/storage/data_stores/main/signatures.py index 556191b76f..563216b63c 100644 --- a/synapse/storage/data_stores/main/signatures.py +++ b/synapse/storage/data_stores/main/signatures.py @@ -48,7 +48,7 @@ class SignatureWorkerStore(SQLBaseStore): for event_id in event_ids } - return self.runInteraction("get_event_reference_hashes", f) + return self.db.runInteraction("get_event_reference_hashes", f) @defer.inlineCallbacks def add_event_hashes(self, event_ids): @@ -98,4 +98,4 @@ class SignatureStore(SignatureWorkerStore): } ) - self._simple_insert_many_txn(txn, table="event_reference_hashes", values=vals) + self.db.simple_insert_many_txn(txn, table="event_reference_hashes", values=vals) diff --git a/synapse/storage/data_stores/main/state.py b/synapse/storage/data_stores/main/state.py index 180b75c710..33bebd1c48 100644 --- a/synapse/storage/data_stores/main/state.py +++ b/synapse/storage/data_stores/main/state.py @@ -12,13 +12,12 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. - +import collections.abc import logging from collections import namedtuple from typing import Iterable, Tuple -from six import iteritems, itervalues -from six.moves import range +from six import iteritems from twisted.internet import defer @@ -27,13 +26,11 @@ from synapse.api.errors import NotFoundError from synapse.events import EventBase from synapse.events.snapshot import EventContext from synapse.storage._base import SQLBaseStore -from synapse.storage.background_updates import BackgroundUpdateStore from synapse.storage.data_stores.main.events_worker import EventsWorkerStore -from synapse.storage.engines import PostgresEngine +from synapse.storage.database import Database from synapse.storage.state import StateFilter from synapse.util.caches import intern_string from synapse.util.caches.descriptors import cached, cachedList -from synapse.util.caches.dictionary_cache import DictionaryCache from synapse.util.stringutils import to_ascii logger = logging.getLogger(__name__) @@ -55,202 +52,13 @@ class _GetStateGroupDelta( return len(self.delta_ids) if self.delta_ids else 0 -class StateGroupBackgroundUpdateStore(SQLBaseStore): - """Defines functions related to state groups needed to run the state backgroud - updates. - """ - - def _count_state_group_hops_txn(self, txn, state_group): - """Given a state group, count how many hops there are in the tree. - - This is used to ensure the delta chains don't get too long. - """ - if isinstance(self.database_engine, PostgresEngine): - sql = """ - WITH RECURSIVE state(state_group) AS ( - VALUES(?::bigint) - UNION ALL - SELECT prev_state_group FROM state_group_edges e, state s - WHERE s.state_group = e.state_group - ) - SELECT count(*) FROM state; - """ - - txn.execute(sql, (state_group,)) - row = txn.fetchone() - if row and row[0]: - return row[0] - else: - return 0 - else: - # We don't use WITH RECURSIVE on sqlite3 as there are distributions - # that ship with an sqlite3 version that doesn't support it (e.g. wheezy) - next_group = state_group - count = 0 - - while next_group: - next_group = self._simple_select_one_onecol_txn( - txn, - table="state_group_edges", - keyvalues={"state_group": next_group}, - retcol="prev_state_group", - allow_none=True, - ) - if next_group: - count += 1 - - return count - - def _get_state_groups_from_groups_txn( - self, txn, groups, state_filter=StateFilter.all() - ): - results = {group: {} for group in groups} - - where_clause, where_args = state_filter.make_sql_filter_clause() - - # Unless the filter clause is empty, we're going to append it after an - # existing where clause - if where_clause: - where_clause = " AND (%s)" % (where_clause,) - - if isinstance(self.database_engine, PostgresEngine): - # Temporarily disable sequential scans in this transaction. This is - # a temporary hack until we can add the right indices in - txn.execute("SET LOCAL enable_seqscan=off") - - # The below query walks the state_group tree so that the "state" - # table includes all state_groups in the tree. It then joins - # against `state_groups_state` to fetch the latest state. - # It assumes that previous state groups are always numerically - # lesser. - # The PARTITION is used to get the event_id in the greatest state - # group for the given type, state_key. - # This may return multiple rows per (type, state_key), but last_value - # should be the same. - sql = """ - WITH RECURSIVE state(state_group) AS ( - VALUES(?::bigint) - UNION ALL - SELECT prev_state_group FROM state_group_edges e, state s - WHERE s.state_group = e.state_group - ) - SELECT DISTINCT type, state_key, last_value(event_id) OVER ( - PARTITION BY type, state_key ORDER BY state_group ASC - ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING - ) AS event_id FROM state_groups_state - WHERE state_group IN ( - SELECT state_group FROM state - ) - """ - - for group in groups: - args = [group] - args.extend(where_args) - - txn.execute(sql + where_clause, args) - for row in txn: - typ, state_key, event_id = row - key = (typ, state_key) - results[group][key] = event_id - else: - max_entries_returned = state_filter.max_entries_returned() - - # We don't use WITH RECURSIVE on sqlite3 as there are distributions - # that ship with an sqlite3 version that doesn't support it (e.g. wheezy) - for group in groups: - next_group = group - - while next_group: - # We did this before by getting the list of group ids, and - # then passing that list to sqlite to get latest event for - # each (type, state_key). However, that was terribly slow - # without the right indices (which we can't add until - # after we finish deduping state, which requires this func) - args = [next_group] - args.extend(where_args) - - txn.execute( - "SELECT type, state_key, event_id FROM state_groups_state" - " WHERE state_group = ? " + where_clause, - args, - ) - results[group].update( - ((typ, state_key), event_id) - for typ, state_key, event_id in txn - if (typ, state_key) not in results[group] - ) - - # If the number of entries in the (type,state_key)->event_id dict - # matches the number of (type,state_keys) types we were searching - # for, then we must have found them all, so no need to go walk - # further down the tree... UNLESS our types filter contained - # wildcards (i.e. Nones) in which case we have to do an exhaustive - # search - if ( - max_entries_returned is not None - and len(results[group]) == max_entries_returned - ): - break - - next_group = self._simple_select_one_onecol_txn( - txn, - table="state_group_edges", - keyvalues={"state_group": next_group}, - retcol="prev_state_group", - allow_none=True, - ) - - return results - - # this inherits from EventsWorkerStore because it calls self.get_events -class StateGroupWorkerStore( - EventsWorkerStore, StateGroupBackgroundUpdateStore, SQLBaseStore -): +class StateGroupWorkerStore(EventsWorkerStore, SQLBaseStore): """The parts of StateGroupStore that can be called from workers. """ - STATE_GROUP_DEDUPLICATION_UPDATE_NAME = "state_group_state_deduplication" - STATE_GROUP_INDEX_UPDATE_NAME = "state_group_state_type_index" - CURRENT_STATE_INDEX_UPDATE_NAME = "current_state_members_idx" - - def __init__(self, db_conn, hs): - super(StateGroupWorkerStore, self).__init__(db_conn, hs) - - # Originally the state store used a single DictionaryCache to cache the - # event IDs for the state types in a given state group to avoid hammering - # on the state_group* tables. - # - # The point of using a DictionaryCache is that it can cache a subset - # of the state events for a given state group (i.e. a subset of the keys for a - # given dict which is an entry in the cache for a given state group ID). - # - # However, this poses problems when performing complicated queries - # on the store - for instance: "give me all the state for this group, but - # limit members to this subset of users", as DictionaryCache's API isn't - # rich enough to say "please cache any of these fields, apart from this subset". - # This is problematic when lazy loading members, which requires this behaviour, - # as without it the cache has no choice but to speculatively load all - # state events for the group, which negates the efficiency being sought. - # - # Rather than overcomplicating DictionaryCache's API, we instead split the - # state_group_cache into two halves - one for tracking non-member events, - # and the other for tracking member_events. This means that lazy loading - # queries can be made in a cache-friendly manner by querying both caches - # separately and then merging the result. So for the example above, you - # would query the members cache for a specific subset of state keys - # (which DictionaryCache will handle efficiently and fine) and the non-members - # cache for all state (which DictionaryCache will similarly handle fine) - # and then just merge the results together. - # - # We size the non-members cache to be smaller than the members cache as the - # vast majority of state in Matrix (today) is member events. - - # TODO: this hasn't been tuned yet - self._state_group_cache = DictionaryCache("*stateGroupCache*", 50000) - self._state_group_members_cache = DictionaryCache( - "*stateGroupMembersCache*", 500000 - ) + def __init__(self, database: Database, db_conn, hs): + super(StateGroupWorkerStore, self).__init__(database, db_conn, hs) @defer.inlineCallbacks def get_room_version(self, room_id): @@ -274,7 +82,7 @@ class StateGroupWorkerStore( @defer.inlineCallbacks def get_room_predecessor(self, room_id): - """Get the predecessor room of an upgraded room if one exists. + """Get the predecessor of an upgraded room if it exists. Otherwise return None. Args: @@ -287,14 +95,22 @@ class StateGroupWorkerStore( * room_id (str): The room ID of the predecessor room * event_id (str): The ID of the tombstone event in the predecessor room + None if a predecessor key is not found, or is not a dictionary. + Raises: - NotFoundError if the room is unknown + NotFoundError if the given room is unknown """ # Retrieve the room's create event create_event = yield self.get_create_event_for_room(room_id) - # Return predecessor if present - return create_event.content.get("predecessor", None) + # Retrieve the predecessor key of the create event + predecessor = create_event.content.get("predecessor", None) + + # Ensure the key is a dictionary + if not isinstance(predecessor, collections.abc.Mapping): + return None + + return predecessor @defer.inlineCallbacks def get_create_event_for_room(self, room_id): @@ -314,7 +130,7 @@ class StateGroupWorkerStore( # If we can't find the create event, assume we've hit a dead end if not create_id: - raise NotFoundError("Unknown room %s" % (room_id)) + raise NotFoundError("Unknown room %s" % (room_id,)) # Retrieve the room's create event and return create_event = yield self.get_event(create_id) @@ -344,22 +160,25 @@ class StateGroupWorkerStore( (intern_string(r[0]), intern_string(r[1])): to_ascii(r[2]) for r in txn } - return self.runInteraction("get_current_state_ids", _get_current_state_ids_txn) + return self.db.runInteraction( + "get_current_state_ids", _get_current_state_ids_txn + ) # FIXME: how should this be cached? - def get_filtered_current_state_ids(self, room_id, state_filter=StateFilter.all()): + def get_filtered_current_state_ids( + self, room_id: str, state_filter: StateFilter = StateFilter.all() + ): """Get the current state event of a given type for a room based on the current_state_events table. This may not be as up-to-date as the result of doing a fresh state resolution as per state_handler.get_current_state Args: - room_id (str) - state_filter (StateFilter): The state filter used to fetch state + room_id + state_filter: The state filter used to fetch state from the database. Returns: - Deferred[dict[tuple[str, str], str]]: Map from type/state_key to - event ID. + defer.Deferred[StateMap[str]]: Map from type/state_key to event ID. """ where_clause, where_args = state_filter.make_sql_filter_clause() @@ -388,7 +207,7 @@ class StateGroupWorkerStore( return results - return self.runInteraction( + return self.db.runInteraction( "get_filtered_current_state_ids", _get_filtered_current_state_ids_txn ) @@ -417,230 +236,9 @@ class StateGroupWorkerStore( return event.content.get("canonical_alias") - @cached(max_entries=10000, iterable=True) - def get_state_group_delta(self, state_group): - """Given a state group try to return a previous group and a delta between - the old and the new. - - Returns: - (prev_group, delta_ids), where both may be None. - """ - - def _get_state_group_delta_txn(txn): - prev_group = self._simple_select_one_onecol_txn( - txn, - table="state_group_edges", - keyvalues={"state_group": state_group}, - retcol="prev_state_group", - allow_none=True, - ) - - if not prev_group: - return _GetStateGroupDelta(None, None) - - delta_ids = self._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}, - ) - - return self.runInteraction("get_state_group_delta", _get_state_group_delta_txn) - - @defer.inlineCallbacks - def get_state_groups_ids(self, _room_id, event_ids): - """Get the event IDs of all the state for the state groups for the given events - - Args: - _room_id (str): id of the room for these events - event_ids (iterable[str]): ids of the events - - Returns: - Deferred[dict[int, dict[tuple[str, str], str]]]: - dict of state_group_id -> (dict of (type, state_key) -> event id) - """ - if not event_ids: - return {} - - event_to_groups = yield self._get_state_group_for_events(event_ids) - - groups = set(itervalues(event_to_groups)) - group_to_state = yield self._get_state_for_groups(groups) - - return group_to_state - - @defer.inlineCallbacks - def get_state_ids_for_group(self, state_group): - """Get the event IDs of all the state in the given state group - - Args: - state_group (int) - - Returns: - Deferred[dict]: Resolves to a map of (type, state_key) -> event_id - """ - group_to_state = yield self._get_state_for_groups((state_group,)) - - return group_to_state[state_group] - - @defer.inlineCallbacks - def get_state_groups(self, room_id, event_ids): - """ Get the state groups for the given list of event_ids - - Returns: - Deferred[dict[int, list[EventBase]]]: - dict of state_group_id -> list of state events. - """ - if not event_ids: - return {} - - group_to_ids = yield self.get_state_groups_ids(room_id, event_ids) - - state_event_map = yield self.get_events( - [ - ev_id - for group_ids in itervalues(group_to_ids) - for ev_id in itervalues(group_ids) - ], - get_prev_content=False, - ) - - return { - group: [ - state_event_map[v] - for v in itervalues(event_id_map) - if v in state_event_map - ] - for group, event_id_map in iteritems(group_to_ids) - } - - @defer.inlineCallbacks - def _get_state_groups_from_groups(self, groups, state_filter): - """Returns the state groups for a given set of groups, filtering on - types of state events. - - Args: - groups(list[int]): list of state group IDs to query - state_filter (StateFilter): The state filter used to fetch state - from the database. - Returns: - Deferred[dict[int, dict[tuple[str, str], str]]]: - dict of state_group_id -> (dict of (type, state_key) -> event id) - """ - results = {} - - chunks = [groups[i : i + 100] for i in range(0, len(groups), 100)] - for chunk in chunks: - res = yield self.runInteraction( - "_get_state_groups_from_groups", - self._get_state_groups_from_groups_txn, - chunk, - state_filter, - ) - results.update(res) - - return results - - @defer.inlineCallbacks - def get_state_for_events(self, event_ids, state_filter=StateFilter.all()): - """Given a list of event_ids and type tuples, return a list of state - dicts for each event. - - Args: - event_ids (list[string]) - state_filter (StateFilter): The state filter used to fetch state - from the database. - - Returns: - deferred: A dict of (event_id) -> (type, state_key) -> [state_events] - """ - event_to_groups = yield self._get_state_group_for_events(event_ids) - - groups = set(itervalues(event_to_groups)) - group_to_state = yield self._get_state_for_groups(groups, state_filter) - - state_event_map = yield self.get_events( - [ev_id for sd in itervalues(group_to_state) for ev_id in itervalues(sd)], - get_prev_content=False, - ) - - event_to_state = { - event_id: { - k: state_event_map[v] - for k, v in iteritems(group_to_state[group]) - if v in state_event_map - } - for event_id, group in iteritems(event_to_groups) - } - - return {event: event_to_state[event] for event in event_ids} - - @defer.inlineCallbacks - def get_state_ids_for_events(self, event_ids, state_filter=StateFilter.all()): - """ - Get the state dicts corresponding to a list of events, containing the event_ids - of the state events (as opposed to the events themselves) - - Args: - event_ids(list(str)): events whose state should be returned - state_filter (StateFilter): The state filter used to fetch state - from the database. - - Returns: - A deferred dict from event_id -> (type, state_key) -> event_id - """ - event_to_groups = yield self._get_state_group_for_events(event_ids) - - groups = set(itervalues(event_to_groups)) - group_to_state = yield self._get_state_for_groups(groups, state_filter) - - event_to_state = { - event_id: group_to_state[group] - for event_id, group in iteritems(event_to_groups) - } - - return {event: event_to_state[event] for event in event_ids} - - @defer.inlineCallbacks - def get_state_for_event(self, event_id, state_filter=StateFilter.all()): - """ - Get the state dict corresponding to a particular event - - Args: - event_id(str): event whose state should be returned - state_filter (StateFilter): The state filter used to fetch state - from the database. - - Returns: - A deferred dict from (type, state_key) -> state_event - """ - state_map = yield self.get_state_for_events([event_id], state_filter) - return state_map[event_id] - - @defer.inlineCallbacks - def get_state_ids_for_event(self, event_id, state_filter=StateFilter.all()): - """ - Get the state dict corresponding to a particular event - - Args: - event_id(str): event whose state should be returned - state_filter (StateFilter): The state filter used to fetch state - from the database. - - Returns: - A deferred dict from (type, state_key) -> state_event - """ - state_map = yield self.get_state_ids_for_events([event_id], state_filter) - return state_map[event_id] - @cached(max_entries=50000) def _get_state_group_for_event(self, event_id): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="event_to_state_groups", keyvalues={"event_id": event_id}, retcol="state_group", @@ -657,7 +255,7 @@ class StateGroupWorkerStore( def _get_state_group_for_events(self, event_ids): """Returns mapping event_id -> state_group """ - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="event_to_state_groups", column="event_id", iterable=event_ids, @@ -668,329 +266,6 @@ class StateGroupWorkerStore( return {row["event_id"]: row["state_group"] for row in rows} - def _get_state_for_group_using_cache(self, cache, group, state_filter): - """Checks if group is in cache. See `_get_state_for_groups` - - Args: - cache(DictionaryCache): the state group cache to use - group(int): The state group to lookup - state_filter (StateFilter): The state filter used to fetch state - from the database. - - Returns 2-tuple (`state_dict`, `got_all`). - `got_all` is a bool indicating if we successfully retrieved all - requests state from the cache, if False we need to query the DB for the - missing state. - """ - is_all, known_absent, state_dict_ids = cache.get(group) - - if is_all or state_filter.is_full(): - # Either we have everything or want everything, either way - # `is_all` tells us whether we've gotten everything. - return state_filter.filter_state(state_dict_ids), is_all - - # tracks whether any of our requested types are missing from the cache - missing_types = False - - if state_filter.has_wildcards(): - # We don't know if we fetched all the state keys for the types in - # the filter that are wildcards, so we have to assume that we may - # have missed some. - missing_types = True - else: - # There aren't any wild cards, so `concrete_types()` returns the - # complete list of event types we're wanting. - for key in state_filter.concrete_types(): - if key not in state_dict_ids and key not in known_absent: - missing_types = True - break - - return state_filter.filter_state(state_dict_ids), not missing_types - - @defer.inlineCallbacks - def _get_state_for_groups(self, groups, state_filter=StateFilter.all()): - """Gets the state at each of a list of state groups, optionally - filtering by type/state_key - - Args: - groups (iterable[int]): list of state groups for which we want - to get the state. - state_filter (StateFilter): The state filter used to fetch state - from the database. - Returns: - Deferred[dict[int, dict[tuple[str, str], str]]]: - dict of state_group_id -> (dict of (type, state_key) -> event id) - """ - - member_filter, non_member_filter = state_filter.get_member_split() - - # Now we look them up in the member and non-member caches - ( - non_member_state, - incomplete_groups_nm, - ) = yield self._get_state_for_groups_using_cache( - groups, self._state_group_cache, state_filter=non_member_filter - ) - - ( - member_state, - incomplete_groups_m, - ) = yield self._get_state_for_groups_using_cache( - groups, self._state_group_members_cache, state_filter=member_filter - ) - - state = dict(non_member_state) - for group in groups: - state[group].update(member_state[group]) - - # Now fetch any missing groups from the database - - incomplete_groups = incomplete_groups_m | incomplete_groups_nm - - if not incomplete_groups: - return state - - cache_sequence_nm = self._state_group_cache.sequence - cache_sequence_m = self._state_group_members_cache.sequence - - # Help the cache hit ratio by expanding the filter a bit - db_state_filter = state_filter.return_expanded() - - group_to_state_dict = yield self._get_state_groups_from_groups( - list(incomplete_groups), state_filter=db_state_filter - ) - - # Now lets update the caches - self._insert_into_cache( - group_to_state_dict, - db_state_filter, - cache_seq_num_members=cache_sequence_m, - cache_seq_num_non_members=cache_sequence_nm, - ) - - # And finally update the result dict, by filtering out any extra - # stuff we pulled