diff options
Diffstat (limited to 'synapse/handlers')
| -rw-r--r-- | synapse/handlers/__init__.py | 4 | ||||
| -rw-r--r-- | synapse/handlers/_base.py | 3 | ||||
| -rw-r--r-- | synapse/handlers/appservice.py | 5 | ||||
| -rw-r--r-- | synapse/handlers/federation.py | 293 | ||||
| -rw-r--r-- | synapse/handlers/initial_sync.py | 38 | ||||
| -rw-r--r-- | synapse/handlers/message.py | 304 | ||||
| -rw-r--r-- | synapse/handlers/pagination.py | 265 | ||||
| -rw-r--r-- | synapse/handlers/profile.py | 9 | ||||
| -rw-r--r-- | synapse/handlers/room.py | 42 | ||||
| -rw-r--r-- | synapse/handlers/room_member.py | 13 | ||||
| -rw-r--r-- | synapse/handlers/search.py | 2 | ||||
| -rw-r--r-- | synapse/handlers/sync.py | 173 |
12 files changed, 595 insertions, 556 deletions
diff --git a/synapse/handlers/__init__.py b/synapse/handlers/__init__.py index 4b9923d8c0..413425fed1 100644 --- a/synapse/handlers/__init__.py +++ b/synapse/handlers/__init__.py @@ -17,9 +17,7 @@ from .admin import AdminHandler from .directory import DirectoryHandler from .federation import FederationHandler from .identity import IdentityHandler -from .message import MessageHandler from .register import RegistrationHandler -from .room import RoomContextHandler from .search import SearchHandler @@ -44,10 +42,8 @@ class Handlers(object): def __init__(self, hs): self.registration_handler = RegistrationHandler(hs) - self.message_handler = MessageHandler(hs) self.federation_handler = FederationHandler(hs) self.directory_handler = DirectoryHandler(hs) self.admin_handler = AdminHandler(hs) self.identity_handler = IdentityHandler(hs) self.search_handler = SearchHandler(hs) - self.room_context_handler = RoomContextHandler(hs) diff --git a/synapse/handlers/_base.py b/synapse/handlers/_base.py index b6a8b3aa3b..704181d2d3 100644 --- a/synapse/handlers/_base.py +++ b/synapse/handlers/_base.py @@ -112,8 +112,9 @@ 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 = yield self.store.get_events( - list(context.current_state_ids.values()) + list(current_state_ids.values()) ) else: current_state = yield self.state_handler.get_current_state( diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index ec9fe01a5a..ee41aed69e 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -23,6 +23,7 @@ from twisted.internet import defer import synapse from synapse.api.constants import EventTypes +from synapse.metrics.background_process_metrics import run_as_background_process from synapse.util.logcontext import make_deferred_yieldable, run_in_background from synapse.util.metrics import Measure @@ -106,7 +107,9 @@ class ApplicationServicesHandler(object): yield self._check_user_exists(event.state_key) if not self.started_scheduler: - self.scheduler.start().addErrback(log_failure) + def start_scheduler(): + return self.scheduler.start().addErrback(log_failure) + run_as_background_process("as_scheduler", start_scheduler) self.started_scheduler = True # Fork off pushes to these services diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index d3ecebd29f..91d8def08b 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -21,8 +21,8 @@ import logging import sys import six -from six import iteritems -from six.moves import http_client +from six import iteritems, itervalues +from six.moves import http_client, zip from signedjson.key import decode_verify_key_bytes from signedjson.sign import verify_signed_json @@ -43,7 +43,6 @@ from synapse.crypto.event_signing import ( add_hashes_and_signatures, compute_event_signature, ) -from synapse.events.utils import prune_event from synapse.events.validator import EventValidator from synapse.state import resolve_events_with_factory from synapse.types import UserID, get_domain_from_id @@ -52,8 +51,8 @@ from synapse.util.async import Linearizer from synapse.util.distributor import user_joined_room from synapse.util.frozenutils import unfreeze from synapse.util.logutils import log_function -from synapse.util.metrics import measure_func from synapse.util.retryutils import NotRetryingDestination +from synapse.visibility import filter_events_for_server from ._base import BaseHandler @@ -77,7 +76,7 @@ class FederationHandler(BaseHandler): self.hs = hs self.store = hs.get_datastore() - self.replication_layer = hs.get_federation_client() + self.federation_client = hs.get_federation_client() self.state_handler = hs.get_state_handler() self.server_name = hs.hostname self.keyring = hs.get_keyring() @@ -256,7 +255,7 @@ class FederationHandler(BaseHandler): # know about for p in prevs - seen: state, got_auth_chain = ( - yield self.replication_layer.get_state_for_room( + yield self.federation_client.get_state_for_room( origin, pdu.room_id, p ) ) @@ -339,7 +338,7 @@ class FederationHandler(BaseHandler): # # see https://github.com/matrix-org/synapse/pull/1744 - missing_events = yield self.replication_layer.get_missing_events( + missing_events = yield self.federation_client.get_missing_events( origin, pdu.room_id, earliest_events_ids=list(latest), @@ -487,7 +486,10 @@ class FederationHandler(BaseHandler): # joined the room. Don't bother if the user is just # changing their profile info. newly_joined = True - prev_state_id = context.prev_state_ids.get( + + prev_state_ids = yield context.get_prev_state_ids(self.store) + + prev_state_id = prev_state_ids.get( (event.type, event.state_key) ) if prev_state_id: @@ -501,137 +503,6 @@ class FederationHandler(BaseHandler): user = UserID.from_string(event.state_key) yield user_joined_room(self.distributor, user, event.room_id) - @measure_func("_filter_events_for_server") - @defer.inlineCallbacks - def _filter_events_for_server(self, server_name, room_id, events): - """Filter the given events for the given server, redacting those the - server can't see. - - Assumes the server is currently in the room. - - Returns - list[FrozenEvent] - """ - # First lets check to see if all the events have a history visibility - # of "shared" or "world_readable". If thats the case then we don't - # need to check membership (as we know the server is in the room). - event_to_state_ids = yield self.store.get_state_ids_for_events( - frozenset(e.event_id for e in events), - types=( - (EventTypes.RoomHistoryVisibility, ""), - ) - ) - - visibility_ids = set() - for sids in event_to_state_ids.itervalues(): - hist = sids.get((EventTypes.RoomHistoryVisibility, "")) - if hist: - visibility_ids.add(hist) - - # If we failed to find any history visibility events then the default - # is "shared" visiblity. - if not visibility_ids: - defer.returnValue(events) - - event_map = yield self.store.get_events(visibility_ids) - all_open = all( - e.content.get("history_visibility") in (None, "shared", "world_readable") - for e in event_map.itervalues() - ) - - if all_open: - defer.returnValue(events) - - # Ok, so we're dealing with events that have non-trivial visibility - # rules, so we need to also get the memberships of the room. - - event_to_state_ids = yield self.store.get_state_ids_for_events( - frozenset(e.event_id for e in events), - types=( - (EventTypes.RoomHistoryVisibility, ""), - (EventTypes.Member, None), - ) - ) - - # We only want to pull out member events that correspond to the - # server's