diff options
Diffstat (limited to 'synapse/handlers')
28 files changed, 4766 insertions, 1225 deletions
diff --git a/synapse/handlers/__init__.py b/synapse/handlers/__init__.py index 5ad408f549..8f8fd82eb0 100644 --- a/synapse/handlers/__init__.py +++ b/synapse/handlers/__init__.py @@ -17,10 +17,8 @@ from .register import RegistrationHandler from .room import ( RoomCreationHandler, RoomContextHandler, ) -from .room_member import RoomMemberHandler from .message import MessageHandler from .federation import FederationHandler -from .profile import ProfileHandler from .directory import DirectoryHandler from .admin import AdminHandler from .identity import IdentityHandler @@ -50,9 +48,7 @@ class Handlers(object): self.registration_handler = RegistrationHandler(hs) self.message_handler = MessageHandler(hs) self.room_creation_handler = RoomCreationHandler(hs) - self.room_member_handler = RoomMemberHandler(hs) self.federation_handler = FederationHandler(hs) - self.profile_handler = ProfileHandler(hs) self.directory_handler = DirectoryHandler(hs) self.admin_handler = AdminHandler(hs) self.identity_handler = IdentityHandler(hs) diff --git a/synapse/handlers/_base.py b/synapse/handlers/_base.py index e83adc8339..e089e66fde 100644 --- a/synapse/handlers/_base.py +++ b/synapse/handlers/_base.py @@ -53,7 +53,20 @@ class BaseHandler(object): self.event_builder_factory = hs.get_event_builder_factory() - def ratelimit(self, requester): + @defer.inlineCallbacks + def ratelimit(self, requester, update=True): + """Ratelimits requests. + + Args: + requester (Requester) + update (bool): Whether to record that a request is being processed. + Set to False when doing multiple checks for one request (e.g. + to check up front if we would reject the request), and set to + True for the last call for a given request. + + Raises: + LimitExceededError if the request should be ratelimited + """ time_now = self.clock.time() user_id = requester.user.to_string() @@ -67,10 +80,25 @@ class BaseHandler(object): if requester.app_service and not requester.app_service.is_rate_limited(): return + # Check if there is a per user override in the DB. + override = yield self.store.get_ratelimit_for_user(user_id) + if override: + # If overriden with a null Hz then ratelimiting has been entirely + # disabled for the user + if not override.messages_per_second: + return + + messages_per_second = override.messages_per_second + burst_count = override.burst_count + else: + messages_per_second = self.hs.config.rc_messages_per_second + burst_count = self.hs.config.rc_message_burst_count + allowed, time_allowed = self.ratelimiter.send_message( user_id, time_now, - msg_rate_hz=self.hs.config.rc_messages_per_second, - burst_count=self.hs.config.rc_message_burst_count, + msg_rate_hz=messages_per_second, + burst_count=burst_count, + update=update, ) if not allowed: raise LimitExceededError( @@ -130,7 +158,7 @@ class BaseHandler(object): # homeserver. requester = synapse.types.create_requester( target_user, is_guest=True) - handler = self.hs.get_handlers().room_member_handler + handler = self.hs.get_room_member_handler() yield handler.update_membership( requester, target_user, diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index 05af54d31b..b596f098fd 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -15,14 +15,21 @@ from twisted.internet import defer +import synapse from synapse.api.constants import EventTypes from synapse.util.metrics import Measure -from synapse.util.logcontext import preserve_fn, preserve_context_over_deferred +from synapse.util.logcontext import ( + make_deferred_yieldable, run_in_background, +) import logging logger = logging.getLogger(__name__) +metrics = synapse.metrics.get_metrics_for(__name__) + +events_processed_counter = metrics.register_counter("events_processed") + def log_failure(failure): logger.error( @@ -70,21 +77,25 @@ class ApplicationServicesHandler(object): with Measure(self.clock, "notify_interested_services"): self.is_processing = True try: - upper_bound = self.current_max limit = 100 while True: upper_bound, events = yield self.store.get_new_events_for_appservice( - upper_bound, limit + self.current_max, limit ) if not events: break + events_by_room = {} for event in events: + events_by_room.setdefault(event.room_id, []).append(event) + + @defer.inlineCallbacks + def handle_event(event): # Gather interested services services = yield self._get_services_for_event(event) if len(services) == 0: - continue # no services need notifying + return # no services need notifying # Do we know this user exists? If not, poke the user # query API for all services which match that user regex. @@ -100,14 +111,35 @@ class ApplicationServicesHandler(object): # Fork off pushes to these services for service in services: - preserve_fn(self.scheduler.submit_event_for_as)( - service, event - ) + self.scheduler.submit_event_for_as(service, event) + + @defer.inlineCallbacks + def handle_room_events(events): + for event in events: + yield handle_event(event) + + yield make_deferred_yieldable(defer.gatherResults([ + run_in_background(handle_room_events, evs) + for evs in events_by_room.itervalues() + ], consumeErrors=True)) yield self.store.set_appservice_last_pos(upper_bound) - if len(events) < limit: - break + now = self.clock.time_msec() + ts = yield self.store.get_received_ts(events[-1].event_id) + + synapse.metrics.event_processing_positions.set( + upper_bound, "appservice_sender", + ) + + events_processed_counter.inc_by(len(events)) + + synapse.metrics.event_processing_lag.set( + now - ts, "appservice_sender", + ) + synapse.metrics.event_processing_last_ts.set( + ts, "appservice_sender", + ) finally: self.is_processing = False @@ -163,8 +195,11 @@ class ApplicationServicesHandler(object): def query_3pe(self, kind, protocol, fields): services = yield self._get_services_for_3pn(protocol) - results = yield preserve_context_over_deferred(defer.DeferredList([ - preserve_fn(self.appservice_api.query_3pe)(service, kind, protocol, fields) + results = yield make_deferred_yieldable(defer.DeferredList([ + run_in_background( + self.appservice_api.query_3pe, + service, kind, protocol, fields, + ) for service in services ], consumeErrors=True)) @@ -225,11 +260,15 @@ class ApplicationServicesHandler(object): event based on the service regex. """ services = self.store.get_app_services() - interested_list = [ - s for s in services if ( - yield s.is_interested(event, self.store) - ) - ] + + # we can't use a list comprehension here. Since python 3, list + # comprehensions use a generator internally. This means you can't yield + # inside of a list comprehension anymore. + interested_list = [] + for s in services: + if (yield s.is_interested(event, self.store)): + interested_list.append(s) + defer.returnValue(interested_list) def _get_services_for_user(self, user_id): diff --git a/synapse/handlers/auth.py b/synapse/handlers/auth.py index fffba34383..a5365c4fe4 100644 --- a/synapse/handlers/auth.py +++ b/synapse/handlers/auth.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2014 - 2016 OpenMarket Ltd +# Copyright 2017 Vector Creations Ltd # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -12,14 +13,19 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. - -from twisted.internet import defer +from twisted.internet import defer, threads from ._base import BaseHandler from synapse.api.constants import LoginType +from synapse.api.errors import ( + AuthError, Codes, InteractiveAuthIncompleteError, LoginError, StoreError, + SynapseError, +) +from synapse.module_api import ModuleApi from synapse.types import UserID -from synapse.api.errors import AuthError, LoginError, Codes, StoreError, SynapseError from synapse.util.async import run_on_reactor +from synapse.util.caches.expiringcache import ExpiringCache +from synapse.util.logcontext import make_deferred_yieldable from twisted.web.client import PartialDownloadError @@ -44,18 +50,23 @@ class AuthHandler(BaseHandler): """ super(AuthHandler, self).__init__(hs) self.checkers = { - LoginType.PASSWORD: self._check_password_auth, LoginType.RECAPTCHA: self._check_recaptcha, LoginType.EMAIL_IDENTITY: self._check_email_identity, + LoginType.MSISDN: self._check_msisdn, LoginType.DUMMY: self._check_dummy_auth, } self.bcrypt_rounds = hs.config.bcrypt_rounds - self.sessions = {} - account_handler = _AccountHandler( - hs, check_user_exists=self.check_user_exists + # This is not a cache per se, but a store of all current sessions that + # expire after N hours + self.sessions = ExpiringCache( + cache_name="register_sessions", + clock=hs.get_clock(), + expiry_ms=self.SESSION_EXPIRE_MS, + reset_expiry_on_get=True, ) + account_handler = ModuleApi(hs, self) self.password_providers = [ module(config=config, account_handler=account_handler) for module, config in hs.config.password_providers @@ -64,39 +75,120 @@ class AuthHandler(BaseHandler): logger.info("Extra password_providers: %r", self.password_providers) self.hs = hs # FIXME better possibility to access registrationHandler later? - self.device_handler = hs.get_device_handler() self.macaroon_gen = hs.get_macaroon_generator() + self._password_enabled = hs.config.password_enabled + + # we keep this as a list despite the O(N^2) implication so that we can + # keep PASSWORD first and avoid confusing clients which pick the first + # type in the list. (NB that the spec doesn't require us to do so and + # clients which favour types that they don't understand over those that + # they do are technically broken) + login_types = [] + if self._password_enabled: + login_types.append(LoginType.PASSWORD) + for provider in self.password_providers: + if hasattr(provider, "get_supported_login_types"): + for t in provider.get_supported_login_types().keys(): + if t not in login_types: + login_types.append(t) + self._supported_login_types = login_types + + @defer.inlineCallbacks + def validate_user_via_ui_auth(self, requester, request_body, clientip): + """ + Checks that the user is who they claim to be, via a UI auth. + + This is used for things like device deletion and password reset where + the user already has a valid access token, but we want to double-check + that it isn't stolen by re-authenticating them. + + Args: + requester (Requester): The user, as given by the access token + + request_body (dict): The body of the request sent by the client + + clientip (str): The IP address of the client. + + Returns: + defer.Deferred[dict]: the parameters for this request (which may + have been given only in a previous call). + + Raises: + InteractiveAuthIncompleteError if the client has not yet completed + any of the permitted login flows + + AuthError if the client has completed a login flow, and it gives + a different user to `requester` + """ + + # build a list of supported flows + flows = [ + [login_type] for login_type in self._supported_login_types + ] + + result, params, _ = yield self.check_auth( + flows, request_body, clientip, + ) + + # find the completed login type + for login_type in self._supported_login_types: + if login_type not in result: + continue + + user_id = result[login_type] + break + else: + # this can't happen + raise Exception( + "check_auth returned True but no successful login type", + ) + + # check that the UI auth matched the access token + if user_id != requester.user.to_string(): + raise AuthError(403, "Invalid auth") + + defer.returnValue(params) @defer.inlineCallbacks def check_auth(self, flows, clientdict, clientip): """ Takes a dictionary sent by the client in the login / registration - protocol and handles the login flow. + protocol and handles the User-Interactive Auth flow. As a side effect, this function fills in the 'creds' key on the user's session with a map, which maps each auth-type (str) to the relevant identity authenticated by that auth-type (mostly str, but for captcha, bool). + If no auth flows have been completed successfully, raises an + InteractiveAuthIncompleteError. To handle this, you can use + synapse.rest.client.v2_alpha._base.interactive_auth_handler as a + decorator. + Args: flows (list): A list of login flows. Each flow is an ordered list of strings representing auth-types. At least one full flow must be completed in order for auth to be successful. + clientdict: The dictionary from the client root level, not the 'auth' key: this method prompts for auth if none is sent. + clientip (str): The IP address of the client. + Returns: - A tuple of (authed, dict, dict, session_id) where authed is true if - the client has successfully completed an auth flow. If it is true - the first dict contains the authenticated credentials of each stage. + defer.Deferred[dict, dict, str]: a deferred tuple of + (creds, params, session_id). + + 'creds' contains the authenticated credentials of each stage. - If authed is false, the first dictionary is the server response to - the login request and should be passed back to the client. + 'params' contains the parameters for this request (which may + have been given only in a previous call). - In either case, the second dict contains the parameters for this - request (which may have been given only in a previous call). + 'session_id' is the ID of this session, either passed in by the + client or assigned by this call - session_id is the ID of this session, either passed in by the client - or assigned by the call to check_auth + Raises: + InteractiveAuthIncompleteError if the client has not yet completed + all the stages in any of the permitted flows. """ authdict = None @@ -124,11 +216,8 @@ class AuthHandler(BaseHandler): clientdict = session['clientdict'] if not authdict: - defer.returnValue( - ( - False, self._auth_dict_for_flows(flows, session), - clientdict, session['id'] - ) + raise InteractiveAuthIncompleteError( + self._auth_dict_for_flows(flows, session), ) if 'creds' not in session: @@ -139,14 +228,12 @@ class AuthHandler(BaseHandler): errordict = {} if 'type' in authdict: login_type = authdict['type'] - if login_type not in self.checkers: - raise LoginError(400, "", Codes.UNRECOGNIZED) try: - result = yield self.checkers[login_type](authdict, clientip) + result = yield self._check_auth_dict(authdict, clientip) if result: creds[login_type] = result self._save_session(session) - except LoginError, e: + except LoginError as e: if login_type == LoginType.EMAIL_IDENTITY: # riot used to have a bug where it would request a new # validation token (thus sending a new email) each time it @@ -155,7 +242,7 @@ class AuthHandler(BaseHandler): # # Grandfather in the old behaviour for now to avoid # breaking old riot deployments. - raise e + raise # this step failed. Merge the error dict into the response # so that the client can have another go. @@ -172,12 +259,14 @@ class AuthHandler(BaseHandler): "Auth completed with creds: %r. Client dict has keys: %r", creds, clientdict.keys() ) - defer.returnValue((True, creds, clientdict, session['id'])) + defer.returnValue((creds, clientdict, session['id'])) ret = self._auth_dict_for_flows(flows, session) ret['completed'] = creds.keys() ret.update(errordict) - defer.returnValue((False, ret, clientdict, session['id'])) + raise InteractiveAuthIncompleteError( + ret, + ) @defer.inlineCallbacks def add_oob_auth(self, stagetype, authdict, clientip): @@ -249,16 +338,37 @@ class AuthHandler(BaseHandler): sess = self._get_session_info(session_id) return sess.setdefault('serverdict', {}).get(key, default) - def _check_password_auth(self, authdict, _): - if "user" not in authdict or "password" not in authdict: - raise LoginError(400, "", Codes.MISSING_PARAM) + @defer.inlineCallbacks + def _check_auth_dict(self, authdict, clientip): + """Attempt to validate the auth dict provided by a client - user_id = authdict["user"] - password = authdict["password"] - if not user_id.startswith('@'): - user_id = UserID.create(user_id, self.hs.hostname).to_string() + Args: + authdict (object): auth dict provided by the client + clientip (str): IP address of the client - return self._check_password(user_id, password) + Returns: + Deferred: result of the stage verification. + + Raises: + StoreError if there was a problem accessing the database + SynapseError if there was a problem with the request + LoginError if there was an authentication problem. + """ + login_type = authdict['type'] + checker = self.checkers.get(login_type) + if checker is not None: + res = yield checker(authdict, clientip) + defer.returnValue(res) + + # build a v1-login-style dict out of the authdict and fall back to the + # v1 code + user_id = authdict.get("user") + + if user_id is None: + raise SynapseError(400, "", Codes.MISSING_PARAM) + + (canonical_id, callback) = yield self.validate_login(user_id, authdict) + defer.returnValue(canonical_id) @defer.inlineCallbacks def _check_recaptcha(self, authdict, clientip): @@ -307,31 +417,47 @@ class AuthHandler(BaseHandler): defer.returnValue(True) raise LoginError(401, "", errcode=Codes.UNAUTHORIZED) - @defer.inlineCallbacks def _check_email_identity(self, authdict, _): + return self._check_threepid('email', authdict) + + def _check_msisdn(self, authdict, _): + return self._check_threepid('msisdn', authdict) + + @defer.inlineCallbacks + def _check_dummy_auth(self, authdict, _): + yield run_on_reactor() + defer.returnValue(True) + + @defer.inlineCallbacks + def _check_threepid(self, medium, authdict): yield run_on_reactor() if 'threepid_creds' not in authdict: raise LoginError(400, "Missing threepid_creds", Codes.MISSING_PARAM) threepid_creds = authdict['threepid_creds'] + identity_handler = self.hs.get_handlers().identity_handler - logger.info("Getting validated threepid. threepidcreds: %r" % (threepid_creds,)) + logger.info("Getting validated threepid. threepidcreds: %r", (threepid_creds,)) threepid = yield identity_handler.threepid_from_creds(threepid_creds) if not threepid: raise LoginError(401, "", errcode=Codes.UNAUTHORIZED) + if threepid['medium'] != medium: + raise LoginError( + 401, + "Expecting threepid of type '%s', got '%s'" % ( + medium, threepid['medium'], + ), + errcode=Codes.UNAUTHORIZED + ) + threepid['threepid_creds'] = authdict['threepid_creds'] defer.returnValue(threepid) - @defer.inlineCallbacks - def _check_dummy_auth(self, authdict, _): - yield run_on_reactor() - defer.returnValue(True) - def _get_params_recaptcha(self): return {"public_key": self.hs.config.recaptcha_public_key} @@ -371,26 +497,8 @@ class AuthHandler(BaseHandler): return self.sessions[session_id] - def validate_password_login(self, user_id, password): - """ - Authenticates the user with their username and password. - - Used only by the v1 login API. - - Args: - user_id (str): complete @user:id - password (str): Password - Returns: - defer.Deferred: (str) canonical user id - Raises: - StoreError if there was a problem accessing the database - LoginError if there was an authentication problem. - """ - return self._check_password(user_id, password) - @defer.inlineCallbacks - def get_access_token_for_user_id(self, user_id, device_id=None, - initial_display_name=None): + def get_access_token_for_user_id(self, user_id, device_id=None): """ Creates a new access token for the user with the given user ID. @@ -404,13 +512,10 @@ class AuthHandler(BaseHandler): device_id (str|None): the device ID to associate with the tokens. None to leave the tokens unassociated with a device (deprecated: we should always have a device ID) - initial_display_name (str): display name to associate with the - device if it needs re-registering Returns: The access token for the user's session. Raises: StoreError if there was a problem storing the token. - LoginError if there was an authentication problem. """ logger.info("Logging in user %s on device %s", user_id, device_id) access_token = yield self.issue_access_token(user_id, device_id) @@ -420,9 +525,11 @@ class AuthHandler(BaseHandler): # really don't want is active access_tokens without a record of the # device, so we double-check it here. if device_id is not None: - yield self.device_handler.check_device_registered( - user_id, device_id, initial_display_name - ) + try: + yield self.store.get_device(user_id, device_id) + except StoreError: + yield self.store.delete_access_token(access_token) + raise StoreError(400, "Login raced against device deletion") defer.returnValue(access_token) @@ -474,29 +581,115 @@ class AuthHandler(BaseHandler): ) defer.returnValue(result) + def get_supported_login_types(self): + """Get a the login types supported for the /login API + + By default this is just 'm.login.password' (unless password_enabled is + False in the config file), but password auth providers can provide + other login types. + + Returns: + Iterable[str]: login types + """ + return self._supported_login_types + @defer.inlineCallbacks - def _check_password(self, user_id, password): - """Authenticate a user against the LDAP and local databases. + def validate_login(self, username, login_submission): + """Authenticates the user for the /login API - user_id is checked case insensitively against the local database, but - will throw if there are multiple inexact matches. + Also used by the user-interactive auth flow to validate + m.login.password auth types. Args: - user_id (str): complete @user:id + username (str): username supplied by the user + login_submission (dict): the whole of the login submission + (including 'type' and other relevant fields) Returns: - (str) the canonical_user_id + Deferred[str, func]: canonical user id, and optional callback + to be called once the access token and device id are issued Raises: - LoginError if login fails + StoreError if there was a problem accessing the database + SynapseError if there was a problem with the request + LoginError if there was an authentication problem. """ + + if username.startswith('@'): + qualified_user_id = username + else: + qualified_user_id = UserID( + username, self.hs.hostname + ).to_string() + + login_type = login_submission.get("type") + known_login_type = False + + # special case to check for "password" for the check_password interface + # for the auth providers + password = login_submission.get("password") + if login_type == LoginType.PASSWORD: + if not self._password_enabled: + raise SynapseError(400, "Password login has been disabled.") + if not password: + raise SynapseError(400, "Missing parameter: password") + for provider in self.password_providers: - is_valid = yield provider.check_password(user_id, password) - if is_valid: - defer.returnValue(user_id) + if (hasattr(provider, "check_password") + and login_type == LoginType.PASSWORD): + known_login_type = True + is_valid = yield provider.check_password( + qualified_user_id, password, + ) + if is_valid: + defer.returnValue((qualified_user_id, None)) + + if (not hasattr(provider, "get_supported_login_types") + or not hasattr(provider, "check_auth")): + # this password provider doesn't understand custom login types + continue + + supported_login_types = provider.get_supported_login_types() + if login_type not in supported_login_types: + # this password provider doesn't understand this login type + continue + + known_login_type = True + login_fields = supported_login_types[login_type] + + missing_fields = [] + login_dict = {} + for f in login_fields: + if f not in login_submission: + missing_fields.append(f) + else: + login_dict[f] = login_submission[f] + if missing_fields: + raise SynapseError( + 400, "Missing parameters for login type %s: %s" % ( + login_type, + missing_fields, + ), + ) - canonical_user_id = yield self._check_local_password(user_id, password) + result = yield provider.check_auth( + username, login_type, login_dict, + ) + if result: + if isinstance(result, str): + result = (result, None) + defer.returnValue(result) + + if login_type == LoginType.PASSWORD: + known_login_type = True + + canonical_user_id = yield self._check_local_password( + qualified_user_id, password, + ) - if canonical_user_id: - defer.returnValue(canonical_user_id) + if canonical_user_id: + defer.returnValue((canonical_user_id, None)) + + if not known_login_type: + raise SynapseError(400, "Unknown login type %s" % login_type) # unknown username or invalid password. We raise a 403 here, but note # that if we're doing user-interactive login, it turns all LoginErrors @@ -522,7 +715,7 @@ class AuthHandler(BaseHandler): if not lookupres: defer.returnValue(None) (user_id, password_hash) = lookupres - result = self.validate_hash(password, password_hash) + result = yield self.validate_hash(password, password_hash) if not result: logger.warn("Failed password login for user %s", user_id) defer.returnValue(None) @@ -546,22 +739,65 @@ class AuthHandler(BaseHandler): raise AuthError(403, "Invalid token", errcode=Codes.FORBIDDEN) @defer.inlineCallbacks - def set_password(self, user_id, newpassword, requester=None): - password_hash = self.hash(newpassword) + def delete_access_token(self, access_token): + """Invalidate a single access token - except_access_token_id = requester.access_token_id if requester else None + Args: + access_token (str): access token to be deleted - try: - yield self.store.user_set_password_hash(user_id, password_hash) - except StoreError as e: - if e.code == 404: - raise SynapseError(404, "Unknown user", Codes.NOT_FOUND) - raise e - yield self.store.user_delete_access_tokens( - user_id, except_access_token_id + Returns: + Deferred + """ + user_info = yield self.auth.get_user_by_access_token(access_token) + yield self.store.delete_access_token(access_token) + + # see if any of our auth providers want to know about this + for provider in self.password_providers: + if hasattr(provider, "on_logged_out"): + yield provider.on_logged_out( + user_id=str(user_info["user"]), + device_id=user_info["device_id"], + access_token=access_token, + ) + + # delete pushers associated with this access token + if user_info["token_id"] is not None: + yield self.hs.get_pusherpool().remove_pushers_by_access_token( + str(user_info["user"]), (user_info["token_id"], ) + ) + + @defer.inlineCallbacks + def delete_access_tokens_for_user(self, user_id, except_token_id=None, + device_id=None): + """Invalidate access tokens belonging to a user + + Args: + user_id (str): ID of user the tokens belong to + except_token_id (str|None): access_token ID which should *not* be + deleted + device_id (str|None): ID of device the tokens are associated with. + If None, tokens associated with any device (or no device) will + be deleted + Returns: + Deferred + """ + tokens_and_devices = yield self.store.user_delete_access_tokens( + user_id, except_token_id=except_token_id, device_id=device_id, ) - yield self.hs.get_pusherpool().remove_pushers_by_user( - user_id, except_access_token_id + + # see if any of our auth providers want to know about this + for provider in self.password_providers: + if hasattr(provider, "on_logged_out"): + for token, token_id, device_id in tokens_and_devices: + yield provider.on_logged_out( + user_id=user_id, + device_id=device_id, + access_token=token, + ) + + # delete pushers associated with the access tokens + yield self.hs.get_pusherpool().remove_pushers_by_access_token( + user_id, (token_id for _, token_id, _ in tokens_and_devices), ) @defer.inlineCallbacks @@ -599,16 +835,6 @@ class AuthHandler(BaseHandler): logger.debug("Saving session %s", session) session["last_used"] = self.hs.get_clock().time_msec() self.sessions[session["id"]] = session - self._prune_sessions() - - def _prune_sessions(self): - for sid, sess in self.sessions.items(): - last_used = 0 - if 'last_used' in sess: - last_used = sess['last_used'] - now = self.hs.get_clock().time_msec() - if last_used < now - AuthHandler.SESSION_EXPIRE_MS: - del self.sessions[sid] def hash(self, password): """Computes a secure hash of password. @@ -617,10 +843,13 @@ class AuthHandler(BaseHandler): password (str): Password to hash. Returns: - Hashed password (str). + Deferred(str): Hashed password. """ - return bcrypt.hashpw(password.encode('utf8') + self.hs.config.password_pepper, - bcrypt.gensalt(self.bcrypt_rounds)) + def _do_hash(): + return bcrypt.hashpw(password.encode('utf8') + self.hs.config.password_pepper, + bcrypt.gensalt(self.bcrypt_rounds)) + + return make_deferred_yieldable(threads.deferToThread(_do_hash)) def validate_hash(self, password, stored_hash): """Validates that self.hash(password) == stored_hash. @@ -630,13 +859,19 @@ class AuthHandler(BaseHandler): stored_hash (str): Expected hash value. Returns: - Whether self.hash(password) == stored_hash (bool). + Deferred(bool): Whether self.hash(password) == stored_hash. """ + + def _do_validate_hash(): + return bcrypt.checkpw( + password.encode('utf8') + self.hs.config.password_pepper, + stored_hash.encode('utf8') + ) + if stored_hash: - return bcrypt.hashpw(password.encode('utf8') + self.hs.config.password_pepper, - stored_hash.encode('utf8')) == stored_hash + return make_deferred_yieldable(threads.deferToThread(_do_validate_hash)) else: - return False + return defer.succeed(False) class MacaroonGeneartor(object): @@ -679,30 +914,3 @@ class MacaroonGeneartor(object): macaroon.add_first_party_caveat("gen = 1") macaroon.add_first_party_caveat("user_id = %s" % (user_id,)) return macaroon - - -class _AccountHandler(object): - """A proxy object that gets passed to password auth providers so they - can register new users etc if necessary. - """ - def __init__(self, hs, check_user_exists): - self.hs = hs - - self._check_user_exists = check_user_exists - - def check_user_exists(self, user_id): - """Check if user exissts. - - Returns: - Deferred(bool) - """ - return self._check_user_exists(user_id) - - def register(self, localpart): - """Registers a new user with given localpart - - Returns: - Deferred: a 2-tuple of (user_id, access_token) - """ - reg = self.hs.get_handlers().registration_handler - return reg.register(localpart=localpart) diff --git a/synapse/handlers/deactivate_account.py b/synapse/handlers/deactivate_account.py new file mode 100644 index 0000000000..b1d3814909 --- /dev/null +++ b/synapse/handlers/deactivate_account.py @@ -0,0 +1,52 @@ +# -*- coding: utf-8 -*- +# Copyright 2017 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. +from twisted.internet import defer + +from ._base import BaseHandler + +import logging + +logger = logging.getLogger(__name__) + + +class DeactivateAccountHandler(BaseHandler): + """Handler which deals with deactivating user accounts.""" + def __init__(self, hs): + super(DeactivateAccountHandler, self).