out of the database. - for group, group_state_dict in iteritems(group_to_state_dict): - # We just replace any existing entries, as we will have loaded - # everything we need from the database anyway. - state[group] = state_filter.filter_state(group_state_dict) - - return state - - def _get_state_for_groups_using_cache(self, groups, cache, state_filter): - """Gets the state at each of a list of state groups, optionally - filtering by type/state_key, querying from a specific cache. - - Args: - groups (iterable[int]): list of state groups for which we want - to get the state. - cache (DictionaryCache): the cache of group ids to state dicts which - we will pass through - either the normal state cache or the specific - members state cache. - state_filter (StateFilter): The state filter used to fetch state - from the database. - - Returns: - tuple[dict[int, dict[tuple[str, str], str]], set[int]]: Tuple of - dict of state_group_id -> (dict of (type, state_key) -> event id) - of entries in the cache, and the state group ids either missing - from the cache or incomplete. - """ - results = {} - incomplete_groups = set() - for group in set(groups): - state_dict_ids, got_all = self._get_state_for_group_using_cache( - cache, group, state_filter - ) - results[group] = state_dict_ids - - if not got_all: - incomplete_groups.add(group) - - return results, incomplete_groups - - def _insert_into_cache( - self, - group_to_state_dict, - state_filter, - cache_seq_num_members, - cache_seq_num_non_members, - ): - """Inserts results from querying the database into the relevant cache. - - Args: - group_to_state_dict (dict): The new entries pulled from database. - Map from state group to state dict - state_filter (StateFilter): The state filter used to fetch state - from the database. - cache_seq_num_members (int): Sequence number of member cache since - last lookup in cache - cache_seq_num_non_members (int): Sequence number of member cache since - last lookup in cache - """ - - # We need to work out which types we've fetched from the DB for the - # member vs non-member caches. This should be as accurate as possible, - # but can be an underestimate (e.g. when we have wild cards) - - member_filter, non_member_filter = state_filter.get_member_split() - if member_filter.is_full(): - # We fetched all member events - member_types = None - else: - # `concrete_types()` will only return a subset when there are wild - # cards in the filter, but that's fine. - member_types = member_filter.concrete_types() - - if non_member_filter.is_full(): - # We fetched all non member events - non_member_types = None - else: - non_member_types = non_member_filter.concrete_types() - - for group, group_state_dict in iteritems(group_to_state_dict): - state_dict_members = {} - state_dict_non_members = {} - - for k, v in iteritems(group_state_dict): - if k[0] == EventTypes.Member: - state_dict_members[k] = v - else: - state_dict_non_members[k] = v - - self._state_group_members_cache.update( - cache_seq_num_members, - key=group, - value=state_dict_members, - fetched_keys=member_types, - ) - - self._state_group_cache.update( - cache_seq_num_non_members, - key=group, - value=state_dict_non_members, - fetched_keys=non_member_types, - ) - - def store_state_group( - self, event_id, room_id, prev_group, delta_ids, current_state_ids - ): - """Store a new set of state, returning a newly assigned state group. - - Args: - event_id (str): The event ID for which the state was calculated - room_id (str) - prev_group (int|None): A previous state group for the room, optional. - delta_ids (dict|None): The delta between state at `prev_group` and - `current_state_ids`, if `prev_group` was given. Same format as - `current_state_ids`. - current_state_ids (dict): The state to store. Map of (type, state_key) - to event_id. - - Returns: - Deferred[int]: The state group ID - """ - - def _store_state_group_txn(txn): - if current_state_ids is None: - # AFAIK, this can never happen - raise Exception("current_state_ids cannot be None") - - state_group = self.database_engine.get_next_state_group_id(txn) - - self._simple_insert_txn( - txn, - table="state_groups", - values={"id": state_group, "room_id": room_id, "event_id": event_id}, - ) - - # We persist as a delta if we can, while also ensuring the chain - # of deltas isn't tooo long, as otherwise read performance degrades. - if prev_group: - is_in_db = self._simple_select_one_onecol_txn( - txn, - table="state_groups", - keyvalues={"id": prev_group}, - retcol="id", - allow_none=True, - ) - if not is_in_db: - raise Exception( - "Trying to persist state with unpersisted prev_group: %r" - % (prev_group,) - ) - - potential_hops = self._count_state_group_hops_txn(txn, prev_group) - if prev_group and potential_hops < MAX_STATE_DELTA_HOPS: - self._simple_insert_txn( - txn, - table="state_group_edges", - values={"state_group": state_group, "prev_state_group": prev_group}, - ) - - self._simple_insert_many_txn( - txn, - table="state_groups_state", - values=[ - { - "state_group": state_group, - "room_id": room_id, - "type": key[0], - "state_key": key[1], - "event_id": state_id, - } - for key, state_id in iteritems(delta_ids) - ], - ) - else: - self._simple_insert_many_txn( - txn, - table="state_groups_state", - values=[ - { - "state_group": state_group, - "room_id": room_id, - "type": key[0], - "state_key": key[1], - "event_id": state_id, - } - for key, state_id in iteritems(current_state_ids) - ], - ) - - # Prefill the state group caches with this group. - # It's fine to use the sequence like this as the state group map - # is immutable. (If the map wasn't immutable then this prefill could - # race with another update) - - current_member_state_ids = { - s: ev - for (s, ev) in iteritems(current_state_ids) - if s[0] == EventTypes.Member - } - txn.call_after( - self._state_group_members_cache.update, - self._state_group_members_cache.sequence, - key=state_group, - value=dict(current_member_state_ids), - ) - - current_non_member_state_ids = { - s: ev - for (s, ev) in iteritems(current_state_ids) - if s[0] != EventTypes.Member - } - txn.call_after( - self._state_group_cache.update, - self._state_group_cache.sequence, - key=state_group, - value=dict(current_non_member_state_ids), - ) - - return state_group - - return self.runInteraction("store_state_group", _store_state_group_txn) - @defer.inlineCallbacks def get_referenced_state_groups(self, state_groups): """Check if the state groups are referenced by events. @@ -1003,7 +278,7 @@ class StateGroupWorkerStore( referenced. """ - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="event_to_state_groups", column="state_group", iterable=state_groups, @@ -1015,213 +290,30 @@ class StateGroupWorkerStore( return set(row["state_group"] for row in rows) -class StateBackgroundUpdateStore( - StateGroupBackgroundUpdateStore, BackgroundUpdateStore -): +class MainStateBackgroundUpdateStore(SQLBaseStore): - STATE_GROUP_DEDUPLICATION_UPDATE_NAME = "state_group_state_deduplication" - STATE_GROUP_INDEX_UPDATE_NAME = "state_group_state_type_index" CURRENT_STATE_INDEX_UPDATE_NAME = "current_state_members_idx" EVENT_STATE_GROUP_INDEX_UPDATE_NAME = "event_to_state_groups_sg_index" - def __init__(self, db_conn, hs): - super(StateBackgroundUpdateStore, self).__init__(db_conn, hs) - self.register_background_update_handler( - self.STATE_GROUP_DEDUPLICATION_UPDATE_NAME, - self._background_deduplicate_state, - ) - self.register_background_update_handler( - self.STATE_GROUP_INDEX_UPDATE_NAME, self._background_index_state - ) - self.register_background_index_update( + def __init__(self, database: Database, db_conn, hs): + super(MainStateBackgroundUpdateStore, self).__init__(database, db_conn, hs) + + self.db.updates.register_background_index_update( self.CURRENT_STATE_INDEX_UPDATE_NAME, index_name="current_state_events_member_index", table="current_state_events", columns=["state_key"], where_clause="type='m.room.member'", ) - self.register_background_index_update( + self.db.updates.register_background_index_update( self.EVENT_STATE_GROUP_INDEX_UPDATE_NAME, index_name="event_to_state_groups_sg_index", table="event_to_state_groups", columns=["state_group"], ) - @defer.inlineCallbacks - def _background_deduplicate_state(self, progress, batch_size): - """This background update will slowly deduplicate state by reencoding - them as deltas. - """ - last_state_group = progress.get("last_state_group", 0) - rows_inserted = progress.get("rows_inserted", 0) - max_group = progress.get("max_group", None) - - BATCH_SIZE_SCALE_FACTOR = 100 - - batch_size = max(1, int(batch_size / BATCH_SIZE_SCALE_FACTOR)) - - if max_group is None: - rows = yield self._execute( - "_background_deduplicate_state", - None, - "SELECT coalesce(max(id), 0) FROM state_groups", - ) - max_group = rows[0][0] - - def reindex_txn(txn): - new_last_state_group = last_state_group - for count in range(batch_size): - txn.execute( - "SELECT id, room_id FROM state_groups" - " WHERE ? < id AND id <= ?" - " ORDER BY id ASC" - " LIMIT 1", - (new_last_state_group, max_group), - ) - row = txn.fetchone() - if row: - state_group, room_id = row - - if not row or not state_group: - return True, count - - txn.execute( - "SELECT state_group FROM state_group_edges" - " WHERE state_group = ?", - (state_group,), - ) - - # If we reach a point where we've already started inserting - # edges we should stop. - if txn.fetchall(): - return True, count - - txn.execute( - "SELECT coalesce(max(id), 0) FROM state_groups" - " WHERE id < ? AND room_id = ?", - (state_group, room_id), - ) - (prev_group,) = txn.fetchone() - new_last_state_group = state_group - - if prev_group: - potential_hops = self._count_state_group_hops_txn(txn, prev_group) - if potential_hops >= MAX_STATE_DELTA_HOPS: - # We want to ensure chains are at most this long,# - # otherwise read performance degrades. - continue - - prev_state = self._get_state_groups_from_groups_txn( - txn, [prev_group] - ) - prev_state = prev_state[prev_group] - - curr_state = self._get_state_groups_from_groups_txn( - txn, [state_group] - ) - curr_state = curr_state[state_group] - - if not set(prev_state.keys()) - set(curr_state.keys()): - # We can only do a delta if the current has a strict super set - # of keys - - delta_state = { - key: value - for key, value in iteritems(curr_state) - if prev_state.get(key, None) != value - } - - self._simple_delete_txn( - txn, - table="state_group_edges", - keyvalues={"state_group": state_group}, - ) - - self._simple_insert_txn( - txn, - table="state_group_edges", - values={ - "state_group": state_group, - "prev_state_group": prev_group, - }, - ) - - self._simple_delete_txn( - txn, - table="state_groups_state", - keyvalues={"state_group": state_group}, - ) - - self._simple_insert_many_txn( - txn, - table="state_groups_state", - values=[ - { - "state_group": state_group, - "room_id": room_id, - "type": key[0], - "state_key": key[1], - "event_id": state_id, - } - for key, state_id in iteritems(delta_state) - ], - ) - - progress = { - "last_state_group": state_group, - "rows_inserted": rows_inserted + batch_size, - "max_group": max_group, - } - - self._background_update_progress_txn( - txn, self.STATE_GROUP_DEDUPLICATION_UPDATE_NAME, progress - ) - - return False, batch_size - - finished, result = yield self.runInteraction( - self.STATE_GROUP_DEDUPLICATION_UPDATE_NAME, reindex_txn - ) - - if finished: - yield self._end_background_update( - self.STATE_GROUP_DEDUPLICATION_UPDATE_NAME - ) - - return result * BATCH_SIZE_SCALE_FACTOR - @defer.inlineCallbacks - def _background_index_state(self, progress, batch_size): - def reindex_txn(conn): - conn.rollback() - if isinstance(self.database_engine, PostgresEngine): - # postgres insists on autocommit for the index - conn.set_session(autocommit=True) - try: - txn = conn.cursor() - txn.execute( - "CREATE INDEX CONCURRENTLY state_groups_state_type_idx" - " ON state_groups_state(state_group, type, state_key)" - ) - txn.execute("DROP INDEX IF EXISTS state_groups_state_id") - finally: - conn.set_session(autocommit=False) - else: - txn = conn.cursor() - txn.execute( - "CREATE INDEX state_groups_state_type_idx" - " ON state_groups_state(state_group, type, state_key)" - ) - txn.execute("DROP INDEX IF EXISTS state_groups_state_id") - - yield self.runWithConnection(reindex_txn) - - yield self._end_background_update(self.STATE_GROUP_INDEX_UPDATE_NAME) - - return 1 - - -class StateStore(StateGroupWorkerStore, StateBackgroundUpdateStore): +class StateStore(StateGroupWorkerStore, MainStateBackgroundUpdateStore): """ Keeps track of the state at a given event. This is done by the concept of `state groups`. Every event is a assigned @@ -1240,8 +332,8 @@ class StateStore(StateGroupWorkerStore, StateBackgroundUpdateStore): * `state_groups_state`: Maps state group to state events. """ - def __init__(self, db_conn, hs): - super(StateStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(StateStore, self).__init__(database, db_conn, hs) def _store_event_state_mappings_txn( self, txn, events_and_contexts: Iterable[Tuple[EventBase, EventContext]] @@ -1259,7 +351,7 @@ class StateStore(StateGroupWorkerStore, StateBackgroundUpdateStore): state_groups[event.event_id] = context.state_group - self._simple_insert_many_txn( + self.db.simple_insert_many_txn( txn, table="event_to_state_groups", values=[ diff --git a/synapse/storage/data_stores/main/state_deltas.py b/synapse/storage/data_stores/main/state_deltas.py index 28f33ec18f..12c982cb26 100644 --- a/synapse/storage/data_stores/main/state_deltas.py +++ b/synapse/storage/data_stores/main/state_deltas.py @@ -98,14 +98,14 @@ class StateDeltasStore(SQLBaseStore): ORDER BY stream_id ASC """ txn.execute(sql, (prev_stream_id, clipped_stream_id)) - return clipped_stream_id, self.cursor_to_dict(txn) + return clipped_stream_id, self.db.cursor_to_dict(txn) - return self.runInteraction( + return self.db.runInteraction( "get_current_state_deltas", get_current_state_deltas_txn ) def _get_max_stream_id_in_current_state_deltas_txn(self, txn): - return self._simple_select_one_onecol_txn( + return self.db.simple_select_one_onecol_txn( txn, table="current_state_delta_stream", keyvalues={}, @@ -113,7 +113,7 @@ class StateDeltasStore(SQLBaseStore): ) def get_max_stream_id_in_current_state_deltas(self): - return self.runInteraction( + return self.db.runInteraction( "get_max_stream_id_in_current_state_deltas", self._get_max_stream_id_in_current_state_deltas_txn, ) diff --git a/synapse/storage/data_stores/main/stats.py b/synapse/storage/data_stores/main/stats.py index 45b3de7d56..7bc186e9a1 100644 --- a/synapse/storage/data_stores/main/stats.py +++ b/synapse/storage/data_stores/main/stats.py @@ -22,6 +22,7 @@ from twisted.internet.defer import DeferredLock from synapse.api.constants import EventTypes, Membership from synapse.storage.data_stores.main.state_deltas import StateDeltasStore +from synapse.storage.database import Database from synapse.storage.engines import PostgresEngine from synapse.util.caches.descriptors import cached @@ -58,8 +59,8 @@ TYPE_TO_ORIGIN_TABLE = {"room": ("rooms", "room_id"), "user": ("users", "name")} class StatsStore(StateDeltasStore): - def __init__(self, db_conn, hs): - super(StatsStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(StatsStore, self).__init__(database, db_conn, hs) self.server_name = hs.hostname self.clock = self.hs.get_clock() @@ -68,17 +69,17 @@ class StatsStore(StateDeltasStore): self.stats_delta_processing_lock = DeferredLock() - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "populate_stats_process_rooms", self._populate_stats_process_rooms ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "populate_stats_process_users", self._populate_stats_process_users ) # we no longer need to perform clean-up, but we will give ourselves # the potential to reintroduce it in the future – so documentation # will still encourage the use of this no-op handler. - self.register_noop_background_update("populate_stats_cleanup") - self.register_noop_background_update("populate_stats_prepare") + self.db.updates.register_noop_background_update("populate_stats_cleanup") + self.db.updates.register_noop_background_update("populate_stats_prepare") def quantise_stats_time(self, ts): """ @@ -102,7 +103,7 @@ class StatsStore(StateDeltasStore): This is a background update which regenerates statistics for users. """ if not self.stats_enabled: - yield self._end_background_update("populate_stats_process_users") + yield self.db.updates._end_background_update("populate_stats_process_users") return 1 last_user_id = progress.get("last_user_id", "") @@ -117,22 +118,22 @@ class StatsStore(StateDeltasStore): txn.execute(sql, (last_user_id, batch_size)) return [r for r, in txn] - users_to_work_on = yield self.runInteraction( + users_to_work_on = yield self.db.runInteraction( "_populate_stats_process_users", _get_next_batch ) # No more rooms -- complete the transaction. if not users_to_work_on: - yield self._end_background_update("populate_stats_process_users") + yield self.db.updates._end_background_update("populate_stats_process_users") return 1 for user_id in users_to_work_on: yield self._calculate_and_set_initial_state_for_user(user_id) progress["last_user_id"] = user_id - yield self.runInteraction( + yield self.db.runInteraction( "populate_stats_process_users", - self._background_update_progress_txn, + self.db.updates._background_update_progress_txn, "populate_stats_process_users", progress, ) @@ -145,7 +146,7 @@ class StatsStore(StateDeltasStore): This is a background update which regenerates statistics for rooms. """ if not self.stats_enabled: - yield self._end_background_update("populate_stats_process_rooms") + yield self.db.updates._end_background_update("populate_stats_process_rooms") return 1 last_room_id = progress.get("last_room_id", "") @@ -160,22 +161,22 @@ class StatsStore(StateDeltasStore): txn.execute(sql, (last_room_id, batch_size)) return [r for r, in txn] - rooms_to_work_on = yield self.runInteraction( + rooms_to_work_on = yield self.db.runInteraction( "populate_stats_rooms_get_batch", _get_next_batch ) # No more rooms -- complete the transaction. if not rooms_to_work_on: - yield self._end_background_update("populate_stats_process_rooms") + yield self.db.updates._end_background_update("populate_stats_process_rooms") return 1 for room_id in rooms_to_work_on: yield self._calculate_and_set_initial_state_for_room(room_id) progress["last_room_id"] = room_id - yield self.runInteraction( + yield self.db.runInteraction( "_populate_stats_process_rooms", - self._background_update_progress_txn, + self.db.updates._background_update_progress_txn, "populate_stats_process_rooms", progress, ) @@ -186,7 +187,7 @@ class StatsStore(StateDeltasStore): """ Returns the stats processor positions. """ - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="stats_incremental_position", keyvalues={}, retcol="stream_id", @@ -215,7 +216,7 @@ class StatsStore(StateDeltasStore): if field and "\0" in field: fields[col] = None - return self._simple_upsert( + return self.db.simple_upsert( table="room_stats_state", keyvalues={"room_id": room_id}, values=fields, @@ -236,7 +237,7 @@ class StatsStore(StateDeltasStore): Deferred[list[dict]], where the dict has the keys of ABSOLUTE_STATS_FIELDS[stats_type], and "bucket_size" and "end_ts". """ - return self.runInteraction( + return self.db.runInteraction( "get_statistics_for_subject", self._get_statistics_for_subject_txn, stats_type, @@ -257,14 +258,14 @@ class StatsStore(StateDeltasStore): ABSOLUTE_STATS_FIELDS[stats_type] + PER_SLICE_FIELDS[stats_type] ) - slice_list = self._simple_select_list_paginate_txn( + slice_list = self.db.simple_select_list_paginate_txn( txn, table + "_historical", - {id_col: stats_id}, "end_ts", start, size, retcols=selected_columns + ["bucket_size", "end_ts"], + keyvalues={id_col: stats_id}, order_direction="DESC", ) @@ -282,7 +283,7 @@ class StatsStore(StateDeltasStore): "name", "topic", "canonical_alias", "avatar", "join_rules", "history_visibility" """ - return self._simple_select_one( + return self.db.simple_select_one( "room_stats_state", {"room_id": room_id}, retcols=( @@ -308,7 +309,7 @@ class StatsStore(StateDeltasStore): """ table, id_col = TYPE_TO_TABLE[stats_type] - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( "%s_current" % (table,), keyvalues={id_col: id}, retcol="completed_delta_stream_id", @@ -344,14 +345,14 @@ class StatsStore(StateDeltasStore): complete_with_stream_id=stream_id, ) - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, table="stats_incremental_position", keyvalues={}, updatevalues={"stream_id": stream_id}, ) - return self.runInteraction( + return self.db.runInteraction( "bulk_update_stats_delta", _bulk_update_stats_delta_txn ) @@ -382,7 +383,7 @@ class StatsStore(StateDeltasStore): Does not work with per-slice fields. """ - return self.runInteraction( + return self.db.runInteraction( "update_stats_delta", self._update_stats_delta_txn, ts, @@ -517,17 +518,17 @@ class StatsStore(StateDeltasStore): else: self.database_engine.lock_table(txn, table) retcols = list(chain(absolutes.keys(), additive_relatives.keys())) - current_row = self._simple_select_one_txn( + current_row = self.db.simple_select_one_txn( txn, table, keyvalues, retcols, allow_none=True ) if current_row is None: merged_dict = {**keyvalues, **absolutes, **additive_relatives} - self._simple_insert_txn(txn, table, merged_dict) + self.db.simple_insert_txn(txn, table, merged_dict) else: for (key, val) in additive_relatives.items(): current_row[key] += val current_row.update(absolutes) - self._simple_update_one_txn(txn, table, keyvalues, current_row) + self.db.simple_update_one_txn(txn, table, keyvalues, current_row) def _upsert_copy_from_table_with_additive_relatives_txn( self, @@ -614,11 +615,11 @@ class StatsStore(StateDeltasStore): txn.execute(sql, qargs) else: self.database_engine.lock_table(txn, into_table) - src_row = self._simple_select_one_txn( + src_row = self.db.simple_select_one_txn( txn, src_table, keyvalues, copy_columns ) all_dest_keyvalues = {**keyvalues, **extra_dst_keyvalues} - dest_current_row = self._simple_select_one_txn( + dest_current_row = self.db.simple_select_one_txn( txn, into_table, keyvalues=all_dest_keyvalues, @@ -634,11 +635,11 @@ class StatsStore(StateDeltasStore): **src_row, **additive_relatives, } - self._simple_insert_txn(txn, into_table, merged_dict) + self.db.simple_insert_txn(txn, into_table, merged_dict) else: for (key, val) in additive_relatives.items(): src_row[key] = dest_current_row[key] + val - self._simple_update_txn(txn, into_table, all_dest_keyvalues, src_row) + self.db.simple_update_txn(txn, into_table, all_dest_keyvalues, src_row) def get_changes_room_total_events_and_bytes(self, min_pos, max_pos): """Fetches the counts of events in the given range of stream IDs. @@ -652,7 +653,7 @@ class StatsStore(StateDeltasStore): changes. """ - return self.runInteraction( + return self.db.runInteraction( "stats_incremental_total_events_and_bytes", self.get_changes_room_total_events_and_bytes_txn, min_pos, @@ -735,7 +736,7 @@ class StatsStore(StateDeltasStore): def _fetch_current_state_stats(txn): pos = self.get_room_max_stream_ordering() - rows = self._simple_select_many_txn( + rows = self.db.simple_select_many_txn( txn, table="current_state_events", column="type", @@ -791,7 +792,7 @@ class StatsStore(StateDeltasStore): current_state_events_count, users_in_room, pos, - ) = yield self.runInteraction( + ) = yield self.db.runInteraction( "get_initial_state_for_room", _fetch_current_state_stats ) @@ -866,7 +867,7 @@ class StatsStore(StateDeltasStore): (count,) = txn.fetchone() return count, pos - joined_rooms, pos = yield self.runInteraction( + joined_rooms, pos = yield self.db.runInteraction( "calculate_and_set_initial_state_for_user", _calculate_and_set_initial_state_for_user_txn, ) diff --git a/synapse/storage/data_stores/main/stream.py b/synapse/storage/data_stores/main/stream.py index 8780fdd989..140da8dad6 100644 --- a/synapse/storage/data_stores/main/stream.py +++ b/synapse/storage/data_stores/main/stream.py @@ -1,5 +1,8 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 OpenMarket Ltd +# Copyright 2017 Vector Creations Ltd +# Copyright 2018-2019 New Vector Ltd +# Copyright 2019 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -44,6 +47,7 @@ from twisted.internet import defer from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.storage._base import SQLBaseStore from synapse.storage.data_stores.main.events_worker import EventsWorkerStore +from synapse.storage.database import Database from synapse.storage.engines import PostgresEngine from synapse.types import RoomStreamToken from synapse.util.caches.stream_change_cache import StreamChangeCache @@ -248,11 +252,11 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): __metaclass__ = abc.ABCMeta - def __init__(self, db_conn, hs): - super(StreamWorkerStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(StreamWorkerStore, self).