domain. - - def check_match(id): - try: - return server_name == get_domain_from_id(id) - except Exception: - return False - - # Parses mapping `event_id -> (type, state_key) -> state event_id` - # to get all state ids that we're interested in. - event_map = yield self.store.get_events([ - e_id - for key_to_eid in list(event_to_state_ids.values()) - for key, e_id in key_to_eid.items() - if key[0] != EventTypes.Member or check_match(key[1]) - ]) - - event_to_state = { - e_id: { - key: event_map[inner_e_id] - for key, inner_e_id in key_to_eid.iteritems() - if inner_e_id in event_map - } - for e_id, key_to_eid in event_to_state_ids.iteritems() - } - - erased_senders = yield self.store.are_users_erased( - e.sender for e in events, - ) - - def redact_disallowed(event, state): - # if the sender has been gdpr17ed, always return a redacted - # copy of the event. - if erased_senders[event.sender]: - logger.info( - "Sender of %s has been erased, redacting", - event.event_id, - ) - return prune_event(event) - - if not state: - return event - - history = state.get((EventTypes.RoomHistoryVisibility, ''), None) - if history: - visibility = history.content.get("history_visibility", "shared") - if visibility in ["invited", "joined"]: - # We now loop through all state events looking for - # membership states for the requesting server to determine - # if the server is either in the room or has been invited - # into the room. - for ev in state.itervalues(): - if ev.type != EventTypes.Member: - continue - try: - domain = get_domain_from_id(ev.state_key) - except Exception: - continue - - if domain != server_name: - continue - - memtype = ev.membership - if memtype == Membership.JOIN: - return event - elif memtype == Membership.INVITE: - if visibility == "invited": - return event - else: - return prune_event(event) - - return event - - defer.returnValue([ - redact_disallowed(e, event_to_state[e.event_id]) - for e in events - ]) - @log_function @defer.inlineCallbacks def backfill(self, dest, room_id, limit, extremities): @@ -651,7 +522,7 @@ class FederationHandler(BaseHandler): if dest == self.server_name: raise SynapseError(400, "Can't backfill from self.") - events = yield self.replication_layer.backfill( + events = yield self.federation_client.backfill( dest, room_id, limit=limit, @@ -699,7 +570,7 @@ class FederationHandler(BaseHandler): state_events = {} events_to_state = {} for e_id in edges: - state, auth = yield self.replication_layer.get_state_for_room( + state, auth = yield self.federation_client.get_state_for_room( destination=dest, room_id=room_id, event_id=e_id @@ -741,7 +612,7 @@ class FederationHandler(BaseHandler): results = yield logcontext.make_deferred_yieldable(defer.gatherResults( [ logcontext.run_in_background( - self.replication_layer.get_pdu, + self.federation_client.get_pdu, [dest], event_id, outlier=True, @@ -863,7 +734,7 @@ class FederationHandler(BaseHandler): """ joined_users = [ (state_key, int(event.depth)) - for (e_type, state_key), event in state.iteritems() + for (e_type, state_key), event in iteritems(state) if e_type == EventTypes.Member and event.membership == Membership.JOIN ] @@ -880,7 +751,7 @@ class FederationHandler(BaseHandler): except Exception: pass - return sorted(joined_domains.iteritems(), key=lambda d: d[1]) + return sorted(joined_domains.items(), key=lambda d: d[1]) curr_domains = get_domains_from_state(curr_state) @@ -943,7 +814,7 @@ class FederationHandler(BaseHandler): tried_domains = set(likely_domains) tried_domains.add(self.server_name) - event_ids = list(extremities.iterkeys()) + event_ids = list(extremities.keys()) logger.debug("calling resolve_state_groups in _maybe_backfill") resolve = logcontext.preserve_fn( @@ -959,15 +830,15 @@ class FederationHandler(BaseHandler): states = dict(zip(event_ids, [s.state for s in states])) state_map = yield self.store.get_events( - [e_id for ids in states.itervalues() for e_id in ids.itervalues()], + [e_id for ids in itervalues(states) for e_id in itervalues(ids)], get_prev_content=False ) states = { key: { k: state_map[e_id] - for k, e_id in state_dict.iteritems() + for k, e_id in iteritems(state_dict) if e_id in state_map - } for key, state_dict in states.iteritems() + } for key, state_dict in iteritems(states) } for e_id, _ in sorted_extremeties_tuple: @@ -1022,7 +893,7 @@ class FederationHandler(BaseHandler): Invites must be signed by the invitee's server before distribution. """ - pdu = yield self.replication_layer.send_invite( + pdu = yield self.federation_client.send_invite( destination=target_host, room_id=event.room_id, event_id=event.event_id, @@ -1038,16 +909,6 @@ class FederationHandler(BaseHandler): [auth_id for auth_id, _ in event.auth_events], include_given=True ) - - for event in auth: - event.signatures.update( - compute_event_signature( - event, - self.hs.hostname, - self.hs.config.signing_key[0] - ) - ) - defer.returnValue([e for e in auth]) @log_function @@ -1094,7 +955,7 @@ class FederationHandler(BaseHandler): target_hosts.insert(0, origin) except ValueError: pass - ret = yield self.replication_layer.send_join(target_hosts, event) + ret = yield self.federation_client.send_join(target_hosts, event) origin = ret["origin"] state = ret["state"] @@ -1248,10 +1109,12 @@ class FederationHandler(BaseHandler): user = UserID.from_string(event.state_key) yield user_joined_room(self.distributor, user, event.room_id) - state_ids = list(context.prev_state_ids.values()) + prev_state_ids = yield context.get_prev_state_ids(self.store) + + state_ids = list(prev_state_ids.values()) auth_chain = yield self.store.get_auth_chain(state_ids) - state = yield self.store.get_events(list(context.prev_state_ids.values())) + state = yield self.store.get_events(list(prev_state_ids.values())) defer.returnValue({ "state": list(state.values()), @@ -1348,7 +1211,7 @@ class FederationHandler(BaseHandler): except ValueError: pass - yield self.replication_layer.send_leave( + yield self.federation_client.send_leave( target_hosts, event ) @@ -1371,7 +1234,7 @@ class FederationHandler(BaseHandler): @defer.inlineCallbacks def _make_and_verify_event(self, target_hosts, room_id, user_id, membership, content={},): - origin, pdu = yield self.replication_layer.make_membership_event( + origin, pdu = yield self.federation_client.make_membership_event( target_hosts, room_id, user_id, @@ -1416,7 +1279,7 @@ class FederationHandler(BaseHandler): @log_function def on_make_leave_request(self, room_id, user_id): """ We've received a /make_leave/ request, so we create a partial - join event for the room and return that. We do *not* persist or + leave event for the room and return that. We do *not* persist or process it until the other server has signed it and sent it back. """ builder = self.event_builder_factory.new({ @@ -1503,18 +1366,6 @@ class FederationHandler(BaseHandler): del results[(event.type, event.state_key)] res = list(results.values()) - for event in res: - # We sign these again because there was a bug where we - # incorrectly signed things the first time round - if self.is_mine_id(event.event_id): - event.signatures.update( - compute_event_signature( - event, - self.hs.hostname, - self.hs.config.signing_key[0] - ) - ) - defer.returnValue(res) else: defer.returnValue([]) @@ -1558,7 +1409,7 @@ class FederationHandler(BaseHandler): limit ) - events = yield self._filter_events_for_server(origin, room_id, events) + events = yield filter_events_for_server(self.store, origin, events) defer.returnValue(events) @@ -1586,18 +1437,6 @@ class FederationHandler(BaseHandler): ) if event: - if self.is_mine_id(event.event_id): - # FIXME: This is a temporary work around where we occasionally - # return events slightly differently than when they were - # originally signed - event.signatures.update( - compute_event_signature( - event, - self.hs.hostname, - self.hs.config.signing_key[0] - ) - ) - in_room = yield self.auth.check_host_in_room( event.room_id, origin @@ -1605,8 +1444,8 @@ class FederationHandler(BaseHandler): if not in_room: raise AuthError(403, "Host not in room.") - events = yield self._filter_events_for_server( - origin, event.room_id, [event] + events = yield filter_events_for_server( + self.store, origin, [event], ) event = events[0] defer.returnValue(event) @@ -1681,7 +1520,7 @@ class FederationHandler(BaseHandler): yield self.store.persist_events( [ (ev_info["event"], context) - for ev_info, context in itertools.izip(event_infos, contexts) + for ev_info, context in zip(event_infos, contexts) ], backfilled=backfilled, ) @@ -1728,7 +1567,7 @@ class FederationHandler(BaseHandler): missing_auth_events.add(e_id) for e_id in missing_auth_events: - m_ev = yield self.replication_layer.get_pdu( + m_ev = yield self.federation_client.get_pdu( [origin], e_id, outlier=True, @@ -1801,8 +1640,9 @@ class FederationHandler(BaseHandler): ) if not auth_events: + prev_state_ids = yield context.get_prev_state_ids(self.store) auth_events_ids = yield self.auth.compute_auth_events( - event, context.prev_state_ids, for_verification=True, + event, prev_state_ids, for_verification=True, ) auth_events = yield self.store.get_events(auth_events_ids) auth_events = { @@ -1862,15 +1702,6 @@ class FederationHandler(BaseHandler): local_auth_chain, remote_auth_chain ) - for event in ret["auth_chain"]: - event.signatures.update( - compute_event_signature( - event, - self.hs.hostname, - self.hs.config.signing_key[0] - ) - ) - logger.debug("on_query_auth returning: %s", ret) defer.returnValue(ret) @@ -1896,8 +1727,8 @@ class FederationHandler(BaseHandler): min_depth=min_depth, ) - missing_events = yield self._filter_events_for_server( - origin, room_id, missing_events, + missing_events = yield filter_events_for_server( + self.store, origin, missing_events, ) defer.returnValue(missing_events) @@ -1946,7 +1777,7 @@ class FederationHandler(BaseHandler): logger.info("Missing auth: %s", missing_auth) # If we don't have all the auth events, we need to get them. try: - remote_auth_chain = yield self.replication_layer.get_event_auth( + remote_auth_chain = yield self.federation_client.get_event_auth( origin, event.room_id, event.event_id ) @@ -2051,9 +1882,10 @@ class FederationHandler(BaseHandler): break if do_resolution: + prev_state_ids = yield context.get_prev_state_ids(self.store) # 1. Get what we think is the auth chain. auth_ids = yield self.auth.compute_auth_events( - event, context.prev_state_ids + event, prev_state_ids ) local_auth_chain = yield self.store.get_auth_chain( auth_ids, include_given=True @@ -2061,7 +1893,7 @@ class FederationHandler(BaseHandler): try: # 2. Get remote difference. - result = yield self.replication_layer.query_auth( + result = yield self.federation_client.query_auth( origin, event.room_id, event.event_id, @@ -2143,21 +1975,34 @@ class FederationHandler(BaseHandler): k: a.event_id for k, a in iteritems(auth_events) if k != event_key } - context.current_state_ids = dict(context.current_state_ids) - context.current_state_ids.update(state_updates) - if context.delta_ids is not None: - context.delta_ids = dict(context.delta_ids) - context.delta_ids.update(state_updates) - context.prev_state_ids = dict(context.prev_state_ids) - context.prev_state_ids.update({ + current_state_ids = yield context.get_current_state_ids(self.store) + 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 = dict(prev_state_ids) + + prev_state_ids.update({ k: a.event_id for k, a in iteritems(auth_events) }) - context.state_group = yield self.store.store_state_group( + + # create a new state group as a delta from the existing one. + prev_group = context.state_group + state_group = yield self.store.store_state_group( event.event_id, event.room_id, - prev_group=context.prev_group, - delta_ids=context.delta_ids, - current_state_ids=context.current_state_ids, + prev_group=prev_group, + delta_ids=state_updates, + current_state_ids=current_state_ids, + ) + + yield context.update_state( + state_group=state_group, + current_state_ids=current_state_ids, + prev_state_ids=prev_state_ids, + prev_group=prev_group, + delta_ids=state_updates, ) @defer.inlineCallbacks @@ -2347,7 +2192,7 @@ class FederationHandler(BaseHandler): yield member_handler.send_membership_event(None, event, context) else: destinations = set(x.split(":", 1)[-1] for x in (sender_user_id, room_id)) - yield self.replication_layer.forward_third_party_invite( + yield self.federation_client.forward_third_party_invite( destinations, room_id, event_dict, @@ -2397,7 +2242,8 @@ class FederationHandler(BaseHandler): event.content["third_party_invite"]["signed"]["token"] ) original_invite = None - original_invite_id = context.prev_state_ids.get(key) + prev_state_ids = yield context.get_prev_state_ids(self.store) + original_invite_id = prev_state_ids.get(key) if original_invite_id: original_invite = yield self.store.get_event( original_invite_id, allow_none=True @@ -2439,7 +2285,8 @@ class FederationHandler(BaseHandler): signed = event.content["third_party_invite"]["signed"] token = signed["token"] - invite_event_id = context.prev_state_ids.get( + prev_state_ids = yield context.get_prev_state_ids(self.store) + invite_event_id = prev_state_ids.get( (EventTypes.ThirdPartyInvite, token,) ) diff --git a/synapse/handlers/initial_sync.py b/synapse/handlers/initial_sync.py index fb11716eb8..40e7580a61 100644 --- a/synapse/handlers/initial_sync.py +++ b/synapse/handlers/initial_sync.py @@ -148,13 +148,15 @@ class InitialSyncHandler(BaseHandler): try: if event.membership == Membership.JOIN: room_end_token = now_token.room_key - deferred_room_state = self.state_handler.get_current_state( - event.room_id + deferred_room_state = run_in_background( + self.state_handler.get_current_state, + event.room_id, ) elif event.membership == Membership.LEAVE: room_end_token = "s%d" % (event.stream_ordering,) - deferred_room_state = self.store.get_state_for_events( - [event.event_id], None + deferred_room_state = run_in_background( + self.store.get_state_for_events, + [event.event_id], None, ) deferred_room_state.addCallback( lambda states: states[event.event_id] @@ -387,19 +389,21 @@ class InitialSyncHandler(BaseHandler): receipts = [] defer.returnValue(receipts) - presence, receipts, (messages, token) = yield defer.gatherResults( - [ - run_in_background(get_presence), - run_in_background(get_receipts), - run_in_background( - self.store.get_recent_events_for_room, - room_id, - limit=limit, - end_token=now_token.room_key, - ) - ], - consumeErrors=True, - ).addErrback(unwrapFirstError) + presence, receipts, (messages, token) = yield make_deferred_yieldable( + defer.gatherResults( + [ + run_in_background(get_presence), + run_in_background(get_receipts), + run_in_background( + self.store.get_recent_events_for_room, + room_id, + limit=limit, + end_token=now_token.room_key, + ) + ], + consumeErrors=True, + ).addErrback(unwrapFirstError), + ) messages = yield filter_events_for_client( self.store, user_id, messages, is_peeking=is_peeking, diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index a39b852ceb..39d7724778 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -23,7 +23,6 @@ from canonicaljson import encode_canonical_json, json from twisted.internet import defer from twisted.internet.defer import succeed -from twisted.python.failure import Failure from synapse.api.constants import MAX_DEPTH, EventTypes, Membership from synapse.api.errors import AuthError, Codes, ConsentNotGivenError, SynapseError @@ -32,247 +31,26 @@ from synapse.crypto.event_signing import add_hashes_and_signatures from synapse.events.utils import serialize_event from synapse.events.validator import EventValidator from synapse.replication.http.send_event import send_event_to_master -from synapse.types import RoomAlias, RoomStreamToken, UserID -from synapse.util.async import Limiter, ReadWriteLock +from synapse.types import RoomAlias, UserID +from synapse.util.async import Linearizer from synapse.util.frozenutils import frozendict_json_encoder from synapse.util.logcontext import run_in_background from synapse.util.metrics import measure_func -from synapse.util.stringutils import random_string -from synapse.visibility import filter_events_for_client from ._base import BaseHandler logger = logging.getLogger(__name__) -class PurgeStatus(object): - """Object tracking the status of a purge request - - This class contains information on the progress of a purge request, for - return by get_purge_status. - - Attributes: - status (int): Tracks whether this request has completed. One of - STATUS_{ACTIVE,COMPLETE,FAILED} +class MessageHandler(object): + """Contains some read only APIs to get state about a room """ - STATUS_ACTIVE = 0 - STATUS_COMPLETE = 1 - STATUS_FAILED = 2 - - STATUS_TEXT = { - STATUS_ACTIVE: "active", - STATUS_COMPLETE: "complete", - STATUS_FAILED: "failed", - } - - def __init__(self): - self.status = PurgeStatus.STATUS_ACTIVE - - def asdict(self): - return { - "status": PurgeStatus.STATUS_TEXT[self.status] - } - - -class MessageHandler(BaseHandler): - def __init__(self, hs): - super(MessageHandler, self).__init__(hs) - self.hs = hs - self.state = hs.get_state_handler() + self.auth = hs.get_auth() self.clock = hs.get_clock() - - self.pagination_lock = ReadWriteLock() - self._purges_in_progress_by_room = set() - # map from purge id to PurgeStatus - self._purges_by_id = {} - - def start_purge_history(self, room_id, token, - delete_local_events=False): - """Start off a history purge on a room. - - Args: - room_id (str): The room to purge from - - token (str): topological token to delete events before - delete_local_events (bool): True to delete local events as well as - remote ones - - Returns: - str: unique ID for this purge transaction. - """ - if room_id in self._purges_in_progress_by_room: - raise SynapseError( - 400, - "History purge already in progress for %s" % (room_id, ), - ) - - purge_id = random_string(16) - - # we log the purge_id here so that it can be tied back to the - # request id in the log lines. - logger.info("[purge] starting purge_id %s", purge_id) - - self._purges_by_id[purge_id] = PurgeStatus() - run_in_background( - self._purge_history, - purge_id, room_id, token, delete_local_events, - ) - return purge_id - - @defer.inlineCallbacks - def _purge_history(self, purge_id, room_id, token, - delete_local_events): - """Carry out a history purge on a room. - - Args: - purge_id (str): The id for this purge - room_id (str): The room to purge from - token (str): topological token to delete events before - delete_local_events (bool): True to delete local events as well as - remote ones - - Returns: - Deferred - """ - self._purges_in_progress_by_room.add(room_id) - try: - with (yield self.pagination_lock.write(room_id)): - yield self.store.purge_history( - room_id, token, delete_local_events, - ) - logger.info("[purge] complete") - self._purges_by_id[purge_id].status = PurgeStatus.STATUS_COMPLETE - except Exception: - logger.error("[purge] failed: %s", Failure().getTraceback().rstrip()) - self._purges_by_id[purge_id].status = PurgeStatus.STATUS_FAILED - finally: - self._purges_in_progress_by_room.discard(room_id) - - # remove the purge from the list 24 hours after it completes - def clear_purge(): - del self._purges_by_id[purge_id] - self.hs.get_reactor().callLater(24 * 3600, clear_purge) - - def get_purge_status(self, purge_id): - """Get the current status of an active purge - - Args: - purge_id (str): purge_id returned by start_purge_history - - Returns: - PurgeStatus|None - """ - return self._purges_by_id.get(purge_id) - - @defer.inlineCallbacks - def get_messages(self, requester, room_id=None, pagin_config=None, - as_client_event=True, event_filter=None): - """Get messages in a room. - - Args: - requester (Requester): The user requesting messages. - room_id (str): The room they want messages from. - pagin_config (synapse.api.streams.PaginationConfig): The pagination - config rules to apply, if any. - as_client_event (bool): True to get events in client-server format. - event_filter (Filter): Filter to apply to results or None - Returns: - dict: Pagination API results - """ - user_id = requester.user.to_string() - - if pagin_config.from_token: - room_token = pagin_config.from_token.room_key - else: - pagin_config.from_token = ( - yield self.hs.get_event_sources().get_current_token_for_room( - room_id=room_id - ) - ) - room_token = pagin_config.from_token.room_key - - room_token = RoomStreamToken.parse(room_token) - - pagin_config.from_token = pagin_config.from_token.copy_and_replace( - "room_key", str(room_token) - ) - - source_config = pagin_config.get_source_config("room") - - with (yield self.pagination_lock.read(room_id)): - membership, member_event_id = yield self._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 - # requires that we have a topo token. - if room_token.topological: - max_topo = room_token.topological - else: - max_topo = yield self.store.get_max_topological_token( - room_id, room_token.stream - ) - - if membership == Membership.LEAVE: - # 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( - 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( - room_id, max_topo - ) - - events, next_key = yield self.store.paginate_room_events( - room_id=room_id, - from_key=source_config.from_key, - to_key=source_config.to_key, - direction=source_config.direction, - limit=source_config.limit, - event_filter=event_filter, - ) - - next_token = pagin_config.from_token.copy_and_replace( - "room_key", next_key - ) - - if not events: - defer.returnValue({ - "chunk": [], - "start": pagin_config.from_token.to_string(), - "end": next_token.to_string(), - }) - - if event_filter: - events = event_filter.filter(events) - - events = yield filter_events_for_client( - self.store, - user_id, - events, - is_peeking=(member_event_id is None), - ) - - time_now = self.clock.time_msec() - - chunk = { - "chunk": [ - serialize_event(e, time_now, as_client_event) - for e in events - ], - "start": pagin_config.from_token.to_string(), - "end": next_token.to_string(), - } - - defer.returnValue(chunk) + self.state = hs.get_state_handler() + self.store = hs.get_datastore() @defer.inlineCallbacks def get_room_data(self, user_id=None, room_id=None, @@ -286,12 +64,12 @@ class MessageHandler(BaseHandler): Raises: SynapseError if something went wrong. """ - membership, membership_event_id = yield self._check_in_room_or_world_readable( + membership, membership_event_id = yield self.auth.check_in_room_or_world_readable( room_id, user_id ) if membership == Membership.JOIN: - data = yield self.state_handler.get_current_state( + data = yield self.state.get_current_state( room_id, event_type, state_key ) elif membership == Membership.LEAVE: @@ -304,31 +82,6 @@ class MessageHandler(BaseHandler): defer.returnValue(data) @defer.inlineCallbacks - 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) - defer.returnValue((member_event.membership, member_event.event_id)) - return - except AuthError: - visibility = yield self.state_handler.get_current_state( - room_id, EventTypes.RoomHistoryVisibility, "" - ) - if ( - visibility and - visibility.content["history_visibility"] == "world_readable" - ): - defer.returnValue((Membership.JOIN, None)) - return - raise AuthError( - 403, "Guest access not allowed", errcode=Codes.GUEST_ACCESS_FORBIDDEN - ) - - @defer.inlineCallbacks def get_state_events(self, user_id, room_id, is_guest=False): """Retrieve all state events for a given room. If the user is joined to the room then return the current state. If the user has @@ -340,12 +93,12 @@ class MessageHandler(BaseHandler): Returns: A list of dicts representing state events. [{}, {}, {}] """ - membership, membership_event_id = yield self._check_in_room_or_world_readable( + membership, membership_event_id = yield self.auth.check_in_room_or_world_readable( room_id, user_id ) if membership == Membership.JOIN: - room_state = yield self.state_handler.get_current_state(room_id) + room_state = yield self.state.get_current_state(room_id) elif membership == Membership.LEAVE: room_state = yield self.store.get_state_for_events( [membership_event_id], None @@ -373,7 +126,7 @@ class MessageHandler(BaseHandler): if not requester.app_service: # We check AS auth after fetching the room membership, as it # requires us to pull out all joined members anyway. - membership, _ = yield self._check_in_room_or_world_readable( + membership, _ = yield self.auth.check_in_room_or_world_readable( room_id, user_id ) if membership != Membership.JOIN: @@ -427,7 +180,7 @@ class EventCreationHandler(object): # We arbitrarily limit concurrent event creation for a room to 5. # This is to stop us from diverging history *too* much. - self.limiter = Limiter(max_count=5) + self.limiter = Linearizer(max_count=5, name="room_event_creation_limit") self.action_generator = hs.get_action_generator() @@ -630,7 +383,8 @@ class EventCreationHandler(object): If so, returns the version of the event in context. Otherwise, returns None. """ - prev_event_id = context.prev_state_ids.get((event.type, event.state_key)) + prev_state_ids = yield context.get_prev_state_ids(self.store) + prev_event_id = prev_state_ids.get((event.type, event.state_key)) prev_event = yield self.store.get_event(prev_event_id, allow_none=True) if not prev_event: return @@ -752,8 +506,8 @@ class EventCreationHandler(object): event = builder.build() logger.debug( - "Created event %s with state: %s", - event.event_id, context.prev_state_ids, + "Created event %s", + event.event_id, ) defer.returnValue( @@ -806,8 +560,9 @@ class EventCreationHandler(object): # If we're a worker we need to hit out to the master. if self.config.worker_app: yield send_event_to_master( - self.hs.get_clock(), - self.http_client, + clock=self.hs.get_clock(), + store=self.store, + client=self.http_client, host=self.config.worker_replication_host, port=self.config.worker_replication_http_port, requester=requester, @@ -884,9 +639,11 @@ class EventCreationHandler(object): e.sender == event.sender ) + current_state_ids = yield context.get_current_state_ids(self.store) + state_to_include_ids = [ e_id - for k, e_id in iteritems(context.current_state_ids) + for k, e_id in iteritems(current_state_ids) if k[0] in self.hs.config.room_invite_state_types or k == (EventTypes.Member, event.sender) ] @@ -922,8 +679,9 @@ class EventCreationHandler(object): ) if event.type == EventTypes.Redaction: + prev_state_ids = yield context.get_prev_state_ids(self.store) auth_events_ids = yield self.auth.compute_auth_events( - event, context.prev_state_ids, for_verification=True, + event, prev_state_ids, for_verification=True, ) auth_events = yield self.store.get_events(auth_events_ids) auth_events = { @@ -943,11 +701,13 @@ class EventCreationHandler(object): "You don't have permission to redact events" ) - if event.type == EventTypes.Create and context.prev_state_ids: - raise AuthError( - 403, - "Changing the room create event is forbidden", - ) + if event.type == EventTypes.Create: + prev_state_ids = yield context.get_prev_state_ids(self.store) + if prev_state_ids: + raise AuthError( + 403, + "Changing the room create event is forbidden", + ) (event_stream_id, max_stream_id) = yield self.store.persist_event( event, context=context diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py new file mode 100644 index 0000000000..b2849783ed --- /dev/null +++ b/synapse/handlers/pagination.py @@ -0,0 +1,265 @@ +# -*- coding: utf-8 -*- +# Copyright 2014 - 2016 OpenMarket Ltd +# Copyright 2017 - 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 twisted.internet import defer +from twisted.python.failure import Failure + +from synapse.api.constants import Membership +from synapse.api.errors import SynapseError +from synapse.events.utils import serialize_event +from synapse.types import RoomStreamToken +from synapse.util.async import ReadWriteLock +from synapse.util.logcontext import run_in_background +from synapse.util.stringutils import random_string +from synapse.visibility import filter_events_for_client + +logger = logging.getLogger(__name__) + + +class PurgeStatus(object): + """Object tracking the status of a purge request + + This class contains information on the progress of a purge request, for + return by get_purge_status. + + Attributes: + status (int): Tracks whether this request has completed. One of + STATUS_{ACTIVE,COMPLETE,FAILED} + """ + + STATUS_ACTIVE = 0 + STATUS_COMPLETE = 1 + STATUS_FAILED = 2 + + STATUS_TEXT = { + STATUS_ACTIVE: "active", + STATUS_COMPLETE: "complete", + STATUS_FAILED: "failed", + } + + def __init__(self): + self.status = PurgeStatus.STATUS_ACTIVE + + def asdict(self): + return { + "status": PurgeStatus.STATUS_TEXT[self.status] + } + + +class PaginationHandler(object): + """Handles