__init__(hs) + self._auth_handler = hs.get_auth_handler() + self._device_handler = hs.get_device_handler() + + @defer.inlineCallbacks + def deactivate_account(self, user_id): + """Deactivate a user's account + + Args: + user_id (str): ID of user to be deactivated + + Returns: + Deferred + """ + # FIXME: Theoretically there is a race here wherein user resets + # password using threepid. + + # first 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) + # then delete any remaining access tokens which weren't associated with + # a device. + yield self._auth_handler.delete_access_tokens_for_user(user_id) + + yield self.store.user_delete_threepids(user_id) + yield self.store.user_set_password_hash(user_id, None) diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index ca7137f315..f7457a7082 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -14,8 +14,11 @@ # limitations under the License. from synapse.api import errors from synapse.api.constants import EventTypes +from synapse.api.errors import FederationDeniedError from synapse.util import stringutils from synapse.util.async import Linearizer +from synapse.util.caches.expiringcache import ExpiringCache +from synapse.util.retryutils import NotRetryingDestination from synapse.util.metrics import measure_func from synapse.types import get_domain_from_id, RoomStreamToken from twisted.internet import defer @@ -32,14 +35,17 @@ class DeviceHandler(BaseHandler): self.hs = hs self.state = hs.get_state_handler() + self._auth_handler = hs.get_auth_handler() self.federation_sender = hs.get_federation_sender() - self.federation = hs.get_replication_layer() - self._remote_edue_linearizer = Linearizer(name="remote_device_list") - self.federation.register_edu_handler( - "m.device_list_update", self._incoming_device_list_update, + self._edu_updater = DeviceListEduUpdater(hs, self) + + federation_registry = hs.get_federation_registry() + + federation_registry.register_edu_handler( + "m.device_list_update", self._edu_updater.incoming_device_list_update, ) - self.federation.register_query_handler( + federation_registry.register_query_handler( "user_devices", self.on_federation_query_user_devices, ) @@ -103,7 +109,7 @@ class DeviceHandler(BaseHandler): device_map = yield self.store.get_devices_by_user(user_id) ips = yield self.store.get_last_client_ip_by_device( - devices=((user_id, device_id) for device_id in device_map.keys()) + user_id, device_id=None ) devices = device_map.values() @@ -130,7 +136,7 @@ class DeviceHandler(BaseHandler): except errors.StoreError: raise errors.NotFoundError ips = yield self.store.get_last_client_ip_by_device( - devices=((user_id, device_id),) + user_id, device_id, ) _update_device_from_client_ips(device, ips) defer.returnValue(device) @@ -149,16 +155,15 @@ class DeviceHandler(BaseHandler): try: yield self.store.delete_device(user_id, device_id) - except errors.StoreError, e: + except errors.StoreError as e: if e.code == 404: # no match pass else: raise - yield self.store.user_delete_access_tokens( + yield self._auth_handler.delete_access_tokens_for_user( user_id, device_id=device_id, - delete_refresh_tokens=True, ) yield self.store.delete_e2e_keys_by_device( @@ -168,6 +173,57 @@ class DeviceHandler(BaseHandler): yield self.notify_device_update(user_id, [device_id]) @defer.inlineCallbacks + def delete_all_devices_for_user(self, user_id, except_device_id=None): + """Delete all of the user's devices + + Args: + user_id (str): + except_device_id (str|None): optional device id which should not + be deleted + + Returns: + defer.Deferred: + """ + device_map = yield self.store.get_devices_by_user(user_id) + device_ids = device_map.keys() + if except_device_id is not None: + device_ids = [d for d in device_ids if d != except_device_id] + yield self.delete_devices(user_id, device_ids) + + @defer.inlineCallbacks + def delete_devices(self, user_id, device_ids): + """ Delete several devices + + Args: + user_id (str): + device_ids (List[str]): The list of device IDs to delete + + Returns: + defer.Deferred: + """ + + try: + yield self.store.delete_devices(user_id, device_ids) + except errors.StoreError as e: + if e.code == 404: + # no match + pass + else: + raise + + # Delete access tokens and e2e keys for each device. Not optimised as it is not + # considered as part of a critical path. + for device_id in device_ids: + yield self._auth_handler.delete_access_tokens_for_user( + user_id, device_id=device_id, + ) + yield self.store.delete_e2e_keys_by_device( + user_id=user_id, device_id=device_id + ) + + yield self.notify_device_update(user_id, device_ids) + + @defer.inlineCallbacks def update_device(self, user_id, device_id, content): """ Update the given device @@ -187,7 +243,7 @@ class DeviceHandler(BaseHandler): new_display_name=content.get("display_name") ) yield self.notify_device_update(user_id, [device_id]) - except errors.StoreError, e: + except errors.StoreError as e: if e.code == 404: raise errors.NotFoundError() else: @@ -212,8 +268,7 @@ class DeviceHandler(BaseHandler): user_id, device_ids, list(hosts) ) - rooms = yield self.store.get_rooms_for_user(user_id) - room_ids = [r.room_id for r in rooms] + room_ids = yield self.store.get_rooms_for_user(user_id) yield self.notifier.on_new_event( "device_list_key", position, rooms=room_ids, @@ -234,8 +289,9 @@ class DeviceHandler(BaseHandler): user_id (str) from_token (StreamToken) """ - rooms = yield self.store.get_rooms_for_user(user_id) - room_ids = set(r.room_id for r in rooms) + now_token = yield self.hs.get_event_sources().get_current_token() + + room_ids = yield self.store.get_rooms_for_user(user_id) # First we check if any devices have changed changed = yield self.store.get_user_whose_devices_changed( @@ -245,11 +301,30 @@ class DeviceHandler(BaseHandler): # Then work out if any users have since joined rooms_changed = self.store.get_rooms_that_changed(room_ids, from_token.room_key) + member_events = yield self.store.get_membership_changes_for_user( + user_id, from_token.room_key, now_token.room_key + ) + rooms_changed.update(event.room_id for event in member_events) + stream_ordering = RoomStreamToken.parse_stream_token( - from_token.room_key).stream + from_token.room_key + ).stream possibly_changed = set(changed) + possibly_left = set() for room_id in rooms_changed: + current_state_ids = yield self.store.get_current_state_ids(room_id) + + # The user may have left the room + # TODO: Check if they actually did or if we were just invited. + if room_id not in room_ids: + for key, event_id in current_state_ids.iteritems(): + etype, state_key = key + if etype != EventTypes.Member: + continue + possibly_left.add(state_key) + continue + # Fetch the current state at the time. try: event_ids = yield self.store.get_forward_extremeties_for_room( @@ -260,8 +335,6 @@ class DeviceHandler(BaseHandler): # ordering: treat it the same as a new room event_ids = [] - current_state_ids = yield self.state.get_current_state_ids(room_id) - # special-case for an empty prev state: include all members # in the changed list if not event_ids: @@ -272,9 +345,25 @@ class DeviceHandler(BaseHandler): possibly_changed.add(state_key) continue + current_member_id = current_state_ids.get((EventTypes.Member, user_id)) + if not current_member_id: + continue + # mapping from event_id -> state_dict prev_state_ids = yield self.store.get_state_ids_for_events(event_ids) + # Check if we've joined the room? If so we just blindly add all the users to + # the "possibly changed" users. + for state_dict in prev_state_ids.itervalues(): + member_event = state_dict.get((EventTypes.Member, user_id), None) + if not member_event or member_event != current_member_id: + for key, event_id in current_state_ids.iteritems(): + etype, state_key = key + if etype != EventTypes.Member: + continue + possibly_changed.add(state_key) + break + # If there has been any change in membership, include them in the # possibly changed list. We'll check if they are joined below, # and we're not toooo worried about spuriously adding users. @@ -285,94 +374,208 @@ class DeviceHandler(BaseHandler): # check if this member has changed since any of the extremities # at the stream_ordering, and add them to the list if so. - for state_dict in prev_state_ids.values(): + for state_dict in prev_state_ids.itervalues(): prev_event_id = state_dict.get(key, None) if not prev_event_id or prev_event_id != event_id: - possibly_changed.add(state_key) + if state_key != user_id: + possibly_changed.add(state_key) break - users_who_share_room = yield self.store.get_users_who_share_room_with_user( - user_id - ) + if possibly_changed or possibly_left: + users_who_share_room = yield self.store.get_users_who_share_room_with_user( + user_id + ) - # Take the intersection of the users whose devices may have changed - # and those that actually still share a room with the user - defer.returnValue(users_who_share_room & possibly_changed) + # Take the intersection of the users whose devices may have changed + # and those that actually still share a room with the user + possibly_joined = possibly_changed & users_who_share_room + possibly_left = (possibly_changed | possibly_left) - users_who_share_room + else: + possibly_joined = [] + possibly_left = [] + + defer.returnValue({ + "changed": list(possibly_joined), + "left": list(possibly_left), + }) - @measure_func("_incoming_device_list_update") @defer.inlineCallbacks - def _incoming_device_list_update(self, origin, edu_content): - user_id = edu_content["user_id"] - device_id = edu_content["device_id"] - stream_id = edu_content["stream_id"] - prev_ids = edu_content.get("prev_id", []) + def on_federation_query_user_devices(self, user_id): + stream_id, devices = yield self.store.get_devices_with_keys_by_user(user_id) + defer.returnValue({ + "user_id": user_id, + "stream_id": stream_id, + "devices": devices, + }) + + @defer.inlineCallbacks + def user_left_room(self, user, room_id): + user_id = user.to_string() + room_ids = yield self.store.get_rooms_for_user(user_id) + if not room_ids: + # We no longer share rooms with this user, so we'll no longer + # receive device updates. Mark this in DB. + yield self.store.mark_remote_user_device_list_as_unsubscribed(user_id) + + +def _update_device_from_client_ips(device, client_ips): + ip = client_ips.get((device["user_id"], device["device_id"]), {}) + device.update({ + "last_seen_ts": ip.get("last_seen"), + "last_seen_ip": ip.get("ip"), + }) + + +class DeviceListEduUpdater(object): + "Handles incoming device list updates from federation and updates the DB" + + def __init__(self, hs, device_handler): + self.store = hs.get_datastore() + self.federation = hs.get_federation_client() + self.clock = hs.get_clock() + self.device_handler = device_handler + + self._remote_edu_linearizer = Linearizer(name="remote_device_list") + + # user_id -> list of updates waiting to be handled. + self._pending_updates = {} + + # Recently seen stream ids. We don't bother keeping these in the DB, + # but they're useful to have them about to reduce the number of spurious + # resyncs. + self._seen_updates = ExpiringCache( + cache_name="device_update_edu", + clock=self.clock, + max_len=10000, + expiry_ms=30 * 60 * 1000, + iterable=True, + ) + + @defer.inlineCallbacks + def incoming_device_list_update(self, origin, edu_content): + """Called on incoming device list update from federation. Responsible + for parsing the EDU and adding to pending updates list. + """ + + user_id = edu_content.pop("user_id") + device_id = edu_content.pop("device_id") + stream_id = str(edu_content.pop("stream_id")) # They may come as ints + prev_ids = edu_content.pop("prev_id", []) + prev_ids = [str(p) for p in prev_ids] # They may come as ints if get_domain_from_id(user_id) != origin: # TODO: Raise? logger.warning("Got device list update edu for %r from %r", user_id, origin) return - rooms = yield self.store.get_rooms_for_user(user_id) - if not rooms: + room_ids = yield self.store.get_rooms_for_user(user_id) + if not room_ids: # We don't share any rooms with this user. Ignore update, as we # probably won't get any further updates. return - with (yield self._remote_edue_linearizer.queue(user_id)): - # If the prev id matches whats in our cache table, then we don't need - # to resync the users device list, otherwise we do. - resync = True - if len(prev_ids) == 1: - extremity = yield self.store.get_device_list_last_stream_id_for_remote( - user_id - ) - logger.info("Extrem: %r, prev_ids: %r", extremity, prev_ids) - if str(extremity) == str(prev_ids[0]): - resync = False + self._pending_updates.setdefault(user_id, []).append( + (device_id, stream_id, prev_ids, edu_content) + ) + + yield self._handle_device_updates(user_id) + + @measure_func("_incoming_device_list_update") + @defer.inlineCallbacks + def _handle_device_updates(self, user_id): + "Actually handle pending updates." + + with (yield self._remote_edu_linearizer.queue(user_id)): + pending_updates = self._pending_updates.pop(user_id, []) + if not pending_updates: + # This can happen since we batch updates + return + + # Given a list of updates we check if we need to resync. This + # happens if we've missed updates. + resync = yield self._need_to_do_resync(user_id, pending_updates) if resync: # Fetch all devices for the user. - result = yield self.federation.query_user_devices(origin, user_id) + origin = get_domain_from_id(user_id) + try: + result = yield self.federation.query_user_devices(origin, user_id) + except NotRetryingDestination: + # TODO: Remember that we are now out of sync and try again + # later + logger.warn( + "Failed to handle device list update for %s," + " we're not retrying the remote", + user_id, + ) + # We abort on exceptions rather than accepting the update + # as otherwise synapse will 'forget' that its device list + # is out of date. If we bail then we will retry the resync + # next time we get a device list update for this user_id. + # This makes it more likely that the device lists will + # eventually become consistent. + return + except FederationDeniedError as e: + logger.info(e) + return + except Exception: + # TODO: Remember that we are now out of sync and try again + # later + logger.exception( + "Failed to handle device list update for %s", user_id + ) + return + stream_id = result["stream_id"] devices = result["devices"] yield self.store.update_remote_device_list_cache( user_id, devices, stream_id, ) device_ids = [device["device_id"] for device in devices] - yield self.notify_device_update(user_id, device_ids) + yield self.device_handler.notify_device_update(user_id, device_ids) else: # Simply update the single device, since we know that is the only # change (becuase of the single prev_id matching the current cache) - content = dict(edu_content) - for key in ("user_id", "device_id", "stream_id", "prev_ids"): - content.pop(key, None) - yield self.store.update_remote_device_list_cache_entry( - user_id, device_id, content, stream_id, + for device_id, stream_id, prev_ids, content in pending_updates: + yield self.store.update_remote_device_list_cache_entry( + user_id, device_id, content, stream_id, + ) + + yield self.device_handler.notify_device_update( + user_id, [device_id for device_id, _, _, _ in pending_updates] ) - yield self.notify_device_update(user_id, [device_id]) - @defer.inlineCallbacks - def on_federation_query_user_devices(self, user_id): - stream_id, devices = yield self.store.get_devices_with_keys_by_user(user_id) - defer.returnValue({ - "user_id": user_id, - "stream_id": stream_id, - "devices": devices, - }) + self._seen_updates.setdefault(user_id, set()).update( + stream_id for _, stream_id, _, _ in pending_updates + ) @defer.inlineCallbacks - def user_left_room(self, user, room_id): - user_id = user.to_string() - rooms = yield self.store.get_rooms_for_user(user_id) - if not rooms: - # We no longer share rooms with this user, so we'll no longer - # receive device updates. Mark this in DB. - yield self.store.mark_remote_user_device_list_as_unsubscribed(user_id) + def _need_to_do_resync(self, user_id, updates): + """Given a list of updates for a user figure out if we need to do a full + resync, or whether we have enough data that we can just apply the delta. + """ + seen_updates = self._seen_updates.get(user_id, set()) + extremity = yield self.store.get_device_list_last_stream_id_for_remote( + user_id + ) -def _update_device_from_client_ips(device, client_ips): - ip = client_ips.get((device["user_id"], device["device_id"]), {}) - device.update({ - "last_seen_ts": ip.get("last_seen"), - "last_seen_ip": ip.get("ip"), - }) + stream_id_in_updates = set() # stream_ids in updates list + for _, stream_id, prev_ids, _ in updates: + if not prev_ids: + # We always do a resync if there are no previous IDs + defer.returnValue(True) + + for prev_id in prev_ids: + if prev_id == extremity: + continue + elif prev_id in seen_updates: + continue + elif prev_id in stream_id_in_updates: + continue + else: + defer.returnValue(True) + + stream_id_in_updates.add(stream_id) + + defer.returnValue(False) diff --git a/synapse/handlers/devicemessage.py b/synapse/handlers/devicemessage.py index f7fad15c62..f147a20b73 100644 --- a/synapse/handlers/devicemessage.py +++ b/synapse/handlers/devicemessage.py @@ -17,7 +17,8 @@ import logging from twisted.internet import defer -from synapse.types import get_domain_from_id +from synapse.api.errors import SynapseError +from synapse.types import get_domain_from_id, UserID from synapse.util.stringutils import random_string @@ -33,10 +34,10 @@ class DeviceMessageHandler(object): """ self.store = hs.get_datastore() self.notifier = hs.get_notifier() - self.is_mine_id = hs.is_mine_id + self.is_mine = hs.is_mine self.federation = hs.get_federation_sender() - hs.get_replication_layer().register_edu_handler( + hs.get_federation_registry().register_edu_handler( "m.direct_to_device", self.on_direct_to_device_edu ) @@ -52,6 +53,12 @@ class DeviceMessageHandler(object): message_type = content["type"] message_id = content["message_id"] for user_id, by_device in content["messages"].items(): + # we use UserID.from_string to catch invalid user ids + if not self.is_mine(UserID.from_string(user_id)): + logger.warning("Request for keys for non-local user %s", + user_id) + raise SynapseError(400, "Not a user here") + messages_by_device = { device_id: { "content": message_content, @@ -77,7 +84,8 @@ class DeviceMessageHandler(object): local_messages = {} remote_messages = {} for user_id, by_device in messages.items(): - if self.is_mine_id(user_id): + # we use UserID.from_string to catch invalid user ids + if self.is_mine(UserID.from_string(user_id)): messages_by_device = { device_id: { "content": message_content, diff --git a/synapse/handlers/directory.py b/synapse/handlers/directory.py index 1b5317edf5..c5b6e75e03 100644 --- a/synapse/handlers/directory.py +++ b/synapse/handlers/directory.py @@ -34,12 +34,15 @@ class DirectoryHandler(BaseHandler): self.state = hs.get_state_handler() self.appservice_handler = hs.get_application_service_handler() + self.event_creation_handler = hs.get_event_creation_handler() - self.federation = hs.get_replication_layer() - self.federation.register_query_handler( + self.federation = hs.get_federation_client() + hs.get_federation_registry().register_query_handler( "directory", self.on_directory_query ) + self.spam_checker = hs.get_spam_checker() + @defer.inlineCallbacks def _create_association(self, room_alias, room_id, servers=None, creator=None): # general association creation for both human users and app services @@ -73,6 +76,11 @@ class DirectoryHandler(BaseHandler): # association creation for human users # TODO(erikj): Do user auth. + if not self.spam_checker.user_may_create_room_alias(user_id, room_alias): + raise SynapseError( + 403, "This user is not permitted to create this alias", + ) + can_create = yield self.can_modify_alias( room_alias, user_id=user_id @@ -175,6 +183,7 @@ class DirectoryHandler(BaseHandler): "room_alias": room_alias.to_string(), }, retry_on_dns_fail=False, + ignore_backoff=True, ) except CodeMessageException as e: logging.warn("Error retrieving alias") @@ -241,8 +250,7 @@ class DirectoryHandler(BaseHandler): def send_room_alias_update_event(self, requester, user_id, room_id): aliases = yield self.store.get_aliases_for_room(room_id) - msg_handler = self.hs.get_handlers().message_handler - yield msg_handler.create_and_send_nonmember_event( + yield self.event_creation_handler.create_and_send_nonmember_event( requester, { "type": EventTypes.Aliases, @@ -264,8 +272,7 @@ class DirectoryHandler(BaseHandler): if not alias_event or alias_event.content.get("alias", "") != alias_str: return - msg_handler = self.hs.get_handlers().message_handler - yield msg_handler.create_and_send_nonmember_event( + yield self.event_creation_handler.create_and_send_nonmember_event( requester, { "type": EventTypes.CanonicalAlias, @@ -326,6 +333,14 @@ class DirectoryHandler(BaseHandler): room_id (str) visibility (str): "public" or "private" """ + if not self.spam_checker.user_may_publish_room( + requester.user.to_string(), room_id + ): + raise AuthError( + 403, + "This user is not permitted to publish rooms to the room list" + ) + if requester.is_guest: raise AuthError(403, "Guests cannot edit the published room list") diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index e40495d1ab..25aec624af 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 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. @@ -13,16 +14,18 @@ # See the License for the specific language governing permissions and # limitations under the License. -import ujson as json +import simplejson as json import logging from canonicaljson import encode_canonical_json from twisted.internet import defer -from synapse.api.errors import SynapseError, CodeMessageException -from synapse.types import get_domain_from_id -from synapse.util.logcontext import preserve_fn, preserve_context_over_deferred -from synapse.util.retryutils import get_retry_limiter, NotRetryingDestination +from synapse.api.errors import ( + SynapseError, CodeMessageException, FederationDeniedError, +) +from synapse.types import get_domain_from_id, UserID +from synapse.util.logcontext import make_deferred_yieldable, run_in_background +from synapse.util.retryutils import NotRetryingDestination logger = logging.getLogger(__name__) @@ -30,15 +33,15 @@ logger = logging.getLogger(__name__) class E2eKeysHandler(object): def __init__(self, hs): self.store = hs.get_datastore() - self.federation = hs.get_replication_layer() + self.federation = hs.get_federation_client() self.device_handler = hs.get_device_handler() - self.is_mine_id = hs.is_mine_id + self.is_mine = hs.is_mine self.clock = hs.get_clock() # doesn't really work as part of the generic query API, because the # query request requires an object POST, but we abuse the # "query handler" interface. - self.federation.register_query_handler( + hs.get_federation_registry().register_query_handler( "client_keys", self.on_federation_query_client_keys ) @@ -70,7 +73,8 @@ class E2eKeysHandler(object): remote_queries = {} for user_id, device_ids in device_keys_query.items(): - if self.is_mine_id(user_id): + # we use UserID.from_string to catch invalid user ids + if self.is_mine(UserID.from_string(user_id)): local_query[user_id] = device_ids else: remote_queries[user_id] = device_ids @@ -121,38 +125,23 @@ class E2eKeysHandler(object): def do_remote_query(destination): destination_query = remote_queries_not_in_cache[destination] try: - limiter = yield get_retry_limiter( - destination, self.clock, self.store + remote_result = yield self.federation.query_client_keys( + destination, + {"device_keys": destination_query}, + timeout=timeout ) - with limiter: - remote_result = yield self.federation.query_client_keys( - destination, - {"device_keys": destination_query}, - timeout=timeout - ) for user_id, keys in remote_result["device_keys"].items(): if user_id in destination_query: results[user_id] = keys - except CodeMessageException as e: - failures[destination] = { - "status": e.code, "message": e.message - } - except NotRetryingDestination as e: - failures[destination] = { - "status": 503, "message": "Not ready for retry", - } except Exception as e: - # include ConnectionRefused and other errors - failures[destination] = { - "status": 503, "message": e.message - } + failures[destination] = _exception_to_failure(e) - yield preserve_context_over_deferred(defer.gatherResults([ - preserve_fn(do_remote_query)(destination) + yield make_deferred_yieldable(defer.gatherResults([ + run_in_background(do_remote_query, destination) for destination in remote_queries_not_in_cache - ])) + ], consumeErrors=True)) defer.returnValue({ "device_keys": results, "failures": failures, @@ -174,7 +163,8 @@ class E2eKeysHandler(object): result_dict = {} for user_id, device_ids in query.items(): - if not self.is_mine_id(user_id): + # we use UserID.from_string to catch invalid user ids + if not self.is_mine(UserID.from_string(user_id)): logger.warning("Request for keys for non-local user %s", user_id) raise SynapseError(400, "Not a user here") @@ -217,7 +207,8 @@ class E2eKeysHandler(object): remote_queries = {} for user_id, device_keys in query.get("one_time_keys", {}).items(): - if self.is_mine_id(user_id): + # we use UserID.from_string to catch invalid user ids + if self.is_mine(UserID.from_string(user_id)): for device_id, algorithm in device_keys.items(): local_query.append((user_id, device_id, algorithm)) else: @@ -239,36 +230,31 @@ class E2eKeysHandler(object): def claim_client_keys(destination): device_keys = remote_queries[destination] try: - limiter = yield get_retry_limiter( - destination, self.clock, self.store + remote_result = yield self.federation.claim_client_keys( + destination, + {"one_time_keys": device_keys}, + timeout=timeout ) - with limiter: - remote_result = yield self.federation.claim_client_keys( - destination, - {"one_time_keys": device_keys}, - timeout=timeout - ) - for user_id, keys in remote_result["one_time_keys"].items(): - if user_id in device_keys: - json_result[user_id] = keys - except CodeMessageException as e: - failures[destination] = { - "status": e.code, "message": e.message - } - except NotRetryingDestination as e: - failures[destination] = { - "status": 503, "message": "Not ready for retry", - } + for user_id, keys in remote_result["one_time_keys"].items(): + if user_id in device_keys: + json_result[user_id] = keys except Exception as e: - # include ConnectionRefused and other errors - failures[destination] = { - "status": 503, "message": e.message - } + failures[destination] = _exception_to_failure(e) - yield preserve_context_over_deferred(defer.gatherResults([ - preserve_fn(claim_client_keys)(destination) + yield make_deferred_yieldable(defer.gatherResults([ + run_in_background(claim_client_keys, destination) for destination in remote_queries - ])) + ], consumeErrors=True)) + + logger.info( + "Claimed one-time-keys: %s", + ",".join(( + "%s for %s:%s" % (key_id, user_id, device_id) + for user_id, user_keys in json_result.iteritems() + for device_id, device_keys in user_keys.iteritems() + for key_id, _ in device_keys.iteritems() + )), + ) defer.returnValue({ "one_time_keys": json_result, @@ -296,19 +282,8 @@ class E2eKeysHandler(object): one_time_keys = keys.get("one_time_keys", None) if one_time_keys: - logger.info( - "Adding %d one_time_keys for device %r for user %r at %d", - len(one_time_keys), device_id, user_id, time_now - ) - key_list = [] - for key_id, key_json in one_time_keys.items(): - algorithm, key_id = key_id.split(":") - key_list.append(( - algorithm, key_id, encode_canonical_json(key_json) - )) - - yield self.store.add_e2e_one_time_keys( - user_id, device_id, time_now, key_list + yield self._upload_one_time_keys_for_user( + user_id, device_id, time_now, one_time_keys, ) # the device should have been registered already, but it may have been @@ -316,8 +291,88 @@ class E2eKeysHandler(object): # old access_token without an associated device_id. Either way, we # need to double-check the device is registered to avoid ending up with # keys without a corresponding device. - self.device_handler.check_device_registered(user_id, device_id) + yield self.device_handler.check_device_registered(user_id, device_id) result = yield self.store.count_e2e_one_time_keys(user_id, device_id) defer.returnValue({"one_time_key_counts": result}) + + @defer.inlineCallbacks + def _upload_one_time_keys_for_user(self, user_id, device_id, time_now, + one_time_keys): + logger.info( + "Adding one_time_keys %r for device %r for user %r at %d", + one_time_keys.keys(), device_id, user_id, time_now, + ) + + # make a list of (alg, id, key) tuples + key_list = [] + for key_id, key_obj in one_time_keys.items(): + algorithm, key_id = key_id.split(":") + key_list.append(( + algorithm, key_id, key_obj + )) + + # First we check if we have already persisted any of the keys. + existing_key_map = yield self.store.get_e2e_one_time_keys( + user_id, device_id, [k_id for _, k_id, _ in key_list] + ) + + new_keys = [] # Keys that we need to insert. (alg, id, json) tuples. + for algorithm, key_id, key in key_list: + ex_json = existing_key_map.get((algorithm, key_id), None) + if ex_json: + if not _one_time_keys_match(ex_json, key): + raise SynapseError( + 400, + ("One time key %s:%s already exists. " + "Old key: %s; new key: %r") % + (algorithm, key_id, ex_json, key) + ) + else: + new_keys.append((algorithm, key_id, encode_canonical_json(key))) + + yield self.store.add_e2e_one_time_keys( + user_id, device_id, time_now, new_keys + ) + + +def _exception_to_failure(e): + if isinstance(e, CodeMessageException): + return { + "status": e.code, "message": e.message, + } + + if isinstance(e, NotRetryingDestination): + return { + "status": 503, "message": "Not ready for retry", + } + + if isinstance(e, FederationDeniedError): + return { + "status": 403, "message": "Federation Denied", + } + + # include ConnectionRefused and other errors + # + # Note that some Exceptions (notably twisted's ResponseFailed etc) don't + # give a string for e.message, which simplejson then fails to serialize. + return { + "status": 503, "message": str(e.message), + } + + +def _one_time_keys_match(old_key_json, new_key): + old_key = json.loads(old_key_json) + + # if either is a string rather than an object, they must match exactly + if not isinstance(old_key, dict) or not isinstance(new_key, dict): + return old_key == new_key + + # otherwise, we strip off the 'signatures' if any, because it's legitimate + # for different upload attempts to have different signatures. + old_key.pop("signatures", None) + new_key_copy = dict(new_key) + new_key_copy.pop("signatures", None) + + return old_key == new_key_copy diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 996bfd0e23..f39233d846 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 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. @@ -14,24 +15,30 @@ # limitations under the License. """Contains handlers for federation events.""" + +import itertools +import logging +import sys + from signedjson.key import decode_verify_key_bytes from signedjson.sign import verify_signed_json +import six +from six.moves import http_client +from twisted.internet import defer from unpaddedbase64 import decode_base64 from ._base import BaseHandler from synapse.api.errors import ( AuthError, FederationError, StoreError, CodeMessageException, SynapseError, + FederationDeniedError, ) from synapse.api.constants import EventTypes, Membership, RejectedReason from synapse.events.validator import EventValidator -from synapse.util import unwrapFirstError -from synapse.util.logcontext import ( - PreserveLoggingContext, preserve_fn, preserve_context_over_deferred -) +from synapse.util import unwrapFirstError, logcontext from synapse.util.metrics import measure_func from synapse.util.logutils import log_function -from synapse.util.async import run_on_reactor +from synapse.util.async import run_on_reactor, Linearizer from synapse.util.frozenutils import unfreeze from synapse.crypto.event_signing import ( compute_event_signature, add_hashes_and_signatures, @@ -42,13 +49,8 @@ from synapse.events.utils import prune_event from synapse.util.retryutils import NotRetryingDestination -from synapse.push.action_generator import ActionGenerator from synapse.util.distributor import user_joined_room -from twisted.internet import defer - -import itertools -import logging logger = logging.getLogger(__name__) @@ -70,38 +72,268 @@ class FederationHandler(BaseHandler): self.hs = hs self.store = hs.get_datastore() - self.replication_layer = hs.get_replication_layer() + self.replication_layer = hs.get_federation_client() self.state_handler = hs.get_state_handler() self.server_name = hs.hostname self.keyring = hs.get_keyring() - - self.replication_layer.set_handler(self) + self.action_generator = hs.get_action_generator() + self.is_mine_id = hs.is_mine_id + self.pusher_pool = hs.get_pusherpool() + self.spam_checker = hs.get_spam_checker() + self.event_creation_handler = hs.get_event_creation_handler() # When joining a room we need to queue any events for that room up self.room_queues = {} + self._room_pdu_linearizer = Linearizer("fed_room_pdu") - @log_function @defer.inlineCallbacks - def on_receive_pdu(self, origin, pdu, state=None, auth_chain=None): - """ Called by the ReplicationLayer when we have a new pdu. We need to - do auth checks and put it through the StateHandler. + @log_function + def on_receive_pdu(self, origin, pdu, get_missing=True): + """ Process a PDU received via a federation /send/ transaction, or + via backfill of missing prev_events + + Args: + origin (str): server which initiated the /send/ transaction. Will + be used to fetch missing events or state. + pdu (FrozenEvent): received PDU + get_missing (bool): True if we should fetch missing prev_events - auth_chain and state are None if we already have the necessary state - and prev_events in the db + Returns (Deferred): completes with None """ - event = pdu - logger.debug("Got event: %s", event.event_id) + # We reprocess pdus when we have seen them only as outliers + existing = yield self.get_persisted_pdu( + origin, pdu.event_id, do_auth=False + ) + + # FIXME: Currently we fetch an event again when we already have it + # if it has been marked as an outlier. + + already_seen = ( + existing and ( + not existing.internal_metadata.is_outlier() + or pdu.internal_metadata.is_outlier() + ) + ) + if already_seen: + logger.debug("Already seen pdu %s", pdu.event_id) + return + + # do some initial sanity-checking of the event. In particular, make + # sure it doesn't have hundreds of prev_events or auth_events, which + # could cause a huge state resolution or cascade of event fetches. + try: + self._sanity_check_event(pdu) + except SynapseError as err: + raise FederationError( + "ERROR", + err.code, + err.msg, + affected=pdu.event_id, + ) # If we are currently in the process of joining this room, then we # queue up events for later processing. - if event.room_id in self.room_queues: - self.room_queues[event.room_id].append((pdu, origin)) + if pdu.room_id in self.room_queues: + logger.info("Ignoring PDU %s for room %s from %s for now; join " + "in progress", pdu.event_id, pdu.room_id, origin) + self.room_queues[pdu.room_id].append((pdu, origin)) return - logger.debug("Processing event: %s", event.event_id) + # If we're no longer in the room just ditch the event entirely. This + # is probably an old server that has come back and thinks we're still + # in the room (or we've been rejoined to the room by a state reset). + # + # If we were never in the room then maybe our database got vaped and + # we should check if we *are* in fact in the room. If we are then we + # can magically rejoin the room. + is_in_room = yield self.auth.check_host_in_room( + pdu.room_id, + self.server_name + ) + if not is_in_room: + was_in_room = yield self.store.was_host_joined( + pdu.room_id, self.server_name, + ) + if was_in_room: + logger.info( + "Ignoring PDU %s for room %s from %s as we've left the room!", + pdu.event_id, pdu.room_id, origin, + ) + return + + state = None + + auth_chain = [] + + fetch_state = False + + # 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 + ) + + logger.debug( + "_handle_new_pdu min_depth for %s: %d", + pdu.room_id, min_depth + ) + + prevs = {e_id for e_id, _ in pdu.prev_events} + seen = yield self.store.have_seen_events(prevs) + + if min_depth 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: + if get_missing and prevs - seen: + # If we're missing stuff, ensure we only fetch stuff one + # at a time. + logger.info( + "Acquiring lock for room %r to fetch %d missing events: %r...", + pdu.room_id, len(prevs - seen), list(prevs - seen)[:5], + ) + with (yield self._room_pdu_linearizer.queue(pdu.room_id)): + logger.info( + "Acquired lock for room %r to fetch %d missing events", + pdu.room_id, len(prevs - seen), + ) + + yield self._get_missing_events_for_pdu( + origin, pdu, prevs, min_depth + ) + + # Update the set of things we've seen after trying to + # fetch the missing stuff + seen = yield self.store.have_seen_events(prevs) + + if not prevs - seen: + logger.info( + "Found all missing prev events for %s", pdu.event_id + ) + elif prevs - seen: + logger.info( + "Not fetching %d missing events for room %r,event %s: %r...", + len(prevs - seen), pdu.room_id, pdu.event_id, + list(prevs - seen)[:5], + ) + + if prevs - seen: + logger.info( + "Still missing %d events for room %r: %r...", + len(prevs - seen), pdu.room_id, list(prevs - seen)[:5] + ) + fetch_state = True - logger.debug("Event: %s", event) + if fetch_state: + # We need to get the state at this event, since we haven't + # processed all the prev events. + logger.debug( + "_handle_new_pdu getting state for %s", + pdu.room_id + ) + try: + state, auth_chain = yield self.replication_layer.get_state_for_room( + origin, pdu.room_id, pdu.event_id, + ) + except Exception: + logger.exception("Failed to get state for event: %s", pdu.event_id) + + yield self._process_received_pdu( + origin, + pdu, + state=state, + auth_chain=auth_chain, + ) + + @defer.inlineCallbacks + 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 + pdu: received pdu + prevs (set(str)): List of event ids which we are missing + min_depth (int): Minimum depth of events to return. + """ + # We recalculate seen, since it may have changed. + seen = yield self.store.have_seen_events(prevs) + + if not prevs - seen: + return + + latest = yield self.store.get_latest_event_ids_in_room( + pdu.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 + latest = set(latest) + latest |= seen + + logger.info( + "Missing %d events for room %r pdu %s: %r...", + len(prevs - seen), pdu.room_id, pdu.event_id, list(prevs - seen)[:5] + ) + + # XXX: we set timeout to 10s to help workaround + # https://github.com/matrix-org/synapse/issues/1733. + # The reason is to avoid holding the linearizer lock + # whilst processing inbound /send transactions, causing + # FDs to stack up and block other inbound transactions + # which empirically can currently take up to 30 minutes. + # + # N.B. this explicitly disables retry attempts. + # + # N.B. this also increases our chances of falling back to + # fetching fresh state for the room if the missing event + # can't be found, which slightly reduces our security. + # it may also increase our DAG extremity count for the room, + # causing additional state resolution? See #1760. + # However, fetching state doesn't hold the linearizer lock + # apparently. + # + # see https://github.com/matrix-org/synapse/pull/1744 + + missing_events = yield self.replication_layer.get_missing_events( + origin, + pdu.room_id, + earliest_events_ids=list(latest), + latest_events=[pdu], + limit=10, + min_depth=min_depth, + timeout=10000, + ) + + logger.info( + "Got %d events: %r...", + len(missing_events), [e.event_id for e in missing_events[:5]] + ) + + # We want to sort these by depth so we process them and + # tell clients about them in order. + missing_events.sort(key=lambda x: x.depth) + + for e in missing_events: + logger.info("Handling found event %s", e.event_id) + yield self.on_receive_pdu( + origin, + e, + get_missing=False + ) + + @log_function + @defer.inlineCallbacks + def _process_received_pdu(self, origin, pdu, state, auth_chain): + """ Called when we have a new pdu. We need to do auth checks and put it + through the StateHandler. + """ + event = pdu + + logger.debug("Processing event: %s", event) # FIXME (erikj): Awful hack to make the case where we are not currently # in the room work @@ -140,9 +372,7 @@ class FederationHandler(BaseHandler): if auth_chain: event_ids |= {e.event_id for e in auth_chain} - seen_ids = set( - (yield self.store.have_events(event_ids)).keys() - ) + seen_ids = yield self.store.have_seen_events(event_ids) if state and auth_chain is not None: # If we have any state or auth_chain given to us by the replication @@ -181,13 +411,6 @@ class FederationHandler(BaseHandler): affected=event.event_id, ) - # if we're receiving valid events from an origin, - # it's probably a good idea to mark it as not in retry-state - # for sending (although this is a bit of a leap) - retry_timings = yield self.store.get_destination_retry_timings(origin) - if retry_timings and retry_timings["retry_last_ts"]: - self.store.set_destination_retry_timings(origin, 0, 0) - room = yield self.store.get_room(event.room_id) if not room: @@ -206,11 +429,10 @@ class FederationHandler(BaseHandler): target_user = UserID.from_string(target_user_id) extra_users.append(target_user) - with PreserveLoggingContext(): - self.notifier.on_new_room_event( - event, event_stream_id, max_stream_id, - extra_users=extra_users - ) + self.notifier.on_new_room_event( + event, event_stream_id, max_stream_id, + extra_users=extra_users + ) if event.type == EventTypes.Member: if event.membership == Membership.JOIN: @@ -249,7 +471,7 @@ class FederationHandler(BaseHandler): def check_match(id): try: return server_name == get_domain_from_id(id) - except: + except Exception: return False # Parses mapping `event_id -> (type, state_key) -> state event_id` @@ -287,7 +509,7 @@ class FederationHandler(BaseHandler): continue try: domain = get_domain_from_id(ev.state_key) - except: + except Exception: continue if domain != server_name: @@ -314,9 +536,16 @@ class FederationHandler(BaseHandler): 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. This may return - be successfull and still return no events if the other side has no new - events to offer. + This will attempt to get more events from the remote. If the other side + has no new events to offer, this will return an empty list. + + As the events are received, we check their signatures, and also do some + sanity-checking on them. If any of the backfilled events are invalid, + this method throws a SynapseError. + + TODO: make this more useful to distinguish failures of the remote + server from invalid events (there is probably no point in trying to + re-fetch invalid events from every other HS in the room.) """ if dest == self.server_name: raise SynapseError(400, "Can't backfill from self.") @@ -328,6 +557,16 @@ class FederationHandler(BaseHandler): extremities=extremities, ) + # ideally we'd sanity check the events here for excess prev_events etc, + # but it's hard to reject events at this point without completely + # breaking backfill in the same way that it is currently broken by + # events whose signature we cannot verify (#3121). + # + # So for now we accept the events anyway. #3124 tracks this. + # + # for ev in events: + # self._sanity_check_event(ev) + # Don't bother processing events we already have. seen_events = yield self.store.have_events_in_timeline( set(e.event_id for e in events) @@ -398,9 +637,10 @@ class FederationHandler(BaseHandler): missing_auth - failed_to_fetch ) - results = yield preserve_context_over_deferred(defer.gatherResults( + results = yield logcontext.make_deferred_yieldable(defer.gatherResults( [ - preserve_fn(self.replication_layer.get_pdu)( + logcontext.run_in_background( + self.replication_layer.get_pdu, [dest], event_id, outlier=True, @@ -420,7 +660,7 @@ class FederationHandler(BaseHandler): failed_to_fetch = missing_auth - set(auth_events) - seen_events = yield self.store.have_events( + seen_events = yield self.store.have_seen_events( set(auth_events.keys()) | set(state_events.keys()) ) @@ -526,7 +766,7 @@ class FederationHandler(BaseHandler): joined_domains[dom] = min(d, old_d) else: joined_domains[dom] = d - except: + except Exception: pass return sorted(joined_domains.items(), key=lambda d: d[1]) @@ -570,6 +810,9 @@ class FederationHandler(BaseHandler): except NotRetryingDestination as e: logger.info(e.message) continue + except FederationDeniedError as e: + logger.info(e) + continue except Exception as e: logger.exception( "Failed to backfill from %s because %s", @@ -592,10 +835,13 @@ class FederationHandler(BaseHandler): event_ids = list(extremities.keys()) logger.debug("calling resolve_state_groups in _maybe_backfill") - states = yield preserve_context_over_deferred(defer.gatherResults([ - preserve_fn(self.state_handler.resolve_state_groups)(room_id, [e]) - for e in event_ids - ])) + resolve = logcontext.preserve_fn( + self.state_handler.resolve_state_groups_for_events + ) + states = yield logcontext.make_deferred_yieldable(defer.gatherResults( + [resolve(room_id, [e]) for e in event_ids], + consumeErrors=True, + )) states = dict(zip(event_ids, [s.state for s in states])) state_map = yield self.store.get_events( @@ -624,6 +870,38 @@ class FederationHandler(BaseHandler): defer.returnValue(False) + def _sanity_check_event(self, ev): + """ + Do some early sanity checks of a received event + + In particular, checks it doesn't have an excessive number of + prev_events or auth_events, which could cause a huge state resolution + or cascade of event fetches. + + Args: + ev (synapse.events.EventBase): event to be checked + + Returns: None + + Raises: + SynapseError if the event does not pass muster + """ + if len(ev.prev_events) > 20: + logger.warn("Rejecting event %s which has %i prev_events", + ev.event_id, len(ev.prev_events)) + raise SynapseError( + http_client.BAD_REQUEST, + "Too many prev_events", + ) + + if len(ev.auth_events) > 10: + logger.warn("Rejecting event %s which has %i auth_events", + ev.event_id, len(ev.auth_events)) + raise SynapseError( + http_client.BAD_REQUEST, + "Too many auth_events", + ) + @defer.inlineCallbacks def send_invite(self, target_host, event): """ Sends the invite to the remote server for signing. @@ -641,7 +919,11 @@ class FederationHandler(BaseHandler): @defer.inlineCallbacks def on_event_auth(self, event_id): - auth = yield self.store.get_auth_chain([event_id]) + event = yield self.store.get_event(event_id) + auth = yield self.store.get_auth_chain( + [auth_id for auth_id, _ in event.auth_events], + include_given=True + ) for event in auth: event.signatures.update( @@ -670,8 +952,6 @@ class FederationHandler(BaseHandler): """ logger.debug("Joining %s to %s", joinee, room_id) - yield self.store.clean_room_for_join(room_id) - origin, event = yield self._make_and_verify_event( target_hosts, room_id, @@ -680,7 +960,15 @@ class FederationHandler(BaseHandler): content, ) + # This shouldn't happen, because the RoomMemberHandler has a + # linearizer lock which only allows one operation per user per room + # at a time - so this is just paranoia. + assert (room_id not in self.room_queues) + self.room_queues[room_id] = [] + + yield self.store.clean_room_for_join(room_id) + handled_events = set() try: @@ -714,7 +1002,7 @@ class FederationHandler(BaseHandler): room_creator_user_id="", is_public=False ) - except: + except Exception: # FIXME pass @@ -722,29 +1010,45 @@ class FederationHandler(BaseHandler): origin, auth_chain, state, event ) - with PreserveLoggingContext(): - self.notifier.on_new_room_event( - event, event_stream_id, max_stream_id, - extra_users=[joinee] - ) + self.notifier.on_new_room_event( + event, event_stream_id, max_stream_id, + extra_users=[joinee] + ) logger.debug("Finished joining %s to %s", joinee, room_id) finally: room_queue = self.room_queues[room_id] del self.room_queues[room_id] - for p, origin in room_queue: - if p.event_id in handled_events: - continue + # we don't need to wait for the queued events to be processed - + # it's just a best-effort thing at this point. We do want to do + # them roughly in order, though, otherwise we'll end up making + # lots of requests for missing prev_events which we do actually + # have. Hence we fire off the deferred, but don't wait for it. - try: - self.on_receive_pdu(origin, p) - except: - logger.exception("Couldn't handle pdu") + logcontext.run_in_background(self._handle_queued_pdus, room_queue) defer.returnValue(True) @defer.inlineCallbacks + def _handle_queued_pdus(self, room_queue): + """Process PDUs which got queued up while we were busy send_joining. + + Args: + room_queue (list[FrozenEvent, str]): list of PDUs to be processed + and the servers that sent them + """ + for p, origin in room_queue: + try: + logger.info("Processing queued PDU %s which was received " + "while we were joining %s", p.event_id, p.room_id) + yield self.on_receive_pdu(origin, p) + except Exception as e: + logger.warn( + "Error handling queued PDU %s from %s: %s", + p.event_id, origin, e) + + @defer.inlineCallbacks @log_function def on_make_join_request(self, room_id, user_id): """ We've received a /make_join/ request, so we create a partial @@ -762,8 +1066,7 @@ class FederationHandler(BaseHandler): }) try: - message_handler = self.hs.get_handlers().message_handler - event, context = yield message_handler._create_new_client_event( + event, context = yield self.event_creation_handler.create_new_client_event( builder=builder, ) except AuthError as e: @@ -791,9 +1094,19 @@ class FederationHandler(BaseHandler): ) event.internal_metadata.outlier = False - # Send this event on behalf of the origin server since they may not - # have an up to data view of the state of the room at this event so - # will not know which servers to send the event to. + # Send this event on behalf of the origin server. + # + # The reasons we have the destination server rather than the origin + # server send it are slightly mysterious: the origin server should have + # all the neccessary state once it gets the response to the send_join, + # so it could send the event itself if it wanted to. It may be that + # doing it this way reduces failure modes, or avoids certain attacks + # where a new server selectively tells a subset of the federation that + # it has joined. + # + # The fact is that, as of the current writing, Synapse doesn't send out + # the join event over federation after joining, and changing it now + # would introduce the danger of backwards-compatibility problems. event.internal_metadata.send_on_behalf_of = origin context, event_stream_id, max_stream_id = yield self._handle_new_event( @@ -812,10 +1125,9 @@ class FederationHandler(BaseHandler): target_user = UserID.from_string(target_user_id) extra_users.append(target_user) - with PreserveLoggingContext(): - self.notifier.on_new_room_event( - event, event_stream_id, max_stream_id, extra_users=extra_users - ) + self.notifier.on_new_room_event( + event, event_stream_id, max_stream_id, extra_users=extra_users + ) if event.type == EventTypes.Member: if event.content["membership"] == Membership.JOIN: @@ -823,9 +1135,7 @@ class FederationHandler(BaseHandler): yield user_joined_room(self.distributor, user, event.room_id) state_ids = context.prev_state_ids.values() - auth_chain = yield self.store.get_auth_chain(set( - [event.event_id] + state_ids - )) + auth_chain = yield self.store.get_auth_chain(state_ids) state = yield self.store.get_events(context.prev_state_ids.values()) @@ -842,6 +1152,34 @@ class FederationHandler(BaseHandler): """ event = pdu + if event.state_key is None: + raise SynapseError(400, "The invite event did not have a state key") + + is_blocked = yield self.store.is_room_blocked(event.room_id) + if is_blocked: + raise SynapseError(403, "This room has been blocked on this server") + + if self.hs.config.block_non_admin_invites: + raise SynapseError(403, "This server does not accept room invites") + + if not self.spam_checker.user_may_invite( + event.sender, event.state_key, event.room_id, + ): + raise SynapseError( + 403, "This user is not permitted to send invites to this server/user" + ) + + membership = event.content.get("membership") + if event.type != EventTypes.Member or membership != Membership.INVITE: + raise SynapseError(400, "The event was not an m.room.member invite event") + + sender_domain = get_domain_from_id(event.sender) + if sender_domain != origin: + raise SynapseError(400, "The invite event was not from the server sending it") + + if not self.is_mine_id(event.state_key): + raise SynapseError(400, "The invite event must be for this server") + event.internal_metadata.outlier = True event.internal_metadata.invite_from_remote = True @@ -861,48 +1199,38 @@ class FederationHandler(BaseHandler): ) target_user = UserID.from_string(event.state_key) - with PreserveLoggingContext(): - self.notifier.on_new_room_event( - event, event_stream_id, max_stream_id, - extra_users=[target_user], - ) + self.notifier.on_new_room_event( + event, event_stream_id, max_stream_id, + extra_users=[target_user], + ) defer.returnValue(event) @defer.inlineCallbacks def do_remotely_reject_invite(self, target_hosts, room_id, user_id): - try: - origin, event = yield self._make_and_verify_event( - target_hosts, - room_id, - user_id, - "leave" - ) - signed_event = self._sign_event(event) - except SynapseError: - raise - except CodeMessageException as e: - logger.warn("Failed to reject invite: %s", e) - raise SynapseError(500, "Failed to reject invite") - - # Try the host we successfully got a response to /make_join/ - # request first. + origin, event = yield self._make_and_verify_event( + target_hosts, + room_id, + user_id, + "leave" + ) + # Mark as outlier as we don't have any state for this event; we're not + # even in the room. + event.internal_metadata.outlier = True + event = self._sign_event(event) + + # Try the host that we succesfully called /make_leave/ on first for + # the /send_leave/ request. try: target_hosts.remove(origin) target_hosts.insert(0, origin) except ValueError: pass - try: - yield self.replication_layer.send_leave( - target_hosts, - signed_event - ) - except SynapseError: - raise - except CodeMessageException as e: - logger.warn("Failed to reject invite: %s", e) - raise SynapseError(500, "Failed to reject invite") + yield self.replication_layer.send_leave( + target_hosts, + event + ) context = yield self.state_handler.compute_event_context(event) @@ -978,8 +1306,7 @@ class FederationHandler(BaseHandler): "state_key": user_id, }) - message_handler = self.hs.get_handlers().message_handler - event, context = yield message_handler._create_new_client_event( + event, context = yield self.event_creation_handler.create_new_client_event( builder=builder, ) @@ -1023,10 +1350,9 @@ class FederationHandler(BaseHandler): target_user = UserID.from_string(target_user_id) extra_users.append(target_user) - with PreserveLoggingContext(): - self.notifier.on_new_room_event( - event, event_stream_id, max_stream_id, extra_users=extra_users - ) + self.notifier.on_new_room_event( + event, event_stream_id, max_stream_id, extra_users=extra_users + ) defer.returnValue(None) @@ -1061,7 +1387,7 @@ class FederationHandler(BaseHandler): for event in res: # We sign these again because there was a bug where we # incorrectly signed things the first time round - if self.hs.is_mine_id(event.event_id): + if self.is_mine_id(event.event_id): event.signatures.update( compute_event_signature( event, @@ -1096,7 +1422,7 @@ class FederationHandler(BaseHandler): if prev_id != event.event_id: results[(event.type, event.state_key)] = prev_id else: - del results[(event.type, event.state_key)] + results.pop((event.type, event.state_key), None) defer.returnValue(results.values()) else: @@ -1134,7 +1460,7 @@ class FederationHandler(BaseHandler): ) if event: - if self.hs.is_mine_id(event.event_id): + 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 @@ -1178,23 +1504,33 @@ class FederationHandler(BaseHandler): auth_events=auth_events, ) - if not event.internal_metadata.is_outlier(): - action_generator = ActionGenerator(self.hs) - yield action_generator.handle_push_actions_for_event( - event, context + try: + if not event.internal_metadata.is_outlier() and not backfilled: + yield self.action_generator.handle_push_actions_for_event( + event, context + ) + + event_stream_id, max_stream_id = yield self.store.persist_event( + event, + context=context, + backfilled=backfilled, ) + except: # noqa: E722, as we reraise the exception this is fine. + tp, value, tb = sys.exc_info() - event_stream_id, max_stream_id = yield self.store.persist_event( - event, - context=context, - backfilled=backfilled, - ) + logcontext.run_in_background( + self.store.remove_push_actions_from_staging, + event.event_id, + ) + + six.reraise(tp, value, tb) if not backfilled: # this intentionally does not yield: we don't care about the result # and don't need to wait for it. - preserve_fn(self.hs.get_pusherpool().on_new_notifications)( - event_stream_id, max_stream_id + logcontext.run_in_background( + self.pusher_pool.on_new_notifications, + event_stream_id, max_stream_id, ) defer.returnValue((context, event_stream_id, max_stream_id)) @@ -1206,16 +1542,17 @@ class FederationHandler(BaseHandler): a bunch of outliers, but not a chunk of individual events that depend on each other for state calculations. """ - contexts = yield preserve_context_over_deferred(defer.gatherResults( + contexts = yield logcontext.make_deferred_yieldable(defer.gatherResults( [ - preserve_fn(self._prep_event)( + logcontext.run_in_background( + self._prep_event, origin, ev_info["event"], state=ev_info.get("state"), auth_events=ev_info.get("auth_events"), ) for ev_info in event_infos - ] + ], consumeErrors=True, )) yield self.store.persist_events( @@ -1325,7 +1662,17 @@ class FederationHandler(BaseHandler): @defer.inlineCallbacks def _prep_event(self, origin, event, state=None, auth_events=None): + """ + + Args: + origin: + event: + state: + auth_events: + Returns: + Deferred, which resolves to synapse.events.snapshot.EventContext + """ context = yield self.state_handler.compute_event_context( event, old_state=state, ) @@ -1362,7 +1709,7 @@ class FederationHandler(BaseHandler): context.rejected = RejectedReason.AUTH_ERROR - if event.type == EventTypes.GuestAccess: + if event.type == EventTypes.GuestAccess and not context.rejected: yield self.maybe_kick_guest_users(event) defer.returnValue(context) @@ -1379,7 +1726,11 @@ class FederationHandler(BaseHandler): pass # Now get the current auth_chain for the event. - local_auth_chain = yield self.store.get_auth_chain([event_id]) + event = yield self.store.get_event(event_id) + local_auth_chain = yield self.store.get_auth_chain( + [auth_id for auth_id, _ in event.auth_events], + include_given=True + ) # TODO: Check if we would now reject event_id. If so we need to tell # everyone. @@ -1427,6 +1778,17 @@ class FederationHandler(BaseHandler): @defer.inlineCallbacks @log_function def do_auth(self, origin, event, context, auth_events): + """ + + Args: + origin (str): + event (synapse.events.FrozenEvent): + context (synapse.events.snapshot.EventContext): + auth_events (dict[(str, str)->str]): + + Returns: + defer.Deferred[None] + """ # Check if we have all the auth events. current_state = set(e.event_id for e in auth_events.values()) event_auth_events = set(e_id for e_id, _ in event.auth_events) @@ -1437,7 +1799,8 @@ class FederationHandler(BaseHandler): event_key = None if event_auth_events - current_state: - have_events = yield self.store.have_events( + # TODO: can we use store.have_seen_events here instead? + have_events = yield self.store.get_seen_events_with_rejections( event_auth_events - current_state ) else: @@ -1460,12 +1823,12 @@ class FederationHandler(BaseHandler): origin, event.room_id, event.event_id ) - seen_remotes = yield self.store.have_events( + seen_remotes = yield self.store.have_seen_events( [e.event_id for e in remote_auth_chain] ) for e in remote_auth_chain: - if e.event_id in seen_remotes.keys(): + if e.event_id in seen_remotes: continue if e.event_id == event.event_id: @@ -1492,11 +1855,11 @@ class FederationHandler(BaseHandler): except AuthError: pass - have_events = yield self.store.have_events( + have_events = yield self.store.get_seen_events_with_rejections( [e_id for e_id, _ in event.auth_events] ) seen_events = set(have_events.keys()) - except: + except Exception: # FIXME: logger.exception("Failed to get auth chain") @@ -1509,18 +1872,18 @@ class FederationHandler(BaseHandler): # Do auth conflict res. logger.info("Different auth: %s", different_auth) - different_events = yield preserve_context_over_deferred(defer.gatherResults( - [ - preserve_fn(self.store.get_event)( + different_events = yield logcontext.make_deferred_yieldable( + defer.gatherResults([ + logcontext.