__init__(database, db_conn, hs) events_max = self.get_room_max_stream_ordering() - event_cache_prefill, min_event_val = self._get_cache_dict( + event_cache_prefill, min_event_val = self.db.get_cache_dict( db_conn, "events", entity_column="room_id", @@ -397,7 +401,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): rows = [_EventDictReturn(row[0], None, row[1]) for row in txn] return rows - rows = yield self.runInteraction("get_room_events_stream_for_room", f) + rows = yield self.db.runInteraction("get_room_events_stream_for_room", f) ret = yield self.get_events_as_list( [r.event_id for r in rows], get_prev_content=True @@ -447,7 +451,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): return rows - rows = yield self.runInteraction("get_membership_changes_for_user", f) + rows = yield self.db.runInteraction("get_membership_changes_for_user", f) ret = yield self.get_events_as_list( [r.event_id for r in rows], get_prev_content=True @@ -508,7 +512,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): end_token = RoomStreamToken.parse(end_token) - rows, token = yield self.runInteraction( + rows, token = yield self.db.runInteraction( "get_recent_event_ids_for_room", self._paginate_room_events_txn, room_id, @@ -545,7 +549,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): txn.execute(sql, (room_id, stream_ordering)) return txn.fetchone() - return self.runInteraction("get_room_event_after_stream_ordering", _f) + return self.db.runInteraction("get_room_event_after_stream_ordering", _f) @defer.inlineCallbacks def get_room_events_max_id(self, room_id=None): @@ -559,7 +563,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): if room_id is None: return "s%d" % (token,) else: - topo = yield self.runInteraction( + topo = yield self.db.runInteraction( "_get_max_topological_txn", self._get_max_topological_txn, room_id ) return "t%d-%d" % (topo, token) @@ -573,7 +577,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): Returns: A deferred "s%d" stream token. """ - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="events", keyvalues={"event_id": event_id}, retcol="stream_ordering" ).addCallback(lambda row: "s%d" % (row,)) @@ -586,7 +590,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): Returns: A deferred "t%d-%d" topological token. """ - return self._simple_select_one( + return self.db.simple_select_one( table="events", keyvalues={"event_id": event_id}, retcols=("stream_ordering", "topological_ordering"), @@ -610,13 +614,13 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): "SELECT coalesce(max(topological_ordering), 0) FROM events" " WHERE room_id = ? AND stream_ordering < ?" ) - return self._execute( + return self.db.execute( "get_max_topological_token", None, sql, room_id, stream_key ).addCallback(lambda r: r[0][0] if r else 0) def _get_max_topological_txn(self, txn, room_id): txn.execute( - "SELECT MAX(topological_ordering) FROM events" " WHERE room_id = ?", + "SELECT MAX(topological_ordering) FROM events WHERE room_id = ?", (room_id,), ) @@ -664,7 +668,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): dict """ - results = yield self.runInteraction( + results = yield self.db.runInteraction( "get_events_around", self._get_events_around_txn, room_id, @@ -706,7 +710,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): dict """ - results = self._simple_select_one_txn( + results = self.db.simple_select_one_txn( txn, "events", keyvalues={"event_id": event_id, "room_id": room_id}, @@ -785,7 +789,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): return upper_bound, [row[1] for row in rows] - upper_bound, event_ids = yield self.runInteraction( + upper_bound, event_ids = yield self.db.runInteraction( "get_all_new_events_stream", get_all_new_events_stream_txn ) @@ -794,7 +798,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): return upper_bound, events def get_federation_out_pos(self, typ): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="federation_stream_position", retcol="stream_id", keyvalues={"type": typ}, @@ -802,7 +806,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): ) def update_federation_out_pos(self, typ, stream_id): - return self._simple_update_one( + return self.db.simple_update_one( table="federation_stream_position", keyvalues={"type": typ}, updatevalues={"stream_id": stream_id}, @@ -953,7 +957,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): if to_key: to_key = RoomStreamToken.parse(to_key) - rows, token = yield self.runInteraction( + rows, token = yield self.db.runInteraction( "paginate_room_events", self._paginate_room_events_txn, room_id, diff --git a/synapse/storage/data_stores/main/tags.py b/synapse/storage/data_stores/main/tags.py index 10d1887f75..2aa1bafd48 100644 --- a/synapse/storage/data_stores/main/tags.py +++ b/synapse/storage/data_stores/main/tags.py @@ -41,7 +41,7 @@ class TagsWorkerStore(AccountDataWorkerStore): tag strings to tag content. """ - deferred = self._simple_select_list( + deferred = self.db.simple_select_list( "room_tags", {"user_id": user_id}, ["room_id", "tag", "content"] ) @@ -78,14 +78,12 @@ class TagsWorkerStore(AccountDataWorkerStore): txn.execute(sql, (last_id, current_id, limit)) return txn.fetchall() - tag_ids = yield self.runInteraction( + tag_ids = yield self.db.runInteraction( "get_all_updated_tags", get_all_updated_tags_txn ) def get_tag_content(txn, tag_ids): - sql = ( - "SELECT tag, content" " FROM room_tags" " WHERE user_id=? AND room_id=?" - ) + sql = "SELECT tag, content FROM room_tags WHERE user_id=? AND room_id=?" results = [] for stream_id, user_id, room_id in tag_ids: txn.execute(sql, (user_id, room_id)) @@ -100,7 +98,7 @@ class TagsWorkerStore(AccountDataWorkerStore): batch_size = 50 results = [] for i in range(0, len(tag_ids), batch_size): - tags = yield self.runInteraction( + tags = yield self.db.runInteraction( "get_all_updated_tag_content", get_tag_content, tag_ids[i : i + batch_size], @@ -137,7 +135,9 @@ class TagsWorkerStore(AccountDataWorkerStore): if not changed: return {} - room_ids = yield self.runInteraction("get_updated_tags", get_updated_tags_txn) + room_ids = yield self.db.runInteraction( + "get_updated_tags", get_updated_tags_txn + ) results = {} if room_ids: @@ -155,7 +155,7 @@ class TagsWorkerStore(AccountDataWorkerStore): Returns: A deferred list of string tags. """ - return self._simple_select_list( + return self.db.simple_select_list( table="room_tags", keyvalues={"user_id": user_id, "room_id": room_id}, retcols=("tag", "content"), @@ -180,7 +180,7 @@ class TagsStore(TagsWorkerStore): content_json = json.dumps(content) def add_tag_txn(txn, next_id): - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="room_tags", keyvalues={"user_id": user_id, "room_id": room_id, "tag": tag}, @@ -189,7 +189,7 @@ class TagsStore(TagsWorkerStore): self._update_revision_txn(txn, user_id, room_id, next_id) with self._account_data_id_gen.get_next() as next_id: - yield self.runInteraction("add_tag", add_tag_txn, next_id) + yield self.db.runInteraction("add_tag", add_tag_txn, next_id) self.get_tags_for_user.invalidate((user_id,)) @@ -212,7 +212,7 @@ class TagsStore(TagsWorkerStore): self._update_revision_txn(txn, user_id, room_id, next_id) with self._account_data_id_gen.get_next() as next_id: - yield self.runInteraction("remove_tag", remove_tag_txn, next_id) + yield self.db.runInteraction("remove_tag", remove_tag_txn, next_id) self.get_tags_for_user.invalidate((user_id,)) diff --git a/synapse/storage/data_stores/main/transactions.py b/synapse/storage/data_stores/main/transactions.py index 01b1be5e14..5b07c2fbc0 100644 --- a/synapse/storage/data_stores/main/transactions.py +++ b/synapse/storage/data_stores/main/transactions.py @@ -24,6 +24,7 @@ from twisted.internet import defer from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage._base import SQLBaseStore, db_to_json +from synapse.storage.database import Database from synapse.util.caches.expiringcache import ExpiringCache # py2 sqlite has buffer hardcoded as only binary type, so we must use it, @@ -52,8 +53,8 @@ class TransactionStore(SQLBaseStore): """A collection of queries for handling PDUs. """ - def __init__(self, db_conn, hs): - super(TransactionStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(TransactionStore, self).__init__(database, db_conn, hs) self._clock.looping_call(self._start_cleanup_transactions, 30 * 60 * 1000) @@ -77,7 +78,7 @@ class TransactionStore(SQLBaseStore): this transaction or a 2-tuple of (int, dict) """ - return self.runInteraction( + return self.db.runInteraction( "get_received_txn_response", self._get_received_txn_response, transaction_id, @@ -85,7 +86,7 @@ class TransactionStore(SQLBaseStore): ) def _get_received_txn_response(self, txn, transaction_id, origin): - result = self._simple_select_one_txn( + result = self.db.simple_select_one_txn( txn, table="received_transactions", keyvalues={"transaction_id": transaction_id, "origin": origin}, @@ -119,7 +120,7 @@ class TransactionStore(SQLBaseStore): response_json (str) """ - return self._simple_insert( + return self.db.simple_insert( table="received_transactions", values={ "transaction_id": transaction_id, @@ -148,7 +149,7 @@ class TransactionStore(SQLBaseStore): if result is not SENTINEL: return result - result = yield self.runInteraction( + result = yield self.db.runInteraction( "get_destination_retry_timings", self._get_destination_retry_timings, destination, @@ -160,7 +161,7 @@ class TransactionStore(SQLBaseStore): return result def _get_destination_retry_timings(self, txn, destination): - result = self._simple_select_one_txn( + result = self.db.simple_select_one_txn( txn, table="destinations", keyvalues={"destination": destination}, @@ -187,7 +188,7 @@ class TransactionStore(SQLBaseStore): """ self._destination_retry_cache.pop(destination, None) - return self.runInteraction( + return self.db.runInteraction( "set_destination_retry_timings", self._set_destination_retry_timings, destination, @@ -227,7 +228,7 @@ class TransactionStore(SQLBaseStore): # We need to be careful here as the data may have changed from under us # due to a worker setting the timings. - prev_row = self._simple_select_one_txn( + prev_row = self.db.simple_select_one_txn( txn, table="destinations", keyvalues={"destination": destination}, @@ -236,7 +237,7 @@ class TransactionStore(SQLBaseStore): ) if not prev_row: - self._simple_insert_txn( + self.db.simple_insert_txn( txn, table="destinations", values={ @@ -247,7 +248,7 @@ class TransactionStore(SQLBaseStore): }, ) elif retry_interval == 0 or prev_row["retry_interval"] < retry_interval: - self._simple_update_one_txn( + self.db.simple_update_one_txn( txn, "destinations", keyvalues={"destination": destination}, @@ -270,4 +271,6 @@ class TransactionStore(SQLBaseStore): def _cleanup_transactions_txn(txn): txn.execute("DELETE FROM received_transactions WHERE ts < ?", (month_ago,)) - return self.runInteraction("_cleanup_transactions", _cleanup_transactions_txn) + return self.db.runInteraction( + "_cleanup_transactions", _cleanup_transactions_txn + ) diff --git a/synapse/storage/data_stores/main/user_directory.py b/synapse/storage/data_stores/main/user_directory.py index 652abe0e6a..90c180ec6d 100644 --- a/synapse/storage/data_stores/main/user_directory.py +++ b/synapse/storage/data_stores/main/user_directory.py @@ -19,9 +19,9 @@ import re from twisted.internet import defer from synapse.api.constants import EventTypes, JoinRules -from synapse.storage.background_updates import BackgroundUpdateStore from synapse.storage.data_stores.main.state import StateFilter from synapse.storage.data_stores.main.state_deltas import StateDeltasStore +from synapse.storage.database import Database from synapse.storage.engines import PostgresEngine, Sqlite3Engine from synapse.types import get_domain_from_id, get_localpart_from_id from synapse.util.caches.descriptors import cached @@ -32,30 +32,30 @@ logger = logging.getLogger(__name__) TEMP_TABLE = "_temp_populate_user_directory" -class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore): +class UserDirectoryBackgroundUpdateStore(StateDeltasStore): # How many records do we calculate before sending it to # add_users_who_share_private_rooms? SHARE_PRIVATE_WORKING_SET = 500 - def __init__(self, db_conn, hs): - super(UserDirectoryBackgroundUpdateStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(UserDirectoryBackgroundUpdateStore, self).__init__(database, db_conn, hs) self.server_name = hs.hostname - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "populate_user_directory_createtables", self._populate_user_directory_createtables, ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "populate_user_directory_process_rooms", self._populate_user_directory_process_rooms, ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "populate_user_directory_process_users", self._populate_user_directory_process_users, ) - self.register_background_update_handler( + self.db.updates.register_background_update_handler( "populate_user_directory_cleanup", self._populate_user_directory_cleanup ) @@ -85,7 +85,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore """ txn.execute(sql) rooms = [{"room_id": x[0], "events": x[1]} for x in txn.fetchall()] - self._simple_insert_many_txn(txn, TEMP_TABLE + "_rooms", rooms) + self.db.simple_insert_many_txn(txn, TEMP_TABLE + "_rooms", rooms) del rooms # If search all users is on, get all the users we want to add. @@ -100,15 +100,17 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore txn.execute("SELECT name FROM users") users = [{"user_id": x[0]} for x in txn.fetchall()] - self._simple_insert_many_txn(txn, TEMP_TABLE + "_users", users) + self.db.simple_insert_many_txn(txn, TEMP_TABLE + "_users", users) new_pos = yield self.get_max_stream_id_in_current_state_deltas() - yield self.runInteraction( + yield self.db.runInteraction( "populate_user_directory_temp_build", _make_staging_area ) - yield self._simple_insert(TEMP_TABLE + "_position", {"position": new_pos}) + yield self.db.simple_insert(TEMP_TABLE + "_position", {"position": new_pos}) - yield self._end_background_update("populate_user_directory_createtables") + yield self.db.updates._end_background_update( + "populate_user_directory_createtables" + ) return 1 @defer.inlineCallbacks @@ -116,7 +118,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore """ Update the user directory stream position, then clean up the old tables. """ - position = yield self._simple_select_one_onecol( + position = yield self.db.simple_select_one_onecol( TEMP_TABLE + "_position", None, "position" ) yield self.update_user_directory_stream_pos(position) @@ -126,11 +128,11 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore txn.execute("DROP TABLE IF EXISTS " + TEMP_TABLE + "_users") txn.execute("DROP TABLE IF EXISTS " + TEMP_TABLE + "_position") - yield self.runInteraction( + yield self.db.runInteraction( "populate_user_directory_cleanup", _delete_staging_area ) - yield self._end_background_update("populate_user_directory_cleanup") + yield self.db.updates._end_background_update("populate_user_directory_cleanup") return 1 @defer.inlineCallbacks @@ -170,13 +172,15 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore return rooms_to_work_on - rooms_to_work_on = yield self.runInteraction( + rooms_to_work_on = yield self.db.runInteraction( "populate_user_directory_temp_read", _get_next_batch ) # No more rooms -- complete the transaction. if not rooms_to_work_on: - yield self._end_background_update("populate_user_directory_process_rooms") + yield self.db.updates._end_background_update( + "populate_user_directory_process_rooms" + ) return 1 logger.info( @@ -243,12 +247,12 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore to_insert.clear() # We've finished a room. Delete it from the table. - yield self._simple_delete_one(TEMP_TABLE + "_rooms", {"room_id": room_id}) + yield self.db.simple_delete_one(TEMP_TABLE + "_rooms", {"room_id": room_id}) # Update the remaining counter. progress["remaining"] -= 1 - yield self.runInteraction( + yield self.db.runInteraction( "populate_user_directory", - self._background_update_progress_txn, + self.db.updates._background_update_progress_txn, "populate_user_directory_process_rooms", progress, ) @@ -267,7 +271,9 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore If search_all_users is enabled, add all of the users to the user directory. """ if not self.hs.config.user_directory_search_all_users: - yield self._end_background_update("populate_user_directory_process_users") + yield self.db.updates._end_background_update( + "populate_user_directory_process_users" + ) return 1 def _get_next_batch(txn): @@ -291,13 +297,15 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore return users_to_work_on - users_to_work_on = yield self.runInteraction( + users_to_work_on = yield self.db.runInteraction( "populate_user_directory_temp_read", _get_next_batch ) # No more users -- complete the transaction. if not users_to_work_on: - yield self._end_background_update("populate_user_directory_process_users") + yield self.db.updates._end_background_update( + "populate_user_directory_process_users" + ) return 1 logger.info( @@ -312,12 +320,12 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore ) # We've finished processing a user. Delete it from the table. - yield self._simple_delete_one(TEMP_TABLE + "_users", {"user_id": user_id}) + yield self.db.simple_delete_one(TEMP_TABLE + "_users", {"user_id": user_id}) # Update the remaining counter. progress["remaining"] -= 1 - yield self.runInteraction( + yield self.db.runInteraction( "populate_user_directory", - self._background_update_progress_txn, + self.db.updates._background_update_progress_txn, "populate_user_directory_process_users", progress, ) @@ -361,7 +369,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore """ def _update_profile_in_user_dir_txn(txn): - new_entry = self._simple_upsert_txn( + new_entry = self.db.simple_upsert_txn( txn, table="user_directory", keyvalues={"user_id": user_id}, @@ -435,7 +443,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore ) elif isinstance(self.database_engine, Sqlite3Engine): value = "%s %s" % (user_id, display_name) if display_name else user_id - self._simple_upsert_txn( + self.db.simple_upsert_txn( txn, table="user_directory_search", keyvalues={"user_id": user_id}, @@ -448,7 +456,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore txn.call_after(self.get_user_in_directory.invalidate, (user_id,)) - return self.runInteraction( + return self.db.runInteraction( "update_profile_in_user_dir", _update_profile_in_user_dir_txn ) @@ -462,7 +470,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore """ def _add_users_who_share_room_txn(txn): - self._simple_upsert_many_txn( + self.db.simple_upsert_many_txn( txn, table="users_who_share_private_rooms", key_names=["user_id", "other_user_id", "room_id"], @@ -474,7 +482,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore value_values=None, ) - return self.runInteraction( + return self.db.runInteraction( "add_users_who_share_room", _add_users_who_share_room_txn ) @@ -489,7 +497,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore def _add_users_in_public_rooms_txn(txn): - self._simple_upsert_many_txn( + self.db.simple_upsert_many_txn( txn, table="users_in_public_rooms", key_names=["user_id", "room_id"], @@ -498,7 +506,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore value_values=None, ) - return self.runInteraction( + return self.db.runInteraction( "add_users_in_public_rooms", _add_users_in_public_rooms_txn ) @@ -513,13 +521,13 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore txn.execute("DELETE FROM users_who_share_private_rooms") txn.call_after(self.get_user_in_directory.invalidate_all) - return self.runInteraction( + return self.db.runInteraction( "delete_all_from_user_dir", _delete_all_from_user_dir_txn ) @cached() def get_user_in_directory(self, user_id): - return self._simple_select_one( + return self.db.simple_select_one( table="user_directory", keyvalues={"user_id": user_id}, retcols=("display_name", "avatar_url"), @@ -528,7 +536,7 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore, BackgroundUpdateStore ) def update_user_directory_stream_pos(self, stream_id): - return self._simple_update_one( + return self.db.simple_update_one( table="user_directory_stream_pos", keyvalues={}, updatevalues={"stream_id": stream_id}, @@ -542,47 +550,47 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): # add_users_who_share_private_rooms? SHARE_PRIVATE_WORKING_SET = 500 - def __init__(self, db_conn, hs): - super(UserDirectoryStore, self).