pagination and purge history requests. + + These are in the same handler due to the fact we need to block clients + paginating during a purge. + """ + + def __init__(self, hs): + self.hs = hs + self.auth = hs.get_auth() + self.store = hs.get_datastore() + self.clock = hs.get_clock() + + self.pagination_lock = ReadWriteLock() + self._purges_in_progress_by_room = set() + # map from purge id to PurgeStatus + self._purges_by_id = {} + + def start_purge_history(self, room_id, token, + delete_local_events=False): + """Start off a history purge on a room. + + Args: + room_id (str): The room to purge from + + token (str): topological token to delete events before + delete_local_events (bool): True to delete local events as well as + remote ones + + Returns: + str: unique ID for this purge transaction. + """ + if room_id in self._purges_in_progress_by_room: + raise SynapseError( + 400, + "History purge already in progress for %s" % (room_id, ), + ) + + purge_id = random_string(16) + + # we log the purge_id here so that it can be tied back to the + # request id in the log lines. + logger.info("[purge] starting purge_id %s", purge_id) + + self._purges_by_id[purge_id] = PurgeStatus() + run_in_background( + self._purge_history, + purge_id, room_id, token, delete_local_events, + ) + return purge_id + + @defer.inlineCallbacks + def _purge_history(self, purge_id, room_id, token, + delete_local_events): + """Carry out a history purge on a room. + + Args: + purge_id (str): The id for this purge + room_id (str): The room to purge from + token (str): topological token to delete events before + delete_local_events (bool): True to delete local events as well as + remote ones + + Returns: + Deferred + """ + self._purges_in_progress_by_room.add(room_id) + try: + with (yield self.pagination_lock.write(room_id)): + yield self.store.purge_history( + room_id, token, delete_local_events, + ) + logger.info("[purge] complete") + self._purges_by_id[purge_id].status = PurgeStatus.STATUS_COMPLETE + except Exception: + logger.error("[purge] failed: %s", Failure().getTraceback().rstrip()) + self._purges_by_id[purge_id].status = PurgeStatus.STATUS_FAILED + finally: + self._purges_in_progress_by_room.discard(room_id) + + # remove the purge from the list 24 hours after it completes + def clear_purge(): + del self._purges_by_id[purge_id] + self.hs.get_reactor().callLater(24 * 3600, clear_purge) + + def get_purge_status(self, purge_id): + """Get the current status of an active purge + + Args: + purge_id (str): purge_id returned by start_purge_history + + Returns: + PurgeStatus|None + """ + return self._purges_by_id.get(purge_id) + + @defer.inlineCallbacks + def get_messages(self, requester, room_id=None, pagin_config=None, + as_client_event=True, event_filter=None): + """Get messages in a room. + + Args: + requester (Requester): The user requesting messages. + room_id (str): The room they want messages from. + pagin_config (synapse.api.streams.PaginationConfig): The pagination + config rules to apply, if any. + as_client_event (bool): True to get events in client-server format. + event_filter (Filter): Filter to apply to results or None + Returns: + dict: Pagination API results + """ + user_id = requester.user.to_string() + + if pagin_config.from_token: + room_token = pagin_config.from_token.room_key + else: + pagin_config.from_token = ( + yield self.hs.get_event_sources().get_current_token_for_room( + room_id=room_id + ) + ) + room_token = pagin_config.from_token.room_key + + room_token = RoomStreamToken.parse(room_token) + + pagin_config.from_token = pagin_config.from_token.copy_and_replace( + "room_key", str(room_token) + ) + + source_config = pagin_config.get_source_config("room") + + with (yield self.pagination_lock.read(room_id)): + membership, member_event_id = yield 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 + # requires that we have a topo token. + if room_token.topological: + max_topo = room_token.topological + else: + max_topo = yield self.store.get_max_topological_token( + room_id, room_token.stream + ) + + if membership == Membership.LEAVE: + # 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( + 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( + room_id, max_topo + ) + + events, next_key = yield self.store.paginate_room_events( + room_id=room_id, + from_key=source_config.from_key, + to_key=source_config.to_key, + direction=source_config.direction, + limit=source_config.limit, + event_filter=event_filter, + ) + + next_token = pagin_config.from_token.copy_and_replace( + "room_key", next_key + ) + + if not events: + defer.returnValue({ + "chunk": [], + "start": pagin_config.from_token.to_string(), + "end": next_token.to_string(), + }) + + if event_filter: + events = event_filter.filter(events) + + events = yield filter_events_for_client( + self.store, + user_id, + events, + is_peeking=(member_event_id is None), + ) + + time_now = self.clock.time_msec() + + chunk = { + "chunk": [ + serialize_event(e, time_now, as_client_event) + for e in events + ], + "start": pagin_config.from_token.to_string(), + "end": next_token.to_string(), + } + + defer.returnValue(chunk) diff --git a/synapse/handlers/profile.py b/synapse/handlers/profile.py index 859f6d2b2e..cb5c6d587e 100644 --- a/synapse/handlers/profile.py +++ b/synapse/handlers/profile.py @@ -18,6 +18,7 @@ import logging from twisted.internet import defer from synapse.api.errors import AuthError, CodeMessageException, SynapseError +from synapse.metrics.background_process_metrics import run_as_background_process from synapse.types import UserID, get_domain_from_id from ._base import BaseHandler @@ -41,7 +42,7 @@ class ProfileHandler(BaseHandler): if hs.config.worker_app is None: self.clock.looping_call( - self._update_remote_profile_cache, self.PROFILE_UPDATE_MS, + self._start_update_remote_profile_cache, self.PROFILE_UPDATE_MS, ) @defer.inlineCallbacks @@ -254,6 +255,12 @@ class ProfileHandler(BaseHandler): room_id, str(e.message) ) + def _start_update_remote_profile_cache(self): + return run_as_background_process( + "Update remote profile", self._update_remote_profile_cache, + ) + + @defer.inlineCallbacks def _update_remote_profile_cache(self): """Called periodically to check profiles of remote users we haven't checked in a while. diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index f67512078b..7b7804d9b2 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -15,6 +15,7 @@ # limitations under the License. """Contains functions for performing events on rooms.""" +import itertools import logging import math import string @@ -24,7 +25,7 @@ from twisted.internet import defer from synapse.api.constants import EventTypes, JoinRules, RoomCreationPreset from synapse.api.errors import AuthError, Codes, StoreError, SynapseError -from synapse.types import RoomAlias, RoomID, RoomStreamToken, UserID +from synapse.types import RoomAlias, RoomID, RoomStreamToken, StreamToken, UserID from synapse.util import stringutils from synapse.visibility import filter_events_for_client @@ -395,9 +396,13 @@ class RoomCreationHandler(BaseHandler): ) -class RoomContextHandler(BaseHandler): +class RoomContextHandler(object): + def __init__(self, hs): + self.hs = hs + self.store = hs.get_datastore() + @defer.inlineCallbacks - def get_event_context(self, user, room_id, event_id, limit): + def get_event_context(self, user, room_id, event_id, limit, event_filter): """Retrieves events, pagination tokens and state around a given event in a room. @@ -407,6 +412,8 @@ class RoomContextHandler(BaseHandler): event_id (str) limit (int): The maximum number of events to return in total (excluding state). + event_filter (Filter|None): the filter to apply to the events returned + (excluding the target event_id) Returns: dict, or None if the event isn't found @@ -414,8 +421,6 @@ class RoomContextHandler(BaseHandler): before_limit = math.floor(limit / 2.) after_limit = limit - before_limit - now_token = yield self.hs.get_event_sources().get_current_token() - users = yield self.store.get_users_in_room(room_id) is_peeking = user.to_string() not in users @@ -441,7 +446,7 @@ class RoomContextHandler(BaseHandler): ) results = yield self.store.get_events_around( - room_id, event_id, before_limit, after_limit + room_id, event_id, before_limit, after_limit, event_filter ) results["events_before"] = yield filter_evts(results["events_before"]) @@ -453,16 +458,35 @@ class RoomContextHandler(BaseHandler): else: last_event_id = event_id + types = None + filtered_types = None + if event_filter and event_filter.lazy_load_members(): + members = set(ev.sender for ev in itertools.chain( + results["events_before"], + (results["event"],), + results["events_after"], + )) + filtered_types = [EventTypes.Member] + types = [(EventTypes.Member, member) for member in members] + + # XXX: why do we return the state as of the last event rather than the + # first? Shouldn't we be consistent with /sync? + # https://github.com/matrix-org/matrix-doc/issues/687 + state = yield self.store.get_state_for_events( - [last_event_id], None + [last_event_id], types, filtered_types=filtered_types, ) results["state"] = list(state[last_event_id].values()) - results["start"] = now_token.copy_and_replace( + # We use a dummy token here as we only care about the room portion of + # the token, which we replace. + token = StreamToken.START + + results["start"] = token.copy_and_replace( "room_key", results["start"] ).to_string() - results["end"] = now_token.copy_and_replace( + results["end"] = token.copy_and_replace( "room_key", results["end"] ).to_string() diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index 00f2e279bc..0d4a3f4677 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -201,7 +201,9 @@ class RoomMemberHandler(object): ratelimit=ratelimit, ) - prev_member_event_id = context.prev_state_ids.get( + prev_state_ids = yield context.get_prev_state_ids(self.store) + + prev_member_event_id = prev_state_ids.get( (EventTypes.Member, target.to_string()), None ) @@ -496,9 +498,10 @@ class RoomMemberHandler(object): if prev_event is not None: return + prev_state_ids = yield context.get_prev_state_ids(self.store) if event.membership == Membership.JOIN: if requester.is_guest: - guest_can_join = yield self._can_guest_join(context.prev_state_ids) + guest_can_join = yield self._can_guest_join(prev_state_ids) if not guest_can_join: # This should be an auth check, but guests are a local concept, # so don't really fit into the general auth process. @@ -517,7 +520,7 @@ class RoomMemberHandler(object): ratelimit=ratelimit, ) - prev_member_event_id = context.prev_state_ids.get( + prev_member_event_id = prev_state_ids.get( (EventTypes.Member, event.state_key), None ) @@ -705,6 +708,10 @@ class RoomMemberHandler(object): inviter_display_name = member_event.content.get("displayname", "") inviter_avatar_url = member_event.content.get("avatar_url", "") + # if user has no display name, default to their MXID + if not inviter_display_name: + inviter_display_name = user.to_string() + canonical_room_alias = "" canonical_alias_event = room_state.get((EventTypes.CanonicalAlias, "")) if canonical_alias_event: diff --git a/synapse/handlers/search.py b/synapse/handlers/search.py index 69ae9731d5..c464adbd0b 100644 --- a/synapse/handlers/search.py +++ b/synapse/handlers/search.py @@ -287,7 +287,7 @@ class SearchHandler(BaseHandler): contexts = {} for event in allowed_events: res = yield self.store.get_events_around( - event.room_id, event.event_id, before_limit, after_limit + event.room_id, event.event_id, before_limit, after_limit, ) logger.info( diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index c24e35362a..dff1f67dcb 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- -# Copyright 2015 - 2016 OpenMarket Ltd +# Copyright 2015, 2016 OpenMarket Ltd +# 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. @@ -25,6 +26,8 @@ from synapse.api.constants import EventTypes, Membership from synapse.push.clientformat import format_push_rules_for_user from synapse.types import RoomStreamToken from synapse.util.async import concurrently_execute +from synapse.util.caches.expiringcache import ExpiringCache +from synapse.util.caches.lrucache import LruCache from synapse.util.caches.response_cache import ResponseCache from synapse.util.logcontext import LoggingContext from synapse.util.metrics import Measure, measure_func @@ -32,6 +35,14 @@ from synapse.visibility import filter_events_for_client logger = logging.getLogger(__name__) +# Store the cache that tracks which lazy-loaded members have been sent to a given +# client for no more than 30 minutes. +LAZY_LOADED_MEMBERS_CACHE_MAX_AGE = 30 * 60 * 1000 + +# Remember the last 100 members we sent to a client for the purposes of +# avoiding redundantly sending the same lazy-loaded members to the client +LAZY_LOADED_MEMBERS_CACHE_MAX_SIZE = 100 + SyncConfig = collections.namedtuple("SyncConfig", [ "user", @@ -181,6 +192,12 @@ class SyncHandler(object): self.response_cache = ResponseCache(hs, "sync") self.state = hs.get_state_handler() + # ExpiringCache((User, Device)) -> LruCache(state_key => event_id) + self.lazy_loaded_members_cache = ExpiringCache( + "lazy_loaded_members_cache", self.clock, + max_len=0, expiry_ms=LAZY_LOADED_MEMBERS_CACHE_MAX_AGE, + ) + 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 @@ -416,29 +433,44 @@ class SyncHandler(object): )) @defer.inlineCallbacks - def get_state_after_event(self, event): + def get_state_after_event(self, event, types=None, filtered_types=None): """ Get the room state after the given event Args: event(synapse.events.EventBase): event of interest + types(list[(str, str|None)]|None): List of (type, state_key) tuples + which are used to filter the state fetched. If `state_key` is None, + all events are returned of the given type. + May be None, which matches any key. + filtered_types(list[str]|None): Only apply filtering via `types` to this + list of event types. Other types of events are returned unfiltered. + If None, `types` filtering is applied to all events. Returns: A Deferred map from ((type, state_key)->Event) """ - state_ids = yield self.store.get_state_ids_for_event(event.event_id) + state_ids = yield self.store.get_state_ids_for_event( + event.event_id, types, filtered_types=filtered_types, + ) if event.is_state(): state_ids = state_ids.copy() state_ids[(event.type, event.state_key)] = event.event_id defer.returnValue(state_ids) @defer.inlineCallbacks - def get_state_at(self, room_id, stream_position): + def get_state_at(self, room_id, stream_position, types=None, filtered_types=None): """ Get the room state at a particular stream position Args: room_id(str): room for which to get state stream_position(StreamToken): point at which to get state + types(list[(str, str|None)]|None): List of (type, state_key) tuples + which are used to filter the state fetched. If `state_key` is None, + all events are returned of the given type. + filtered_types(list[str]|None): Only apply filtering via `types` to this + list of event types. Other types of events are returned unfiltered. + If None, `types` filtering is applied to all events. Returns: A Deferred map from ((type, state_key)->Event) @@ -453,7 +485,9 @@ class SyncHandler(object): if last_events: last_event = last_events[-1] - state = yield self.get_state_after_event(last_event) + state = yield self.get_state_after_event( + last_event, types, filtered_types=filtered_types, + ) else: # no events in this room - so presumably no state @@ -485,59 +519,129 @@ class SyncHandler(object): # TODO(mjark) Check for new redactions in the state events. with Measure(self.clock, "compute_state_delta"): + + types = None + filtered_types = None + + lazy_load_members = sync_config.filter_collection.lazy_load_members() + include_redundant_members = ( + sync_config.filter_collection.include_redundant_members() + ) + + if lazy_load_members: + # We only request state for the members needed to display the + # timeline: + + types = [ + (EventTypes.Member, state_key) + for state_key in set( + event.sender # FIXME: we also care about invite targets etc. + for event in batch.events + ) + ] + + # only apply the filtering to room members + filtered_types = [EventTypes.Member] + + timeline_state = { + (event.type, event.state_key): event.event_id + for event in batch.events if event.is_state() + } + if full_state: if batch: current_state_ids = yield self.store.get_state_ids_for_event( - batch.events[-1].event_id + batch.events[-1].event_id, types=types, + filtered_types=filtered_types, ) state_ids = yield self.store.get_state_ids_for_event( - batch.events[0].event_id + batch.events[0].event_id, types=types, + filtered_types=filtered_types, ) + else: current_state_ids = yield self.get_state_at( - room_id, stream_position=now_token + room_id, stream_position=now_token, types=types, + filtered_types=filtered_types, ) state_ids = current_state_ids - timeline_state = { - (event.type, event.state_key): event.event_id - for event in batch.events if event.is_state() - } - state_ids = _calculate_state( timeline_contains=timeline_state, timeline_start=state_ids, previous={}, current=current_state_ids, + lazy_load_members=lazy_load_members, ) elif batch.limited: state_at_previous_sync = yield self.get_state_at( - room_id, stream_position=since_token + room_id, stream_position=since_token, types=types, + filtered_types=filtered_types, ) current_state_ids = yield self.store.get_state_ids_for_event( - batch.events[-1].event_id + batch.events[-1].event_id, types=types, + filtered_types=filtered_types, ) state_at_timeline_start = yield self.store.get_state_ids_for_event( - batch.events[0].event_id + batch.events[0].event_id, types=types, + filtered_types=filtered_types, ) - timeline_state = { - (event.type, event.state_key): event.event_id - for event in batch.events if event.is_state() - } - state_ids = _calculate_state( timeline_contains=timeline_state, timeline_start=state_at_timeline_start, previous=state_at_previous_sync, current=current_state_ids, + lazy_load_members=lazy_load_members, ) else: state_ids = {} + if lazy_load_members: + if types: + state_ids = yield self.store.get_state_ids_for_event( + batch.events[0].event_id, types=types, + filtered_types=filtered_types, + ) + + if lazy_load_members and not include_redundant_members: + cache_key = (sync_config.user.to_string(), sync_config.device_id) + cache = self.lazy_loaded_members_cache.get(cache_key) + if cache is None: + logger.debug("creating LruCache for %r", cache_key) + cache = LruCache(LAZY_LOADED_MEMBERS_CACHE_MAX_SIZE) + self.lazy_loaded_members_cache[cache_key] = cache + else: + logger.debug("found LruCache for %r", cache_key) + + # if it's a new sync sequence, then assume the client has had + # amnesia and doesn't want any recent lazy-loaded members + # de-duplicated. + if since_token is None: + logger.debug("clearing LruCache for %r", cache_key) + cache.clear() + else: + # only send members which aren't in our LruCache (either + # because they're new to this client or have been pushed out + # of the cache) + logger.debug("filtering state from %r...", state_ids) + state_ids = { + t: event_id + for t, event_id in state_ids.iteritems() + if cache.get(t[1]) != event_id + } + logger.debug("...to %r", state_ids) + + # add any member IDs we are about to send into our LruCache + for t, event_id in itertools.chain( + state_ids.items(), + timeline_state.items(), + ): + if t[0] == EventTypes.Member: + cache.set(t[1], event_id) state = {} if state_ids: @@ -1448,7 +1552,9 @@ def _action_has_highlight(actions): return False -def _calculate_state(timeline_contains, timeline_start, previous, current): +def _calculate_state( + timeline_contains, timeline_start, previous, current, lazy_load_members, +): """Works out what state to include in a sync response. Args: @@ -1457,6 +1563,9 @@ def _calculate_state(timeline_contains, timeline_start, previous, current): previous (dict): state at the end of the previous sync (or empty dict if this is an initial sync) current (dict): state at the end of the timeline + lazy_load_members (bool): whether to return members from timeline_start + or not. assumes that timeline_start has already been filtered to + include only the members the client needs to know about. Returns: dict @@ -1472,9 +1581,25 @@ def _calculate_state(timeline_contains, timeline_start, previous, current): } c_ids = set(e for e in current.values()) - tc_ids = set(e for e in timeline_contains.values()) - p_ids = set(e for e in previous.values()) ts_ids = set(e for e in timeline_start.values()) + p_ids = set(e for e in previous.values()) + tc_ids = set(e for e in timeline_contains.values()) + + # If we are lazyloading room members, we explicitly add the membership events + # for the senders in the timeline into the state block returned by /sync, + # as we may not have sent them to the client before. We find these membership + # events by filtering them out of timeline_start, which has already been filtered + # to only include membership events for the senders in the timeline. + # In practice, we can do this by removing them from the p_ids list, + # which is the list of relevant state we know we have already sent to the client. + # see https://github.com/matrix-org/synapse/pull/2970 + # /files/efcdacad7d1b7f52f879179701c7e0d9b763511f#r204732809 + + if lazy_load_members: + p_ids.difference_update( + e for t, e in timeline_start.iteritems() + if t[0] == EventTypes.Member + ) state_ids = ((c_ids | ts_ids) - p_ids) - tc_ids |