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) + ], consumeErrors=True) + ).addErrback(unwrapFirstError) if different_events: local_view = dict(auth_events) @@ -1539,16 +1902,9 @@ class FederationHandler(BaseHandler): current_state = set(e.event_id for e in auth_events.values()) different_auth = event_auth_events - current_state - context.current_state_ids = dict(context.current_state_ids) - context.current_state_ids.update({ - k: a.event_id for k, a in auth_events.items() - if k != event_key - }) - context.prev_state_ids = dict(context.prev_state_ids) - context.prev_state_ids.update({ - k: a.event_id for k, a in auth_events.items() - }) - context.state_group = self.store.get_next_state_group() + yield self._update_context_for_auth_events( + event, context, auth_events, event_key, + ) if different_auth and not event.internal_metadata.is_outlier(): logger.info("Different auth after resolution: %s", different_auth) @@ -1572,7 +1928,9 @@ class FederationHandler(BaseHandler): auth_ids = yield self.auth.compute_auth_events( event, context.prev_state_ids ) - local_auth_chain = yield self.store.get_auth_chain(auth_ids) + local_auth_chain = yield self.store.get_auth_chain( + auth_ids, include_given=True + ) try: # 2. Get remote difference. @@ -1583,13 +1941,13 @@ class FederationHandler(BaseHandler): local_auth_chain, ) - seen_remotes = yield self.store.have_events( + seen_remotes = yield self.store.have_seen_events( [e.event_id for e in result["auth_chain"]] ) # 3. Process any remote auth chain events we haven't seen. for ev in result["auth_chain"]: - if ev.event_id in seen_remotes.keys(): + if ev.event_id in seen_remotes: continue if ev.event_id == event.event_id: @@ -1619,23 +1977,16 @@ class FederationHandler(BaseHandler): except AuthError: pass - except: + except Exception: # FIXME: logger.exception("Failed to query auth chain") # 4. Look at rejects and their proofs. # TODO. - context.current_state_ids = dict(context.current_state_ids) - context.current_state_ids.update({ - k: a.event_id for k, a in auth_events.items() - if k != event_key - }) - context.prev_state_ids = dict(context.prev_state_ids) - context.prev_state_ids.update({ - k: a.event_id for k, a in auth_events.items() - }) - context.state_group = self.store.get_next_state_group() + yield self._update_context_for_auth_events( + event, context, auth_events, event_key, + ) try: self.auth.check(event, auth_events=auth_events) @@ -1644,6 +1995,45 @@ class FederationHandler(BaseHandler): raise e @defer.inlineCallbacks + def _update_context_for_auth_events(self, event, context, auth_events, + event_key): + """Update the state_ids in an event context after auth event resolution, + storing the changes as a new state group. + + Args: + event (Event): The event we're handling the context for + + context (synapse.events.snapshot.EventContext): event context + to be updated + + auth_events (dict[(str, str)->str]): 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. + """ + state_updates = { + k: a.event_id for k, a in auth_events.iteritems() + 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({ + k: a.event_id for k, a in auth_events.iteritems() + }) + context.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, + ) + + @defer.inlineCallbacks def construct_auth_difference(self, local_auth, remote_auth): """ Given a local and remote auth chain, find the differences. This assumes that we have already processed all events in remote_auth @@ -1686,7 +2076,7 @@ class FederationHandler(BaseHandler): def get_next(it, opt=None): try: return it.next() - except: + except Exception: return opt current_local = get_next(local_iter) @@ -1811,8 +2201,7 @@ class FederationHandler(BaseHandler): if (yield self.auth.check_host_in_room(room_id, self.hs.hostname)): builder = self.event_builder_factory.new(event_dict) EventValidator().validate_new(builder) - message_handler = self.hs.get_handlers().message_handler - event, context = yield message_handler._create_new_client_event( + event, context = yield self.event_creation_handler.create_new_client_event( builder=builder ) @@ -1827,7 +2216,7 @@ class FederationHandler(BaseHandler): raise e yield self._check_signature(event, context) - member_handler = self.hs.get_handlers().room_member_handler + member_handler = self.hs.get_room_member_handler() yield member_handler.send_membership_event(None, event, context) else: destinations = set(x.split(":", 1)[-1] for x in (sender_user_id, room_id)) @@ -1840,10 +2229,17 @@ class FederationHandler(BaseHandler): @defer.inlineCallbacks @log_function def on_exchange_third_party_invite_request(self, origin, room_id, event_dict): + """Handle an exchange_third_party_invite request from a remote server + + The remote server will call this when it wants to turn a 3pid invite + into a normal m.room.member invite. + + Returns: + Deferred: resolves (to None) + """ builder = self.event_builder_factory.new(event_dict) - message_handler = self.hs.get_handlers().message_handler - event, context = yield message_handler._create_new_client_event( + event, context = yield self.event_creation_handler.create_new_client_event( builder=builder, ) @@ -1858,10 +2254,13 @@ class FederationHandler(BaseHandler): raise e yield self._check_signature(event, context) + # XXX we send the invite here, but send_membership_event also sends it, + # so we end up making two requests. I think this is redundant. returned_invite = yield self.send_invite(origin, event) # TODO: Make sure the signatures actually are correct. event.signatures.update(returned_invite.signatures) - member_handler = self.hs.get_handlers().room_member_handler + + member_handler = self.hs.get_room_member_handler() yield member_handler.send_membership_event(None, event, context) @defer.inlineCallbacks @@ -1890,8 +2289,9 @@ class FederationHandler(BaseHandler): builder = self.event_builder_factory.new(event_dict) EventValidator().validate_new(builder) - message_handler = self.hs.get_handlers().message_handler - event, context = yield message_handler._create_new_client_event(builder=builder) + event, context = yield self.event_creation_handler.create_new_client_event( + builder=builder, + ) defer.returnValue((event, context)) @defer.inlineCallbacks diff --git a/synapse/handlers/groups_local.py b/synapse/handlers/groups_local.py new file mode 100644 index 0000000000..977993e7d4 --- /dev/null +++ b/synapse/handlers/groups_local.py @@ -0,0 +1,471 @@ +# -*- coding: utf-8 -*- +# Copyright 2017 Vector Creations 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. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from twisted.internet import defer + +from synapse.api.errors import SynapseError +from synapse.types import get_domain_from_id + +import logging + +logger = logging.getLogger(__name__) + + +def _create_rerouter(func_name): + """Returns a function that looks at the group id and calls the function + on federation or the local group server if the group is local + """ + def f(self, group_id, *args, **kwargs): + if self.is_mine_id(group_id): + return getattr(self.groups_server_handler, func_name)( + group_id, *args, **kwargs + ) + else: + destination = get_domain_from_id(group_id) + return getattr(self.transport_client, func_name)( + destination, group_id, *args, **kwargs + ) + return f + + +class GroupsLocalHandler(object): + def __init__(self, hs): + self.hs = hs + self.store = hs.get_datastore() + self.room_list_handler = hs.get_room_list_handler() + self.groups_server_handler = hs.get_groups_server_handler() + self.transport_client = hs.get_federation_transport_client() + self.auth = hs.get_auth() + self.clock = hs.get_clock() + self.keyring = hs.get_keyring() + self.is_mine_id = hs.is_mine_id + self.signing_key = hs.config.signing_key[0] + self.server_name = hs.hostname + self.notifier = hs.get_notifier() + self.attestations = hs.get_groups_attestation_signing() + + self.profile_handler = hs.get_profile_handler() + + # Ensure attestations get renewed + hs.get_groups_attestation_renewer() + + # The following functions merely route the query to the local groups server + # or federation depending on if the group is local or remote + + get_group_profile = _create_rerouter("get_group_profile") + update_group_profile = _create_rerouter("update_group_profile") + get_rooms_in_group = _create_rerouter("get_rooms_in_group") + + get_invited_users_in_group = _create_rerouter("get_invited_users_in_group") + + add_room_to_group = _create_rerouter("add_room_to_group") + update_room_in_group = _create_rerouter("update_room_in_group") + remove_room_from_group = _create_rerouter("remove_room_from_group") + + update_group_summary_room = _create_rerouter("update_group_summary_room") + delete_group_summary_room = _create_rerouter("delete_group_summary_room") + + update_group_category = _create_rerouter("update_group_category") + delete_group_category = _create_rerouter("delete_group_category") + get_group_category = _create_rerouter("get_group_category") + get_group_categories = _create_rerouter("get_group_categories") + + update_group_summary_user = _create_rerouter("update_group_summary_user") + delete_group_summary_user = _create_rerouter("delete_group_summary_user") + + update_group_role = _create_rerouter("update_group_role") + delete_group_role = _create_rerouter("delete_group_role") + get_group_role = _create_rerouter("get_group_role") + get_group_roles = _create_rerouter("get_group_roles") + + set_group_join_policy = _create_rerouter("set_group_join_policy") + + @defer.inlineCallbacks + def get_group_summary(self, group_id, requester_user_id): + """Get the group summary for a group. + + If the group is remote we check that the users have valid attestations. + """ + if self.is_mine_id(group_id): + res = yield self.groups_server_handler.get_group_summary( + group_id, requester_user_id + ) + else: + res = yield self.transport_client.get_group_summary( + get_domain_from_id(group_id), group_id, requester_user_id, + ) + + group_server_name = get_domain_from_id(group_id) + + # Loop through the users and validate the attestations. + chunk = res["users_section"]["users"] + valid_users = [] + for entry in chunk: + g_user_id = entry["user_id"] + attestation = entry.pop("attestation", {}) + try: + if get_domain_from_id(g_user_id) != group_server_name: + yield self.attestations.verify_attestation( + attestation, + group_id=group_id, + user_id=g_user_id, + server_name=get_domain_from_id(g_user_id), + ) + valid_users.append(entry) + except Exception as e: + logger.info("Failed to verify user is in group: %s", e) + + res["users_section"]["users"] = valid_users + + res["users_section"]["users"].sort(key=lambda e: e.get("order", 0)) + res["rooms_section"]["rooms"].sort(key=lambda e: e.get("order", 0)) + + # Add `is_publicised` flag to indicate whether the user has publicised their + # membership of the group on their profile + result = yield self.store.get_publicised_groups_for_user(requester_user_id) + is_publicised = group_id in result + + res.setdefault("user", {})["is_publicised"] = is_publicised + + defer.returnValue(res) + + @defer.inlineCallbacks + def create_group(self, group_id, user_id, content): + """Create a group + """ + + logger.info("Asking to create group with ID: %r", group_id) + + if self.is_mine_id(group_id): + res = yield self.groups_server_handler.create_group( + group_id, user_id, content + ) + local_attestation = None + remote_attestation = None + else: + local_attestation = self.attestations.create_attestation(group_id, user_id) + content["attestation"] = local_attestation + + content["user_profile"] = yield self.profile_handler.get_profile(user_id) + + res = yield self.transport_client.create_group( + get_domain_from_id(group_id), group_id, user_id, content, + ) + + remote_attestation = res["attestation"] + yield self.attestations.verify_attestation( + remote_attestation, + group_id=group_id, + user_id=user_id, + server_name=get_domain_from_id(group_id), + ) + + is_publicised = content.get("publicise", False) + token = yield self.store.register_user_group_membership( + group_id, user_id, + membership="join", + is_admin=True, + local_attestation=local_attestation, + remote_attestation=remote_attestation, + is_publicised=is_publicised, + ) + self.notifier.on_new_event( + "groups_key", token, users=[user_id], + ) + + defer.returnValue(res) + + @defer.inlineCallbacks + def get_users_in_group(self, group_id, requester_user_id): + """Get users in a group + """ + if self.is_mine_id(group_id): + res = yield self.groups_server_handler.get_users_in_group( + group_id, requester_user_id + ) + defer.returnValue(res) + + group_server_name = get_domain_from_id(group_id) + + res = yield self.transport_client.get_users_in_group( + get_domain_from_id(group_id), group_id, requester_user_id, + ) + + chunk = res["chunk"] + valid_entries = [] + for entry in chunk: + g_user_id = entry["user_id"] + attestation = entry.pop("attestation", {}) + try: + if get_domain_from_id(g_user_id) != group_server_name: + yield self.attestations.verify_attestation( + attestation, + group_id=group_id, + user_id=g_user_id, + server_name=get_domain_from_id(g_user_id), + ) + valid_entries.append(entry) + except Exception as e: + logger.info("Failed to verify user is in group: %s", e) + + res["chunk"] = valid_entries + + defer.returnValue(res) + + @defer.inlineCallbacks + def join_group(self, group_id, user_id, content): + """Request to join a group + """ + if self.is_mine_id(group_id): + yield self.groups_server_handler.join_group( + group_id, user_id, content + ) + local_attestation = None + remote_attestation = None + else: + local_attestation = self.attestations.create_attestation(group_id, user_id) + content["attestation"] = local_attestation + + res = yield self.transport_client.join_group( + get_domain_from_id(group_id), group_id, user_id, content, + ) + + remote_attestation = res["attestation"] + + yield self.attestations.verify_attestation( + remote_attestation, + group_id=group_id, + user_id=user_id, + server_name=get_domain_from_id(group_id), + ) + + # TODO: Check that the group is public and we're being added publically + is_publicised = content.get("publicise", False) + + token = yield self.store.register_user_group_membership( + group_id, user_id, + membership="join", + is_admin=False, + local_attestation=local_attestation, + remote_attestation=remote_attestation, + is_publicised=is_publicised, + ) + self.notifier.on_new_event( + "groups_key", token, users=[user_id], + ) + + defer.returnValue({}) + + @defer.inlineCallbacks + def accept_invite(self, group_id, user_id, content): + """Accept an invite to a group + """ + if self.is_mine_id(group_id): + yield self.groups_server_handler.accept_invite( + group_id, user_id, content + ) + local_attestation = None + remote_attestation = None + else: + local_attestation = self.attestations.create_attestation(group_id, user_id) + content["attestation"] = local_attestation + + res = yield self.transport_client.accept_group_invite( + get_domain_from_id(group_id), group_id, user_id, content, + ) + + remote_attestation = res["attestation"] + + yield self.attestations.verify_attestation( + remote_attestation, + group_id=group_id, + user_id=user_id, + server_name=get_domain_from_id(group_id), + ) + + # TODO: Check that the group is public and we're being added publically + is_publicised = content.get("publicise", False) + + token = yield self.store.register_user_group_membership( + group_id, user_id, + membership="join", + is_admin=False, + local_attestation=local_attestation, + remote_attestation=remote_attestation, + is_publicised=is_publicised, + ) + self.notifier.on_new_event( + "groups_key", token, users=[user_id], + ) + + defer.returnValue({}) + + @defer.inlineCallbacks + def invite(self, group_id, user_id, requester_user_id, config): + """Invite a user to a group + """ + content = { + "requester_user_id": requester_user_id, + "config": config, + } + if self.is_mine_id(group_id): + res = yield self.groups_server_handler.invite_to_group( + group_id, user_id, requester_user_id, content, + ) + else: + res = yield self.transport_client.invite_to_group( + get_domain_from_id(group_id), group_id, user_id, requester_user_id, + content, + ) + + defer.returnValue(res) + + @defer.inlineCallbacks + def on_invite(self, group_id, user_id, content): + """One of our users were invited to a group + """ + # TODO: Support auto join and rejection + + if not self.is_mine_id(user_id): + raise SynapseError(400, "User not on this server") + + local_profile = {} + if "profile" in content: + if "name" in content["profile"]: + local_profile["name"] = content["profile"]["name"] + if "avatar_url" in content["profile"]: + local_profile["avatar_url"] = content["profile"]["avatar_url"] + + token = yield self.store.register_user_group_membership( + group_id, user_id, + membership="invite", + content={"profile": local_profile, "inviter": content["inviter"]}, + ) + self.notifier.on_new_event( + "groups_key", token, users=[user_id], + ) + try: + user_profile = yield self.profile_handler.get_profile(user_id) + except Exception as e: + logger.warn("No profile for user %s: %s", user_id, e) + user_profile = {} + + defer.returnValue({"state": "invite", "user_profile": user_profile}) + + @defer.inlineCallbacks + def remove_user_from_group(self, group_id, user_id, requester_user_id, content): + """Remove a user from a group + """ + if user_id == requester_user_id: + token = yield self.store.register_user_group_membership( + group_id, user_id, + membership="leave", + ) + self.notifier.on_new_event( + "groups_key", token, users=[user_id], + ) + + # TODO: Should probably remember that we tried to leave so that we can + # retry if the group server is currently down. + + if self.is_mine_id(group_id): + res = yield self.groups_server_handler.remove_user_from_group( + group_id, user_id, requester_user_id, content, + ) + else: + content["requester_user_id"] = requester_user_id + res = yield self.transport_client.remove_user_from_group( + get_domain_from_id(group_id), group_id, requester_user_id, + user_id, content, + ) + + defer.returnValue(res) + + @defer.inlineCallbacks + def user_removed_from_group(self, group_id, user_id, content): + """One of our users was removed/kicked from a group + """ + # TODO: Check if user in group + token = yield self.store.register_user_group_membership( + group_id, user_id, + membership="leave", + ) + self.notifier.on_new_event( + "groups_key", token, users=[user_id], + ) + + @defer.inlineCallbacks + def get_joined_groups(self, user_id): + group_ids = yield self.store.get_joined_groups(user_id) + defer.returnValue({"groups": group_ids}) + + @defer.inlineCallbacks + def get_publicised_groups_for_user(self, user_id): + if self.hs.is_mine_id(user_id): + result = yield self.store.get_publicised_groups_for_user(user_id) + + # Check AS associated groups for this user - this depends on the + # RegExps in the AS registration file (under `users`) + for app_service in self.store.get_app_services(): + result.extend(app_service.get_groups_for_user(user_id)) + + defer.returnValue({"groups": result}) + else: + bulk_result = yield self.transport_client.bulk_get_publicised_groups( + get_domain_from_id(user_id), [user_id], + ) + result = bulk_result.get("users", {}).get(user_id) + # TODO: Verify attestations + defer.returnValue({"groups": result}) + + @defer.inlineCallbacks + def bulk_get_publicised_groups(self, user_ids, proxy=True): + destinations = {} + local_users = set() + + for user_id in user_ids: + if self.hs.is_mine_id(user_id): + local_users.add(user_id) + else: + destinations.setdefault( + get_domain_from_id(user_id), set() + ).add(user_id) + + if not proxy and destinations: + raise SynapseError(400, "Some user_ids are not local") + + results = {} + failed_results = [] + for destination, dest_user_ids in destinations.iteritems(): + try: + r = yield self.transport_client.bulk_get_publicised_groups( + destination, list(dest_user_ids), + ) + results.update(r["users"]) + except Exception: + failed_results.extend(dest_user_ids) + + for uid in local_users: + results[uid] = yield self.store.get_publicised_groups_for_user( + uid + ) + + # Check AS associated groups for this user - this depends on the + # RegExps in the AS registration file (under `users`) + for app_service in self.store.get_app_services(): + results[uid].extend(app_service.get_groups_for_user(uid)) + + defer.returnValue({"users": results}) diff --git a/synapse/handlers/identity.py b/synapse/handlers/identity.py index 559e5d5a71..91a0898860 100644 --- a/synapse/handlers/identity.py +++ b/synapse/handlers/identity.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2015, 2016 OpenMarket Ltd +# Copyright 2017 Vector Creations Ltd # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -14,18 +15,20 @@ # limitations under the License. """Utilities for interacting with Identity Servers""" + +import logging + +import simplejson as json + from twisted.internet import defer from synapse.api.errors import ( - CodeMessageException + MatrixCodeMessageException, CodeMessageException ) from ._base import BaseHandler from synapse.util.async import run_on_reactor from synapse.api.errors import SynapseError, Codes -import json -import logging - logger = logging.getLogger(__name__) @@ -89,6 +92,9 @@ class IdentityHandler(BaseHandler): ), {'sid': creds['sid'], 'client_secret': client_secret} ) + except MatrixCodeMessageException as e: + logger.info("getValidated3pid failed with Matrix error: %r", e) + raise SynapseError(e.code, e.msg, e.errcode) except CodeMessageException as e: data = json.loads(e.msg) @@ -150,7 +156,7 @@ class IdentityHandler(BaseHandler): params.update(kwargs) try: - data = yield self.http_client.post_urlencoded_get_json( + data = yield self.http_client.post_json_get_json( "https://%s%s" % ( id_server, "/_matrix/identity/api/v1/validate/email/requestToken" @@ -158,6 +164,46 @@ class IdentityHandler(BaseHandler): params ) defer.returnValue(data) + except MatrixCodeMessageException as e: + logger.info("Proxied requestToken failed with Matrix error: %r", e) + raise SynapseError(e.code, e.msg, e.errcode) + except CodeMessageException as e: + logger.info("Proxied requestToken failed: %r", e) + raise e + + @defer.inlineCallbacks + def requestMsisdnToken( + self, id_server, country, phone_number, + client_secret, send_attempt, **kwargs + ): + yield run_on_reactor() + + if not self._should_trust_id_server(id_server): + raise SynapseError( + 400, "Untrusted ID server '%s'" % id_server, + Codes.SERVER_NOT_TRUSTED + ) + + params = { + 'country': country, + 'phone_number': phone_number, + 'client_secret': client_secret, + 'send_attempt': send_attempt, + } + params.update(kwargs) + + try: + data = yield self.http_client.post_json_get_json( + "https://%s%s" % ( + id_server, + "/_matrix/identity/api/v1/validate/msisdn/requestToken" + ), + params + ) + defer.returnValue(data) + except MatrixCodeMessageException as e: + logger.info("Proxied requestToken failed with Matrix error: %r", e) + raise SynapseError(e.code, e.msg, e.errcode) except CodeMessageException as e: logger.info("Proxied requestToken failed: %r", e) raise e diff --git a/synapse/handlers/initial_sync.py b/synapse/handlers/initial_sync.py index e0ade4c164..cd33a86599 100644 --- a/synapse/handlers/initial_sync.py +++ b/synapse/handlers/initial_sync.py @@ -19,6 +19,7 @@ from synapse.api.constants import EventTypes, Membership from synapse.api.errors import AuthError, Codes from synapse.events.utils import serialize_event from synapse.events.validator import EventValidator +from synapse.handlers.presence import format_user_presence_state from synapse.streams.config import PaginationConfig from synapse.types import ( UserID, StreamToken, @@ -26,7 +27,7 @@ from synapse.types import ( from synapse.util import unwrapFirstError from synapse.util.async import concurrently_execute from synapse.util.caches.snapshot_cache import SnapshotCache -from synapse.util.logcontext import preserve_fn, preserve_context_over_deferred +from synapse.util.logcontext import make_deferred_yieldable, run_in_background from synapse.visibility import filter_events_for_client from ._base import BaseHandler @@ -162,10 +163,11 @@ class InitialSyncHandler(BaseHandler): lambda states: states[event.event_id] ) - (messages, token), current_state = yield preserve_context_over_deferred( + (messages, token), current_state = yield make_deferred_yieldable( defer.gatherResults( [ - preserve_fn(self.store.get_recent_events_for_room)( + run_in_background( + self.store.get_recent_events_for_room, event.room_id, limit=limit, end_token=room_end_token, @@ -213,7 +215,7 @@ class InitialSyncHandler(BaseHandler): }) d["account_data"] = account_data_events - except: + except Exception: logger.exception("Failed to get snapshot") yield concurrently_execute(handle_room, room_list, 10) @@ -225,9 +227,17 @@ class InitialSyncHandler(BaseHandler): "content": content, }) + now = self.clock.time_msec() + ret = { "rooms": rooms_ret, - "presence": presence, + "presence": [ + { + "type": "m.presence", + "content": format_user_presence_state(event, now), + } + for event in presence + ], "account_data": account_data_events, "receipts": receipt, "end": now_token.to_string(), @@ -382,9 +392,10 @@ class InitialSyncHandler(BaseHandler): presence, receipts, (messages, token) = yield defer.gatherResults( [ - preserve_fn(get_presence)(), - preserve_fn(get_receipts)(), - preserve_fn(self.store.get_recent_events_for_room)( + 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, diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 7a498af5a2..b793fc4df7 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -1,5 +1,6 @@ # -*- 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. @@ -12,31 +13,64 @@ # 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 simplejson +import sys -from twisted.internet import defer +from canonicaljson import encode_canonical_json +import six +from twisted.internet import defer, reactor +from twisted.python.failure import Failure -from synapse.api.constants import EventTypes, Membership -from synapse.api.errors import AuthError, Codes, SynapseError, LimitExceededError +from synapse.api.constants import EventTypes, Membership, MAX_DEPTH +from synapse.api.errors import AuthError, Codes, SynapseError 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.push.action_generator import ActionGenerator from synapse.types import ( UserID, RoomAlias, RoomStreamToken, ) from synapse.util.async import run_on_reactor, ReadWriteLock, Limiter -from synapse.util.logcontext import preserve_fn +from synapse.util.logcontext import run_in_background from synapse.util.metrics import measure_func +from synapse.util.frozenutils import frozendict_json_encoder +from synapse.util.stringutils import random_string from synapse.visibility import filter_events_for_client +from synapse.replication.http.send_event import send_event_to_master from ._base import BaseHandler -from canonicaljson import encode_canonical_json +logger = logging.getLogger(__name__) -import logging -import random -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 MessageHandler(BaseHandler): @@ -46,25 +80,89 @@ class MessageHandler(BaseHandler): self.hs = hs self.state = hs.get_state_handler() self.clock = hs.get_clock() - self.validator = EventValidator() self.pagination_lock = ReadWriteLock() + self._purges_in_progress_by_room = set() + # map from purge id to PurgeStatus + self._purges_by_id = {} - # 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) + def start_purge_history(self, room_id, topological_ordering, + delete_local_events=False): + """Start off a history purge on a room. + + Args: + room_id (str): The room to purge from + + topological_ordering (int): minimum topo ordering to preserve + 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, topological_ordering, delete_local_events, + ) + return purge_id @defer.inlineCallbacks - def purge_history(self, room_id, event_id): - event = yield self.store.get_event(event_id) + def _purge_history(self, purge_id, room_id, topological_ordering, + delete_local_events): + """Carry out a history purge on a room. - if event.room_id != room_id: - raise SynapseError(400, "Event is for wrong room.") + Args: + purge_id (str): The id for this purge + room_id (str): The room to purge from + topological_ordering (int): minimum topo ordering to preserve + delete_local_events (bool): True to delete local events as well as + remote ones - depth = event.depth + 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, topological_ordering, 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] + reactor.callLater(24 * 3600, clear_purge) + + def get_purge_status(self, purge_id): + """Get the current status of an active purge - with (yield self.pagination_lock.write(room_id)): - yield self.store.delete_old_state(room_id, depth) + 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, @@ -175,7 +273,167 @@ class