__init__(db_conn, hs) + def __init__(self, database: Database, db_conn, hs): + super(UserDirectoryStore, self).__init__(database, db_conn, hs) def remove_from_user_dir(self, user_id): def _remove_from_user_dir_txn(txn): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="user_directory", keyvalues={"user_id": user_id} ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="user_directory_search", keyvalues={"user_id": user_id} ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="users_in_public_rooms", keyvalues={"user_id": user_id} ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="users_who_share_private_rooms", keyvalues={"user_id": user_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="users_who_share_private_rooms", keyvalues={"other_user_id": user_id}, ) txn.call_after(self.get_user_in_directory.invalidate, (user_id,)) - return self.runInteraction("remove_from_user_dir", _remove_from_user_dir_txn) + return self.db.runInteraction("remove_from_user_dir", _remove_from_user_dir_txn) @defer.inlineCallbacks def get_users_in_dir_due_to_room(self, room_id): """Get all user_ids that are in the room directory because they're in the given room_id """ - user_ids_share_pub = yield self._simple_select_onecol( + user_ids_share_pub = yield self.db.simple_select_onecol( table="users_in_public_rooms", keyvalues={"room_id": room_id}, retcol="user_id", desc="get_users_in_dir_due_to_room", ) - user_ids_share_priv = yield self._simple_select_onecol( + user_ids_share_priv = yield self.db.simple_select_onecol( table="users_who_share_private_rooms", keyvalues={"room_id": room_id}, retcol="other_user_id", @@ -605,23 +613,23 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): """ def _remove_user_who_share_room_txn(txn): - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="users_who_share_private_rooms", keyvalues={"user_id": user_id, "room_id": room_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="users_who_share_private_rooms", keyvalues={"other_user_id": user_id, "room_id": room_id}, ) - self._simple_delete_txn( + self.db.simple_delete_txn( txn, table="users_in_public_rooms", keyvalues={"user_id": user_id, "room_id": room_id}, ) - return self.runInteraction( + return self.db.runInteraction( "remove_user_who_share_room", _remove_user_who_share_room_txn ) @@ -636,14 +644,14 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): Returns: list: user_id """ - rows = yield self._simple_select_onecol( + rows = yield self.db.simple_select_onecol( table="users_who_share_private_rooms", keyvalues={"user_id": user_id}, retcol="room_id", desc="get_rooms_user_is_in", ) - pub_rows = yield self._simple_select_onecol( + pub_rows = yield self.db.simple_select_onecol( table="users_in_public_rooms", keyvalues={"user_id": user_id}, retcol="room_id", @@ -674,14 +682,14 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): ) f2 USING (room_id) """ - rows = yield self._execute( + rows = yield self.db.execute( "get_rooms_in_common_for_users", None, sql, user_id, other_user_id ) return [room_id for room_id, in rows] def get_user_directory_stream_pos(self): - return self._simple_select_one_onecol( + return self.db.simple_select_one_onecol( table="user_directory_stream_pos", keyvalues={}, retcol="stream_id", @@ -786,8 +794,8 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): # This should be unreachable. raise Exception("Unrecognized database engine") - results = yield self._execute( - "search_user_dir", self.cursor_to_dict, sql, *args + results = yield self.db.execute( + "search_user_dir", self.db.cursor_to_dict, sql, *args ) limited = len(results) > limit diff --git a/synapse/storage/data_stores/main/user_erasure_store.py b/synapse/storage/data_stores/main/user_erasure_store.py index aa4f0da5f0..af8025bc17 100644 --- a/synapse/storage/data_stores/main/user_erasure_store.py +++ b/synapse/storage/data_stores/main/user_erasure_store.py @@ -31,7 +31,7 @@ class UserErasureWorkerStore(SQLBaseStore): Returns: Deferred[bool]: True if the user has requested erasure """ - return self._simple_select_onecol( + return self.db.simple_select_onecol( table="erased_users", keyvalues={"user_id": user_id}, retcol="1", @@ -56,7 +56,7 @@ class UserErasureWorkerStore(SQLBaseStore): # iterate it multiple times, and (b) avoiding duplicates. user_ids = tuple(set(user_ids)) - rows = yield self._simple_select_many_batch( + rows = yield self.db.simple_select_many_batch( table="erased_users", column="user_id", iterable=user_ids, @@ -88,4 +88,4 @@ class UserErasureStore(UserErasureWorkerStore): self._invalidate_cache_and_stream(txn, self.is_user_erased, (user_id,)) - return self.runInteraction("mark_user_erased", f) + return self.db.runInteraction("mark_user_erased", f) diff --git a/synapse/storage/data_stores/state/__init__.py b/synapse/storage/data_stores/state/__init__.py new file mode 100644 index 0000000000..86e09f6229 --- /dev/null +++ b/synapse/storage/data_stores/state/__init__.py @@ -0,0 +1,16 @@ +# -*- coding: utf-8 -*- +# Copyright 2019 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from synapse.storage.data_stores.state.store import StateGroupDataStore # noqa: F401 diff --git a/synapse/storage/data_stores/state/bg_updates.py b/synapse/storage/data_stores/state/bg_updates.py new file mode 100644 index 0000000000..e8edaf9f7b --- /dev/null +++ b/synapse/storage/data_stores/state/bg_updates.py @@ -0,0 +1,374 @@ +# -*- coding: utf-8 -*- +# Copyright 2014-2016 OpenMarket Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging + +from six import iteritems + +from twisted.internet import defer + +from synapse.storage._base import SQLBaseStore +from synapse.storage.database import Database +from synapse.storage.engines import PostgresEngine +from synapse.storage.state import StateFilter + +logger = logging.getLogger(__name__) + + +MAX_STATE_DELTA_HOPS = 100 + + +class StateGroupBackgroundUpdateStore(SQLBaseStore): + """Defines functions related to state groups needed to run the state backgroud + updates. + """ + + def _count_state_group_hops_txn(self, txn, state_group): + """Given a state group, count how many hops there are in the tree. + + This is used to ensure the delta chains don't get too long. + """ + if isinstance(self.database_engine, PostgresEngine): + sql = """ + WITH RECURSIVE state(state_group) AS ( + VALUES(?::bigint) + UNION ALL + SELECT prev_state_group FROM state_group_edges e, state s + WHERE s.state_group = e.state_group + ) + SELECT count(*) FROM state; + """ + + txn.execute(sql, (state_group,)) + row = txn.fetchone() + if row and row[0]: + return row[0] + else: + return 0 + else: + # We don't use WITH RECURSIVE on sqlite3 as there are distributions + # that ship with an sqlite3 version that doesn't support it (e.g. wheezy) + next_group = state_group + count = 0 + + while next_group: + next_group = self.db.simple_select_one_onecol_txn( + txn, + table="state_group_edges", + keyvalues={"state_group": next_group}, + retcol="prev_state_group", + allow_none=True, + ) + if next_group: + count += 1 + + return count + + def _get_state_groups_from_groups_txn( + self, txn, groups, state_filter=StateFilter.all() + ): + results = {group: {} for group in groups} + + where_clause, where_args = state_filter.make_sql_filter_clause() + + # Unless the filter clause is empty, we're going to append it after an + # existing where clause + if where_clause: + where_clause = " AND (%s)" % (where_clause,) + + if isinstance(self.database_engine, PostgresEngine): + # Temporarily disable sequential scans in this transaction. This is + # a temporary hack until we can add the right indices in + txn.execute("SET LOCAL enable_seqscan=off") + + # The below query walks the state_group tree so that the "state" + # table includes all state_groups in the tree. It then joins + # against `state_groups_state` to fetch the latest state. + # It assumes that previous state groups are always numerically + # lesser. + # The PARTITION is used to get the event_id in the greatest state + # group for the given type, state_key. + # This may return multiple rows per (type, state_key), but last_value + # should be the same. + sql = """ + WITH RECURSIVE state(state_group) AS ( + VALUES(?::bigint) + UNION ALL + SELECT prev_state_group FROM state_group_edges e, state s + WHERE s.state_group = e.state_group + ) + SELECT DISTINCT type, state_key, last_value(event_id) OVER ( + PARTITION BY type, state_key ORDER BY state_group ASC + ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING + ) AS event_id FROM state_groups_state + WHERE state_group IN ( + SELECT state_group FROM state + ) + """ + + for group in groups: + args = [group] + args.extend(where_args) + + txn.execute(sql + where_clause, args) + for row in txn: + typ, state_key, event_id = row + key = (typ, state_key) + results[group][key] = event_id + else: + max_entries_returned = state_filter.max_entries_returned() + + # We don't use WITH RECURSIVE on sqlite3 as there are distributions + # that ship with an sqlite3 version that doesn't support it (e.g. wheezy) + for group in groups: + next_group = group + + while next_group: + # We did this before by getting the list of group ids, and + # then passing that list to sqlite to get latest event for + # each (type, state_key). However, that was terribly slow + # without the right indices (which we can't add until + # after we finish deduping state, which requires this func) + args = [next_group] + args.extend(where_args) + + txn.execute( + "SELECT type, state_key, event_id FROM state_groups_state" + " WHERE state_group = ? " + where_clause, + args, + ) + results[group].update( + ((typ, state_key), event_id) + for typ, state_key, event_id in txn + if (typ, state_key) not in results[group] + ) + + # If the number of entries in the (type,state_key)->event_id dict + # matches the number of (type,state_keys) types we were searching + # for, then we must have found them all, so no need to go walk + # further down the tree... UNLESS our types filter contained + # wildcards (i.e. Nones) in which case we have to do an exhaustive + # search + if ( + max_entries_returned is not None + and len(results[group]) == max_entries_returned + ): + break + + next_group = self.db.simple_select_one_onecol_txn( + txn, + table="state_group_edges", + keyvalues={"state_group": next_group}, + retcol="prev_state_group", + allow_none=True, + ) + + return results + + +class StateBackgroundUpdateStore(StateGroupBackgroundUpdateStore): + + STATE_GROUP_DEDUPLICATION_UPDATE_NAME = "state_group_state_deduplication" + STATE_GROUP_INDEX_UPDATE_NAME = "state_group_state_type_index" + STATE_GROUPS_ROOM_INDEX_UPDATE_NAME = "state_groups_room_id_idx" + + def __init__(self, database: Database, db_conn, hs): + super(StateBackgroundUpdateStore, self).__init__(database, db_conn, hs) + self.db.updates.register_background_update_handler( + self.STATE_GROUP_DEDUPLICATION_UPDATE_NAME, + self._background_deduplicate_state, + ) + self.db.updates.register_background_update_handler( + self.STATE_GROUP_INDEX_UPDATE_NAME, self._background_index_state + ) + self.db.updates.register_background_index_update( + self.STATE_GROUPS_ROOM_INDEX_UPDATE_NAME, + index_name="state_groups_room_id_idx", + table="state_groups", + columns=["room_id"], + ) + + @defer.inlineCallbacks + def _background_deduplicate_state(self, progress, batch_size): + """This background update will slowly deduplicate state by reencoding + them as deltas. + """ + last_state_group = progress.get("last_state_group", 0) + rows_inserted = progress.get("rows_inserted", 0) + max_group = progress.get("max_group", None) + + BATCH_SIZE_SCALE_FACTOR = 100 + + batch_size = max(1, int(batch_size / BATCH_SIZE_SCALE_FACTOR)) + + if max_group is None: + rows = yield self.db.execute( + "_background_deduplicate_state", + None, + "SELECT coalesce(max(id), 0) FROM state_groups", + ) + max_group = rows[0][0] + + def reindex_txn(txn): + new_last_state_group = last_state_group + for count in range(batch_size): + txn.execute( + "SELECT id, room_id FROM state_groups" + " WHERE ? < id AND id <= ?" + " ORDER BY id ASC" + " LIMIT 1", + (new_last_state_group, max_group), + ) + row = txn.fetchone() + if row: + state_group, room_id = row + + if not row or not state_group: + return True, count + + txn.execute( + "SELECT state_group FROM state_group_edges" + " WHERE state_group = ?", + (state_group,), + ) + + # If we reach a point where we've already started inserting + # edges we should stop. + if txn.fetchall(): + return True, count + + txn.execute( + "SELECT coalesce(max(id), 0) FROM state_groups" + " WHERE id < ? AND room_id = ?", + (state_group, room_id), + ) + (prev_group,) = txn.fetchone() + new_last_state_group = state_group + + if prev_group: + potential_hops = self._count_state_group_hops_txn(txn, prev_group) + if potential_hops >= MAX_STATE_DELTA_HOPS: + # We want to ensure chains are at most this long,# + # otherwise read performance degrades. + continue + + prev_state = self._get_state_groups_from_groups_txn( + txn, [prev_group] + ) + prev_state = prev_state[prev_group] + + curr_state = self._get_state_groups_from_groups_txn( + txn, [state_group] + ) + curr_state = curr_state[state_group] + + if not set(prev_state.keys()) - set(curr_state.keys()): + # We can only do a delta if the current has a strict super set + # of keys + + delta_state = { + key: value + for key, value in iteritems(curr_state) + if prev_state.get(key, None) != value + } + + self.db.simple_delete_txn( + txn, + table="state_group_edges", + keyvalues={"state_group": state_group}, + ) + + self.db.simple_insert_txn( + txn, + table="state_group_edges", + values={ + "state_group": state_group, + "prev_state_group": prev_group, + }, + ) + + self.db.simple_delete_txn( + txn, + table="state_groups_state", + keyvalues={"state_group": state_group}, + ) + + self.db.simple_insert_many_txn( + txn, + table="state_groups_state", + values=[ + { + "state_group": state_group, + "room_id": room_id, + "type": key[0], + "state_key": key[1], + "event_id": state_id, + } + for key, state_id in iteritems(delta_state) + ], + ) + + progress = { + "last_state_group": state_group, + "rows_inserted": rows_inserted + batch_size, + "max_group": max_group, + } + + self.db.updates._background_update_progress_txn( + txn, self.STATE_GROUP_DEDUPLICATION_UPDATE_NAME, progress + ) + + return False, batch_size + + finished, result = yield self.db.runInteraction( + self.STATE_GROUP_DEDUPLICATION_UPDATE_NAME, reindex_txn + ) + + if finished: + yield self.db.updates._end_background_update( + self.STATE_GROUP_DEDUPLICATION_UPDATE_NAME + ) + + return result * BATCH_SIZE_SCALE_FACTOR + + @defer.inlineCallbacks + def _background_index_state(self, progress, batch_size): + def reindex_txn(conn): + conn.rollback() + if isinstance(self.database_engine, PostgresEngine): + # postgres insists on autocommit for the index + conn.set_session(autocommit=True) + try: + txn = conn.cursor() + txn.execute( + "CREATE INDEX CONCURRENTLY state_groups_state_type_idx" + " ON state_groups_state(state_group, type, state_key)" + ) + txn.execute("DROP INDEX IF EXISTS state_groups_state_id") + finally: + conn.set_session(autocommit=False) + else: + txn = conn.cursor() + txn.execute( + "CREATE INDEX state_groups_state_type_idx" + " ON state_groups_state(state_group, type, state_key)" + ) + txn.execute("DROP INDEX IF EXISTS state_groups_state_id") + + yield self.db.runWithConnection(reindex_txn) + + yield self.db.updates._end_background_update(self.STATE_GROUP_INDEX_UPDATE_NAME) + + return 1 diff --git a/synapse/storage/data_stores/main/schema/delta/23/drop_state_index.sql b/synapse/storage/data_stores/state/schema/delta/23/drop_state_index.sql index ae09fa0065..ae09fa0065 100644 --- a/synapse/storage/data_stores/main/schema/delta/23/drop_state_index.sql +++ b/synapse/storage/data_stores/state/schema/delta/23/drop_state_index.sql diff --git a/synapse/storage/data_stores/main/schema/delta/30/state_stream.sql b/synapse/storage/data_stores/state/schema/delta/30/state_stream.sql index e85699e82e..e85699e82e 100644 --- a/synapse/storage/data_stores/main/schema/delta/30/state_stream.sql +++ b/synapse/storage/data_stores/state/schema/delta/30/state_stream.sql diff --git a/synapse/storage/data_stores/state/schema/delta/32/remove_state_indices.sql b/synapse/storage/data_stores/state/schema/delta/32/remove_state_indices.sql new file mode 100644 index 0000000000..1450313bfa --- /dev/null +++ b/synapse/storage/data_stores/state/schema/delta/32/remove_state_indices.sql @@ -0,0 +1,19 @@ +/* Copyright 2016 OpenMarket Ltd + * + * 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. + */ + + +-- The following indices are redundant, other indices are equivalent or +-- supersets +DROP INDEX IF EXISTS state_groups_id; -- Duplicate of PRIMARY KEY diff --git a/synapse/storage/data_stores/main/schema/delta/35/add_state_index.sql b/synapse/storage/data_stores/state/schema/delta/35/add_state_index.sql index 33980d02f0..33980d02f0 100644 --- a/synapse/storage/data_stores/main/schema/delta/35/add_state_index.sql +++ b/synapse/storage/data_stores/state/schema/delta/35/add_state_index.sql diff --git a/synapse/storage/data_stores/main/schema/delta/35/state.sql b/synapse/storage/data_stores/state/schema/delta/35/state.sql index 0f1fa68a89..0f1fa68a89 100644 --- a/synapse/storage/data_stores/main/schema/delta/35/state.sql +++ b/synapse/storage/data_stores/state/schema/delta/35/state.sql diff --git a/synapse/storage/data_stores/main/schema/delta/35/state_dedupe.sql b/synapse/storage/data_stores/state/schema/delta/35/state_dedupe.sql index 97e5067ef4..97e5067ef4 100644 --- a/synapse/storage/data_stores/main/schema/delta/35/state_dedupe.sql +++ b/synapse/storage/data_stores/state/schema/delta/35/state_dedupe.sql diff --git a/synapse/storage/data_stores/main/schema/delta/47/state_group_seq.py b/synapse/storage/data_stores/state/schema/delta/47/state_group_seq.py index 9fd1ccf6f7..9fd1ccf6f7 100644 --- a/synapse/storage/data_stores/main/schema/delta/47/state_group_seq.py +++ b/synapse/storage/data_stores/state/schema/delta/47/state_group_seq.py diff --git a/synapse/storage/data_stores/state/schema/delta/56/state_group_room_idx.sql b/synapse/storage/data_stores/state/schema/delta/56/state_group_room_idx.sql new file mode 100644 index 0000000000..7916ef18b2 --- /dev/null +++ b/synapse/storage/data_stores/state/schema/delta/56/state_group_room_idx.sql @@ -0,0 +1,17 @@ +/* Copyright 2019 The Matrix.org Foundation C.I.C. + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +INSERT INTO background_updates (update_name, progress_json) VALUES + ('state_groups_room_id_idx', '{}'); diff --git a/synapse/storage/data_stores/state/schema/full_schemas/54/full.sql b/synapse/storage/data_stores/state/schema/full_schemas/54/full.sql new file mode 100644 index 0000000000..35f97d6b3d --- /dev/null +++ b/synapse/storage/data_stores/state/schema/full_schemas/54/full.sql @@ -0,0 +1,37 @@ +/* Copyright 2019 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +CREATE TABLE state_groups ( + id BIGINT PRIMARY KEY, + room_id TEXT NOT NULL, + event_id TEXT NOT NULL +); + +CREATE TABLE state_groups_state ( + state_group BIGINT NOT NULL, + room_id TEXT NOT NULL, + type TEXT NOT NULL, + state_key TEXT NOT NULL, + event_id TEXT NOT NULL +); + +CREATE TABLE state_group_edges ( + state_group BIGINT NOT NULL, + prev_state_group BIGINT NOT NULL +); + +CREATE INDEX state_group_edges_idx ON state_group_edges (state_group); +CREATE INDEX state_group_edges_prev_idx ON state_group_edges (prev_state_group); +CREATE INDEX state_groups_state_type_idx ON state_groups_state (state_group, type, state_key); diff --git a/synapse/storage/data_stores/state/schema/full_schemas/54/sequence.sql.postgres b/synapse/storage/data_stores/state/schema/full_schemas/54/sequence.sql.postgres new file mode 100644 index 0000000000..fcd926c9fb --- /dev/null +++ b/synapse/storage/data_stores/state/schema/full_schemas/54/sequence.sql.postgres @@ -0,0 +1,21 @@ +/* Copyright 2019 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +CREATE SEQUENCE state_group_id_seq + START WITH 1 + INCREMENT BY 1 + NO MINVALUE + NO MAXVALUE + CACHE 1; diff --git a/synapse/storage/data_stores/state/store.py b/synapse/storage/data_stores/state/store.py new file mode 100644 index 0000000000..c4ee9b7ccb --- /dev/null +++ b/synapse/storage/data_stores/state/store.py @@ -0,0 +1,644 @@ +# -*- coding: utf-8 -*- +# Copyright 2014-2016 OpenMarket Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +from collections import namedtuple +from typing import Dict, Iterable, List, Set, Tuple + +from six import iteritems +from six.moves import range + +from twisted.internet import defer + +from synapse.api.constants import EventTypes +from synapse.storage._base import SQLBaseStore +from synapse.storage.data_stores.state.bg_updates import StateBackgroundUpdateStore +from synapse.storage.database import Database +from synapse.storage.state import StateFilter +from synapse.types import StateMap +from synapse.util.caches import get_cache_factor_for +from synapse.util.caches.descriptors import cached +from synapse.util.caches.dictionary_cache import DictionaryCache + +logger = logging.getLogger(__name__) + + +MAX_STATE_DELTA_HOPS = 100 + + +class _GetStateGroupDelta( + namedtuple("_GetStateGroupDelta", ("prev_group", "delta_ids")) +): + """Return type of get_state_group_delta that implements __len__, which lets + us use the itrable flag when caching + """ + + __slots__ = [] + + def __len__(self): + return len(self.delta_ids) if self.delta_ids else 0 + + +class StateGroupDataStore(StateBackgroundUpdateStore, SQLBaseStore): + """A data store for fetching/storing state groups. + """ + + def __init__(self, database: Database, db_conn, hs): + super(StateGroupDataStore, self).__init__(database, db_conn, hs) + + # Originally the state store used a single DictionaryCache to cache the + # event IDs for the state types in a given state group to avoid hammering + # on the state_group* tables. + # + # The point of using a DictionaryCache is that it can cache a subset + # of the state events for a given state group (i.e. a subset of the keys for a + # given dict which is an entry in the cache for a given state group ID). + # + # However, this poses problems when performing complicated queries + # on the store - for instance: "give me all the state for this group, but + # limit members to this subset of users", as DictionaryCache's API isn't + # rich enough to say "please cache any of these fields, apart from this subset". + # This is problematic when lazy loading members, which requires this behaviour, + # as without it the cache has no choice but to speculatively load all + # state events for the group, which negates the efficiency being sought. + # + # Rather than overcomplicating DictionaryCache's API, we instead split the + # state_group_cache into two halves - one for tracking non-member events, + # and the other for tracking member_events. This means that lazy loading + # queries can be made in a cache-friendly manner by querying both caches + # separately and then merging the result. So for the example above, you + # would query the members cache for a specific subset of state keys + # (which DictionaryCache will handle efficiently and fine) and the non-members + # cache for all state (which DictionaryCache will similarly handle fine) + # and then just merge the results together. + # + # We size the non-members cache to be smaller than the members cache as the + # vast majority of state in Matrix (today) is member events. + + self._state_group_cache = DictionaryCache( + "*stateGroupCache*", + # TODO: this hasn't been tuned yet + 50000 * get_cache_factor_for("stateGroupCache"), + ) + self._state_group_members_cache = DictionaryCache( + "*stateGroupMembersCache*", + 500000 * get_cache_factor_for("stateGroupMembersCache"), + ) + + @cached(max_entries=10000, iterable=True) + def get_state_group_delta(self, state_group): + """Given a state group try to return a previous group and a delta between + the old and the new. + + Returns: + (prev_group, delta_ids), where both may be None. + """ + + def _get_state_group_delta_txn(txn): + prev_group = self.db.simple_select_one_onecol_txn( + txn, + table="state_group_edges", + keyvalues={"state_group": state_group}, + retcol="prev_state_group", + allow_none=True, + ) + + if not prev_group: + return _GetStateGroupDelta(None, None) + + delta_ids = self.db.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}, + ) + + return self.db.runInteraction( + "get_state_group_delta", _get_state_group_delta_txn + ) + + @defer.inlineCallbacks + def _get_state_groups_from_groups( + self, groups: List[int], state_filter: StateFilter + ): + """Returns the state groups for a given set of groups from the + database, filtering on types of state events. + + Args: + groups: list of state group IDs to query + state_filter: The state filter used to fetch state + from the database. + Returns: + Deferred[Dict[int, StateMap[str]]]: Dict of state group to state map. + """ + results = {} + + chunks = [groups[i : i + 100] for i in range(0, len(groups), 100)] + for chunk in chunks: + res = yield self.db.runInteraction( + "_get_state_groups_from_groups", + self._get_state_groups_from_groups_txn, + chunk, + state_filter, + ) + results.update(res) + + return results + + def _get_state_for_group_using_cache(self, cache, group, state_filter): + """Checks if group is in cache. See `_get_state_for_groups` + + Args: + cache(DictionaryCache): the state group cache to use + group(int): The state group to lookup + state_filter (StateFilter): The state filter used to fetch state + from the database. + + Returns 2-tuple (`state_dict`, `got_all`). + `got_all` is a bool indicating if we successfully retrieved all + requests state from the cache, if False we need to query the DB for the + missing state. + """ + is_all, known_absent, state_dict_ids = cache.get(group) + + if is_all or state_filter.is_full(): + # Either we have everything or want everything, either way + # `is_all` tells us whether we've gotten everything. + return state_filter.filter_state(state_dict_ids), is_all + + # tracks whether any of our requested types are missing from the cache + missing_types = False + + if state_filter.has_wildcards(): + # We don't know if we fetched all the state keys for the types in + # the filter that are wildcards, so we have to assume that we may + # have missed some. + missing_types = True + else: + # There aren't any wild cards, so `concrete_types()` returns the + # complete list of event types we're wanting. + for key in state_filter.concrete_types(): + if key not in state_dict_ids and key not in known_absent: + missing_types = True + break + + return state_filter.filter_state(state_dict_ids), not missing_types + + @defer.inlineCallbacks + def _get_state_for_groups( + self, groups: Iterable[int], state_filter: StateFilter = StateFilter.all() + ): + """Gets the state at each of a list of state groups, optionally + filtering by type/state_key + + Args: + groups: list of state groups for which we want + to get the state. + state_filter: The state filter used to fetch state + from the database. + Returns: + Deferred[Dict[int, StateMap[str]]]: Dict of state group to state map. + """ + + member_filter, non_member_filter = state_filter.get_member_split() + + # Now we look them up in the member and non-member caches + ( + non_member_state, + incomplete_groups_nm, + ) = yield self._get_state_for_groups_using_cache( + groups, self._state_group_cache, state_filter=non_member_filter + ) + + ( + member_state, + incomplete_groups_m, + ) = yield self._get_state_for_groups_using_cache( + groups, self._state_group_members_cache, state_filter=member_filter + ) + + state = dict(non_member_state) + for group in groups: + state[group].update(member_state[group]) + + # Now fetch any missing groups from the database + + incomplete_groups = incomplete_groups_m | incomplete_groups_nm + + if not incomplete_groups: + return state + + cache_sequence_nm = self._state_group_cache.sequence + cache_sequence_m = self._state_group_members_cache.sequence + + # Help the cache hit ratio by expanding the filter a bit + db_state_filter = state_filter.return_expanded() + + group_to_state_dict = yield self._get_state_groups_from_groups( + list(incomplete_groups), state_filter=db_state_filter + ) + + # Now lets update the caches + self._insert_into_cache( + group_to_state_dict, + db_state_filter, + cache_seq_num_members=cache_sequence_m, + cache_seq_num_non_members=cache_sequence_nm, + ) + + # And finally update the result dict, by filtering out any extra + # stuff we pulled out of the database. + for group, group_state_dict in iteritems(group_to_state_dict): + # We just replace any existing entries, as we will have loaded + # everything we need from the database anyway. + state[group] = state_filter.filter_state(group_state_dict) + + return state + + def _get_state_for_groups_using_cache( + self, groups: Iterable[int], cache: DictionaryCache, state_filter: StateFilter + ) -> Tuple[Dict[int, StateMap[str]], Set[int]]: + """Gets the state at each of a list of state groups, optionally + filtering by type/state_key, querying from a specific cache. + + Args: + groups: list of state groups for which we want to get the state. + cache: the cache of group ids to state dicts which + we will pass through - either the normal state cache or the + specific members state cache. + state_filter: The state filter used to fetch state from the + database. + + Returns: + Tuple of dict of state_group_id to state map of entries in the + cache, and the state group ids either missing from the cache or + incomplete. + """ + results = {} + incomplete_groups = set() + for group in set(groups): + state_dict_ids, got_all = self._get_state_for_group_using_cache( + cache, group, state_filter + ) + results[group] = state_dict_ids + + if not got_all: + incomplete_groups.add(group) + + return results, incomplete_groups + + def _insert_into_cache( + self, + group_to_state_dict, + state_filter, + cache_seq_num_members, + cache_seq_num_non_members, + ): + """Inserts results from querying the database into the relevant cache. + + Args: + group_to_state_dict (dict): The new entries pulled from database. + Map from state group to state dict + state_filter (StateFilter): The state filter used to fetch state + from the database. + cache_seq_num_members (int): Sequence number of member cache since + last lookup in cache + cache_seq_num_non_members (int): Sequence number of member cache since + last lookup in cache + """ + + # We need to work out which types we've fetched from the DB for the + # member vs non-member caches. This should be as accurate as possible, + # but can be an underestimate (e.g. when we have wild cards) + + member_filter, non_member_filter = state_filter.get_member_split() + if member_filter.is_full(): + # We fetched all member events + member_types = None + else: + # `concrete_types()` will only return a subset when there are wild + # cards in the filter, but that's fine. + member_types = member_filter.concrete_types() + + if non_member_filter.is_full(): + # We fetched all non member events + non_member_types = None + else: + non_member_types = non_member_filter.concrete_types() + + for group, group_state_dict in iteritems(group_to_state_dict): + state_dict_members = {} + state_dict_non_members = {} + + for k, v in iteritems(group_state_dict): + if k[0] == EventTypes.Member: + state_dict_members[k] = v + else: + state_dict_non_members[k] = v + + self._state_group_members_cache.update( + cache_seq_num_members, + key=group, + value=state_dict_members, + fetched_keys=member_types, + ) + + self._state_group_cache.update( + cache_seq_num_non_members, + key=group, + value=state_dict_non_members, + fetched_keys=non_member_types, + ) + + def store_state_group( + self, event_id, room_id, prev_group, delta_ids, current_state_ids + ): + """Store a new set of state, returning a newly assigned state group. + + Args: + event_id (str): The event ID for which the state was calculated + room_id (str) + prev_group (int|None): A previous state group for the room, optional. + delta_ids (dict|None): The delta between state at `prev_group` and + `current_state_ids`, if `prev_group` was given. Same format as + `current_state_ids`. + current_state_ids (dict): The state to store. Map of (type, state_key) + to event_id. + + Returns: + Deferred[int]: The state group ID + """ + + def _store_state_group_txn(txn): + if current_state_ids is None: + # AFAIK, this can never happen + raise Exception("current_state_ids cannot be None") + + state_group = self.database_engine.get_next_state_group_id(txn) + + self.db.simple_insert_txn( + txn, + table="state_groups", + values={"id": state_group, "room_id": room_id, "event_id": event_id}, + ) + + # We persist as a delta if we can, while also ensuring the chain + # of deltas isn't tooo long, as otherwise read performance degrades. + if prev_group: + is_in_db = self.db.simple_select_one_onecol_txn( + txn, + table="state_groups", + keyvalues={"id": prev_group}, + retcol="id", + allow_none=True, + ) + if not is_in_db: + raise Exception( + "Trying to persist state with unpersisted prev_group: %r" + % (prev_group,) + ) + + potential_hops = self._count_state_group_hops_txn(txn, prev_group) + if prev_group and potential_hops < MAX_STATE_DELTA_HOPS: + self.db.simple_insert_txn( + txn, + table="state_group_edges", + values={"state_group": state_group, "prev_state_group": prev_group}, + ) + + self.db.simple_insert_many_txn( + txn, + table="state_groups_state", + values=[ + { + "state_group": state_group, + "room_id": room_id, + "type": key[0], + "state_key": key[1], + "event_id": state_id, + } + for key, state_id in iteritems(delta_ids) + ], + ) + else: + self.db.simple_insert_many_txn( + txn, + table="state_groups_state", + values=[ + { + "state_group": state_group, + "room_id": room_id, + "type": key[0], + "state_key": key[1], + "event_id": state_id, + } + for key, state_id in iteritems(current_state_ids) + ], + ) + + # Prefill the state group caches with this group. + # It's fine to use the sequence like this as the state group map + # is immutable. (If the map wasn't immutable then this prefill could + # race with another update) + + current_member_state_ids = { + s: ev + for (s, ev) in iteritems(current_state_ids) + if s[0] == EventTypes.Member + } + txn.call_after( + self._state_group_members_cache.update, + self._state_group_members_cache.sequence, + key=state_group, + value=dict(current_member_state_ids), + ) + + current_non_member_state_ids = { + s: ev + for (s, ev) in iteritems(current_state_ids) + if s[0] != EventTypes.Member + } + txn.call_after( + self._state_group_cache.update, + self._state_group_cache.sequence, + key=state_group, + value=dict(current_non_member_state_ids), + ) + + return state_group + + return self.db.runInteraction("store_state_group", _store_state_group_txn) + + def purge_unreferenced_state_groups( + self, room_id: str, state_groups_to_delete + ) -> defer.Deferred: + """Deletes no longer referenced state groups and de-deltas any state + groups that reference them. + + Args: + room_id: The room the state groups belong to (must all be in the + same room). + state_groups_to_delete (Collection[int]): Set of all state groups + to delete. + """ + + return self.db.runInteraction( + "purge_unreferenced_state_groups", + self._purge_unreferenced_state_groups, + room_id, + state_groups_to_delete, + ) + + def _purge_unreferenced_state_groups(self, txn, room_id, state_groups_to_delete): + logger.info( + "[purge] found %i state groups to delete", len(state_groups_to_delete) + ) + + rows = self.db.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 = set( + row["state_group"] + for row in rows + if row["state_group"] not in state_groups_to_delete + ) + + logger.info( + "[purge] de-delta-ing %i remaining state groups", + len(remaining_state_groups), + ) + + # Now we turn the state groups that reference to-be-deleted state + # groups to non delta versions. + for sg in remaining_state_groups: + logger.info("[purge] de-delta-ing remaining state group %s", sg) + curr_state = self._get_state_groups_from_groups_txn(txn, [sg]) + curr_state = curr_state[sg] + + self.db.simple_delete_txn( + txn, table="state_groups_state", keyvalues={"state_group": sg} + ) + + self.db.simple_delete_txn( + txn, table="state_group_edges", keyvalues={"state_group": sg} + ) + + self.db.simple_insert_many_txn( + txn, + table="state_groups_state", + values=[ + { + "state_group": sg, + "room_id": room_id, + "type": key[0], + "state_key": key[1], + "event_id": state_id, + } + for key, state_id in iteritems(curr_state) + ], + ) + + logger.info("[purge] removing redundant state groups") + txn.executemany( + "DELETE FROM state_groups_state WHERE state_group = ?", + ((sg,) for sg in state_groups_to_delete), + ) + txn.executemany( + "DELETE FROM state_groups WHERE id = ?", + ((sg,) for sg in state_groups_to_delete), + ) + + @defer.inlineCallbacks + def get_previous_state_groups(self, state_groups): + """Fetch the previous groups of the given state groups. + + Args: + state_groups (Iterable[int]) + + Returns: + Deferred[dict[int, int]]: mapping from state group to previous + state group. + """ + + rows = yield self.db.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", + ) + + return {row["state_group"]: row["prev_state_group"] for row in rows} + + def purge_room_state(self, room_id, state_groups_to_delete): + """Deletes all record of a room from state tables + + Args: + room_id (str): + state_groups_to_delete (list[int]): State groups to delete + """ + + return self.db.runInteraction( + "purge_room_state", + self._purge_room_state_txn, + room_id, + state_groups_to_delete, + ) + + def _purge_room_state_txn(self, txn, room_id, state_groups_to_delete): + # first we have to delete the state groups states + logger.info("[purge] removing %s from state_groups_state", room_id) + + self.db.simple_delete_many_txn( + txn, + table="state_groups_state", + column="state_group", + iterable=state_groups_to_delete, + keyvalues={}, + ) + + # ... and the state group edges + logger.info("[purge] removing %s from state_group_edges", room_id) + + self.db.simple_delete_many_txn( + txn, + table="state_group_edges", + column="state_group", + iterable=state_groups_to_delete, + keyvalues={}, + ) + + # ... and the state groups + logger.info("[purge] removing %s from state_groups", room_id) + + self.db.simple_delete_many_txn( + txn, + table="state_groups", + column="id", + iterable=state_groups_to_delete, + keyvalues={}, + ) diff --git a/synapse/storage/database.py b/synapse/storage/database.py new file mode 100644 index 0000000000..1003dd84a5 --- /dev/null +++ b/synapse/storage/database.py @@ -0,0 +1,1529 @@ +# -*- coding: utf-8 -*- +# Copyright 2014-2016 OpenMarket Ltd +# Copyright 2017-2018 New Vector Ltd +# Copyright 2019 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import logging +import sys +import time +from typing import Iterable, Tuple + +from six import iteritems, iterkeys, itervalues +from six.moves import intern, range + +from prometheus_client import Histogram + +from twisted.enterprise import adbapi +from twisted.internet import defer + +from synapse.api.errors import StoreError +from synapse.config.database import DatabaseConnectionConfig +from synapse.logging.context import LoggingContext, make_deferred_yieldable +from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.storage.background_updates import BackgroundUpdater +from synapse.storage.engines import PostgresEngine, Sqlite3Engine +from synapse.util.stringutils import exception_to_unicode + +# import a function which will return a monotonic time, in seconds +try: + # on python 3, use time.monotonic, since time.clock can go backwards + from time import monotonic as monotonic_time +except ImportError: + # ... but python 2 doesn't have it + from time import clock as monotonic_time + +logger = logging.getLogger(__name__) + +try: + MAX_TXN_ID = sys.maxint - 1 +except AttributeError: + # python 3 does not have a maximum int value + MAX_TXN_ID = 2 ** 63 - 1 + +sql_logger = logging.getLogger("synapse.storage.SQL") +transaction_logger = logging.getLogger("synapse.storage.txn") +perf_logger = logging.getLogger("synapse.storage.TIME") + +sql_scheduling_timer = Histogram("synapse_storage_schedule_time", "sec") + +sql_query_timer = Histogram("synapse_storage_query_time", "sec", ["verb"]) +sql_txn_timer = Histogram("synapse_storage_transaction_time", "sec", ["desc"]) + + +# Unique indexes which have been added in background updates. Maps from table name +# to the name of the background update which added the unique index to that table. +# +# This is used by the upsert logic to figure out which tables are safe to do a proper +# UPSERT on: until the relevant background update has completed, we +# have to emulate an upsert by locking the table. +# +UNIQUE_INDEX_BACKGROUND_UPDATES = { + "user_ips": "user_ips_device_unique_index", + "device_lists_remote_extremeties": "device_lists_remote_extremeties_unique_idx", + "device_lists_remote_cache": "device_lists_remote_cache_unique_idx", + "event_search": "event_search_event_id_idx", +} + + +def make_pool( + reactor, db_config: DatabaseConnectionConfig, engine +) -> adbapi.ConnectionPool: + """Get the connection pool for the database. + """ + + return adbapi.ConnectionPool( + db_config.config["name"], + cp_reactor=reactor, + cp_openfun=engine.on_new_connection, + **db_config.config.get("args", {}) + ) + + +def make_conn(db_config: DatabaseConnectionConfig, engine): + """Make a new connection to the database and return it. + + Returns: + Connection + """ + + db_params = { + k: v + for k, v in db_config.config.get("args", {}).items() + if not k.startswith("cp_") + } + db_conn = engine.module.connect(**db_params) + engine.on_new_connection(db_conn) + return db_conn + + +class LoggingTransaction(object): + """An object that almost-transparently proxies for the 'txn' object + passed to the constructor. Adds logging and metrics to the .execute() + method. + + Args: + txn: The database transcation object to wrap. + name (str): The name of this transactions for logging. + database_engine (Sqlite3Engine|PostgresEngine) + after_callbacks(list|None): A list that callbacks will be appended to + that have been added by `call_after` which should be run on + successful completion of the transaction. None indicates that no + callbacks should be allowed to be scheduled to run. + exception_callbacks(list|None): A list that callbacks will be appended + to that have been added by `call_on_exception` which should be run + if transaction ends with an error. None indicates that no callbacks + should be allowed to be scheduled to run. + """ + + __slots__ = [ + "txn", + "name", + "database_engine", + "after_callbacks", + "exception_callbacks", + ] + + def __init__( + self, txn, name, database_engine, after_callbacks=None, exception_callbacks=None + ): + object.__setattr__(self, "txn", txn) + object.__setattr__(self, "name", name) + object.__setattr__(self, "database_engine", database_engine) + object.__setattr__(self, "after_callbacks", after_callbacks) + object.