MessageHandler(BaseHandler): defer.returnValue(chunk) @defer.inlineCallbacks - def create_event(self, event_dict, token_id=None, txn_id=None, prev_event_ids=None): + def get_room_data(self, user_id=None, room_id=None, + event_type=None, state_key="", is_guest=False): + """ Get data from a room. + + Args: + event : The room path event + Returns: + The path data content. + Raises: + SynapseError if something went wrong. + """ + membership, membership_event_id = yield self._check_in_room_or_world_readable( + room_id, user_id + ) + + if membership == Membership.JOIN: + data = yield self.state_handler.get_current_state( + room_id, event_type, state_key + ) + elif membership == Membership.LEAVE: + key = (event_type, state_key) + room_state = yield self.store.get_state_for_events( + [membership_event_id], [key] + ) + data = room_state[membership_event_id].get(key) + + 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 + left the room return the state events from when they left. + + Args: + user_id(str): The user requesting state events. + room_id(str): The room ID to get all state events from. + Returns: + A list of dicts representing state events. [{}, {}, {}] + """ + membership, membership_event_id = yield self._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) + elif membership == Membership.LEAVE: + room_state = yield self.store.get_state_for_events( + [membership_event_id], None + ) + room_state = room_state[membership_event_id] + + now = self.clock.time_msec() + defer.returnValue( + [serialize_event(c, now) for c in room_state.values()] + ) + + @defer.inlineCallbacks + def get_joined_members(self, requester, room_id): + """Get all the joined members in the room and their profile information. + + If the user has left the room return the state events from when they left. + + Args: + requester(Requester): The user requesting state events. + room_id(str): The room ID to get all state events from. + Returns: + A dict of user_id to profile info + """ + user_id = requester.user.to_string() + 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( + room_id, user_id + ) + if membership != Membership.JOIN: + raise NotImplementedError( + "Getting joined members after leaving is not implemented" + ) + + users_with_profile = yield self.state.get_current_user_in_room(room_id) + + # If this is an AS, double check that they are allowed to see the members. + # This can either be because the AS user is in the room or becuase there + # is a user in the room that the AS is "interested in" + if requester.app_service and user_id not in users_with_profile: + for uid in users_with_profile: + if requester.app_service.is_interested_in_user(uid): + break + else: + # Loop fell through, AS has no interested users in room + raise AuthError(403, "Appservice not in room") + + defer.returnValue({ + user_id: { + "avatar_url": profile.avatar_url, + "display_name": profile.display_name, + } + for user_id, profile in users_with_profile.iteritems() + }) + + +class EventCreationHandler(object): + def __init__(self, hs): + self.hs = hs + self.auth = hs.get_auth() + self.store = hs.get_datastore() + self.state = hs.get_state_handler() + self.clock = hs.get_clock() + self.validator = EventValidator() + self.profile_handler = hs.get_profile_handler() + self.event_builder_factory = hs.get_event_builder_factory() + self.server_name = hs.hostname + self.ratelimiter = hs.get_ratelimiter() + self.notifier = hs.get_notifier() + self.config = hs.config + + self.http_client = hs.get_simple_http_client() + + # This is only used to get at ratelimit function, and maybe_kick_guest_users + self.base_handler = BaseHandler(hs) + + self.pusher_pool = hs.get_pusherpool() + + # 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.action_generator = hs.get_action_generator() + + self.spam_checker = hs.get_spam_checker() + + @defer.inlineCallbacks + def create_event(self, requester, event_dict, token_id=None, txn_id=None, + prev_events_and_hashes=None): """ Given a dict from a client, create a new event. @@ -185,49 +443,56 @@ class MessageHandler(BaseHandler): Adds display names to Join membership events. Args: + requester event_dict (dict): An entire event token_id (str) txn_id (str) - prev_event_ids (list): The prev event ids to use when creating the event + + prev_events_and_hashes (list[(str, dict[str, str], int)]|None): + 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. + + If None, they will be requested from the database. Returns: Tuple of created event (FrozenEvent), Context """ builder = self.event_builder_factory.new(event_dict) - with (yield self.limiter.queue(builder.room_id)): - self.validator.validate_new(builder) - - if builder.type == EventTypes.Member: - membership = builder.content.get("membership", None) - target = UserID.from_string(builder.state_key) - - if membership in {Membership.JOIN, Membership.INVITE}: - # If event doesn't include a display name, add one. - profile = self.hs.get_handlers().profile_handler - content = builder.content - - try: - if "displayname" not in content: - content["displayname"] = yield profile.get_displayname(target) - if "avatar_url" not in content: - content["avatar_url"] = yield profile.get_avatar_url(target) - except Exception as e: - logger.info( - "Failed to get profile information for %r: %s", - target, e - ) + self.validator.validate_new(builder) + + if builder.type == EventTypes.Member: + membership = builder.content.get("membership", None) + target = UserID.from_string(builder.state_key) + + if membership in {Membership.JOIN, Membership.INVITE}: + # If event doesn't include a display name, add one. + profile = self.profile_handler + content = builder.content + + try: + if "displayname" not in content: + content["displayname"] = yield profile.get_displayname(target) + if "avatar_url" not in content: + content["avatar_url"] = yield profile.get_avatar_url(target) + except Exception as e: + logger.info( + "Failed to get profile information for %r: %s", + target, e + ) - if token_id is not None: - builder.internal_metadata.token_id = token_id + if token_id is not None: + builder.internal_metadata.token_id = token_id - if txn_id is not None: - builder.internal_metadata.txn_id = txn_id + if txn_id is not None: + builder.internal_metadata.txn_id = txn_id - event, context = yield self._create_new_client_event( - builder=builder, - prev_event_ids=prev_event_ids, - ) + event, context = yield self.create_new_client_event( + builder=builder, + requester=requester, + prev_events_and_hashes=prev_events_and_hashes, + ) defer.returnValue((event, context)) @@ -248,21 +513,6 @@ class MessageHandler(BaseHandler): "Tried to send member event through non-member codepath" ) - # We check here if we are currently being rate limited, so that we - # don't do unnecessary work. We check again just before we actually - # send the event. - time_now = self.clock.time() - allowed, time_allowed = self.ratelimiter.send_message( - event.sender, time_now, - msg_rate_hz=self.hs.config.rc_messages_per_second, - burst_count=self.hs.config.rc_message_burst_count, - update=False, - ) - if not allowed: - raise LimitExceededError( - retry_after_ms=int(1000 * (time_allowed - time_now)), - ) - user = UserID.from_string(event.sender) assert self.hs.is_mine(user), "User must be our own: %s" % (user,) @@ -279,12 +529,6 @@ class MessageHandler(BaseHandler): ratelimit=ratelimit, ) - if event.type == EventTypes.Message: - presence = self.hs.get_presence_handler() - # We don't want to block sending messages on any presence code. This - # matters as sometimes presence code can take a while. - preserve_fn(presence.bump_presence_active_time)(user) - @defer.inlineCallbacks def deduplicate_state_event(self, event, context): """ @@ -318,144 +562,87 @@ class MessageHandler(BaseHandler): See self.create_event and self.send_nonmember_event. """ - event, context = yield self.create_event( - event_dict, - token_id=requester.access_token_id, - txn_id=txn_id - ) - yield self.send_nonmember_event( - requester, - event, - context, - ratelimit=ratelimit, - ) - defer.returnValue(event) - - @defer.inlineCallbacks - def get_room_data(self, user_id=None, room_id=None, - event_type=None, state_key="", is_guest=False): - """ Get data from a room. - - Args: - event : The room path event - Returns: - The path data content. - Raises: - SynapseError if something went wrong. - """ - membership, membership_event_id = yield self._check_in_room_or_world_readable( - room_id, user_id - ) - if membership == Membership.JOIN: - data = yield self.state_handler.get_current_state( - room_id, event_type, state_key + # We limit the number of concurrent event sends in a room so that we + # don't fork the DAG too much. If we don't limit then we can end up in + # a situation where event persistence can't keep up, causing + # extremities to pile up, which in turn leads to state resolution + # taking longer. + with (yield self.limiter.queue(event_dict["room_id"])): + event, context = yield self.create_event( + requester, + event_dict, + token_id=requester.access_token_id, + txn_id=txn_id ) - elif membership == Membership.LEAVE: - key = (event_type, state_key) - room_state = yield self.store.get_state_for_events( - [membership_event_id], [key] - ) - data = room_state[membership_event_id].get(key) - defer.returnValue(data) + spam_error = self.spam_checker.check_event_for_spam(event) + if spam_error: + if not isinstance(spam_error, basestring): + spam_error = "Spam is not permitted here" + raise SynapseError( + 403, spam_error, Codes.FORBIDDEN + ) - @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 + yield self.send_nonmember_event( + requester, + event, + context, + ratelimit=ratelimit, ) + defer.returnValue(event) + @measure_func("create_new_client_event") @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 - left the room return the state events from when they left. + def create_new_client_event(self, builder, requester=None, + prev_events_and_hashes=None): + """Create a new event for a local client Args: - user_id(str): The user requesting state events. - room_id(str): The room ID to get all state events from. + builder (EventBuilder): + + requester (synapse.types.Requester|None): + + prev_events_and_hashes (list[(str, dict[str, str], int)]|None): + 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. + + If None, they will be requested from the database. + Returns: - A list of dicts representing state events. [{}, {}, {}] + Deferred[(synapse.events.EventBase, synapse.events.snapshot.EventContext)] """ - membership, membership_event_id = yield self._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) - elif membership == Membership.LEAVE: - room_state = yield self.store.get_state_for_events( - [membership_event_id], None + if prev_events_and_hashes is not None: + assert len(prev_events_and_hashes) <= 10, \ + "Attempting to create an event with %i prev_events" % ( + len(prev_events_and_hashes), ) - room_state = room_state[membership_event_id] - - now = self.clock.time_msec() - defer.returnValue( - [serialize_event(c, now) for c in room_state.values()] - ) - - @measure_func("_create_new_client_event") - @defer.inlineCallbacks - def _create_new_client_event(self, builder, prev_event_ids=None): - if prev_event_ids: - prev_events = yield self.store.add_event_hashes(prev_event_ids) - prev_max_depth = yield self.store.get_max_depth_of_events(prev_event_ids) - depth = prev_max_depth + 1 else: - latest_ret = yield self.store.get_latest_event_ids_and_hashes_in_room( - builder.room_id, - ) - - # We want to limit the max number of prev events we point to in our - # new event - if len(latest_ret) > 10: - # Sort by reverse depth, so we point to the most recent. - latest_ret.sort(key=lambda a: -a[2]) - new_latest_ret = latest_ret[:5] - - # We also randomly point to some of the older events, to make - # sure that we don't completely ignore the older events. - if latest_ret[5:]: - sample_size = min(5, len(latest_ret[5:])) - new_latest_ret.extend(random.sample(latest_ret[5:], sample_size)) - latest_ret = new_latest_ret - - if latest_ret: - depth = max([d for _, _, d in latest_ret]) + 1 - else: - depth = 1 + prev_events_and_hashes = \ + yield self.store.get_prev_events_for_room(builder.room_id) + + if prev_events_and_hashes: + depth = max([d for _, _, d in prev_events_and_hashes]) + 1 + # we cap depth of generated events, to ensure that they are not + # rejected by other servers (and so that they can be persisted in + # the db) + depth = min(depth, MAX_DEPTH) + else: + depth = 1 - prev_events = [ - (event_id, prev_hashes) - for event_id, prev_hashes, _ in latest_ret - ] + prev_events = [ + (event_id, prev_hashes) + for event_id, prev_hashes, _ in prev_events_and_hashes + ] builder.prev_events = prev_events builder.depth = depth - state_handler = self.state_handler - - context = yield state_handler.compute_event_context(builder) + context = yield self.state.compute_event_context(builder) + if requester: + context.app_service = requester.app_service if builder.is_state(): builder.prev_state = yield self.store.add_event_hashes( @@ -488,12 +675,21 @@ class MessageHandler(BaseHandler): event, context, ratelimit=True, - extra_users=[] + extra_users=[], ): - # We now need to go and hit out to wherever we need to hit out to. + """Processes a new event. This includes checking auth, persisting it, + notifying users, sending to remote servers, etc. - if ratelimit: - self.ratelimit(requester) + If called from a worker will hit out to the master process for final + processing. + + Args: + requester (Requester) + event (FrozenEvent) + context (EventContext) + ratelimit (bool) + extra_users (list(UserID)): Any extra users to notify about event + """ try: yield self.auth.check_from_context(event, context) @@ -501,7 +697,72 @@ class MessageHandler(BaseHandler): logger.warn("Denying new event %r because %s", event, err) raise err - yield self.maybe_kick_guest_users(event, context) + # Ensure that we can round trip before trying to persist in db + try: + dump = frozendict_json_encoder.encode(event.content) + simplejson.loads(dump) + except Exception: + logger.exception("Failed to encode content: %r", event.content) + raise + + yield self.action_generator.handle_push_actions_for_event( + event, context + ) + + try: + # If we're a worker we need to hit out to the master. + if self.config.worker_app: + yield send_event_to_master( + self.http_client, + host=self.config.worker_replication_host, + port=self.config.worker_replication_http_port, + requester=requester, + event=event, + context=context, + ratelimit=ratelimit, + extra_users=extra_users, + ) + return + + yield self.persist_and_notify_client_event( + requester, + event, + context, + ratelimit=ratelimit, + extra_users=extra_users, + ) + except: # noqa: E722, as we reraise the exception this is fine. + # Ensure that we actually remove the entries in the push actions + # staging area, if we calculated them. + tp, value, tb = sys.exc_info() + + run_in_background( + self.store.remove_push_actions_from_staging, + event.event_id, + ) + + six.reraise(tp, value, tb) + + @defer.inlineCallbacks + def persist_and_notify_client_event( + self, + requester, + event, + context, + ratelimit=True, + extra_users=[], + ): + """Called when we have fully built the event, have already + calculated the push actions for the event, and checked auth. + + This should only be run on master. + """ + assert not self.config.worker_app + + if ratelimit: + yield self.base_handler.ratelimit(requester) + + yield self.base_handler.maybe_kick_guest_users(event, context) if event.type == EventTypes.CanonicalAlias: # Check the alias is acually valid (at this time at least) @@ -531,9 +792,9 @@ class MessageHandler(BaseHandler): state_to_include_ids = [ e_id - for k, e_id in context.current_state_ids.items() + for k, e_id in context.current_state_ids.iteritems() if k[0] in self.hs.config.room_invite_state_types - or k[0] == EventTypes.Member and k[1] == event.sender + or k == (EventTypes.Member, event.sender) ] state_to_include = yield self.store.get_events(state_to_include_ids) @@ -545,7 +806,7 @@ class MessageHandler(BaseHandler): "content": e.content, "sender": e.sender, } - for e in state_to_include.values() + for e in state_to_include.itervalues() ] invitee = UserID.from_string(event.state_key) @@ -594,30 +855,39 @@ class MessageHandler(BaseHandler): "Changing the room create event is forbidden", ) - action_generator = ActionGenerator(self.hs) - yield action_generator.handle_push_actions_for_event( - event, context - ) - (event_stream_id, max_stream_id) = yield self.store.persist_event( event, context=context ) # this intentionally does not yield: we don't care about the result # and don't need to wait for it. - preserve_fn(self.hs.get_pusherpool().on_new_notifications)( + run_in_background( + self.pusher_pool.on_new_notifications, event_stream_id, max_stream_id ) @defer.inlineCallbacks def _notify(): yield run_on_reactor() - yield self.notifier.on_new_room_event( - event, event_stream_id, max_stream_id, - extra_users=extra_users - ) + try: + self.notifier.on_new_room_event( + event, event_stream_id, max_stream_id, + extra_users=extra_users + ) + except Exception: + logger.exception("Error notifying about new room event") - preserve_fn(_notify)() + run_in_background(_notify) - # If invite, remove room_state from unsigned before sending. - event.unsigned.pop("invite_room_state", None) + if event.type == EventTypes.Message: + # We don't want to block sending messages on any presence code. This + # matters as sometimes presence code can take a while. + run_in_background(self._bump_active_time, requester.user) + + @defer.inlineCallbacks + def _bump_active_time(self, user): + try: + presence = self.hs.get_presence_handler() + yield presence.bump_presence_active_time(user) + except Exception: + logger.exception("Error bumping presence active time") diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index da610e430f..585f3e4da2 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -29,7 +29,9 @@ from synapse.api.errors import SynapseError from synapse.api.constants import PresenceState from synapse.storage.presence import UserPresenceState -from synapse.util.logcontext import preserve_fn +from synapse.util.caches.descriptors import cachedInlineCallbacks +from synapse.util.async import Linearizer +from synapse.util.logcontext import run_in_background from synapse.util.logutils import log_function from synapse.util.metrics import Measure from synapse.util.wheel_timer import WheelTimer @@ -91,29 +93,30 @@ class PresenceHandler(object): self.store = hs.get_datastore() self.wheel_timer = WheelTimer() self.notifier = hs.get_notifier() - self.replication = hs.get_replication_layer() self.federation = hs.get_federation_sender() self.state = hs.get_state_handler() - self.replication.register_edu_handler( + federation_registry = hs.get_federation_registry() + + federation_registry.register_edu_handler( "m.presence", self.incoming_presence ) - self.replication.register_edu_handler( + federation_registry.register_edu_handler( "m.presence_invite", lambda origin, content: self.invite_presence( observed_user=UserID.from_string(content["observed_user"]), observer_user=UserID.from_string(content["observer_user"]), ) ) - self.replication.register_edu_handler( + federation_registry.register_edu_handler( "m.presence_accept", lambda origin, content: self.accept_presence( observed_user=UserID.from_string(content["observed_user"]), observer_user=UserID.from_string(content["observer_user"]), ) ) - self.replication.register_edu_handler( + federation_registry.register_edu_handler( "m.presence_deny", lambda origin, content: self.deny_presence( observed_user=UserID.from_string(content["observed_user"]), @@ -186,6 +189,7 @@ class PresenceHandler(object): # process_id to millisecond timestamp last updated. self.external_process_to_current_syncs = {} self.external_process_last_updated_ms = {} + self.external_sync_linearizer = Linearizer(name="external_sync_linearizer") # Start a LoopingCall in 30s that fires every 5s. # The initial delay is to allow disconnected clients a chance to @@ -251,6 +255,14 @@ class PresenceHandler(object): logger.info("Finished _persist_unpersisted_changes") @defer.inlineCallbacks + def _update_states_and_catch_exception(self, new_states): + try: + res = yield self._update_states(new_states) + defer.returnValue(res) + except Exception: + logger.exception("Error updating presence") + + @defer.inlineCallbacks def _update_states(self, new_states): """Updates presence of users. Sets the appropriate timeouts. Pokes the notifier and federation if and only if the changed presence state @@ -315,11 +327,7 @@ class PresenceHandler(object): if to_federation_ping: federation_presence_out_counter.inc_by(len(to_federation_ping)) - _, _, hosts_to_states = yield self._get_interested_parties( - to_federation_ping.values() - ) - - self._push_to_remotes(hosts_to_states) + self._push_to_remotes(to_federation_ping.values()) def _handle_timeouts(self): """Checks the presence of users that have timed out and updates as @@ -364,8 +372,8 @@ class PresenceHandler(object): now=now, ) - preserve_fn(self._update_states)(changes) - except: + run_in_background(self._update_states_and_catch_exception, changes) + except Exception: logger.exception("Exception in _handle_timeouts loop") @defer.inlineCallbacks @@ -422,20 +430,23 @@ class PresenceHandler(object): @defer.inlineCallbacks def _end(): - if affect_presence: + try: self.user_to_num_current_syncs[user_id] -= 1 prev_state = yield self.current_state_for_user(user_id) yield self._update_states([prev_state.copy_and_replace( last_user_sync_ts=self.clock.time_msec(), )]) + except Exception: + logger.exception("Error updating presence after sync") @contextmanager def _user_syncing(): try: yield finally: - preserve_fn(_end)() + if affect_presence: + run_in_background(_end) defer.returnValue(_user_syncing()) @@ -508,6 +519,73 @@ class PresenceHandler(object): self.external_process_to_current_syncs[process_id] = syncing_user_ids @defer.inlineCallbacks + def update_external_syncs_row(self, process_id, user_id, is_syncing, sync_time_msec): + """Update the syncing users for an external process as a delta. + + Args: + process_id (str): An identifier for the process the users are + syncing against. This allows synapse to process updates + as user start and stop syncing against a given process. + user_id (str): The user who has started or stopped syncing + is_syncing (bool): Whether or not the user is now syncing + sync_time_msec(int): Time in ms when the user was last syncing + """ + with (yield self.external_sync_linearizer.queue(process_id)): + prev_state = yield self.current_state_for_user(user_id) + + process_presence = self.external_process_to_current_syncs.setdefault( + process_id, set() + ) + + updates = [] + if is_syncing and user_id not in process_presence: + if prev_state.state == PresenceState.OFFLINE: + updates.append(prev_state.copy_and_replace( + state=PresenceState.ONLINE, + last_active_ts=sync_time_msec, + last_user_sync_ts=sync_time_msec, + )) + else: + updates.append(prev_state.copy_and_replace( + last_user_sync_ts=sync_time_msec, + )) + process_presence.add(user_id) + elif user_id in process_presence: + updates.append(prev_state.copy_and_replace( + last_user_sync_ts=sync_time_msec, + )) + + if not is_syncing: + process_presence.discard(user_id) + + if updates: + yield self._update_states(updates) + + self.external_process_last_updated_ms[process_id] = self.clock.time_msec() + + @defer.inlineCallbacks + def update_external_syncs_clear(self, process_id): + """Marks all users that had been marked as syncing by a given process + as offline. + + Used when the process has stopped/disappeared. + """ + with (yield self.external_sync_linearizer.queue(process_id)): + process_presence = self.external_process_to_current_syncs.pop( + process_id, set() + ) + prev_states = yield self.current_state_for_users(process_presence) + time_now_ms = self.clock.time_msec() + + yield self._update_states([ + prev_state.copy_and_replace( + last_user_sync_ts=time_now_ms, + ) + for prev_state in prev_states.itervalues() + ]) + self.external_process_last_updated_ms.pop(process_id, None) + + @defer.inlineCallbacks def current_state_for_user(self, user_id): """Get the current presence state for a user. """ @@ -526,14 +604,14 @@ class PresenceHandler(object): for user_id in user_ids } - missing = [user_id for user_id, state in states.items() if not state] + missing = [user_id for user_id, state in states.iteritems() if not state] if missing: # There are things not in our in memory cache. Lets pull them out of # the database. res = yield self.store.get_presence_for_users(missing) states.update(res) - missing = [user_id for user_id, state in states.items() if not state] + missing = [user_id for user_id, state in states.iteritems() if not state] if missing: new = { user_id: UserPresenceState.default(user_id) @@ -545,89 +623,39 @@ class PresenceHandler(object): defer.returnValue(states) @defer.inlineCallbacks - def _get_interested_parties(self, states, calculate_remote_hosts=True): - """Given a list of states return which entities (rooms, users, servers) - are interested in the given states. - - Returns: - 3-tuple: `(room_ids_to_states, users_to_states, hosts_to_states)`, - with each item being a dict of `entity_name` -> `[UserPresenceState]` - """ - room_ids_to_states = {} - users_to_states = {} - for state in states: - events = yield self.store.get_rooms_for_user(state.user_id) - for e in events: - room_ids_to_states.setdefault(e.room_id, []).append(state) - - plist = yield self.store.get_presence_list_observers_accepted(state.user_id) - for u in plist: - users_to_states.setdefault(u, []).append(state) - - # Always notify self - users_to_states.setdefault(state.user_id, []).append(state) - - hosts_to_states = {} - if calculate_remote_hosts: - for room_id, states in room_ids_to_states.items(): - local_states = filter(lambda s: self.is_mine_id(s.user_id), states) - if not local_states: - continue - - users = yield self.store.get_users_in_room(room_id) - hosts = set(get_domain_from_id(u) for u in users) - - for host in hosts: - hosts_to_states.setdefault(host, []).extend(local_states) - - for user_id, states in users_to_states.items(): - local_states = filter(lambda s: self.is_mine_id(s.user_id), states) - if not local_states: - continue - - host = get_domain_from_id(user_id) - hosts_to_states.setdefault(host, []).extend(local_states) - - # TODO: de-dup hosts_to_states, as a single host might have multiple - # of same presence - - defer.returnValue((room_ids_to_states, users_to_states, hosts_to_states)) - - @defer.inlineCallbacks def _persist_and_notify(self, states): """Persist states in the database, poke the notifier and send to interested remote servers """ stream_id, max_token = yield self.store.update_presence(states) - parties = yield self._get_interested_parties(states) - room_ids_to_states, users_to_states, hosts_to_states = parties + parties = yield get_interested_parties(self.store, states) + room_ids_to_states, users_to_states = parties self.notifier.on_new_event( "presence_key", stream_id, rooms=room_ids_to_states.keys(), - users=[UserID.from_string(u) for u in users_to_states.keys()] + users=[UserID.from_string(u) for u in users_to_states] ) - self._push_to_remotes(hosts_to_states) + self._push_to_remotes(states) @defer.inlineCallbacks def notify_for_states(self, state, stream_id): - parties = yield self._get_interested_parties([state]) - room_ids_to_states, users_to_states, hosts_to_states = parties + parties = yield get_interested_parties(self.store, [state]) + room_ids_to_states, users_to_states = parties self.notifier.on_new_event( "presence_key", stream_id, rooms=room_ids_to_states.keys(), - users=[UserID.from_string(u) for u in users_to_states.keys()] + users=[UserID.from_string(u) for u in users_to_states] ) - def _push_to_remotes(self, hosts_to_states): + def _push_to_remotes(self, states): """Sends state updates to remote servers. Args: - hosts_to_states (dict): Mapping `server_name` -> `[UserPresenceState]` + states (list(UserPresenceState)) """ - for host, states in hosts_to_states.items(): - self.federation.send_presence(host, states) + self.federation.send_presence(states) @defer.inlineCallbacks def incoming_presence(self, origin, content): @@ -719,9 +747,7 @@ class PresenceHandler(object): for state in updates ]) else: - defer.returnValue([ - format_user_presence_state(state, now) for state in updates - ]) + defer.returnValue(updates) @defer.inlineCallbacks def set_state(self, target_user, state, ignore_status_msg=False): @@ -766,18 +792,17 @@ class PresenceHandler(object): # don't need to send to local clients here, as that is done as part # of the event stream/sync. # TODO: Only send to servers not already in the room. - user_ids = yield self.store.get_users_in_room(room_id) if self.is_mine(user): state = yield self.current_state_for_user(user.to_string()) - hosts = set(get_domain_from_id(u) for u in user_ids) - self._push_to_remotes({host: (state,) for host in hosts}) + self._push_to_remotes([state]) else: + user_ids = yield self.store.get_users_in_room(room_id) user_ids = filter(self.is_mine_id, user_ids) states = yield self.current_state_for_users(user_ids) - self._push_to_remotes({user.domain: states.values()}) + self._push_to_remotes(states.values()) @defer.inlineCallbacks def get_presence_list(self, observer_user, accepted=None): @@ -795,6 +820,9 @@ class PresenceHandler(object): as_event=False, ) + now = self.clock.time_msec() + results[:] = [format_user_presence_state(r, now) for r in results] + is_accepted = { row["observed_user_id"]: row["accepted"] for row in presence_list } @@ -847,6 +875,7 @@ class PresenceHandler(object): ) state_dict = yield self.get_state(observed_user, as_event=False) + state_dict = format_user_presence_state(state_dict, self.clock.time_msec()) self.federation.send_edu( destination=observer_user.domain, @@ -910,11 +939,12 @@ class PresenceHandler(object): def is_visible(self, observed_user, observer_user): """Returns whether a user can see another