__setattr__(self, "exception_callbacks", exception_callbacks) + + def call_after(self, callback, *args, **kwargs): + """Call the given callback on the main twisted thread after the + transaction has finished. Used to invalidate the caches on the + correct thread. + """ + self.after_callbacks.append((callback, args, kwargs)) + + def call_on_exception(self, callback, *args, **kwargs): + self.exception_callbacks.append((callback, args, kwargs)) + + def __getattr__(self, name): + return getattr(self.txn, name) + + def __setattr__(self, name, value): + setattr(self.txn, name, value) + + def __iter__(self): + return self.txn.__iter__() + + def execute_batch(self, sql, args): + if isinstance(self.database_engine, PostgresEngine): + from psycopg2.extras import execute_batch + + self._do_execute(lambda *x: execute_batch(self.txn, *x), sql, args) + else: + for val in args: + self.execute(sql, val) + + def execute(self, sql, *args): + self._do_execute(self.txn.execute, sql, *args) + + def executemany(self, sql, *args): + self._do_execute(self.txn.executemany, sql, *args) + + def _make_sql_one_line(self, sql): + "Strip newlines out of SQL so that the loggers in the DB are on one line" + return " ".join(l.strip() for l in sql.splitlines() if l.strip()) + + def _do_execute(self, func, sql, *args): + sql = self._make_sql_one_line(sql) + + # TODO(paul): Maybe use 'info' and 'debug' for values? + sql_logger.debug("[SQL] {%s} %s", self.name, sql) + + sql = self.database_engine.convert_param_style(sql) + if args: + try: + sql_logger.debug("[SQL values] {%s} %r", self.name, args[0]) + except Exception: + # Don't let logging failures stop SQL from working + pass + + start = time.time() + + try: + return func(sql, *args) + except Exception as e: + logger.debug("[SQL FAIL] {%s} %s", self.name, e) + raise + finally: + secs = time.time() - start + sql_logger.debug("[SQL time] {%s} %f sec", self.name, secs) + sql_query_timer.labels(sql.split()[0]).observe(secs) + + +class PerformanceCounters(object): + def __init__(self): + self.current_counters = {} + self.previous_counters = {} + + def update(self, key, duration_secs): + count, cum_time = self.current_counters.get(key, (0, 0)) + count += 1 + cum_time += duration_secs + self.current_counters[key] = (count, cum_time) + + def interval(self, interval_duration_secs, limit=3): + counters = [] + for name, (count, cum_time) in iteritems(self.current_counters): + prev_count, prev_time = self.previous_counters.get(name, (0, 0)) + counters.append( + ( + (cum_time - prev_time) / interval_duration_secs, + count - prev_count, + name, + ) + ) + + self.previous_counters = dict(self.current_counters) + + counters.sort(reverse=True) + + top_n_counters = ", ".join( + "%s(%d): %.3f%%" % (name, count, 100 * ratio) + for ratio, count, name in counters[:limit] + ) + + return top_n_counters + + +class Database(object): + """Wraps a single physical database and connection pool. + + A single database may be used by multiple data stores. + """ + + _TXN_ID = 0 + + def __init__(self, hs, database_config: DatabaseConnectionConfig, engine): + self.hs = hs + self._clock = hs.get_clock() + self._database_config = database_config + self._db_pool = make_pool(hs.get_reactor(), database_config, engine) + + self.updates = BackgroundUpdater(hs, self) + + self._previous_txn_total_time = 0 + self._current_txn_total_time = 0 + self._previous_loop_ts = 0 + + # TODO(paul): These can eventually be removed once the metrics code + # is running in mainline, and we have some nice monitoring frontends + # to watch it + self._txn_perf_counters = PerformanceCounters() + + self.engine = engine + + # A set of tables that are not safe to use native upserts in. + self._unsafe_to_upsert_tables = set(UNIQUE_INDEX_BACKGROUND_UPDATES.keys()) + + # We add the user_directory_search table to the blacklist on SQLite + # because the existing search table does not have an index, making it + # unsafe to use native upserts. + if isinstance(self.engine, Sqlite3Engine): + self._unsafe_to_upsert_tables.add("user_directory_search") + + if self.engine.can_native_upsert: + # Check ASAP (and then later, every 1s) to see if we have finished + # background updates of tables that aren't safe to update. + self._clock.call_later( + 0.0, + run_as_background_process, + "upsert_safety_check", + self._check_safe_to_upsert, + ) + + def is_running(self): + """Is the database pool currently running + """ + return self._db_pool.running + + @defer.inlineCallbacks + def _check_safe_to_upsert(self): + """ + Is it safe to use native UPSERT? + + If there are background updates, we will need to wait, as they may be + the addition of indexes that set the UNIQUE constraint that we require. + + If the background updates have not completed, wait 15 sec and check again. + """ + updates = yield self.simple_select_list( + "background_updates", + keyvalues=None, + retcols=["update_name"], + desc="check_background_updates", + ) + updates = [x["update_name"] for x in updates] + + for table, update_name in UNIQUE_INDEX_BACKGROUND_UPDATES.items(): + if update_name not in updates: + logger.debug("Now safe to upsert in %s", table) + self._unsafe_to_upsert_tables.discard(table) + + # If there's any updates still running, reschedule to run. + if updates: + self._clock.call_later( + 15.0, + run_as_background_process, + "upsert_safety_check", + self._check_safe_to_upsert, + ) + + def start_profiling(self): + self._previous_loop_ts = monotonic_time() + + def loop(): + curr = self._current_txn_total_time + prev = self._previous_txn_total_time + self._previous_txn_total_time = curr + + time_now = monotonic_time() + time_then = self._previous_loop_ts + self._previous_loop_ts = time_now + + duration = time_now - time_then + ratio = (curr - prev) / duration + + top_three_counters = self._txn_perf_counters.interval(duration, limit=3) + + perf_logger.info( + "Total database time: %.3f%% {%s}", ratio * 100, top_three_counters + ) + + self._clock.looping_call(loop, 10000) + + def new_transaction( + self, conn, desc, after_callbacks, exception_callbacks, func, *args, **kwargs + ): + start = monotonic_time() + txn_id = self._TXN_ID + + # We don't really need these to be unique, so lets stop it from + # growing really large. + self._TXN_ID = (self._TXN_ID + 1) % (MAX_TXN_ID) + + name = "%s-%x" % (desc, txn_id) + + transaction_logger.debug("[TXN START] {%s}", name) + + try: + i = 0 + N = 5 + while True: + cursor = LoggingTransaction( + conn.cursor(), + name, + self.engine, + after_callbacks, + exception_callbacks, + ) + try: + r = func(cursor, *args, **kwargs) + conn.commit() + return r + except self.engine.module.OperationalError as e: + # This can happen if the database disappears mid + # transaction. + logger.warning( + "[TXN OPERROR] {%s} %s %d/%d", + name, + exception_to_unicode(e), + i, + N, + ) + if i < N: + i += 1 + try: + conn.rollback() + except self.engine.module.Error as e1: + logger.warning( + "[TXN EROLL] {%s} %s", name, exception_to_unicode(e1) + ) + continue + raise + except self.engine.module.DatabaseError as e: + if self.engine.is_deadlock(e): + logger.warning("[TXN DEADLOCK] {%s} %d/%d", name, i, N) + if i < N: + i += 1 + try: + conn.rollback() + except self.engine.module.Error as e1: + logger.warning( + "[TXN EROLL] {%s} %s", + name, + exception_to_unicode(e1), + ) + continue + raise + finally: + # we're either about to retry with a new cursor, or we're about to + # release the connection. Once we release the connection, it could + # get used for another query, which might do a conn.rollback(). + # + # In the latter case, even though that probably wouldn't affect the + # results of this transaction, python's sqlite will reset all + # statements on the connection [1], which will make our cursor + # invalid [2]. + # + # In any case, continuing to read rows after commit()ing seems + # dubious from the PoV of ACID transactional semantics + # (sqlite explicitly says that once you commit, you may see rows + # from subsequent updates.) + # + # In psycopg2, cursors are essentially a client-side fabrication - + # all the data is transferred to the client side when the statement + # finishes executing - so in theory we could go on streaming results + # from the cursor, but attempting to do so would make us + # incompatible with sqlite, so let's make sure we're not doing that + # by closing the cursor. + # + # (*named* cursors in psycopg2 are different and are proper server- + # side things, but (a) we don't use them and (b) they are implicitly + # closed by ending the transaction anyway.) + # + # In short, if we haven't finished with the cursor yet, that's a + # problem waiting to bite us. + # + # TL;DR: we're done with the cursor, so we can close it. + # + # [1]: https://github.com/python/cpython/blob/v3.8.0/Modules/_sqlite/connection.c#L465 + # [2]: https://github.com/python/cpython/blob/v3.8.0/Modules/_sqlite/cursor.c#L236 + cursor.close() + except Exception as e: + logger.debug("[TXN FAIL] {%s} %s", name, e) + raise + finally: + end = monotonic_time() + duration = end - start + + LoggingContext.current_context().add_database_transaction(duration) + + transaction_logger.debug("[TXN END] {%s} %f sec", name, duration) + + self._current_txn_total_time += duration + self._txn_perf_counters.update(desc, duration) + sql_txn_timer.labels(desc).observe(duration) + + @defer.inlineCallbacks + def runInteraction(self, desc, func, *args, **kwargs): + """Starts a transaction on the database and runs a given function + + Arguments: + desc (str): description of the transaction, for logging and metrics + func (func): callback function, which will be called with a + database transaction (twisted.enterprise.adbapi.Transaction) as + its first argument, followed by `args` and `kwargs`. + + args (list): positional args to pass to `func` + kwargs (dict): named args to pass to `func` + + Returns: + Deferred: The result of func + """ + after_callbacks = [] + exception_callbacks = [] + + if LoggingContext.current_context() == LoggingContext.sentinel: + logger.warning("Starting db txn '%s' from sentinel context", desc) + + try: + result = yield self.runWithConnection( + self.new_transaction, + desc, + after_callbacks, + exception_callbacks, + func, + *args, + **kwargs + ) + + for after_callback, after_args, after_kwargs in after_callbacks: + after_callback(*after_args, **after_kwargs) + except: # noqa: E722, as we reraise the exception this is fine. + for after_callback, after_args, after_kwargs in exception_callbacks: + after_callback(*after_args, **after_kwargs) + raise + + return result + + @defer.inlineCallbacks + def runWithConnection(self, func, *args, **kwargs): + """Wraps the .runWithConnection() method on the underlying db_pool. + + Arguments: + func (func): callback function, which will be called with a + database connection (twisted.enterprise.adbapi.Connection) as + its first argument, followed by `args` and `kwargs`. + args (list): positional args to pass to `func` + kwargs (dict): named args to pass to `func` + + Returns: + Deferred: The result of func + """ + parent_context = LoggingContext.current_context() + if parent_context == LoggingContext.sentinel: + logger.warning( + "Starting db connection from sentinel context: metrics will be lost" + ) + parent_context = None + + start_time = monotonic_time() + + def inner_func(conn, *args, **kwargs): + with LoggingContext("runWithConnection", parent_context) as context: + sched_duration_sec = monotonic_time() - start_time + sql_scheduling_timer.observe(sched_duration_sec) + context.add_database_scheduled(sched_duration_sec) + + if self.engine.is_connection_closed(conn): + logger.debug("Reconnecting closed database connection") + conn.reconnect() + + return func(conn, *args, **kwargs) + + result = yield make_deferred_yieldable( + self._db_pool.runWithConnection(inner_func, *args, **kwargs) + ) + + return result + + @staticmethod + def cursor_to_dict(cursor): + """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. + """ + col_headers = list(intern(str(column[0])) for column in cursor.description) + results = list(dict(zip(col_headers, row)) for row in cursor) + return results + + def execute(self, desc, decoder, query, *args): + """Runs a single query for a result set. + + Args: + 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): + txn.execute(query, args) + if decoder: + return decoder(txn) + else: + return txn.fetchall() + + return self.runInteraction(desc, interaction) + + # "Simple" SQL API methods that operate on a single table with no JOINs, + # no complex WHERE clauses, just a dict of values for columns. + + @defer.inlineCallbacks + def simple_insert(self, table, values, or_ignore=False, desc="simple_insert"): + """Executes an INSERT query on the named table. + + Args: + table : string giving the table name + values : dict of new column names and values for them + or_ignore : bool stating whether an exception should be raised + when a conflicting row already exists. If True, False will be + returned by the function instead + desc : string giving a description of the transaction + + Returns: + bool: Whether the row was inserted or not. Only useful when + `or_ignore` is True + """ + try: + yield self.runInteraction(desc, self.simple_insert_txn, table, values) + except self.engine.module.IntegrityError: + # We have to do or_ignore flag at this layer, since we can't reuse + # a cursor after we receive an error from the db. + if not or_ignore: + raise + return False + return True + + @staticmethod + def simple_insert_txn(txn, table, values): + keys, vals = zip(*values.items()) + + sql = "INSERT INTO %s (%s) VALUES(%s)" % ( + table, + ", ".join(k for k in keys), + ", ".join("?" for _ in keys), + ) + + txn.execute(sql, vals) + + def simple_insert_many(self, table, values, desc): + return self.runInteraction(desc, self.simple_insert_many_txn, table, values) + + @staticmethod + def simple_insert_many_txn(txn, table, values): + if not values: + return + + # This is a *slight* abomination to get a list of tuples of key names + # and a list of tuples of value names. + # + # i.e. [{"a": 1, "b": 2}, {"c": 3, "d": 4}] + # => [("a", "b",), ("c", "d",)] and [(1, 2,), (3, 4,)] + # + # The sort is to ensure that we don't rely on dictionary iteration + # order. + keys, vals = zip( + *[zip(*(sorted(i.items(), key=lambda kv: kv[0]))) for i in values if i] + ) + + for k in keys: + if k != keys[0]: + raise RuntimeError("All items must have the same keys") + + sql = "INSERT INTO %s (%s) VALUES(%s)" % ( + table, + ", ".join(k for k in keys[0]), + ", ".join("?" for _ in keys[0]), + ) + + txn.executemany(sql, vals) + + @defer.inlineCallbacks + def simple_upsert( + self, + table, + keyvalues, + values, + insertion_values={}, + desc="simple_upsert", + lock=True, + ): + """ + + `lock` should generally be set to True (the default), but can be set + to False if either of the following are true: + + * there is a UNIQUE INDEX on the key columns. In this case a conflict + will cause an IntegrityError in which case this function will retry + the update. + + * we somehow know that we are the only thread which will be updating + this table. + + Args: + table (str): The table to upsert into + keyvalues (dict): The unique key columns and their new values + values (dict): The nonunique columns and their new values + insertion_values (dict): additional key/values to use only when + inserting + lock (bool): True to lock the table when doing the upsert. + Returns: + Deferred(None or bool): Native upserts always return None. Emulated + upserts return True if a new entry was created, False if an existing + one was updated. + """ + attempts = 0 + while True: + try: + result = yield self.runInteraction( + desc, + self.simple_upsert_txn, + table, + keyvalues, + values, + insertion_values, + lock=lock, + ) + return result + except self.engine.module.IntegrityError as e: + attempts += 1 + if attempts >= 5: + # don't retry forever, because things other than races + # can cause IntegrityErrors + raise + + # presumably we raced with another transaction: let's retry. + logger.warning( + "IntegrityError when upserting into %s; retrying: %s", table, e + ) + + def simple_upsert_txn( + self, txn, table, keyvalues, values, insertion_values={}, lock=True + ): + """ + Pick the UPSERT method which works best on the platform. Either the + native one (Pg9.5+, recent SQLites), or fall back to an emulated method. + + Args: + txn: The transaction to use. + table (str): The table to upsert into + keyvalues (dict): The unique key tables and their new values + values (dict): The nonunique columns and their new values + insertion_values (dict): additional key/values to use only when + inserting + lock (bool): True to lock the table when doing the upsert. + Returns: + None or bool: Native upserts always return None. Emulated + upserts return True if a new entry was created, False if an existing + one was updated. + """ + if self.engine.can_native_upsert and table not in self._unsafe_to_upsert_tables: + return self.simple_upsert_txn_native_upsert( + txn, table, keyvalues, values, insertion_values=insertion_values + ) + else: + return self.simple_upsert_txn_emulated( + txn, + table, + keyvalues, + values, + insertion_values=insertion_values, + lock=lock, + ) + + def simple_upsert_txn_emulated( + self, txn, table, keyvalues, values, insertion_values={}, lock=True + ): + """ + Args: + table (str): The table to upsert into + keyvalues (dict): The unique key tables and their new values + values (dict): The nonunique columns and their new values + insertion_values (dict): additional key/values to use only when + inserting + lock (bool): True to lock the table when doing the upsert. + Returns: + bool: Return True if a new entry was created, False if an existing + one was updated. + """ + # We need to lock the table :(, unless we're *really* careful + if lock: + self.engine.lock_table(txn, table) + + def _getwhere(key): + # If the value we're passing in is None (aka NULL), we need to use + # IS, not =, as NULL = NULL equals NULL (False). + if keyvalues[key] is None: + return "%s IS ?" % (key,) + else: + return "%s = ?" % (key,) + + if not values: + # If `values` is empty, then all of the values we care about are in + # the unique key, so there is nothing to UPDATE. We can just do a + # SELECT instead to see if it exists. + sql = "SELECT 1 FROM %s WHERE %s" % ( + table, + " AND ".join(_getwhere(k) for k in keyvalues), + ) + sqlargs = list(keyvalues.values()) + txn.execute(sql, sqlargs) + if txn.fetchall(): + # We have an existing record. + return False + else: + # First try to update. + sql = "UPDATE %s SET %s WHERE %s" % ( + table, + ", ".join("%s = ?" % (k,) for k in values), + " AND ".join(_getwhere(k) for k in keyvalues), + ) + sqlargs = list(values.values()) + list(keyvalues.values()) + + txn.execute(sql, sqlargs) + if txn.rowcount > 0: + # successfully updated at least one row. + return False + + # We didn't find any existing rows, so insert a new one + allvalues = {} + allvalues.update(keyvalues) + allvalues.update(values) + allvalues.update(insertion_values) + + sql = "INSERT INTO %s (%s) VALUES (%s)" % ( + table, + ", ".join(k for k in allvalues), + ", ".join("?" for _ in allvalues), + ) + txn.execute(sql, list(allvalues.values())) + # successfully inserted + return True + + def simple_upsert_txn_native_upsert( + self, txn, table, keyvalues, values, insertion_values={} + ): + """ + Use the native UPSERT functionality in recent PostgreSQL versions. + + Args: + table (str): The table to upsert into + keyvalues (dict): The unique key tables and their new values + values (dict): The nonunique columns and their new values + insertion_values (dict): additional key/values to use only when + inserting + Returns: + None + """ + allvalues = {} + allvalues.update(keyvalues) + allvalues.update(insertion_values) + + if not values: + latter = "NOTHING" + else: + allvalues.update(values) + latter = "UPDATE SET " + ", ".join(k + "=EXCLUDED." + k for k in values) + + sql = ("INSERT INTO %s (%s) VALUES (%s) ON CONFLICT (%s) DO %s") % ( + table, + ", ".join(k for k in allvalues), + ", ".join("?" for _ in allvalues), + ", ".join(k for k in keyvalues), + latter, + ) + txn.execute(sql, list(allvalues.values())) + + def simple_upsert_many_txn( + self, txn, table, key_names, key_values, value_names, value_values + ): + """ + Upsert, many times. + + Args: + table (str): The table to upsert into + key_names (list[str]): The key column names. + key_values (list[list]): A list of each row's key column values. + value_names (list[str]): The value column names. If empty, no + values will be used, even if value_values is provided. + value_values (list[list]): A list of each row's value column values. + Returns: + None + """ + if self.engine.can_native_upsert and 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 + ) + else: + return self.simple_upsert_many_txn_emulated( + txn, table, key_names, key_values, value_names, value_values + ) + + def simple_upsert_many_txn_emulated( + self, txn, table, key_names, key_values, value_names, value_values + ): + """ + Upsert, many times, but without native UPSERT support or batching. + + Args: + table (str): The table to upsert into + key_names (list[str]): The key column names. + key_values (list[list]): A list of each row's key column values. + value_names (list[str]): The value column names. If empty, no + values will be used, even if value_values is provided. + value_values (list[list]): A list of each row's value column values. + Returns: + None + """ + # 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))] + + for keyv, valv in zip(key_values, value_values): + _keys = {x: y for x, y in zip(key_names, keyv)} + _vals = {x: y for x, y in zip(value_names, valv)} + + self.simple_upsert_txn_emulated(txn, table, _keys, _vals) + + def simple_upsert_many_txn_native_upsert( + self, txn, table, key_names, key_values, value_names, value_values + ): + """ + Upsert, many times, using batching where possible. + + Args: + table (str): The table to upsert into + key_names (list[str]): The key column names. + key_values (list[list]): A list of each row's key column values. + value_names (list[str]): The value column names. If empty, no + values will be used, even if value_values is provided. + value_values (list[list]): A list of each row's value column values. + Returns: + None + """ + allnames = [] + allnames.extend(key_names) + 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 + ) + + sql = "INSERT INTO %s (%s) VALUES (%s) ON CONFLICT (%s) DO %s" % ( + table, + ", ".join(k for k in allnames), + ", ".join("?" for _ in allnames), + ", ".join(key_names), + latter, + ) + + args = [] + + for x, y in zip(key_values, value_values): + args.append(tuple(x) + tuple(y)) + + return txn.execute_batch(sql, args) + + def simple_select_one( + self, table, keyvalues, retcols, allow_none=False, desc="simple_select_one" + ): + """Executes a SELECT query on the named table, which is expected to + return a single row, returning multiple columns from it. + + Args: + table : string giving the table name + keyvalues : dict of column names and values to select the row with + retcols : list of strings giving the names of the columns to return + + allow_none : If true, return None instead of failing if the SELECT + statement returns no rows + """ + return self.runInteraction( + desc, self.simple_select_one_txn, table, keyvalues, retcols, allow_none + ) + + def simple_select_one_onecol( + self, + table, + keyvalues, + retcol, + allow_none=False, + desc="simple_select_one_onecol", + ): + """Executes a SELECT query on the named table, which is expected to + return a single row, returning a single column from it. + + Args: + table : string giving the table name + keyvalues : dict of column names and values to select the row with + retcol : string giving the name of the column to return + """ + return self.runInteraction( + desc, + self.simple_select_one_onecol_txn, + table, + keyvalues, + retcol, + allow_none=allow_none, + ) + + @classmethod + def simple_select_one_onecol_txn( + cls, txn, table, keyvalues, retcol, allow_none=False + ): + ret = cls.simple_select_onecol_txn( + txn, table=table, keyvalues=keyvalues, retcol=retcol + ) + + if ret: + return ret[0] + else: + if allow_none: + return None + else: + raise StoreError(404, "No row found") + + @staticmethod + def simple_select_onecol_txn(txn, table, keyvalues, retcol): + sql = ("SELECT %(retcol)s FROM %(table)s") % {"retcol": retcol, "table": table} + + if keyvalues: + sql += " WHERE %s" % " AND ".join("%s = ?" % k for k in iterkeys(keyvalues)) + txn.execute(sql, list(keyvalues.values())) + else: + txn.execute(sql) + + return [r[0] for r in txn] + + def simple_select_onecol( + self, table, keyvalues, retcol, desc="simple_select_onecol" + ): + """Executes a SELECT query on the named table, which returns a list + comprising of the values of the named column from the selected rows. + + Args: + table (str): table name + keyvalues (dict|None): column names and values to select the rows with + retcol (str): column whos value we wish to retrieve. + + Returns: + Deferred: Results in a list + """ + return self.runInteraction( + desc, self.simple_select_onecol_txn, table, keyvalues, retcol + ) + + def simple_select_list(self, table, keyvalues, retcols, desc="simple_select_list"): + """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 (str): the table name + keyvalues (dict[str, Any] | None): + column names and values to select the rows with, or None to not + apply a WHERE clause. + retcols (iterable[str]): the names of the columns to return + Returns: + defer.Deferred: resolves to list[dict[str, Any]] + """ + return self.runInteraction( + desc, self.simple_select_list_txn, table, keyvalues, retcols + ) + + @classmethod + def simple_select_list_txn(cls, txn, table, keyvalues, retcols): + """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 (str): the table name + keyvalues (dict[str, T] | None): + column names and values to select the rows with, or None to not + apply a WHERE clause. + retcols (iterable[str]): the names of the columns to return + """ + if keyvalues: + sql = "SELECT %s FROM %s WHERE %s" % ( + ", ".join(retcols), + table, + " AND ".join("%s = ?" % (k,) for k in keyvalues), + ) + txn.execute(sql, list(keyvalues.values())) + else: + sql = "SELECT %s FROM %s" % (", ".join(retcols), table) + txn.execute(sql) + + return cls.cursor_to_dict(txn) + + @defer.inlineCallbacks + def simple_select_many_batch( + self, + table, + column, + iterable, + retcols, + keyvalues={}, + desc="simple_select_many_batch", + batch_size=100, + ): + """Executes a SELECT query on the named table, which may return zero or + more rows, returning the result as a list of dicts. + + Filters rows by if value of `column` is in `iterable`. + + Args: + table : string giving the table name + column : column name to test for inclusion against `iterable` + 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 + """ + results = [] + + if not iterable: + return results + + # iterables can not be sliced, so convert it to a list first + it_list = list(iterable) + + chunks = [ + it_list[i : i + batch_size] for i in range(0, len(it_list), batch_size) + ] + for chunk in chunks: + rows = yield self.runInteraction( + desc, + self.simple_select_many_txn, + table, + column, + chunk, + keyvalues, + retcols, + ) + + results.extend(rows) + + return results + + @classmethod + def simple_select_many_txn(cls, txn, table, column, iterable, keyvalues, retcols): + """Executes a SELECT query on the named table, which may return zero or + more rows, returning the result as a list of dicts. + + Filters rows by if value of `column` is in `iterable`. + + Args: + txn : Transaction object + table : string giving the table name + column : column name to test for inclusion against `iterable` + 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 + """ + if not iterable: + return [] + + clause, values = make_in_list_sql_clause(txn.database_engine, column, iterable) + clauses = [clause] + + for key, value in iteritems(keyvalues): + clauses.append("%s = ?" % (key,)) + values.append(value) + + sql = "SELECT %s FROM %s WHERE %s" % ( + ", ".join(retcols), + table, + " AND ".join(clauses), + ) + + txn.execute(sql, values) + return cls.cursor_to_dict(txn) + + def simple_update(self, table, keyvalues, updatevalues, desc): + return self.runInteraction( + desc, self.simple_update_txn, table, keyvalues, updatevalues + ) + + @staticmethod + def simple_update_txn(txn, table, keyvalues, updatevalues): + if keyvalues: + where = "WHERE %s" % " AND ".join("%s = ?" % k for k in iterkeys(keyvalues)) + else: + where = "" + + update_sql = "UPDATE %s SET %s %s" % ( + table, + ", ".join("%s = ?" % (k,) for k in updatevalues), + where, + ) + + txn.execute(update_sql, list(updatevalues.values()) + list(keyvalues.values())) + + return txn.rowcount + + def simple_update_one( + self, table, keyvalues, updatevalues, desc="simple_update_one" + ): + """Executes an UPDATE query on the named table, setting new values for + columns in a row matching the key values. + + Args: + table : string giving the table name + keyvalues : dict of column names and values to select the row with + updatevalues : dict giving column names and values to update + retcols : optional list of column names to return + + If present, retcols gives a list of column names on which to perform + a SELECT statement *before* performing the UPDATE statement. The values + of these will be returned in a dict. + + These are performed within the same transaction, allowing an atomic + get-and-set. This can be used to implement compare-and-set by putting + the update column in the 'keyvalues' dict as well. + """ + return self.runInteraction( + desc, self.simple_update_one_txn, table, keyvalues, updatevalues + ) + + @classmethod + def simple_update_one_txn(cls, txn, table, keyvalues, updatevalues): + rowcount = cls.simple_update_txn(txn, table, keyvalues, updatevalues) + + if rowcount == 0: + raise StoreError(404, "No row found (%s)" % (table,)) + if rowcount > 1: + raise StoreError(500, "More than one row matched (%s)" % (table,)) + + @staticmethod + def simple_select_one_txn(txn, table, keyvalues, retcols, allow_none=False): + select_sql = "SELECT %s FROM %s WHERE %s" % ( + ", ".join(retcols), + table, + " AND ".join("%s = ?" % (k,) for k in keyvalues), + ) + + txn.execute(select_sql, list(keyvalues.values())) + row = txn.fetchone() + + if not row: + if allow_none: + return None + raise StoreError(404, "No row found (%s)" % (table,)) + if txn.rowcount > 1: + raise StoreError(500, "More than one row matched (%s)" % (table,)) + + return dict(zip(retcols, row)) + + def simple_delete_one(self, table, keyvalues, desc="simple_delete_one"): + """Executes a DELETE query on the named table, expecting to delete a + single row. + + Args: + table : string giving the table name + keyvalues : dict of column names and values to select the row with + """ + return self.runInteraction(desc, self.simple_delete_one_txn, table, keyvalues) + + @staticmethod + def simple_delete_one_txn(txn, table, keyvalues): + """Executes a DELETE query on the named table, expecting to delete a + single row. + + Args: + table : string giving the table name + keyvalues : dict of column names and values to select the row with + """ + sql = "DELETE FROM %s WHERE %s" % ( + table, + " AND ".join("%s = ?" % (k,) for k in keyvalues), + ) + + txn.execute(sql, list(keyvalues.values())) + if txn.rowcount == 0: + raise StoreError(404, "No row found (%s)" % (table,)) + if txn.rowcount > 1: + raise StoreError(500, "More than one row matched (%s)" % (table,)) + + def simple_delete(self, table, keyvalues, desc): + return self.runInteraction(desc, self.simple_delete_txn, table, keyvalues) + + @staticmethod + def simple_delete_txn(txn, table, keyvalues): + sql = "DELETE FROM %s WHERE %s" % ( + table, + " AND ".join("%s = ?" % (k,) for k in keyvalues), + ) + + txn.execute(sql, list(keyvalues.values())) + return txn.rowcount + + def simple_delete_many(self, table, column, iterable, keyvalues, desc): + return self.runInteraction( + desc, self.simple_delete_many_txn, table, column, iterable, keyvalues + ) + + @staticmethod + def simple_delete_many_txn(txn, table, column, iterable, keyvalues): + """Executes a DELETE query on the named table. + + Filters rows by if value of `column` is in `iterable`. + + Args: + txn : Transaction object + table : string giving the table name + column : column name to test for inclusion against `iterable` + iterable : list + keyvalues : dict of column names and values to select the rows with + + Returns: + int: Number rows deleted + """ + if not iterable: + return 0 + + sql = "DELETE FROM %s" % table + + clause, values = make_in_list_sql_clause(txn.database_engine, column, iterable) + clauses = [clause] + + for key, value in iteritems(keyvalues): + clauses.append("%s = ?" % (key,)) + values.append(value) + + if clauses: + sql = "%s WHERE %s" % (sql, " AND ".join(clauses)) + txn.execute(sql, values) + + return txn.rowcount + + def get_cache_dict( + self, db_conn, table, entity_column, stream_column, max_value, limit=100000 + ): + # Fetch a mapping of room_id -> max stream position for "recent" rooms. + # It doesn't really matter how many we get, the StreamChangeCache will + # do the right thing to ensure it respects the max size of cache. + sql = ( + "SELECT %(entity)s, MAX(%(stream)s) FROM %(table)s" + " WHERE %(stream)s > ? - %(limit)s" + " GROUP BY %(entity)s" + ) % { + "table": table, + "entity": entity_column, + "stream": stream_column, + "limit": limit, + } + + sql = self.engine.convert_param_style(sql) + + txn = db_conn.cursor() + txn.execute(sql, (int(max_value),)) + + cache = {row[0]: int(row[1]) for row in txn} + + txn.close() + + if cache: + min_val = min(itervalues(cache)) + else: + min_val = max_value + + return cache, min_val + + def simple_select_list_paginate( + self, + table, + orderby, + start, + limit, + retcols, + filters=None, + keyvalues=None, + order_direction="ASC", + desc="simple_select_list_paginate", + ): + """ + 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, + returning the result as a list of dicts. + + Args: + table (str): the table name + filters (dict[str, T] | None): + column names and values to filter the rows with, or None to not + apply a WHERE ? LIKE ? clause. + keyvalues (dict[str, T] | None): + column names and values to select the rows with, or None to not + apply a WHERE clause. + orderby (str): Column to order the results by. + start (int): Index to begin the query at. + limit (int): Number of results to return. + retcols (iterable[str]): the names of the columns to return + order_direction (str): Whether the results should be ordered "ASC" or "DESC". + Returns: + defer.Deferred: resolves to list[dict[str, Any]] + """ + return self.runInteraction( + desc, + self.simple_select_list_paginate_txn, + table, + orderby, + start, + limit, + retcols, + filters=filters, + keyvalues=keyvalues, + order_direction=order_direction, + ) + + @classmethod + def simple_select_list_paginate_txn( + cls, + txn, + table, + orderby, + start, + limit, + retcols, + filters=None, + keyvalues=None, + order_direction="ASC", + ): + """ + 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, + returning the result as a list of dicts. + + Use `filters` to search attributes using SQL wildcards and/or `keyvalues` to + select attributes with exact matches. All constraints are joined together + using 'AND'. + + Args: + txn : Transaction object + table (str): the table name + orderby (str): Column to order the results by. + start (int): Index to begin the query at. + limit (int): Number of results to return. + retcols (iterable[str]): the names of the columns to return + filters (dict[str, T] | None): + column names and values to filter the rows with, or None to not + apply a WHERE ? LIKE ? clause. + keyvalues (dict[str, T] | None): + column names and values to select the rows with, or None to not + apply a WHERE clause. + order_direction (str): Whether the results should be ordered "ASC" or "DESC". + Returns: + defer.Deferred: resolves to list[dict[str, Any]] + """ + if order_direction not in ["ASC", "DESC"]: + raise ValueError("order_direction must be one of 'ASC' or 'DESC'.") + + where_clause = "WHERE " if filters or keyvalues else "" + arg_list = [] + if filters: + where_clause += " AND ".join("%s LIKE ?" % (k,) for k in filters) + arg_list += list(filters.values()) + where_clause += " AND " if filters and keyvalues else "" + if keyvalues: + where_clause += " AND ".join("%s = ?" % (k,) for k in keyvalues) + arg_list += list(keyvalues.values()) + + sql = "SELECT %s FROM %s %s ORDER BY %s %s LIMIT ? OFFSET ?" % ( + ", ".join(retcols), + table, + where_clause, + orderby, + order_direction, + ) + txn.execute(sql, arg_list + [limit, start]) + + return cls.cursor_to_dict(txn) + + def simple_search_list(self, table, term, col, retcols, desc="simple_search_list"): + """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 (str): the table name + term (str | None): + term for searching the table matched to a column. + col (str): column to query term should be matched to + retcols (iterable[str]): the names of the columns to return + Returns: + defer.Deferred: resolves to list[dict[str, Any]] or None + """ + + return self.runInteraction( + desc, self.simple_search_list_txn, table, term, col, retcols + ) + + @classmethod + def simple_search_list_txn(cls, txn, table, term, col, retcols): + """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 (str): the table name + term (str | None): + term for searching the table matched to a column. + col (str): column to query term should be matched to + retcols (iterable[str]): the names of the columns to return + Returns: + defer.Deferred: resolves to list[dict[str, Any]] or None + """ + if term: + sql = "SELECT %s FROM %s WHERE %s LIKE ?" % (", ".join(retcols), table, col) + termvalues = ["%%" + term + "%%"] + txn.execute(sql, termvalues) + else: + return 0 + + return cls.cursor_to_dict(txn) + + +def make_in_list_sql_clause( + database_engine, column: str, iterable: Iterable +) -> Tuple[str, Iterable]: + """Returns an SQL clause that checks the given column is in the iterable. + + On SQLite this expands to `column IN (?, ?, ...)`, whereas on Postgres + it expands to `column = ANY(?)`. While both DBs support the `IN` form, + using the `ANY` form on postgres means that it views queries with + different length iterables as the same, helping the query stats. + + Args: + database_engine + column: Name of the column + iterable: The values to check the column against. + + Returns: + A tuple of SQL query and the args + """ + + if database_engine.supports_using_any_list: + # This should hopefully be faster, but also makes postgres query + # stats easier to understand. + return "%s = ANY(?)" % (column,), [list(iterable)] + else: + return "%s IN (%s)" % (column, ",".join("?" for _ in iterable)), list(iterable) diff --git a/synapse/storage/engines/postgres.py b/synapse/storage/engines/postgres.py index b7c4eda338..c84cb452b0 100644 --- a/synapse/storage/engines/postgres.py +++ b/synapse/storage/engines/postgres.py @@ -32,20 +32,7 @@ class PostgresEngine(object): self.synchronous_commit = database_config.get("synchronous_commit", True) self._version = None # unknown as yet - def check_database(self, txn): - txn.execute("SHOW SERVER_ENCODING") - rows = txn.fetchall() - if rows and rows[0][0] != "UTF8": - raise IncorrectDatabaseSetup( - "Database has incorrect encoding: '%s' instead of 'UTF8'\n" - "See docs/postgres.rst for more information." % (rows[0][0],) - ) - - def convert_param_style(self, sql): - return sql.replace("?", "%s") - - def on_new_connection(self, db_conn): - + def check_database(self, db_conn, allow_outdated_version: bool = False): # Get the version of PostgreSQL that we're using. As per the psycopg2 # docs: The number is formed by converting the major, minor, and # revision numbers into two-decimal-digit numbers and appending them @@ -53,9 +40,22 @@ class PostgresEngine(object): self._version = db_conn.server_version # Are we on a supported PostgreSQL version? - if self._version < 90500: + if not allow_outdated_version and self._version < 90500: raise RuntimeError("Synapse requires PostgreSQL 9.5+ or above.") + with db_conn.cursor() as txn: + txn.execute("SHOW SERVER_ENCODING") + rows = txn.fetchall() + if rows and rows[0][0] != "UTF8": + raise IncorrectDatabaseSetup( + "Database has incorrect encoding: '%s' instead of 'UTF8'\n" + "See docs/postgres.rst for more information." % (rows[0][0],) + ) + + def convert_param_style(self, sql): + return sql.replace("?", "%s") + + def on_new_connection(self, db_conn): db_conn.set_isolation_level( self.module.extensions.ISOLATION_LEVEL_REPEATABLE_READ ) @@ -119,8 +119,8 @@ class PostgresEngine(object): Returns: string """ - # note that this is a bit of a hack because it relies on on_new_connection - # having been called at least once. Still, that should be a safe bet here. + # note that this is a bit of a hack because it relies on check_database + # having been called. Still, that should be a safe bet here. numver = self._version assert numver is not None diff --git a/synapse/storage/engines/sqlite.py b/synapse/storage/engines/sqlite.py index ddad17dc5a..cbf52f5191 100644 --- a/synapse/storage/engines/sqlite.py +++ b/synapse/storage/engines/sqlite.py @@ -16,8 +16,6 @@ import struct import threading -from synapse.storage.prepare_database import prepare_database - class Sqlite3Engine(object): single_threaded = True @@ -25,6 +23,9 @@ class Sqlite3Engine(object): def __init__(self, database_module, database_config): self.module = database_module + database = database_config.get("args", {}).get("database") + self._is_in_memory = database in (None, ":memory:",) + # The current max state_group, or None if we haven't looked # in the DB yet. self._current_state_group_id = None @@ -52,14 +53,26 @@ class Sqlite3Engine(object): """ return False - def check_database(self, txn): - pass + def check_database(self, db_conn, allow_outdated_version: bool = False): + if not allow_outdated_version: + version = self.module.sqlite_version_info + if version < (3, 11, 0): + raise RuntimeError("Synapse requires sqlite 3.11 or above.") def convert_param_style(self, sql): return sql def on_new_connection(self, db_conn): - prepare_database(db_conn, self, config=None) + + # We need to import here to avoid an import loop. + from synapse.storage.prepare_database import prepare_database + + if self._is_in_memory: + # In memory databases need to be rebuilt each time. Ideally we'd + # reuse the same connection as we do when starting up, but that + # would involve using adbapi before we have started the reactor. + prepare_database(db_conn, self, config=None) + db_conn.create_function("rank", 1, _rank) def is_deadlock(self, error): diff --git a/synapse/storage/persist_events.py b/synapse/storage/persist_events.py index fa03ca9ff7..1ed44925fc 100644 --- a/synapse/storage/persist_events.py +++ b/synapse/storage/persist_events.py @@ -183,7 +183,7 @@ class EventsPersistenceStorage(object): # so we use separate variables here even though they point to the same # store for now. self.main_store = stores.main - self.state_store = stores.main + self.state_store = stores.state self._clock = hs.get_clock() self.is_mine_id = hs.is_mine_id diff --git a/synapse/storage/prepare_database.py b/synapse/storage/prepare_database.py index 2e7753820e..e86984cd50 100644 --- a/synapse/storage/prepare_database.py +++ b/synapse/storage/prepare_database.py @@ -18,6 +18,7 @@ import imp import logging import os import re +from collections import Counter import attr @@ -28,7 +29,7 @@ logger = logging.getLogger(__name__) # Remember to update this number every time a change is made to database # schema files, so the users will be informed on server restarts. -SCHEMA_VERSION = 56 +SCHEMA_VERSION = 57 dir_path = os.path.abspath(os.path.dirname(__file__)) @@ -41,7 +42,7 @@ class UpgradeDatabaseException(PrepareDatabaseException): pass -def prepare_database(db_conn, database_engine, config): +def prepare_database(db_conn, database_engine, config, data_stores=["main", "state"]): """Prepares a database for usage. Will either create all necessary tables or upgrade from an older schema version. @@ -54,11 +55,10 @@ def prepare_database(db_conn, database_engine, config): config (synapse.config.homeserver.HomeServerConfig|None): application config, or None if we are connecting to an existing database which we expect to be configured already + data_stores (list[str]): The name of the data stores that will be used + with this database. Defaults to all data stores. """ - # For now we only have the one datastore. - data_stores = ["main"] - try: cur = db_conn.cursor() version_info = _get_or_create_schema_state(cur, database_engine) @@ -70,7 +70,10 @@ def prepare_database(db_conn, database_engine, config): if user_version != SCHEMA_VERSION: # If we don't pass in a config file then we are expecting to # have already upgraded the DB. - raise UpgradeDatabaseException("Database needs to be upgraded") + raise UpgradeDatabaseException( + "Expected database schema version %i but got %i" + % (SCHEMA_VERSION, user_version) + ) else: _upgrade_existing_database( cur, @@ -313,6 +316,9 @@ def _upgrade_existing_database( ) ) + # Used to check if we have any duplicate file names + file_name_counter = Counter() + # Now find which directories have anything of interest. directory_entries = [] for directory in directories: @@ -323,6 +329,9 @@ def _upgrade_existing_database( _DirectoryListing(file_name, os.path.join(directory, file_name)) for file_name in file_names ) + + for file_name in file_names: + file_name_counter[file_name] += 1 except FileNotFoundError: # Data stores can have empty entries for a given version delta. pass @@ -331,6 +340,17 @@ def _upgrade_existing_database( "Could not open delta dir for version %d: %s" % (v, directory) ) + duplicates = set( + file_name for file_name, count in file_name_counter.items() if count > 1 + ) + if duplicates: + # We don't support using the same file name in the same delta version. + raise PrepareDatabaseException( + "Found multiple delta files with the same name in v%d: %s", + v, + duplicates, + ) + # We sort to ensure that we apply the delta files in a consistent # order (to avoid bugs caused by inconsistent directory listing order) directory_entries.sort() @@ -447,7 +467,7 @@ def _apply_module_schema_files(cur, database_engine, modname, names_and_streams) # Mark as done. cur.execute( database_engine.convert_param_style( - "INSERT INTO applied_module_schemas (module_name, file)" " VALUES (?,?)" + "INSERT INTO applied_module_schemas (module_name, file) VALUES (?,?)" ), (modname, name), ) diff --git a/synapse/storage/purge_events.py b/synapse/storage/purge_events.py index a368182034..fdc0abf5cf 100644 --- a/synapse/storage/purge_events.py +++ b/synapse/storage/purge_events.py @@ -34,7 +34,7 @@ class PurgeEventsStorage(object): """ state_groups_to_delete = yield self.stores.main.purge_room(room_id) - yield self.stores.main.purge_room_state(room_id, state_groups_to_delete) + yield self.stores.state.purge_room_state(room_id, state_groups_to_delete) @defer.inlineCallbacks def purge_history(self, room_id, token, delete_local_events): @@ -58,7 +58,7 @@ class PurgeEventsStorage(object): sg_to_delete = yield self._find_unreferenced_groups(state_groups) - yield self.stores.main.purge_unreferenced_state_groups(room_id, sg_to_delete) + yield self.stores.state.purge_unreferenced_state_groups(room_id, sg_to_delete) @defer.inlineCallbacks def _find_unreferenced_groups(self, state_groups): @@ -102,7 +102,7 @@ class PurgeEventsStorage(object): # groups that are referenced. current_search -= referenced - edges = yield self.stores.main.get_previous_state_groups(current_search) + edges = yield self.stores.state.get_previous_state_groups(current_search) prevs = set(edges.values()) # We don't bother re-handling groups we've already seen diff --git a/synapse/storage/state.py b/synapse/storage/state.py index 3735846899..c522c80922 100644 --- a/synapse/storage/state.py +++ b/synapse/storage/state.py @@ -14,6 +14,7 @@ # limitations under the License. import logging +from typing import Iterable, List, TypeVar from six import iteritems, itervalues @@ -22,9 +23,13 @@ import attr from twisted.internet import defer from synapse.api.constants import EventTypes +from synapse.types import StateMap logger = logging.getLogger(__name__) +# Used for generic functions below +T = TypeVar("T") + @attr.s(slots=True) class StateFilter(object): @@ -233,14 +238,14 @@ class StateFilter(object): return len(self.concrete_types()) - def filter_state(self, state_dict): + def filter_state(self, state_dict: StateMap[T]) -> StateMap[T]: """Returns the state filtered with by this StateFilter Args: - state (dict[tuple[str, str], Any]): The state map to filter + state: The state map to filter Returns: - dict[tuple[str, str], Any]: The filtered state map + The filtered state map """ if self.is_full(): return dict(state_dict) @@ -333,16 +338,16 @@ class StateGroupStorage(object): def __init__(self, hs, stores): self.stores = stores - def get_state_group_delta(self, state_group): + def get_state_group_delta(self, state_group: int): """Given a state group try to return a previous group and a delta between the old and the new. Returns: - Deferred[Tuple[Optional[int], Optional[list[dict[tuple[str, str], str]]]]]): + Deferred[Tuple[Optional[int], Optional[StateMap[str]]]]: (prev_group, delta_ids) """ - return self.stores.main.get_state_group_delta(state_group) + return self.stores.state.get_state_group_delta(state_group) @defer.inlineCallbacks def get_state_groups_ids(self, _room_id, event_ids): @@ -353,7 +358,7 @@ class StateGroupStorage(object): event_ids (iterable[str]): ids of the events Returns: - Deferred[dict[int, dict[tuple[str, str], str]]]: + Deferred[dict[int, StateMap[str]]]: dict of state_group_id -> (dict of (type, state_key) -> event id) """ if not event_ids: @@ -362,7 +367,7 @@ class StateGroupStorage(object): event_to_groups = yield self.stores.main._get_state_group_for_events(event_ids) groups = set(itervalues(event_to_groups)) - group_to_state = yield self.stores.main._get_state_for_groups(groups) + group_to_state = yield self.stores.state._get_state_for_groups(groups) return group_to_state @@ -410,20 +415,21 @@ class StateGroupStorage(object): for group, event_id_map in iteritems(group_to_ids) } - def _get_state_groups_from_groups(self, groups, state_filter): + def _get_state_groups_from_groups( + self, groups: List[int], state_filter: StateFilter + ): """Returns the state groups for a given set of groups, filtering on types of state events. Args: - groups(list[int]): list of state group IDs to query - state_filter (StateFilter): The state filter used to fetch state + groups: list of state group IDs to query + state_filter: The state filter used to fetch state from the database. Returns: - Deferred[dict[int, dict[tuple[str, str], str]]]: - dict of state_group_id -> (dict of (type, state_key) -> event id) + Deferred[Dict[int, StateMap[str]]]: Dict of state group to state map. """ - return self.stores.main._get_state_groups_from_groups(groups, state_filter) + return self.stores.state._get_state_groups_from_groups(groups, state_filter) @defer.inlineCallbacks def get_state_for_events(self, event_ids, state_filter=StateFilter.all()): @@ -439,7 +445,7 @@ class StateGroupStorage(object): event_to_groups = yield self.stores.main._get_state_group_for_events(event_ids) groups = set(itervalues(event_to_groups)) - group_to_state = yield self.stores.main._get_state_for_groups( + group_to_state = yield self.stores.state._get_state_for_groups( groups, state_filter ) @@ -476,7 +482,7 @@ class StateGroupStorage(object): event_to_groups = yield self.stores.main._get_state_group_for_events(event_ids) groups = set(itervalues(event_to_groups)) - group_to_state = yield self.stores.main._get_state_for_groups( + group_to_state = yield self.stores.state._get_state_for_groups( groups, state_filter ) @@ -519,7 +525,9 @@ class StateGroupStorage(object): state_map = yield self.get_state_ids_for_events([event_id], state_filter) return state_map[event_id] - def _get_state_for_groups(self, groups, state_filter=StateFilter.all()): + def _get_state_for_groups( + self, groups: Iterable[int], state_filter: StateFilter = StateFilter.all() + ): """Gets the state at each of a list of state groups, optionally filtering by type/state_key @@ -529,10 +537,9 @@ class StateGroupStorage(object): state_filter (StateFilter): The state filter used to fetch state from the database. Returns: - Deferred[dict[int, dict[tuple[str, str], str]]]: - dict of state_group_id -> (dict of (type, state_key) -> event id) + Deferred[dict[int, StateMap[str]]]: Dict of state group to state map. """ - return self.stores.main._get_state_for_groups(groups, state_filter) + return self.stores.state._get_state_for_groups(groups, state_filter) def store_state_group( self, event_id, room_id, prev_group, delta_ids, current_state_ids @@ -552,6 +559,6 @@ class StateGroupStorage(object): Returns: Deferred[int]: The state group ID """ - return self.stores.main.store_state_group( + return self.stores.state.store_state_group( event_id, room_id, prev_group, delta_ids, current_state_ids ) diff --git a/synapse/streams/config.py b/synapse/streams/config.py index 02994ab2a5..cd56cd91ed 100644 --- a/synapse/streams/config.py +++ b/synapse/streams/config.py @@ -88,9 +88,12 @@ class PaginationConfig(object): raise SynapseError(400, "Invalid request.") def __repr__(self): - return ( - "PaginationConfig(from_tok=%r, to_tok=%r," " direction=%r, limit=%r)" - ) % (self.from_token, self.to_token, self.direction, self.limit) + return ("PaginationConfig(from_tok=%r, to_tok=%r, direction=%r, limit=%r)") % ( + self.from_token, + self.to_token, + self.direction, + self.limit, + ) def get_source_config(self, source_name): keyname = "%s_key" % source_name diff --git a/synapse/streams/events.py b/synapse/streams/events.py index b91fb2db7b..fcd2aaa9c9 100644 --- a/synapse/streams/events.py +++ b/synapse/streams/events.py @@ -13,6 +13,8 @@ # See the License for the specific language governing permissions and # limitations under the License. +from typing import Any, Dict + from twisted.internet import defer from synapse.handlers.account_data import AccountDataEventSource @@ -35,7 +37,7 @@ class EventSources(object): def __init__(self, hs): self.sources = { name: cls(hs) for name, cls in EventSources.SOURCE_TYPES.items() - } + } # type: Dict[str, Any] self.store = hs.get_datastore() @defer.inlineCallbacks diff --git a/synapse/types.py b/synapse/types.py index aafc3ffe74..65e4d8c181 100644 --- a/synapse/types.py +++ b/synapse/types.py @@ -15,7 +15,9 @@ # limitations under the License. import re import string +import sys from collections import namedtuple +from typing import Dict, Tuple, TypeVar import attr from signedjson.key import decode_verify_key_bytes @@ -23,6 +25,23 @@ from unpaddedbase64 import decode_base64 from synapse.api.errors import SynapseError +# define a version of typing.Collection that works on python 3.5 +if sys.version_info[:3] >= (3, 6, 0): + from typing import Collection +else: + from typing import Sized, Iterable, Container + + T_co = TypeVar("T_co", covariant=True) + + class Collection(Iterable[T_co], Container[T_co], Sized): + __slots__ = () + + +# Define a state map type from type/state_key to T (usually an event ID or +# event) +T = TypeVar("T") +StateMap = Dict[Tuple[str, str], T] + class Requester( namedtuple( diff --git a/synapse/util/async_helpers.py b/synapse/util/async_helpers.py index 5c4de2e69f..04b6abdc24 100644 --- a/synapse/util/async_helpers.py +++ b/synapse/util/async_helpers.py @@ -140,8 +140,8 @@ def concurrently_execute(func, args, limit): Args: func (func): Function to execute, should return a deferred or coroutine. - args (list): List of arguments to pass to func, each invocation of func - gets a signle argument. + args (Iterable): List of arguments to pass to func, each invocation of func + gets a single argument. limit (int): Maximum number of conccurent executions. Returns: diff --git a/synapse/util/caches/descriptors.py b/synapse/util/caches/descriptors.py index f1e525399f..c4cf38abc7 100644 --- a/synapse/util/caches/descriptors.py +++ b/synapse/util/caches/descriptors.py @@ -274,7 +274,7 @@ class _CacheDescriptorBase(object): else: self.function_to_call = orig - arg_spec = inspect.getargspec(orig) + arg_spec = inspect.getfullargspec(orig) all_args = arg_spec.args if "cache_context" in all_args: diff --git a/synapse/util/caches/snapshot_cache.py b/synapse/util/caches/snapshot_cache.py deleted file mode 100644 index 8318db8d2c..0000000000 --- a/synapse/util/caches/snapshot_cache.py +++ /dev/null @@ -1,94 +0,0 @@ -# -*- coding: utf-8 -*- -# Copyright 2015, 2016 OpenMarket Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from synapse.util.async_helpers import ObservableDeferred - - -class SnapshotCache(object): - """Cache for snapshots like the response of /initialSync. - The response of initialSync only has to be a recent snapshot of the - server state. It shouldn't matter to clients if it is a few minutes out - of date. - - This caches a deferred response. Until the deferred completes it will be - returned from the cache. This means that if the client retries the request - while the response is still being computed, that original response will be - used rather than trying to compute a new response. - - Once the deferred completes it will removed from the cache after 5 minutes. - We delay removing it from the cache because a client retrying its request - could race with us finishing computing the response. - - Rather than tracking precisely how long something has been in the cache we - keep two generations of completed responses. Every 5 minutes discard the - old generation, move the new generation to the old generation, and set the - new generation to be empty. This means that a result will be in the cache - somewhere between 5 and 10 minutes. - """ - - DURATION_MS = 5 * 60 * 1000 # Cache results for 5 minutes. - - def __init__(self): - self.pending_result_cache = {} # Request that haven't finished yet. - self.prev_result_cache = {} # The older requests that have finished. - self.next_result_cache = {} # The newer requests that have finished. - self.time_last_rotated_ms = 0 - - def rotate(self, time_now_ms): - # Rotate once if the cache duration has passed since the last rotation. - if time_now_ms - self.time_last_rotated_ms >= self.DURATION_MS: - self.prev_result_cache = self.next_result_cache - self.next_result_cache = {} - self.time_last_rotated_ms += self.DURATION_MS - - # Rotate again if the cache duration has passed twice since the last - # rotation. - if time_now_ms - self.time_last_rotated_ms >= self.DURATION_MS: - self.prev_result_cache = self.next_result_cache - self.next_result_cache = {} - self.time_last_rotated_ms = time_now_ms - - def get(self, time_now_ms, key): - self.rotate(time_now_ms) - # This cache is intended to deduplicate requests, so we expect it to be - # missed most of the time. So we just lookup the key in all of the - # dictionaries rather than trying to short circuit the lookup if the - # key is found. - result = self.prev_result_cache.get(key) - result = self.next_result_cache.get(key, result) - result = self.pending_result_cache.get(key, result) - if result is not None: - return result.observe() - else: - return None - - def set(self, time_now_ms, key, deferred): - self.rotate(time_now_ms) - - result = ObservableDeferred(deferred) - - self.pending_result_cache[key] = result - - def shuffle_along(r): - # When the deferred completes we shuffle it along to the first - # generation of the result cache. So that it will eventually - # expire from the rotation of that cache. - self.next_result_cache[key] = result - self.pending_result_cache.pop(key, None) - return r - - result.addBoth(shuffle_along) - - return result.observe() diff --git a/synapse/util/metrics.py b/synapse/util/metrics.py index 3286804322..7b18455469 100644 --- a/synapse/util/metrics.py +++ b/synapse/util/metrics.py @@ -13,6 +13,7 @@ # See the License for the specific language governing permissions and # limitations under the License. +import inspect import logging from functools import wraps @@ -64,12 +65,22 @@ def measure_func(name=None): def wrapper(func): block_name = func.__name__ if name is None else name - @wraps(func) - @defer.inlineCallbacks - def measured_func(self, *args, **kwargs): - with Measure(self.clock, block_name): - r = yield func(self, *args, **kwargs) - return r + if inspect.iscoroutinefunction(func): + + @wraps(func) + async def measured_func(self, *args, **kwargs): + with Measure(self.clock, block_name): + r = await func(self, *args, **kwargs) + return r + + else: + + @wraps(func) + @defer.inlineCallbacks + def measured_func(self, *args, **kwargs): + with Measure(self.clock, block_name): + r = yield func(self, *args, **kwargs) + return r return measured_func @@ -80,72 +91,48 @@ class Measure(object): __slots__ = [ "clock", "name", - "start_context", + "_logging_context", "start", - "created_context", - "start_usage", ] def __init__(self, clock, name): self.clock = clock self.name = name - self.start_context = None + self._logging_context = None self.start = None - self.created_context = False def __enter__(self): - self.start = self.clock.time() - self.start_context = LoggingContext.current_context() - if not self.start_context: - self.start_context = LoggingContext("Measure") - self.start_context.__enter__() - self.created_context = True - - self.start_usage = self.start_context.get_resource_usage() + if self._logging_context: + raise RuntimeError("Measure() objects cannot be re-used") + self.start = self.clock.time() + parent_context = LoggingContext.current_context() + self._logging_context = LoggingContext( + "Measure[%s]" % (self.name,), parent_context + ) + self._logging_context.__enter__() in_flight.register((self.name,), self._update_in_flight) def __exit__(self, exc_type, exc_val, exc_tb): - if isinstance(exc_type, Exception) or not self.start_context: - return - - in_flight.unregister((self.name,), self._update_in_flight) + if not self._logging_context: + raise RuntimeError("Measure() block exited without being entered") duration = self.clock.time() - self.start + usage = self._logging_context.get_resource_usage() - block_counter.labels(self.name).inc() - block_timer.labels(self.name).inc(duration) - - context = LoggingContext.current_context() - - if context != self.start_context: - logger.warning( - "Context has unexpectedly changed from '%s' to '%s'. (%r)", - self.start_context, - context, - self.name, - ) - return - - if not context: - logger.warning("Expected context. (%r)", self.name) - return + in_flight.unregister((self.name,), self._update_in_flight) + self._logging_context.__exit__(exc_type, exc_val, exc_tb) - current = context.get_resource_usage() - usage = current - self.start_usage try: + block_counter.labels(self.name).inc() + block_timer.labels(self.name).inc(duration) block_ru_utime.labels(self.name).inc(usage.ru_utime) block_ru_stime.labels(self.name).inc(usage.ru_stime) block_db_txn_count.labels(self.name).inc(usage.db_txn_count) block_db_txn_duration.labels(self.name).inc(usage.db_txn_duration_sec) block_db_sched_duration.labels(self.name).inc(usage.db_sched_duration_sec) except ValueError: - logger.warning( - "Failed to save metrics! OLD: %r, NEW: %r", self.start_usage, current - ) - - if self.created_context: - self.start_context.__exit__(exc_type, exc_val, exc_tb) + logger.warning("Failed to save metrics! Usage: %s", usage) def _update_in_flight(self, metrics): """Gets called when processing in flight metrics diff --git a/synapse/util/module_loader.py b/synapse/util/module_loader.py index 2705cbe5f8..bb62db4637 100644 --- a/synapse/util/module_loader.py +++ b/synapse/util/module_loader.py @@ -34,7 +34,7 @@ def load_module(provider): provider_class = getattr(module, clz) try: - provider_config = provider_class.parse_config(provider["config"]) + provider_config = provider_class.parse_config(provider.get("config")) except Exception as e: raise ConfigError("Failed to parse config for %r: %r" % (provider["module"], e)) diff --git a/synapse/visibility.py b/synapse/visibility.py index 8c843febd8..100dc47a8a 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -44,10 +44,16 @@ MEMBERSHIP_PRIORITY = ( @defer.inlineCallbacks def filter_events_for_client( - storage: Storage, user_id, events, is_peeking=False, always_include_ids=frozenset() + storage: Storage, + user_id, + events, + is_peeking=False, + always_include_ids=frozenset(), + apply_retention_policies=True, ): """ - Check which events a user is allowed to see + Check which events a user is allowed to see. If the user can see the event but its + sender asked for their data to be erased, prune the content of the event. Args: storage @@ -59,6 +65,10 @@ def filter_events_for_client( events always_include_ids (set(event_id)): set of event ids to specifically include (unless sender is ignored) + apply_retention_policies (bool): Whether to filter out events that's older than + allowed by the room's retention policy. Useful when this function is called + to e.g. check whether a user should be allowed to see the state at a given + event rather than to know if it should send an event to a user's client(s). Returns: Deferred[list[synapse.events.EventBase]] @@ -86,6 +96,15 @@ def filter_events_for_client( erased_senders = yield storage.main.are_users_erased((e.sender for e in events)) + if apply_retention_policies: + room_ids = set(e.room_id for e in events) + retention_policies = {} + + for room_id in room_ids: + retention_policies[ + room_id + ] = yield storage.main.get_retention_policy_for_room(room_id) + def allowed(event): """ Args: @@ -103,6 +122,18 @@ def filter_events_for_client( if not event.is_state() and event.sender in ignore_list: return None + # Don't try to apply the room's retention policy if the event is a state event, as + # MSC1763 states that retention is only considered for non-state events. + if apply_retention_policies and not event.is_state(): + retention_policy = retention_policies[event.room_id] + max_lifetime = retention_policy.get("max_lifetime") + + if max_lifetime is not None: + oldest_allowed_ts = storage.main.clock.time_msec() - max_lifetime + + if event.origin_server_ts < oldest_allowed_ts: + return None + if event.event_id in always_include_ids: return event |