user's presence. """ - observer_rooms = yield self.store.get_rooms_for_user(observer_user.to_string()) - observed_rooms = yield self.store.get_rooms_for_user(observed_user.to_string()) - - observer_room_ids = set(r.room_id for r in observer_rooms) - observed_room_ids = set(r.room_id for r in observed_rooms) + observer_room_ids = yield self.store.get_rooms_for_user( + observer_user.to_string() + ) + observed_room_ids = yield self.store.get_rooms_for_user( + observed_user.to_string() + ) if observer_room_ids & observed_room_ids: defer.returnValue(True) @@ -979,14 +1009,18 @@ def should_notify(old_state, new_state): return False -def format_user_presence_state(state, now): +def format_user_presence_state(state, now, include_user_id=True): """Convert UserPresenceState to a format that can be sent down to clients and to other servers. + + The "user_id" is optional so that this function can be used to format presence + updates for client /sync responses and for federation /send requests. """ content = { "presence": state.state, - "user_id": state.user_id, } + if include_user_id: + content["user_id"] = state.user_id if state.last_active_ts: content["last_active_ago"] = now - state.last_active_ts if state.status_msg and state.state != PresenceState.OFFLINE: @@ -1025,7 +1059,6 @@ class PresenceEventSource(object): # sending down the rare duplicate is not a concern. with Measure(self.clock, "presence.get_new_events"): - user_id = user.to_string() if from_key is not None: from_key = int(from_key) @@ -1034,18 +1067,7 @@ class PresenceEventSource(object): max_token = self.store.get_current_presence_token() - plist = yield self.store.get_presence_list_accepted(user.localpart) - users_interested_in = set(row["observed_user_id"] for row in plist) - users_interested_in.add(user_id) # So that we receive our own presence - - users_who_share_room = yield self.store.get_users_who_share_room_with_user( - user_id - ) - users_interested_in.update(users_who_share_room) - - if explicit_room_id: - user_ids = yield self.store.get_users_in_room(explicit_room_id) - users_interested_in.update(user_ids) + users_interested_in = yield self._get_interested_in(user, explicit_room_id) user_ids_changed = set() changed = None @@ -1073,16 +1095,13 @@ class PresenceEventSource(object): updates = yield presence.current_state_for_users(user_ids_changed) - now = self.clock.time_msec() - - defer.returnValue(([ - { - "type": "m.presence", - "content": format_user_presence_state(s, now), - } - for s in updates.values() - if include_offline or s.state != PresenceState.OFFLINE - ], max_token)) + if include_offline: + defer.returnValue((updates.values(), max_token)) + else: + defer.returnValue(([ + s for s in updates.itervalues() + if s.state != PresenceState.OFFLINE + ], max_token)) def get_current_key(self): return self.store.get_current_presence_token() @@ -1090,6 +1109,31 @@ class PresenceEventSource(object): def get_pagination_rows(self, user, pagination_config, key): return self.get_new_events(user, from_key=None, include_offline=False) + @cachedInlineCallbacks(num_args=2, cache_context=True) + def _get_interested_in(self, user, explicit_room_id, cache_context): + """Returns the set of users that the given user should see presence + updates for + """ + user_id = user.to_string() + plist = yield self.store.get_presence_list_accepted( + user.localpart, on_invalidate=cache_context.invalidate, + ) + users_interested_in = set(row["observed_user_id"] for row in plist) + users_interested_in.add(user_id) # So that we receive our own presence + + users_who_share_room = yield self.store.get_users_who_share_room_with_user( + user_id, on_invalidate=cache_context.invalidate, + ) + users_interested_in.update(users_who_share_room) + + if explicit_room_id: + user_ids = yield self.store.get_users_in_room( + explicit_room_id, on_invalidate=cache_context.invalidate, + ) + users_interested_in.update(user_ids) + + defer.returnValue(users_interested_in) + def handle_timeouts(user_states, is_mine_fn, syncing_user_ids, now): """Checks the presence of users that have timed out and updates as @@ -1157,14 +1201,17 @@ def handle_timeout(state, is_mine, syncing_user_ids, now): # If there are have been no sync for a while (and none ongoing), # set presence to offline if user_id not in syncing_user_ids: - if now - state.last_user_sync_ts > SYNC_ONLINE_TIMEOUT: + # If the user has done something recently but hasn't synced, + # don't set them as offline. + sync_or_active = max(state.last_user_sync_ts, state.last_active_ts) + if now - sync_or_active > SYNC_ONLINE_TIMEOUT: state = state.copy_and_replace( state=PresenceState.OFFLINE, status_msg=None, ) changed = True else: - # We expect to be poked occaisonally by the other side. + # We expect to be poked occasionally by the other side. # This is to protect against forgetful/buggy servers, so that # no one gets stuck online forever. if now - state.last_federation_update_ts > FEDERATION_TIMEOUT: @@ -1255,3 +1302,66 @@ def handle_update(prev_state, new_state, is_mine, wheel_timer, now): persist_and_notify = True return new_state, persist_and_notify, federation_ping + + +@defer.inlineCallbacks +def get_interested_parties(store, states): + """Given a list of states return which entities (rooms, users) + are interested in the given states. + + Args: + states (list(UserPresenceState)) + + Returns: + 2-tuple: `(room_ids_to_states, users_to_states)`, + with each item being a dict of `entity_name` -> `[UserPresenceState]` + """ + room_ids_to_states = {} + users_to_states = {} + for state in states: + room_ids = yield store.get_rooms_for_user(state.user_id) + for room_id in room_ids: + room_ids_to_states.setdefault(room_id, []).append(state) + + plist = yield store.get_presence_list_observers_accepted(state.user_id) + for u in plist: + users_to_states.setdefault(u, []).append(state) + + # Always notify self + users_to_states.setdefault(state.user_id, []).append(state) + + defer.returnValue((room_ids_to_states, users_to_states)) + + +@defer.inlineCallbacks +def get_interested_remotes(store, states, state_handler): + """Given a list of presence states figure out which remote servers + should be sent which. + + All the presence states should be for local users only. + + Args: + store (DataStore) + states (list(UserPresenceState)) + + Returns: + Deferred list of ([destinations], [UserPresenceState]), where for + each row the list of UserPresenceState should be sent to each + destination + """ + hosts_and_states = [] + + # First we look up the rooms each user is in (as well as any explicit + # subscriptions), then for each distinct room we look up the remote + # hosts in those rooms. + room_ids_to_states, users_to_states = yield get_interested_parties(store, states) + + for room_id, states in room_ids_to_states.iteritems(): + hosts = yield state_handler.get_current_hosts_in_room(room_id) + hosts_and_states.append((hosts, states)) + + for user_id, states in users_to_states.iteritems(): + host = get_domain_from_id(user_id) + hosts_and_states.append(([host], states)) + + defer.returnValue(hosts_and_states) diff --git a/synapse/handlers/profile.py b/synapse/handlers/profile.py index 87f74dfb8e..3465a787ab 100644 --- a/synapse/handlers/profile.py +++ b/synapse/handlers/profile.py @@ -17,25 +17,87 @@ import logging from twisted.internet import defer -import synapse.types from synapse.api.errors import SynapseError, AuthError, CodeMessageException -from synapse.types import UserID +from synapse.types import UserID, get_domain_from_id from ._base import BaseHandler - logger = logging.getLogger(__name__) class ProfileHandler(BaseHandler): + PROFILE_UPDATE_MS = 60 * 1000 + PROFILE_UPDATE_EVERY_MS = 24 * 60 * 60 * 1000 def __init__(self, hs): super(ProfileHandler, self).__init__(hs) - self.federation = hs.get_replication_layer() - self.federation.register_query_handler( + self.federation = hs.get_federation_client() + hs.get_federation_registry().register_query_handler( "profile", self.on_profile_query ) + self.user_directory_handler = hs.get_user_directory_handler() + + if hs.config.worker_app is None: + self.clock.looping_call( + self._update_remote_profile_cache, self.PROFILE_UPDATE_MS, + ) + + @defer.inlineCallbacks + def get_profile(self, user_id): + target_user = UserID.from_string(user_id) + if self.hs.is_mine(target_user): + displayname = yield self.store.get_profile_displayname( + target_user.localpart + ) + avatar_url = yield self.store.get_profile_avatar_url( + target_user.localpart + ) + + defer.returnValue({ + "displayname": displayname, + "avatar_url": avatar_url, + }) + else: + try: + result = yield self.federation.make_query( + destination=target_user.domain, + query_type="profile", + args={ + "user_id": user_id, + }, + ignore_backoff=True, + ) + defer.returnValue(result) + except CodeMessageException as e: + if e.code != 404: + logger.exception("Failed to get displayname") + + raise + + @defer.inlineCallbacks + def get_profile_from_cache(self, user_id): + """Get the profile information from our local cache. If the user is + ours then the profile information will always be corect. Otherwise, + it may be out of date/missing. + """ + target_user = UserID.from_string(user_id) + if self.hs.is_mine(target_user): + displayname = yield self.store.get_profile_displayname( + target_user.localpart + ) + avatar_url = yield self.store.get_profile_avatar_url( + target_user.localpart + ) + + defer.returnValue({ + "displayname": displayname, + "avatar_url": avatar_url, + }) + else: + profile = yield self.store.get_from_remote_profile_cache(user_id) + defer.returnValue(profile or {}) + @defer.inlineCallbacks def get_displayname(self, target_user): if self.hs.is_mine(target_user): @@ -52,14 +114,15 @@ class ProfileHandler(BaseHandler): args={ "user_id": target_user.to_string(), "field": "displayname", - } + }, + ignore_backoff=True, ) except CodeMessageException as e: if e.code != 404: logger.exception("Failed to get displayname") raise - except: + except Exception: logger.exception("Failed to get displayname") else: defer.returnValue(result["displayname"]) @@ -81,7 +144,13 @@ class ProfileHandler(BaseHandler): target_user.localpart, new_displayname ) - yield self._update_join_states(requester) + if self.hs.config.user_directory_search_all_users: + profile = yield self.store.get_profileinfo(target_user.localpart) + yield self.user_directory_handler.handle_local_profile_change( + target_user.to_string(), profile + ) + + yield self._update_join_states(requester, target_user) @defer.inlineCallbacks def get_avatar_url(self, target_user): @@ -99,13 +168,14 @@ class ProfileHandler(BaseHandler): args={ "user_id": target_user.to_string(), "field": "avatar_url", - } + }, + ignore_backoff=True, ) except CodeMessageException as e: if e.code != 404: logger.exception("Failed to get avatar_url") raise - except: + except Exception: logger.exception("Failed to get avatar_url") defer.returnValue(result["avatar_url"]) @@ -124,7 +194,13 @@ class ProfileHandler(BaseHandler): target_user.localpart, new_avatar_url ) - yield self._update_join_states(requester) + if self.hs.config.user_directory_search_all_users: + profile = yield self.store.get_profileinfo(target_user.localpart) + yield self.user_directory_handler.handle_local_profile_change( + target_user.to_string(), profile + ) + + yield self._update_join_states(requester, target_user) @defer.inlineCallbacks def on_profile_query(self, args): @@ -149,34 +225,71 @@ class ProfileHandler(BaseHandler): defer.returnValue(response) @defer.inlineCallbacks - def _update_join_states(self, requester): - user = requester.user - if not self.hs.is_mine(user): + def _update_join_states(self, requester, target_user): + if not self.hs.is_mine(target_user): return - self.ratelimit(requester) + yield self.ratelimit(requester) - joins = yield self.store.get_rooms_for_user( - user.to_string(), + room_ids = yield self.store.get_rooms_for_user( + target_user.to_string(), ) - for j in joins: - handler = self.hs.get_handlers().room_member_handler + for room_id in room_ids: + handler = self.hs.get_room_member_handler() try: - # Assume the user isn't a guest because we don't let guests set - # profile or avatar data. - # XXX why are we recreating `requester` here for each room? - # what was wrong with the `requester` we were passed? - requester = synapse.types.create_requester(user) + # Assume the target_user isn't a guest, + # because we don't let guests set profile or avatar data. yield handler.update_membership( requester, - user, - j.room_id, + target_user, + room_id, "join", # We treat a profile update like a join. ratelimit=False, # Try to hide that these events aren't atomic. ) except Exception as e: logger.warn( "Failed to update join event for room %s - %s", - j.room_id, str(e.message) + room_id, str(e.message) ) + + def _update_remote_profile_cache(self): + """Called periodically to check profiles of remote users we haven't + checked in a while. + """ + entries = yield self.store.get_remote_profile_cache_entries_that_expire( + last_checked=self.clock.time_msec() - self.PROFILE_UPDATE_EVERY_MS + ) + + for user_id, displayname, avatar_url in entries: + is_subscribed = yield self.store.is_subscribed_remote_profile_for_user( + user_id, + ) + if not is_subscribed: + yield self.store.maybe_delete_remote_profile_cache(user_id) + continue + + try: + profile = yield self.federation.make_query( + destination=get_domain_from_id(user_id), + query_type="profile", + args={ + "user_id": user_id, + }, + ignore_backoff=True, + ) + except Exception: + logger.exception("Failed to get avatar_url") + + yield self.store.update_remote_profile_cache( + user_id, displayname, avatar_url + ) + continue + + new_name = profile.get("displayname") + new_avatar = profile.get("avatar_url") + + # We always hit update to update the last_check timestamp + yield self.store.update_remote_profile_cache( + user_id, new_name, new_avatar + ) diff --git a/synapse/handlers/read_marker.py b/synapse/handlers/read_marker.py new file mode 100644 index 0000000000..5142ae153d --- /dev/null +++ b/synapse/handlers/read_marker.py @@ -0,0 +1,64 @@ +# -*- coding: utf-8 -*- +# Copyright 2017 Vector Creations 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 ._base import BaseHandler + +from twisted.internet import defer + +from synapse.util.async import Linearizer + +import logging +logger = logging.getLogger(__name__) + + +class ReadMarkerHandler(BaseHandler): + def __init__(self, hs): + super(ReadMarkerHandler, self).__init__(hs) + self.server_name = hs.config.server_name + self.store = hs.get_datastore() + self.read_marker_linearizer = Linearizer(name="read_marker") + self.notifier = hs.get_notifier() + + @defer.inlineCallbacks + def received_client_read_marker(self, room_id, user_id, event_id): + """Updates the read marker for a given user in a given room if the event ID given + is ahead in the stream relative to the current read marker. + + This uses a notifier to indicate that account data should be sent down /sync if + the read marker has changed. + """ + + with (yield self.read_marker_linearizer.queue((room_id, user_id))): + existing_read_marker = yield self.store.get_account_data_for_room_and_type( + user_id, room_id, "m.fully_read", + ) + + should_update = True + + if existing_read_marker: + # Only update if the new marker is ahead in the stream + should_update = yield self.store.is_event_after( + event_id, + existing_read_marker['event_id'] + ) + + if should_update: + content = { + "event_id": event_id + } + max_id = yield self.store.add_account_data_to_room( + user_id, room_id, "m.fully_read", content + ) + self.notifier.on_new_event("account_data_key", max_id, users=[user_id]) diff --git a/synapse/handlers/receipts.py b/synapse/handlers/receipts.py index 50aa513935..2e0672161c 100644 --- a/synapse/handlers/receipts.py +++ b/synapse/handlers/receipts.py @@ -12,6 +12,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +from synapse.util import logcontext from ._base import BaseHandler @@ -34,7 +35,7 @@ class ReceiptsHandler(BaseHandler): self.store = hs.get_datastore() self.hs = hs self.federation = hs.get_federation_sender() - hs.get_replication_layer().register_edu_handler( + hs.get_federation_registry().register_edu_handler( "m.receipt", self._received_remote_receipt ) self.clock = self.hs.get_clock() @@ -59,6 +60,8 @@ class ReceiptsHandler(BaseHandler): is_new = yield self._handle_new_receipts([receipt]) if is_new: + # fire off a process in the background to send the receipt to + # remote servers self._push_remotes([receipt]) @defer.inlineCallbacks @@ -126,42 +129,46 @@ class ReceiptsHandler(BaseHandler): defer.returnValue(True) + @logcontext.preserve_fn # caller should not yield on this @defer.inlineCallbacks def _push_remotes(self, receipts): """Given a list of receipts, works out which remote servers should be poked and pokes them. """ - # TODO: Some of this stuff should be coallesced. - for receipt in receipts: - room_id = receipt["room_id"] - receipt_type = receipt["receipt_type"] - user_id = receipt["user_id"] - event_ids = receipt["event_ids"] - data = receipt["data"] - - users = yield self.state.get_current_user_in_room(room_id) - remotedomains = set(get_domain_from_id(u) for u in users) - remotedomains = remotedomains.copy() - remotedomains.discard(self.server_name) - - logger.debug("Sending receipt to: %r", remotedomains) - - for domain in remotedomains: - self.federation.send_edu( - destination=domain, - edu_type="m.receipt", - content={ - room_id: { - receipt_type: { - user_id: { - "event_ids": event_ids, - "data": data, + try: + # TODO: Some of this stuff should be coallesced. + for receipt in receipts: + room_id = receipt["room_id"] + receipt_type = receipt["receipt_type"] + user_id = receipt["user_id"] + event_ids = receipt["event_ids"] + data = receipt["data"] + + users = yield self.state.get_current_user_in_room(room_id) + remotedomains = set(get_domain_from_id(u) for u in users) + remotedomains = remotedomains.copy() + remotedomains.discard(self.server_name) + + logger.debug("Sending receipt to: %r", remotedomains) + + for domain in remotedomains: + self.federation.send_edu( + destination=domain, + edu_type="m.receipt", + content={ + room_id: { + receipt_type: { + user_id: { + "event_ids": event_ids, + "data": data, + } } - } + }, }, - }, - key=(room_id, receipt_type, user_id), - ) + key=(room_id, receipt_type, user_id), + ) + except Exception: + logger.exception("Error pushing receipts to remote servers") @defer.inlineCallbacks def get_receipts_for_room(self, room_id, to_key): @@ -210,10 +217,9 @@ class ReceiptEventSource(object): else: from_key = None - rooms = yield self.store.get_rooms_for_user(user.to_string()) - rooms = [room.room_id for room in rooms] + room_ids = yield self.store.get_rooms_for_user(user.to_string()) events = yield self.store.get_linearized_receipts_for_rooms( - rooms, + room_ids, from_key=from_key, to_key=to_key, ) diff --git a/synapse/handlers/register.py b/synapse/handlers/register.py index 03c6a85fc6..f83c6b3cf8 100644 --- a/synapse/handlers/register.py +++ b/synapse/handlers/register.py @@ -15,7 +15,6 @@ """Contains functions for registering clients.""" import logging -import urllib from twisted.internet import defer @@ -23,8 +22,10 @@ from synapse.api.errors import ( AuthError, Codes, SynapseError, RegistrationError, InvalidCaptchaError ) from synapse.http.client import CaptchaServerHttpClient -from synapse.types import UserID -from synapse.util.async import run_on_reactor +from synapse import types +from synapse.types import UserID, create_requester, RoomID, RoomAlias +from synapse.util.async import run_on_reactor, Linearizer +from synapse.util.threepids import check_3pid_allowed from ._base import BaseHandler logger = logging.getLogger(__name__) @@ -36,21 +37,33 @@ class RegistrationHandler(BaseHandler): super(RegistrationHandler, self).__init__(hs) self.auth = hs.get_auth() + self._auth_handler = hs.get_auth_handler() + self.profile_handler = hs.get_profile_handler() + self.user_directory_handler = hs.get_user_directory_handler() self.captcha_client = CaptchaServerHttpClient(hs) self._next_generated_user_id = None self.macaroon_gen = hs.get_macaroon_generator() + self._generate_user_id_linearizer = Linearizer( + name="_generate_user_id_linearizer", + ) + @defer.inlineCallbacks def check_username(self, localpart, guest_access_token=None, assigned_user_id=None): - yield run_on_reactor() + if types.contains_invalid_mxid_characters(localpart): + raise SynapseError( + 400, + "User ID can only contain characters a-z, 0-9, or '=_-./'", + Codes.INVALID_USERNAME + ) - if urllib.quote(localpart.encode('utf-8')) != localpart: + if not localpart: raise SynapseError( 400, - "User ID can only contain characters a-z, 0-9, or '_-./'", + "User ID cannot be empty", Codes.INVALID_USERNAME ) @@ -73,7 +86,7 @@ class RegistrationHandler(BaseHandler): "A different user ID has already been registered for this session", ) - yield self.check_user_id_not_appservice_exclusive(user_id) + self.check_user_id_not_appservice_exclusive(user_id) users = yield self.store.get_users_by_id_case_insensitive(user_id) if users: @@ -123,7 +136,7 @@ class RegistrationHandler(BaseHandler): yield run_on_reactor() password_hash = None if password: - password_hash = self.auth_handler().hash(password) + password_hash = yield self.auth_handler().hash(password) if localpart: yield self.check_username(localpart, guest_access_token=guest_access_token) @@ -158,6 +171,13 @@ class RegistrationHandler(BaseHandler): ), admin=admin, ) + + if self.hs.config.user_directory_search_all_users: + profile = yield self.store.get_profileinfo(localpart) + yield self.user_directory_handler.handle_local_profile_change( + user_id, profile + ) + else: # autogen a sequential user ID attempts = 0 @@ -185,10 +205,17 @@ class RegistrationHandler(BaseHandler): token = None attempts += 1 + # auto-join the user to any rooms we're supposed to dump them into + fake_requester = create_requester(user_id) + for r in self.hs.config.auto_join_rooms: + try: + yield self._join_user_to_room(fake_requester, r) + except Exception as e: + logger.error("Failed to join new user to %r: %r", r, e) + # We used to generate default identicons here, but nowadays # we want clients to generate their own as part of their branding # rather than there being consistent matrix-wide ones, so we don't. - defer.returnValue((user_id, token)) @defer.inlineCallbacks @@ -246,11 +273,10 @@ class RegistrationHandler(BaseHandler): """ Registers email_id as SAML2 Based Auth. """ - if urllib.quote(localpart) != localpart: + if types.contains_invalid_mxid_characters(localpart): raise SynapseError( 400, - "User ID must only contain characters which do not" - " require URL encoding." + "User ID can only contain characters a-z, 0-9, or '=_-./'", ) user = UserID(localpart, self.hs.hostname) user_id = user.to_string() @@ -279,12 +305,12 @@ class RegistrationHandler(BaseHandler): """ for c in threepidCreds: - logger.info("validating theeepidcred sid %s on id server %s", + logger.info("validating threepidcred sid %s on id server %s", c['sid'], c['idServer']) try: identity_handler = self.hs.get_handlers().identity_handler threepid = yield identity_handler.threepid_from_creds(c) - except: + except Exception: logger.exception("Couldn't validate 3pid") raise RegistrationError(400, "Couldn't validate 3pid") @@ -293,6 +319,11 @@ class RegistrationHandler(BaseHandler): logger.info("got threepid with medium '%s' and address '%s'", threepid['medium'], threepid['address']) + if not check_3pid_allowed(self.hs, threepid['medium'], threepid['address']): + raise RegistrationError( + 403, "Third party identifier is not allowed" + ) + @defer.inlineCallbacks def bind_emails(self, user_id, threepidCreds): """Links emails with a user ID and informs an identity server. @@ -325,9 +356,11 @@ class RegistrationHandler(BaseHandler): @defer.inlineCallbacks def _generate_user_id(self, reseed=False): if reseed or self._next_generated_user_id is None: - self._next_generated_user_id = ( - yield self.store.find_next_generated_user_id_localpart() - ) + with (yield self._generate_user_id_linearizer.queue(())): + if reseed or self._next_generated_user_id is None: + self._next_generated_user_id = ( + yield self.store.find_next_generated_user_id_localpart() + ) id = self._next_generated_user_id self._next_generated_user_id += 1 @@ -411,13 +444,12 @@ class RegistrationHandler(BaseHandler): create_profile_with_localpart=user.localpart, ) else: - yield self.store.user_delete_access_tokens(user_id=user_id) + yield self._auth_handler.delete_access_tokens_for_user(user_id) yield self.store.add_access_token_to_user(user_id=user_id, token=token) if displayname is not None: logger.info("setting user display name: %s -> %s", user_id, displayname) - profile_handler = self.hs.get_handlers().profile_handler - yield profile_handler.set_displayname( + yield self.profile_handler.set_displayname( user, requester, displayname, by_admin=True, ) @@ -427,16 +459,59 @@ class RegistrationHandler(BaseHandler): return self.hs.get_auth_handler() @defer.inlineCallbacks - def guest_access_token_for(self, medium, address, inviter_user_id): + def get_or_register_3pid_guest(self, medium, address, inviter_user_id): + """Get a guest access token for a 3PID, creating a guest account if + one doesn't already exist. + + Args: + medium (str) + address (str) + inviter_user_id (str): The user ID who is trying to invite the + 3PID + + Returns: + Deferred[(str, str)]: A 2-tuple of `(user_id, access_token)` of the + 3PID guest account. + """ access_token = yield self.store.get_3pid_guest_access_token(medium, address) if access_token: - defer.returnValue(access_token) + user_info = yield self.auth.get_user_by_access_token( + access_token + ) - _, access_token = yield self.register( + defer.returnValue((user_info["user"].to_string(), access_token)) + + user_id, access_token = yield self.register( generate_token=True, make_guest=True ) access_token = yield self.store.save_or_get_3pid_guest_access_token( medium, address, access_token, inviter_user_id ) - defer.returnValue(access_token) + + defer.returnValue((user_id, access_token)) + + @defer.inlineCallbacks + def _join_user_to_room(self, requester, room_identifier): + room_id = None + room_member_handler = self.hs.get_room_member_handler() + if RoomID.is_valid(room_identifier): + room_id = room_identifier + elif RoomAlias.is_valid(room_identifier): + room_alias = RoomAlias.from_string(room_identifier) + room_id, remote_room_hosts = ( + yield room_member_handler.lookup_room_alias(room_alias) + ) + room_id = room_id.to_string() + else: + raise SynapseError(400, "%s was not legal room ID or room alias" % ( + room_identifier, + )) + + yield room_member_handler.update_membership( + requester=requester, + target=requester.user, + room_id=room_id, + remote_room_hosts=remote_room_hosts, + action="join", + ) diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 7e7671c9a2..8df8fcbbad 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2014 - 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. @@ -60,8 +61,14 @@ class RoomCreationHandler(BaseHandler): }, } + def __init__(self, hs): + super(RoomCreationHandler, self).__init__(hs) + + self.spam_checker = hs.get_spam_checker() + self.event_creation_handler = hs.get_event_creation_handler() + @defer.inlineCallbacks - def create_room(self, requester, config): + def create_room(self, requester, config, ratelimit=True): """ Creates a new room. Args: @@ -75,14 +82,18 @@ class RoomCreationHandler(BaseHandler): """ user_id = requester.user.to_string() - self.ratelimit(requester) + if not self.spam_checker.user_may_create_room(user_id): + raise SynapseError(403, "You are not permitted to create rooms") + + if ratelimit: + yield self.ratelimit(requester) if "room_alias_name" in config: for wchar in string.whitespace: if wchar in config["room_alias_name"]: raise SynapseError(400, "Invalid characters in room alias") - room_alias = RoomAlias.create( + room_alias = RoomAlias( config["room_alias_name"], self.hs.hostname, ) @@ -99,7 +110,7 @@ class RoomCreationHandler(BaseHandler): for i in invite_list: try: UserID.from_string(i) - except: + except Exception: raise SynapseError(400, "Invalid user_id: %s" % (i,)) invite_3pid_list = config.get("invite_3pid", []) @@ -114,7 +125,7 @@ class RoomCreationHandler(BaseHandler): while attempts < 5: try: random_string = stringutils.random_string(18) - gen_room_id = RoomID.create( + gen_room_id = RoomID( random_string, self.hs.hostname, ) @@ -154,24 +165,23 @@ class RoomCreationHandler(BaseHandler): creation_content = config.get("creation_content", {}) - msg_handler = self.hs.get_handlers().message_handler - room_member_handler = self.hs.get_handlers().room_member_handler + room_member_handler = self.hs.get_room_member_handler() yield self._send_events_for_new_room( requester, room_id, - msg_handler, room_member_handler, preset_config=preset_config, invite_list=invite_list, initial_state=initial_state, creation_content=creation_content, room_alias=room_alias, + power_level_content_override=config.get("power_level_content_override", {}) ) if "name" in config: name = config["name"] - yield msg_handler.create_and_send_nonmember_event( + yield self.event_creation_handler.create_and_send_nonmember_event( requester, { "type": EventTypes.Name, @@ -184,7 +194,7 @@ class RoomCreationHandler(BaseHandler): if "topic" in config: topic = config["topic"] - yield msg_handler.create_and_send_nonmember_event( + yield self.event_creation_handler.create_and_send_nonmember_event( requester, { "type": EventTypes.Topic, @@ -195,12 +205,12 @@ class RoomCreationHandler(BaseHandler): }, ratelimit=False) - content = {} - is_direct = config.get("is_direct", None) - if is_direct: - content["is_direct"] = is_direct - for invitee in invite_list: + content = {} + is_direct = config.get("is_direct", None) + if is_direct: + content["is_direct"] = is_direct + yield room_member_handler.update_membership( requester, UserID.from_string(invitee), @@ -214,7 +224,7 @@ class RoomCreationHandler(BaseHandler): id_server = invite_3pid["id_server"] address = invite_3pid["address"] medium = invite_3pid["medium"] - yield self.hs.get_handlers().room_member_handler.do_3pid_invite( + yield self.hs.get_room_member_handler().do_3pid_invite( room_id, requester.user, medium, @@ -239,13 +249,13 @@ class RoomCreationHandler(BaseHandler): self, creator, # A Requester object. room_id, - msg_handler, room_member_handler, preset_config, invite_list, initial_state, creation_content, - room_alias + room_alias, + power_level_content_override, ): def create(etype, content, **kwargs): e = { @@ -261,7 +271,7 @@ class RoomCreationHandler(BaseHandler): @defer.inlineCallbacks def send(etype, content, **kwargs): event = create(etype, content, **kwargs) - yield msg_handler.create_and_send_nonmember_event( + yield self.event_creation_handler.create_and_send_nonmember_event( creator, event, ratelimit=False @@ -291,7 +301,15 @@ class RoomCreationHandler(BaseHandler): ratelimit=False, ) - if (EventTypes.PowerLevels, '') not in initial_state: + # We treat the power levels override specially as this needs to be one + # of the first events that get sent into a room. + pl_content = initial_state.pop((EventTypes.PowerLevels, ''), None) + if pl_content is not None: + yield send( + etype=EventTypes.PowerLevels, + content=pl_content, + ) + else: power_level_content = { "users": { creator_id: 100, @@ -316,6 +334,8 @@ class RoomCreationHandler(BaseHandler): for invitee in invite_list: power_level_content["users"][invitee] = 100 + power_level_content.update(power_level_content_override) + yield send( etype=EventTypes.PowerLevels, content=power_level_content, @@ -356,7 +376,7 @@ class RoomCreationHandler(BaseHandler): class RoomContextHandler(BaseHandler): @defer.inlineCallbacks - def get_event_context(self, user, room_id, event_id, limit, is_guest): + def get_event_context(self, user, room_id, event_id, limit): """Retrieves events, pagination tokens and state around a given event in a room. @@ -375,12 +395,15 @@ class RoomContextHandler(BaseHandler): 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 + def filter_evts(events): return filter_events_for_client( self.store, user.to_string(), events, - is_peeking=is_guest + is_peeking=is_peeking ) event = yield self.store.get_event(event_id, get_prev_content=True, @@ -452,12 +475,9 @@ class RoomEventSource(object): user.to_string() ) if app_service: - events, end_key = yield self.store.get_appservice_room_stream( - service=app_service, - from_key=from_key, - to_key=to_key, - limit=limit, - ) + # We no longer support AS users using /sync directly. + # See https://github.com/matrix-org/matrix-doc/issues/1144 + raise NotImplementedError() else: room_events = yield self.store.get_membership_changes_for_user( user.to_string(), from_key, to_key diff --git a/synapse/handlers/room_list.py b/synapse/handlers/room_list.py index 19eebbd43f..5757bb7f8a 100644 --- a/synapse/handlers/room_list.py +++ b/synapse/handlers/room_list.py @@ -15,12 +15,15 @@ from twisted.internet import defer +from six.moves import range + from ._base import BaseHandler from synapse.api.constants import ( EventTypes, JoinRules, ) from synapse.util.async import concurrently_execute +from synapse.util.caches.descriptors import cachedInlineCallbacks from synapse.util.caches.response_cache import ResponseCache from synapse.types import ThirdPartyInstanceID @@ -42,8 +45,9 @@ EMTPY_THIRD_PARTY_ID = ThirdPartyInstanceID(None, None) class RoomListHandler(BaseHandler): def __init__(self, hs): super(RoomListHandler, self).__init__(hs) - self.response_cache = ResponseCache(hs) - self.remote_response_cache = ResponseCache(hs, timeout_ms=30 * 1000) + self.response_cache = ResponseCache(hs, "room_list") + self.remote_response_cache = ResponseCache(hs, "remote_room_list", + timeout_ms=30 * 1000) def get_local_public_room_list(self, limit=None, since_token=None, search_filter=None, @@ -62,23 +66,24 @@ class RoomListHandler(BaseHandler): appservice and network id to use an appservice specific one. Setting to None returns all public rooms across all lists. """ + logger.info( + "Getting public room list: limit=%r, since=%r, search=%r, network=%r", + limit, since_token, bool(search_filter), network_tuple, + ) if search_filter: # We explicitly don't bother caching searches or requests for # appservice specific lists. + logger.info("Bypassing cache as search request.") return self._get_public_room_list( limit, since_token, search_filter, network_tuple=network_tuple, ) key = (limit, since_token, network_tuple) - result = self.response_cache.get(key) - if not result: - result = self.response_cache.set( - key, - self._get_public_room_list( - limit, since_token, network_tuple=network_tuple - ) - ) - return result + return self.response_cache.wrap( + key, + self._get_public_room_list, + limit, since_token, network_tuple=network_tuple, + ) @defer.inlineCallbacks def _get_public_room_list(self, limit=None, since_token=None, @@ -91,7 +96,6 @@ class RoomListHandler(BaseHandler): rooms_to_order_value = {} rooms_to_num_joined = {} - rooms_to_latest_event_ids = {} newly_visible = [] newly_unpublished = [] @@ -116,19 +120,26 @@ class RoomListHandler(BaseHandler): @defer.inlineCallbacks def get_order_for_room(room_id): - latest_event_ids = rooms_to_latest_event_ids.get(room_id, None) - if not latest_event_ids: + # Most of the rooms won't have changed between the since token and + # now (especially if the since token is "now"). So, we can ask what + # the current users are in a room (that will hit a cache) and then + # check if the room has changed since the since token. (We have to + # do it in that order to avoid races). + # If things have changed then fall back to getting the current state + # at the since token. + joined_users = yield self.store.get_users_in_room(room_id) + if self.store.has_room_changed_since(room_id, stream_token): latest_event_ids = yield self.store.get_forward_extremeties_for_room( room_id, stream_token ) - rooms_to_latest_event_ids[room_id] = latest_event_ids - if not latest_event_ids: - return + if not latest_event_ids: + return + + joined_users = yield self.state_handler.get_current_user_in_room( + room_id, latest_event_ids, + ) - joined_users = yield self.state_handler.get_current_user_in_room( - room_id, latest_event_ids, - ) num_joined_users = len(joined_users) rooms_to_num_joined[room_id] = num_joined_users @@ -138,6 +149,8 @@ class RoomListHandler(BaseHandler): # We want larger rooms to be first, hence negating num_joined_users rooms_to_order_value[room_id] = (-num_joined_users, room_id) + logger.info("Getting ordering for %i rooms since %s", + len(room_ids), stream_token) yield concurrently_execute(get_order_for_room, room_ids, 10) sorted_entries = sorted(rooms_to_order_value.items(), key=lambda e: e[1]) @@ -165,34 +178,43 @@ class RoomListHandler(BaseHandler): rooms_to_scan = rooms_to_scan[:since_token.current_limit] rooms_to_scan.reverse() - # Actually generate the entries. _generate_room_entry will append to - # chunk but will stop if len(chunk) > limit - chunk = [] - if limit and not search_filter: + logger.info("After sorting and filtering, %i rooms remain", + len(rooms_to_scan)) + + # _append_room_entry_to_chunk will append to chunk but will stop if + # len(chunk) > limit + # + # Normally we will generate enough results on the first iteration here, + # but if there is a search filter, _append_room_entry_to_chunk may + # filter some results out, in which case we loop again. + # + # We don't want to scan over the entire range either as that + # would potentially waste a lot of work. + # + # XXX if there is no limit, we may end up DoSing the server with + # calls to get_current_state_ids for every single room on the + # server. Surely we should cap this somehow? + # + if limit: step = limit + 1 - for i in xrange(0, len(rooms_to_scan), step): - # We iterate here because the vast majority of cases we'll stop - # at first iteration, but occaisonally _generate_room_entry - # won't append to the chunk and so we need to loop again. - # We don't want to scan over the entire range either as that - # would potentially waste a lot of work. - yield concurrently_execute( - lambda r: self._generate_room_entry( - r, rooms_to_num_joined[r], - chunk, limit, search_filter - ), - rooms_to_scan[i:i + step], 10 - ) - if len(chunk) >= limit + 1: - break else: + # step cannot be zero + step = len(rooms_to_scan) if len(rooms_to_scan) != 0 else 1 + + chunk = [] + for i in range(0, len(rooms_to_scan), step): + batch = rooms_to_scan[i:i + step] + logger.info("Processing %i rooms for result", len(batch)) yield concurrently_execute( - lambda r: self._generate_room_entry( + lambda r: self._append_room_entry_to_chunk( r, rooms_to_num_joined[r], chunk, limit, search_filter ), - rooms_to_scan, 5 + batch, 5, ) + logger.info("Now %i rooms in result", len(chunk)) + if len(chunk) >= limit + 1: + break chunk.sort(key=lambda e: (-e["num_joined_members"], e["room_id"])) @@ -256,21 +278,36 @@ class RoomListHandler(BaseHandler): defer.returnValue(results) @defer.inlineCallbacks - def _generate_room_entry(self, room_id, num_joined_users, chunk, limit, - search_filter): + def _append_room_entry_to_chunk(self, room_id, num_joined_users, chunk, limit, + search_filter): + """Generate the entry for a room in the public room list and append it + to the `chunk` if it matches the search filter + """ if limit and len(chunk) > limit + 1: # We've already got enough, so lets just drop it. return + result = yield self.generate_room_entry(room_id, num_joined_users) + + if result and _matches_room_entry(result, search_filter): + chunk.append(result) + + @cachedInlineCallbacks(num_args=1, cache_context=True) + def generate_room_entry(self, room_id, num_joined_users, cache_context, + with_alias=True, allow_private=False): + """Returns the entry for a room + """ result = { "room_id": room_id, "num_joined_members": num_joined_users, } - current_state_ids = yield self.state_handler.get_current_state_ids(room_id) + current_state_ids = yield self.store.get_current_state_ids( + room_id, on_invalidate=cache_context.invalidate, + ) event_map = yield self.store.get_events([ - event_id for key, event_id in current_state_ids.items() + event_id for key, event_id in current_state_ids.iteritems() if key[0] in ( EventTypes.JoinRules, EventTypes.Name, @@ -291,12 +328,15 @@ class RoomListHandler(BaseHandler): join_rules_event = current_state.get((EventTypes.JoinRules, "")) if join_rules_event: join_rule = join_rules_event.content.get("join_rule", None) - if join_rule and join_rule != JoinRules.PUBLIC: + if not allow_private and join_rule and join_rule != JoinRules.PUBLIC: defer.returnValue(None) - aliases = yield self.store.get_aliases_for_room(room_id) - if aliases: - result["aliases"] = aliases + if with_alias: + aliases = yield self.store.get_aliases_for_room( + room_id, on_invalidate=cache_context.invalidate + ) + if aliases: + result["aliases"] = aliases name_event = yield current_state.get((EventTypes.Name, "")) if name_event: @@ -334,8 +374,7 @@ class RoomListHandler(BaseHandler): if avatar_url: result["avatar_url"] = avatar_url - if _matches_room_entry(result, search_filter): - chunk.append(result) + defer.returnValue(result) @defer.inlineCallbacks def get_remote_public_room_list(self, server_name, limit=None, since_token=None, @@ -365,7 +404,7 @@ class RoomListHandler(BaseHandler): def _get_remote_list_cached(self, server_name, limit=None, since_token=None, search_filter=None, include_all_networks=False, third_party_instance_id=None,): - repl_layer = self.hs.get_replication_layer() + repl_layer = self.hs.get_federation_client() if search_filter: # We can't cache when asking for search return repl_layer.get_public_rooms( @@ -378,18 +417,14 @@ class RoomListHandler(BaseHandler): server_name, limit, since_token, include_all_networks, third_party_instance_id, ) - result = self.remote_response_cache.get(key) - if not result: - result = self.remote_response_cache.set( - key, - repl_layer.get_public_rooms( - server_name, limit=limit, since_token=since_token, - search_filter=search_filter, - include_all_networks=include_all_networks, - third_party_instance_id=third_party_instance_id, - ) - ) - return result + return self.remote_response_cache.wrap( + key, + repl_layer.get_public_rooms, + server_name, limit=limit, since_token=since_token, + search_filter=search_filter, + include_all_networks=include_all_networks, + third_party_instance_id=third_party_instance_id, + ) class RoomListNextBatch(namedtuple("RoomListNextBatch", ( diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index b2806555cf..714583f1d5 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 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. @@ -13,7 +14,7 @@ # See the License for the specific language governing permissions and # limitations under the License. - +import abc import logging from signedjson.key import decode_verify_key_bytes @@ -29,47 +30,139 @@ from synapse.api.errors import AuthError, SynapseError, Codes from synapse.types import UserID, RoomID from synapse.util.async import Linearizer from synapse.util.distributor import user_left_room, user_joined_room -from ._base import BaseHandler + logger = logging.getLogger(__name__) id_server_scheme = "https://" -class RoomMemberHandler(BaseHandler): +class RoomMemberHandler(object): # TODO(paul): This handler currently contains a messy conflation of # low-level API that works on UserID objects and so on, and REST-level # API that takes ID strings and returns pagination chunks. These concerns # ought to be separated out a lot better. + __metaclass__ = abc.ABCMeta + def __init__(self, hs): - super(RoomMemberHandler, self).__init__(hs) + self.hs = hs + self.store = hs.get_datastore() + self.auth = hs.get_auth() + self.state_handler = hs.get_state_handler() + self.config = hs.config + self.simple_http_client = hs.get_simple_http_client() + + self.federation_handler = hs.get_handlers().federation_handler + self.directory_handler = hs.get_handlers().directory_handler + self.registration_handler = hs.get_handlers().registration_handler + self.profile_handler = hs.get_profile_handler() + self.event_creation_hander = hs.get_event_creation_handler() self.member_linearizer = Linearizer(name="member") self.clock = hs.get_clock() + self.spam_checker = hs.get_spam_checker() - self.distributor = hs.get_distributor() - self.distributor.declare("user_joined_room") - self.distributor.declare("user_left_room") + @abc.abstractmethod + def _remote_join(self, requester, remote_room_hosts, room_id, user, content): + """Try and join a room that this server is not in + + Args: + requester (Requester) + remote_room_hosts (list[str]): List of servers that can be used + to join via. + room_id (str): Room that we are trying to join + user (UserID): User who is trying to join + content (dict): A dict that should be used as the content of the + join event. + + Returns: + Deferred + """ + raise NotImplementedError() + + @abc.abstractmethod + def _remote_reject_invite(self, remote_room_hosts, room_id, target): + """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. + + Args: + requester (Requester) + remote_room_hosts (list[str]): List of servers to use to try and + reject invite + room_id (str) + target (UserID): The user rejecting the invite + + Returns: + Deferred[dict]: A dictionary to be returned to the client, may + include event_id etc, or nothing if we locally rejected + """ + raise NotImplementedError() + + @abc.abstractmethod + def get_or_register_3pid_guest(self, requester, medium, address, inviter_user_id): + """Get a guest access token for a 3PID, creating a guest account if + one doesn't already exist. + + Args: + requester (Requester) + medium (str) + address (str) + inviter_user_id (str): The user ID who is trying to invite the + 3PID + + Returns: + Deferred[(str, str)]: A 2-tuple of `(user_id, access_token)` of the + 3PID guest account. + """ + raise NotImplementedError() + + @abc.abstractmethod + def _user_joined_room(self, target, room_id): + """Notifies distributor on master process that the user has joined the + room. + + Args: + target (UserID) + room_id (str) + + Returns: + Deferred|None + """ + raise NotImplementedError() + + @abc.abstractmethod + def _user_left_room(self, target, room_id): + """Notifies distributor on master process that the user has left the + room. + + Args: + target (UserID) + room_id (str) + + Returns: + Deferred|None + """ + raise NotImplementedError() @defer.inlineCallbacks def _local_membership_update( self, requester, target, room_id, membership, - prev_event_ids, + prev_events_and_hashes, txn_id=None, ratelimit=True, content=None, ): if content is None: content = {} - msg_handler = self.hs.get_handlers().message_handler content["membership"] = membership if requester.is_guest: content["kind"] = "guest" - event, context = yield msg_handler.create_event( + event, context = yield self.event_creation_hander.create_event( + requester, { "type": EventTypes.Member, "content": content, @@ -82,16 +175,18 @@ class RoomMemberHandler(BaseHandler): }, token_id=requester.access_token_id, txn_id=txn_id, - prev_event_ids=prev_event_ids, + prev_events_and_hashes=prev_events_and_hashes, ) # Check if this event matches the previous membership event for the user. - duplicate = yield msg_handler.deduplicate_state_event(event, context) + duplicate = yield self.event_creation_hander.deduplicate_state_event( + event, context, + ) if duplicate is not None: # Discard the new event since this membership change is a no-op. defer.returnValue(duplicate) - yield msg_handler.handle_new_client_event( + yield self.event_creation_hander.handle_new_client_event( requester, event, context, @@ -113,40 +208,16 @@ class RoomMemberHandler(BaseHandler): prev_member_event = yield self.store.get_event(prev_member_event_id) newly_joined = prev_member_event.membership != Membership.JOIN if newly_joined: - yield user_joined_room(self.distributor, target, room_id) + yield self._user_joined_room(target, room_id) elif event.membership == Membership.LEAVE: if prev_member_event_id: prev_member_event = yield self.store.get_event(prev_member_event_id) if prev_member_event.membership == Membership.JOIN: - user_left_room(self.distributor, target, room_id) + yield self._user_left_room(target, room_id) defer.returnValue(event) @defer.inlineCallbacks - def remote_join(self, remote_room_hosts, room_id, user, content): - if len(remote_room_hosts) == 0: - raise SynapseError(404, "No known servers") - - # We don't do an auth check if we are doing an invite - # join dance for now, since we're kinda implicitly checking - # that we are allowed to join when we decide whether or not we - # need to do the invite/join dance. - yield self.hs.get_handlers().federation_handler.do_invite_join( - remote_room_hosts, - room_id, - user.to_string(), - content, - ) - yield user_joined_room(self.distributor, user, room_id) - - def reject_remote_invite(self, user_id, room_id, remote_room_hosts): - return self.hs.get_handlers().federation_handler.do_remotely_reject_invite( - remote_room_hosts, - room_id, - user_id - ) - - @defer.inlineCallbacks def update_membership( self, requester, @@ -192,14 +263,19 @@ class RoomMemberHandler(BaseHandler): content_specified = bool(content) if content is None: content = {} + else: + # We do a copy here as we potentially change some keys + # later on. + content = dict(content) effective_membership_state = action if action in ["kick", "unban"]: effective_membership_state = "leave" + # if this is a join with a 3pid signature, we may need to turn a 3pid + # invite into a normal invite before we can handle the join. if third_party_signed is not None: - replication = self.hs.get_replication_layer() - yield replication.exchange_third_party_invite( + yield self.federation_handler.exchange_third_party_invite( third_party_signed["sender"], target.to_string(), room_id, @@ -209,7 +285,41 @@ class RoomMemberHandler(BaseHandler): if not remote_room_hosts: remote_room_hosts = [] - latest_event_ids = yield self.store.get_latest_event_ids_in_room(room_id) + if effective_membership_state not in ("leave", "ban",): + is_blocked = yield self.store.is_room_blocked(room_id) + if is_blocked: + raise SynapseError(403, "This room has been blocked on this server") + + if effective_membership_state == "invite": + block_invite = False + is_requester_admin = yield self.auth.is_server_admin( + requester.user, + ) + if not is_requester_admin: + if self.config.block_non_admin_invites: + logger.info( + "Blocking invite: user is not admin and non-admin " + "invites disabled" + ) + block_invite = True + + if not self.spam_checker.user_may_invite( + requester.user.to_string(), target.to_string(), room_id, + ): + logger.info("Blocking invite due to spam checker") + block_invite = True + + 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 + ) current_state_ids = yield self.state_handler.get_current_state_ids( room_id, latest_event_ids=latest_event_ids, ) @@ -250,13 +360,13 @@ class RoomMemberHandler(BaseHandler): raise AuthError(403, "Guest access not allowed") if not is_host_in_room: - inviter = yield self.get_inviter(target.to_string(), room_id) + inviter = yield self._get_inviter(target.to_string(), room_id) if inviter and not self.hs.is_mine(inviter): remote_room_hosts.append(inviter.domain) content["membership"] = Membership.JOIN - profile = self.hs.get_handlers().profile_handler + profile = self.profile_handler if not content_specified: content["displayname"] = yield profile.get_displayname(target) content["avatar_url"] = yield profile.get_avatar_url(target) @@ -264,15 +374,15 @@ class RoomMemberHandler(BaseHandler): if requester.is_guest: content["kind"] = "guest" - ret = yield self.remote_join( - remote_room_hosts, room_id, target, content + ret = yield self._remote_join( + requester, remote_room_hosts, room_id, target, content ) defer.returnValue(ret) elif effective_membership_state == Membership.LEAVE: if not is_host_in_room: # perhaps we've been invited - inviter = yield self.get_inviter(target.to_string(), room_id) + inviter = yield self._get_inviter(target.to_string(), room_id) if not inviter: raise SynapseError(404, "Not a known room") @@ -286,20 +396,10 @@ class RoomMemberHandler(BaseHandler): else: # send the rejection to the inviter's HS. remote_room_hosts = remote_room_hosts + [inviter.domain] - - try: - ret = yield self.reject_remote_invite( - target.to_string(), room_id, remote_room_hosts - ) - defer.returnValue(ret) - except SynapseError as e: - logger.warn("Failed to reject invite: %s", e) - - yield self.store.locally_reject_invite( - target.to_string(), room_id - ) - - defer.returnValue({}) + res = yield self._remote_reject_invite( + requester, remote_room_hosts, room_id, target, + ) + defer.returnValue(res) res = yield self._local_membership_update( requester=requester, @@ -308,7 +408,7 @@ class RoomMemberHandler(BaseHandler): membership=effective_membership_state, txn_id=txn_id, ratelimit=ratelimit, - prev_event_ids=latest_event_ids, + prev_events_and_hashes=prev_events_and_hashes, content=content, ) defer.returnValue(res) @@ -354,8 +454,9 @@ class RoomMemberHandler(BaseHandler): else: requester = synapse.types.create_requester(target_user) - message_handler = self.hs.get_handlers().message_handler - prev_event = yield message_handler.deduplicate_state_event(event, context) + prev_event = yield self.event_creation_hander.deduplicate_state_event( + event, context, + ) if prev_event is not None: return @@ -367,7 +468,12 @@ class RoomMemberHandler(BaseHandler): # so don't really fit into the general auth process. raise AuthError(403, "Guest access not allowed") - yield message_handler.handle_new_client_event( + if event.membership not in (Membership.LEAVE, Membership.BAN): + is_blocked = yield self.store.is_room_blocked(room_id) + if is_blocked: + raise SynapseError(403, "This room has been blocked on this server") + + yield self.event_creation_hander.handle_new_client_event( requester, event, context, @@ -389,12 +495,12 @@ class RoomMemberHandler(BaseHandler): prev_member_event = yield self.store.get_event(prev_member_event_id) newly_joined = prev_member_event.membership != Membership.JOIN if newly_joined: - yield user_joined_room(self.distributor, target_user, room_id) + yield self._user_joined_room(target_user, room_id) elif event.membership == Membership.LEAVE: if prev_member_event_id: prev_member_event = yield self.store.get_event(prev_member_event_id) if prev_member_event.membership == Membership.JOIN: - user_left_room(self.distributor, target_user, room_id) + yield self._user_left_room(target_user, room_id) @defer.inlineCallbacks def _can_guest_join(self, current_state_ids): @@ -428,7 +534,7 @@ class RoomMemberHandler(BaseHandler): Raises: SynapseError if room alias could not be found. """ - directory_handler = self.hs.get_handlers().directory_handler + directory_handler = self.directory_handler mapping = yield directory_handler.get_association(room_alias) if not mapping: @@ -440,7 +546,7 @@ class RoomMemberHandler(BaseHandler): defer.returnValue((RoomID.from_string(room_id), servers)) @defer.inlineCallbacks - def get_inviter(self, user_id, room_id): + def _get_inviter(self, user_id, room_id): invite = yield self.store.get_invite_for_user_in_room( user_id=user_id, room_id=room_id, @@ -459,6 +565,16 @@ class RoomMemberHandler(BaseHandler): requester, txn_id ): + if self.config.block_non_admin_invites: + is_requester_admin = yield self.auth.is_server_admin( + requester.user, + ) + if not is_requester_admin: + raise SynapseError( + 403, "Invites have been disabled on this server", + Codes.FORBIDDEN, + ) + invitee = yield self._lookup_3pid( id_server, medium, address ) @@ -496,7 +612,7 @@ class RoomMemberHandler(BaseHandler): str: the matrix ID of the 3pid, or None if it is not recognized. """ try: - data = yield self.hs.get_simple_http_client().get_json( + data = yield self.simple_http_client.get_json( "%s%s/_matrix/identity/api/v1/lookup" % (id_server_scheme, id_server,), { "medium": medium, @@ -507,7 +623,7 @@ class RoomMemberHandler(BaseHandler): if "mxid" in data: if "signatures" not in data: raise AuthError(401, "No signatures on 3pid binding") - self.verify_any_signature(data, id_server) + yield self._verify_any_signature(data, id_server) defer.returnValue(data["mxid"]) except IOError as e: @@ -515,11 +631,11 @@ class RoomMemberHandler(BaseHandler): defer.returnValue(None) @defer.inlineCallbacks - def verify_any_signature(self, data, server_hostname): + def _verify_any_signature(self, data, server_hostname): if server_hostname not in data["signatures"]: raise AuthError(401, "No signature from server %s" % (server_hostname,)) for key_name, signature in data["signatures"][server_hostname].items(): - key_data = yield self.hs.get_simple_http_client().get_json( + key_data = yield self.simple_http_client.get_json( "%s%s/_matrix/identity/api/v1/pubkey/%s" % (id_server_scheme, server_hostname, key_name,), ) @@ -544,7 +660,7 @@ class RoomMemberHandler(BaseHandler): user, txn_id ): - room_state = yield self.hs.get_state_handler().get_current_state(room_id) + room_state = yield self.state_handler.get_current_state(room_id) inviter_display_name = "" inviter_avatar_url = "" @@ -575,6 +691,7 @@ class RoomMemberHandler(BaseHandler): token, public_keys, fallback_public_key, display_name = ( yield self._ask_id_server_for_third_party_invite( + requester=requester, id_server=id_server, medium=medium, address=address, @@ -589,8 +706,7 @@ class RoomMemberHandler(BaseHandler): ) ) - msg_handler = self.hs.get_handlers().message_handler - yield msg_handler.create_and_send_nonmember_event( + yield self.event_creation_hander.create_and_send_nonmember_event( requester, { "type": EventTypes.ThirdPartyInvite, @@ -612,6 +728,7 @@ class RoomMemberHandler(BaseHandler): @defer.inlineCallbacks def _ask_id_server_for_third_party_invite( self, + requester, id_server, medium, address, @@ -628,6 +745,7 @@ class RoomMemberHandler(BaseHandler): Asks an identity server for a third party invite. Args: + requester (Requester) id_server (str): hostname + optional port for the identity server. medium (str): The literal string "email". address (str): The third party address being invited. @@ -669,24 +787,20 @@ class RoomMemberHandler(BaseHandler): "sender_avatar_url": inviter_avatar_url, } - if self.hs.config.invite_3pid_guest: - registration_handler = self.hs.get_handlers().registration_handler - guest_access_token = yield registration_handler.guest_access_token_for( + if self.config.invite_3pid_guest: + guest_access_token, guest_user_id = yield self.get_or_register_3pid_guest( + requester=requester, medium=medium, address=address, inviter_user_id=inviter_user_id, ) - guest_user_info = yield self.hs.get_auth().get_user_by_access_token( - guest_access_token - ) - invite_config.update({ "guest_access_token": guest_access_token, - "guest_user_id": guest_user_info["user"].to_string(), + "guest_user_id": guest_user_id, }) - data = yield self.hs.get_simple_http_client().post_urlencoded_get_json( + data = yield self.simple_http_client.post_urlencoded_get_json( is_url, invite_config ) @@ -709,25 +823,6 @@ class RoomMemberHandler(BaseHandler): defer.returnValue((token, public_keys, fallback_public_key, display_name)) @defer.inlineCallbacks - def forget(self, user, room_id): - user_id = user.to_string() - - member = yield self.state_handler.get_current_state( - room_id=room_id, - event_type=EventTypes.Member, - state_key=user_id - ) - membership = member.membership if member else None - - if membership is not None and membership != Membership.LEAVE: - raise SynapseError(400, "User %s in room %s" % ( - user_id, room_id - )) - - if membership: - yield self.store.forget(user_id, room_id) - - @defer.inlineCallbacks def _is_host_in_room(self, current_state_ids): # Have we just created the room, and is this about to be the very # first member event? @@ -735,10 +830,11 @@ class RoomMemberHandler(BaseHandler): if len(current_state_ids) == 1 and create_event_id: defer.returnValue(self.hs.is_mine_id(create_event_id)) - for (etype, state_key), event_id in current_state_ids.items(): + for etype, state_key in current_state_ids: if etype != EventTypes.Member or not self.hs.is_mine_id(state_key): continue + event_id = current_state_ids[(etype, state_key)] event = yield self.store.get_event(event_id, allow_none=True) if not event: continue @@ -747,3 +843,102 @@ class RoomMemberHandler(BaseHandler): defer.returnValue(True) defer.returnValue(False) + + +class RoomMemberMasterHandler(RoomMemberHandler): + def __init__(self, hs): + super(RoomMemberMasterHandler, self).__init__(hs) + + self.distributor = hs.get_distributor() + self.distributor.declare("user_joined_room") + self.distributor.declare("user_left_room") + + @defer.inlineCallbacks + def _remote_join(self, requester, remote_room_hosts, room_id, user, content): + """Implements RoomMemberHandler._remote_join + """ + # filter ourselves out of remote_room_hosts: do_invite_join ignores it + # and if it is the only entry we'd like to return a 404 rather than a + # 500. + + remote_room_hosts = [ + host for host in remote_room_hosts if host != self.hs.hostname + ] + + if len(remote_room_hosts) == 0: + raise SynapseError(404, "No known servers") + + # We don't do an auth check if we are doing an invite + # join dance for now, since we're kinda implicitly checking + # that we are allowed to join when we decide whether or not we + # need to do the invite/join dance. + yield self.federation_handler.do_invite_join( + remote_room_hosts, + room_id, + user.to_string(), + content, + ) + yield self._user_joined_room(user, room_id) + + @defer.inlineCallbacks + def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target): + """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(), + ) + defer.returnValue(ret) + except Exception as e: + # if we were unable to reject the exception, just mark + # it as rejected on our end and plough ahead. + # + # The 'except' clause is very broad, but we need to + # capture everything from DNS failures upwards + # + logger.warn("Failed to reject invite: %s", e) + + yield self.store.locally_reject_invite( + target.to_string(), room_id + ) + defer.returnValue({}) + + def get_or_register_3pid_guest(self, requester, medium, address, inviter_user_id): + """Implements RoomMemberHandler.get_or_register_3pid_guest + """ + rg = self.registration_handler + return rg.get_or_register_3pid_guest(medium, address, inviter_user_id) + + def _user_joined_room(self, target, room_id): + """Implements RoomMemberHandler._user_joined_room + """ + return user_joined_room(self.distributor, target, room_id) + + def _user_left_room(self, target, room_id): + """Implements RoomMemberHandler._user_left_room + """ + return user_left_room(self.distributor, target, room_id) + + @defer.inlineCallbacks + def forget(self, user, room_id): + user_id = user.to_string() + + member = yield self.state_handler.get_current_state( + room_id=room_id, + event_type=EventTypes.Member, + state_key=user_id + ) + membership = member.membership if member else None + + if membership is not None and membership not in [ + Membership.LEAVE, Membership.BAN + ]: + raise SynapseError(400, "User %s in room %s" % ( + user_id, room_id + )) + + if membership: + yield self.store.forget(user_id, room_id) diff --git a/synapse/handlers/room_member_worker.py b/synapse/handlers/room_member_worker.py new file mode 100644 index 0000000000..493aec1e48 --- /dev/null +++ b/synapse/handlers/room_member_worker.py @@ -0,0 +1,102 @@ +# -*- 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 twisted.internet import defer + +from synapse.api.errors import SynapseError +from synapse.handlers.room_member import RoomMemberHandler +from synapse.replication.http.membership import ( + remote_join, remote_reject_invite, get_or_register_3pid_guest, + notify_user_membership_change, +) + + +logger = logging.getLogger(__name__) + + +class RoomMemberWorkerHandler(RoomMemberHandler): + @defer.inlineCallbacks + def _remote_join(self, requester, remote_room_hosts, room_id, user, content): + """Implements RoomMemberHandler._remote_join + """ + if len(remote_room_hosts) == 0: + raise SynapseError(404, "No known servers") + + ret = yield remote_join( + self.simple_http_client, + host=self.config.worker_replication_host, + port=self.config.worker_replication_http_port, + requester=requester, + remote_room_hosts=remote_room_hosts, + room_id=room_id, + user_id=user.to_string(), + content=content, + ) + + yield self._user_joined_room(user, room_id) + + defer.returnValue(ret) + + def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target): + """Implements RoomMemberHandler._remote_reject_invite + """ + return remote_reject_invite( + self.simple_http_client, + host=self.config.worker_replication_host, + port=self.config.worker_replication_http_port, + requester=requester, + remote_room_hosts=remote_room_hosts, + room_id=room_id, + user_id=target.to_string(), + ) + + def _user_joined_room(self, target, room_id): + """Implements RoomMemberHandler._user_joined_room + """ + return notify_user_membership_change( + self.simple_http_client, + host=self.config.worker_replication_host, + port=self.config.worker_replication_http_port, + user_id=target.to_string(), + room_id=room_id, + change="joined", + ) + + def _user_left_room(self, target, room_id): + """Implements RoomMemberHandler._user_left_room + """ + return notify_user_membership_change( + self.simple_http_client, + host=self.config.worker_replication_host, + port=self.config.worker_replication_http_port, + user_id=target.to_string(), + room_id=room_id, + change="left", + ) + + def get_or_register_3pid_guest(self, requester, medium, address, inviter_user_id): + """Implements RoomMemberHandler.get_or_register_3pid_guest + """ + return get_or_register_3pid_guest( + self.simple_http_client, + host=self.config.worker_replication_host, + port=self.config.worker_replication_http_port, + requester=requester, + medium=medium, + address=address, + inviter_user_id=inviter_user_id, + ) diff --git a/synapse/handlers/search.py b/synapse/handlers/search.py index df75d70fac..9772ed1a0e 100644 --- a/synapse/handlers/search.py +++ b/synapse/handlers/search.py @@ -61,7 +61,7 @@ class SearchHandler(BaseHandler): assert batch_group is not None assert batch_group_key is not None assert batch_token is not None - except: + except Exception: raise SynapseError(400, "Invalid batch") try: diff --git a/synapse/handlers/set_password.py b/synapse/handlers/set_password.py new file mode 100644 index 0000000000..e057ae54c9 --- /dev/null +++ b/synapse/handlers/set_password.py @@ -0,0 +1,56 @@ +# -*- coding: utf-8 -*- +# Copyright 2017 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 synapse.api.errors import Codes, StoreError, SynapseError +from ._base import BaseHandler + +logger = logging.getLogger(__name__) + + +class SetPasswordHandler(BaseHandler): + """Handler which deals with changing user account passwords""" + def __init__(self, hs): + super(SetPasswordHandler, self).__init__(hs) + self._auth_handler = hs.get_auth_handler() + self._device_handler = hs.get_device_handler() + + @defer.inlineCallbacks + def set_password(self, user_id, newpassword, requester=None): + password_hash = yield self._auth_handler.hash(newpassword) + + except_device_id = requester.device_id if requester else None + except_access_token_id = requester.access_token_id if requester else None + + try: + yield self.store.user_set_password_hash(user_id, password_hash) + except StoreError as e: + if e.code == 404: + raise SynapseError(404, "Unknown user", Codes.NOT_FOUND) + raise e + + # we want to log out all of the user's other sessions. First delete + # all his other devices. + yield self._device_handler.delete_all_devices_for_user( + user_id, except_device_id=except_device_id, + ) + + # and now delete any access tokens which weren't associated with + # devices (or were associated with this device). + yield self._auth_handler.delete_access_tokens_for_user( + user_id, except_token_id=except_access_token_id, + ) diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index d7dcd1ce5b..b52e4c2aff 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -20,6 +20,7 @@ from synapse.util.metrics import Measure, measure_func from synapse.util.caches.response_cache import ResponseCache from synapse.push.clientformat import format_push_rules_for_user from synapse.visibility import filter_events_for_client +from synapse.types import RoomStreamToken from twisted.internet import defer @@ -51,6 +52,7 @@ class TimelineBatch(collections.namedtuple("TimelineBatch", [ to tell if room needs to be part of the sync result. """ return bool(self.events) + __bool__ = __nonzero__ # python3 class JoinedSyncResult(collections.namedtuple("JoinedSyncResult", [ @@ -75,6 +77,7 @@ class JoinedSyncResult(collections.namedtuple("JoinedSyncResult", [ # nb the notification count does not, er, count: if there's nothing # else in the result, we don't need to send it. ) + __bool__ = __nonzero__ # python3 class ArchivedSyncResult(collections.namedtuple("ArchivedSyncResult", [ @@ -94,6 +97,7 @@ class ArchivedSyncResult(collections.namedtuple("ArchivedSyncResult", [ or self.state or self.account_data ) + __bool__ = __nonzero__ # python3 class InvitedSyncResult(collections.namedtuple("InvitedSyncResult", [ @@ -105,6 +109,30 @@ class InvitedSyncResult(collections.namedtuple("InvitedSyncResult", [ def __nonzero__(self): """Invited rooms should always be reported to the client""" return True + __bool__ = __nonzero__ # python3 + + +class GroupsSyncResult(collections.namedtuple("GroupsSyncResult", [ + "join", + "invite", + "leave", +])): + __slots__ = [] + + def __nonzero__(self): + return bool(self.join or self.invite or self.leave) + __bool__ = __nonzero__ # python3 + + +class DeviceLists(collections.namedtuple("DeviceLists", [ + "changed", # list of user_ids whose devices may have changed + "left", # list of user_ids whose devices we no longer track +])): + __slots__ = [] + + def __nonzero__(self): + return bool(self.changed or self.left) + __bool__ = __nonzero__ # python3 class SyncResult(collections.namedtuple("SyncResult", [ @@ -116,6 +144,9 @@ class SyncResult(collections.namedtuple("SyncResult", [ "archived", # ArchivedSyncResult for each archived room. "to_device", # List of direct messages for the device. "device_lists", # List of user_ids whose devices have chanegd + "device_one_time_keys_count", # Dict of algorithm to count for one time keys + # for this device + "groups", ])): __slots__ = [] @@ -131,8 +162,10 @@ class SyncResult(collections.namedtuple("SyncResult", [ self.archived or self.account_data or self.to_device or - self.device_lists + self.device_lists or + self.groups ) + __bool__ = __nonzero__ # python3 class SyncHandler(object): @@ -143,7 +176,7 @@ class SyncHandler(object): self.presence_handler = hs.get_presence_handler() self.event_sources = hs.get_event_sources() self.clock = hs.get_clock() - self.response_cache = ResponseCache(hs) + self.response_cache = ResponseCache(hs, "sync") self.state = hs.get_state_handler() def wait_for_sync_for_user(self, sync_config, since_token=None, timeout=0, @@ -154,15 +187,11 @@ class SyncHandler(object): Returns: A Deferred SyncResult. """ - result = self.response_cache.get(sync_config.request_key) - if not result: - result = self.response_cache.set( - sync_config.request_key, - self._wait_for_sync_for_user( - sync_config, since_token, timeout, full_state - ) - ) - return result + return self.response_cache.wrap( + sync_config.request_key, + self._wait_for_sync_for_user, + sync_config, since_token, timeout, full_state, + ) @defer.inlineCallbacks def _wait_for_sync_for_user(self, sync_config, since_token, timeout, @@ -209,10 +238,10 @@ class SyncHandler(object): defer.returnValue(rules) @defer.inlineCallbacks - def ephemeral_by_room(self, sync_config, now_token, since_token=None): + 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_config (SyncConfig): The flags, filters and user for the sync. + sync_result_builder(SyncResultBuilder) now_token (StreamToken): Where the server is currently up to. since_token (StreamToken): Where the server was when the client last synced. @@ -222,11 +251,12 @@ class SyncHandler(object): typing events for that room. """ + sync_config = sync_result_builder.sync_config + with Measure(self.clock, "ephemeral_by_room"): typing_key = since_token.typing_key if since_token else "0" - rooms = yield self.store.get_rooms_for_user(sync_config.user.to_string()) - room_ids = [room.room_id for room in rooms] + room_ids = sync_result_builder.joined_room_ids typing_source = self.event_sources.sources["typing"] typing, typing_key = yield typing_source.get_new_events( @@ -288,10 +318,20 @@ class SyncHandler(object): if recents: recents = sync_config.filter_collection.filter_room_timeline(recents) + + # We check if there are any state events, if there are then we pass + # all current state events to the filter_events function. This is to + # 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 = frozenset(current_state_ids.itervalues()) + recents = yield filter_events_for_client( self.store, sync_config.user.to_string(), recents, + always_include_ids=current_state_ids, ) else: recents = [] @@ -323,10 +363,20 @@ class SyncHandler(object): loaded_recents = sync_config.filter_collection.filter_room_timeline( events ) + + # We check if there are any state events, if there are then we pass + # all current state events to the filter_events function. This is to + # 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 = frozenset(current_state_ids.itervalues()) + loaded_recents = yield filter_events_for_client( self.store, sync_config.user.to_string(), loaded_recents, + always_include_ids=current_state_ids, ) loaded_recents.extend(recents) recents = loaded_recents @@ -520,10 +570,22 @@ class SyncHandler(object): # Always use the `now_token` in `SyncResultBuilder` now_token = yield self.event_sources.get_current_token() + user_id = sync_config.user.to_string() + app_service = self.store.get_app_service_by_user_id(user_id) + if app_service: + # We no longer support AS users using /sync directly. + # See https://github.com/matrix-org/matrix-doc/issues/1144 + raise NotImplementedError() + else: + joined_room_ids = yield self.get_rooms_for_user_at( + user_id, now_token.room_stream_id, + ) + sync_result_builder = SyncResultBuilder( sync_config, full_state, since_token=since_token, now_token=now_token, + joined_room_ids=joined_room_ids, ) account_data_by_room = yield self._generate_sync_entry_for_account_data( @@ -533,7 +595,8 @@ class SyncHandler(object): res = yield self._generate_sync_entry_for_rooms( sync_result_builder, account_data_by_room ) - newly_joined_rooms, newly_joined_users = res + newly_joined_rooms, newly_joined_users, _, _ = res + _, _, newly_left_rooms, newly_left_users = res block_all_presence_data = ( since_token is None and @@ -547,9 +610,22 @@ class SyncHandler(object): yield self._generate_sync_entry_for_to_device(sync_result_builder) device_lists = yield self._generate_sync_entry_for_device_list( - sync_result_builder + sync_result_builder, + newly_joined_rooms=newly_joined_rooms, + newly_joined_users=newly_joined_users, + newly_left_rooms=newly_left_rooms, + newly_left_users=newly_left_users, ) + 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( + user_id, device_id + ) + + yield self._generate_sync_entry_for_groups(sync_result_builder) + defer.returnValue(SyncResult( presence=sync_result_builder.presence, account_data=sync_result_builder.account_data, @@ -558,31 +634,103 @@ class SyncHandler(object): archived=sync_result_builder.archived, to_device=sync_result_builder.to_device, device_lists=device_lists, + groups=sync_result_builder.groups, + device_one_time_keys_count=one_time_key_counts, next_batch=sync_result_builder.now_token, )) + @measure_func("_generate_sync_entry_for_groups") + @defer.inlineCallbacks + 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( + user_id, since_token.groups_key, now_token.groups_key, + ) + else: + results = yield self.store.get_all_groups_for_user( + user_id, now_token.groups_key, + ) + + invited = {} + joined = {} + left = {} + for result in results: + membership = result["membership"] + group_id = result["group_id"] + gtype = result["type"] + content = result["content"] + + if membership == "join": + if gtype == "membership": + # TODO: Add profile + content.pop("membership", None) + joined[group_id] = content["content"] + else: + joined.setdefault(group_id, {})[gtype] = content + elif membership == "invite": + if gtype == "membership": + content.pop("membership", None) + invited[group_id] = content["content"] + else: + if gtype == "membership": + left[group_id] = content["content"] + + sync_result_builder.groups = GroupsSyncResult( + join=joined, + invite=invited, + leave=left, + ) + @measure_func("_generate_sync_entry_for_device_list") @defer.inlineCallbacks - def _generate_sync_entry_for_device_list(self, sync_result_builder): + def _generate_sync_entry_for_device_list(self, sync_result_builder, + newly_joined_rooms, newly_joined_users, + newly_left_rooms, newly_left_users): user_id = sync_result_builder.sync_config.user.to_string() since_token = sync_result_builder.since_token if since_token and since_token.device_list_key: - rooms = yield self.store.get_rooms_for_user(user_id) - room_ids = set(r.room_id for r in rooms) - - user_ids_changed = set() changed = yield self.store.get_user_whose_devices_changed( since_token.device_list_key ) - for other_user_id in changed: - other_rooms = yield self.store.get_rooms_for_user(other_user_id) - if room_ids.intersection(e.room_id for e in other_rooms): - user_ids_changed.add(other_user_id) - defer.returnValue(user_ids_changed) + # TODO: Be more clever than this, i.e. remove users who we already + # share a room with? + for room_id in newly_joined_rooms: + joined_users = yield self.state.get_current_user_in_room(room_id) + newly_joined_users.update(joined_users) + + for room_id in newly_left_rooms: + left_users = yield self.state.get_current_user_in_room(room_id) + newly_left_users.update(left_users) + + # TODO: Check that these users are actually new, i.e. either they + # weren't in the previous sync *or* they left and rejoined. + changed.update(newly_joined_users) + + if not changed and not newly_left_users: + defer.returnValue(DeviceLists( + changed=[], + left=newly_left_users, + )) + + users_who_share_room = yield self.store.get_users_who_share_room_with_user( + user_id + ) + + defer.returnValue(DeviceLists( + changed=users_who_share_room & changed, + left=set(newly_left_users) - users_who_share_room, + )) else: - defer.returnValue([]) + defer.returnValue(DeviceLists( + changed=[], + left=[], + )) @defer.inlineCallbacks def _generate_sync_entry_for_to_device(self, sync_result_builder): @@ -609,14 +757,14 @@ class SyncHandler(object): deleted = yield self.store.delete_messages_for_device( user_id, device_id, since_stream_id ) - logger.info("Deleted %d to-device messages up to %d", - deleted, 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( user_id, device_id, since_stream_id, now_token.to_device_key ) - logger.info( + logger.debug( "Returning %d to-device messages between %d and %d (current token: %d)", len(messages), since_stream_id, stream_id, now_token.to_device_key ) @@ -721,14 +869,14 @@ class SyncHandler(object): extra_users_ids.update(users) extra_users_ids.discard(user.to_string()) - states = yield self.presence_handler.get_states( - extra_users_ids, - as_event=True, - ) - presence.extend(states) + if extra_users_ids: + states = yield self.presence_handler.get_states( + extra_users_ids, + ) + presence.extend(states) - # Deduplicate the presence entries so that there's at most one per user - presence = {p["content"]["user_id"]: p for p in presence}.values() + # Deduplicate the presence entries so that there's at most one per user + presence = {p.user_id: p for p in presence}.values() presence = sync_config.filter_collection.filter_presence( presence @@ -746,8 +894,8 @@ class SyncHandler(object): account_data_by_room(dict): Dictionary of per room account data Returns: - Deferred(tuple): Returns a 2-tuple of - `(newly_joined_rooms, newly_joined_users)` + Deferred(tuple): Returns a 4-tuple of + `(newly_joined_rooms, newly_joined_users, newly_left_rooms, newly_left_users)` """ user_id = sync_result_builder.sync_config.user.to_string() block_all_room_ephemeral = ( @@ -759,12 +907,27 @@ class SyncHandler(object): ephemeral_by_room = {} else: now_token, ephemeral_by_room = yield self.ephemeral_by_room( - sync_result_builder.sync_config, + sync_result_builder, now_token=sync_result_builder.now_token, since_token=sync_result_builder.since_token, ) sync_result_builder.now_token = now_token + # We check up front if anything has changed, if it hasn't then there is + # no point in going futher. + 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) + if not have_changed: + tags_by_room = yield self.store.get_updated_tags( + user_id, + since_token.account_data_key, + ) + if not tags_by_room: + logger.debug("no-oping sync") + defer.returnValue(([], [], [], [])) + ignored_account_data = yield self.store.get_global_account_data_by_type_for_user( "m.ignored_user_list", user_id=user_id, ) @@ -774,17 +937,17 @@ class SyncHandler(object): else: ignored_users = frozenset() - if sync_result_builder.since_token: + if since_token: res = yield self._get_rooms_changed(sync_result_builder, ignored_users) - room_entries, invited, newly_joined_rooms = res + room_entries, invited, newly_joined_rooms, newly_left_rooms = res tags_by_room = yield self.store.get_updated_tags( - user_id, - sync_result_builder.since_token.account_data_key, + user_id, since_token.account_data_key, ) else: res = yield 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) @@ -805,17 +968,55 @@ class SyncHandler(object): # Now we want to get any newly joined users newly_joined_users = set() - if sync_result_builder.since_token: + newly_left_users = set() + if since_token: for joined_sync in sync_result_builder.joined: it = itertools.chain( - joined_sync.timeline.events, joined_sync.state.values() + joined_sync.timeline.events, joined_sync.state.itervalues() ) for event in it: if event.type == EventTypes.Member: if event.membership == Membership.JOIN: newly_joined_users.add(event.state_key) + else: + prev_content = event.unsigned.get("prev_content", {}) + prev_membership = prev_content.get("membership", None) + if prev_membership == Membership.JOIN: + newly_left_users.add(event.state_key) + + newly_left_users -= newly_joined_users + + defer.returnValue(( + newly_joined_rooms, + newly_joined_users, + newly_left_rooms, + newly_left_users, + )) + + @defer.inlineCallbacks + 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. + """ + user_id = sync_result_builder.sync_config.user.to_string() + since_token = sync_result_builder.since_token + now_token = sync_result_builder.now_token - defer.returnValue((newly_joined_rooms, newly_joined_users)) + assert since_token + + # Get a list of membership change events that have happened. + rooms_changed = yield self.store.get_membership_changes_for_user( + user_id, since_token.room_key, now_token.room_key + ) + + if rooms_changed: + defer.returnValue(True) + + stream_id = RoomStreamToken.parse_stream_token(since_token.room_key).stream + for room_id in sync_result_builder.joined_room_ids: + if self.store.has_room_changed_since(room_id, stream_id): + defer.returnValue(True) + defer.returnValue(False) @defer.inlineCallbacks def _get_rooms_changed(self, sync_result_builder, ignored_users): @@ -836,14 +1037,6 @@ class SyncHandler(object): assert since_token - app_service = self.store.get_app_service_by_user_id(user_id) - if app_service: - rooms = yield self.store.get_app_service_rooms(app_service) - joined_room_ids = set(r.room_id for r in rooms) - else: - rooms = yield self.store.get_rooms_for_user(user_id) - joined_room_ids = set(r.room_id for r in rooms) - # Get a list of membership change events that have happened. rooms_changed = yield self.store.get_membership_changes_for_user( user_id, since_token.room_key, now_token.room_key @@ -854,16 +1047,29 @@ class SyncHandler(object): mem_change_events_by_room_id.setdefault(event.room_id, []).append(event) newly_joined_rooms = [] + newly_left_rooms = [] room_entries = [] invited = [] - for room_id, events in mem_change_events_by_room_id.items(): + for room_id, events in mem_change_events_by_room_id.iteritems(): non_joins = [e for e in events if e.membership != Membership.JOIN] has_join = len(non_joins) != len(events) # We want to figure out if we joined the room at some point since # the last sync (even if we have since left). This is to make sure # we do send down the room, and with full state, where necessary - if room_id in joined_room_ids or has_join: + + old_state_ids = None + if room_id in sync_result_builder.joined_room_ids and non_joins: + # Always include if the user (re)joined the room, especially + # important so that device list changes are calculated correctly. + # If there are non join member events, but we are still in the room, + # then the user must have left and joined + newly_joined_rooms.append(room_id) + + # User is in the room so we don't need to do the invite/leave checks + 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_mem_ev_id = old_state_ids.get((EventTypes.Member, user_id), None) old_mem_ev = None @@ -874,12 +1080,33 @@ class SyncHandler(object): if not old_mem_ev or old_mem_ev.membership != Membership.JOIN: newly_joined_rooms.append(room_id) - if room_id in joined_room_ids: - continue + # If user is in the room then we don't need to do the invite/leave checks + if room_id in sync_result_builder.joined_room_ids: + continue if not non_joins: continue + # Check if we have left the room. This can either be because we were + # joined before *or* that we since joined and then left. + if events[-1].membership != Membership.JOIN: + if has_join: + 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_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_id, allow_none=True + ) + if old_mem_ev and old_mem_ev.membership == Membership.JOIN: + newly_left_rooms.append(room_id) + # Only bother if we're still currently invited should_invite = non_joins[-1].membership == Membership.INVITE if should_invite: @@ -921,7 +1148,7 @@ class SyncHandler(object): # Get all events for rooms we're currently joined to. room_to_events = yield self.store.get_room_events_stream_for_rooms( - room_ids=joined_room_ids, + room_ids=sync_result_builder.joined_room_ids, from_key=since_token.room_key, to_key=now_token.room_key, limit=timeline_limit + 1, @@ -929,7 +1156,7 @@ class SyncHandler(object): # We loop through all room ids, even if there are no new events, in case # there are non room events taht we need to notify about. - for room_id in joined_room_ids: + for room_id in sync_result_builder.joined_room_ids: room_entry = room_to_events.get(room_id, None) if room_entry: @@ -957,7 +1184,7 @@ class SyncHandler(object): upto_token=since_token, )) - defer.returnValue((room_entries, invited, newly_joined_rooms)) + defer.returnValue((room_entries, invited, newly_joined_rooms, newly_left_rooms)) @defer.inlineCallbacks def _get_all_rooms(self, sync_result_builder, ignored_users): @@ -1137,6 +1364,54 @@ 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): + """Get set of joined rooms for a user at the given stream ordering. + + The stream ordering *must* be recent, otherwise this may throw an + exception if older than a month. (This function is called with the + current token, which should be perfectly fine). + + Args: + user_id (str) + stream_ordering (int) + + ReturnValue: + 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_room_ids = set() + + # We need to check that the stream ordering of the join for each room + # is before the stream_ordering asked for. This might not be the case + # if the user joins a room between us getting the current token and + # calling `get_rooms_for_user_with_stream_ordering`. + # If the membership's stream ordering is after the given stream + # ordering, we need to go and work out if the user was in the room + # before. + for room_id, membership_stream_ordering in joined_rooms: + if membership_stream_ordering <= stream_ordering: + joined_room_ids.add(room_id) + continue + + logger.info("User joined room after current token: %s", room_id) + + extrems = yield self.store.get_forward_extremeties_for_room( + room_id, stream_ordering, + ) + users_in_room = yield self.state.get_current_user_in_room( + room_id, extrems, + ) + if user_id in users_in_room: + joined_room_ids.add(room_id) + + joined_room_ids = frozenset(joined_room_ids) + defer.returnValue(joined_room_ids) + def _action_has_highlight(actions): for action in actions: @@ -1186,7 +1461,8 @@ def _calculate_state(timeline_contains, timeline_start, previous, current): class SyncResultBuilder(object): "Used to help build up a new SyncResult for a user" - def __init__(self, sync_config, full_state, since_token, now_token): + def __init__(self, sync_config, full_state, since_token, now_token, + joined_room_ids): """ Args: sync_config(SyncConfig) @@ -1198,6 +1474,7 @@ class SyncResultBuilder(object): self.full_state = full_state self.since_token = since_token self.now_token = now_token + self.joined_room_ids = joined_room_ids self.presence = [] self.account_data = [] @@ -1205,6 +1482,8 @@ class SyncResultBuilder(object): self.invited = [] self.archived = [] self.device = [] + self.groups = None + self.to_device = [] class RoomSyncResultBuilder(object): diff --git a/synapse/handlers/typing.py b/synapse/handlers/typing.py index 0eea7f8f9c..5d9736e88f 100644 --- a/synapse/handlers/typing.py +++ b/synapse/handlers/typing.py @@ -16,7 +16,7 @@ from twisted.internet import defer from synapse.api.errors import SynapseError, AuthError -from synapse.util.logcontext import preserve_fn +from synapse.util.logcontext import run_in_background from synapse.util.metrics import Measure from synapse.util.wheel_timer import WheelTimer from synapse.types import UserID, get_domain_from_id @@ -24,7 +24,6 @@ from synapse.types import UserID, get_domain_from_id import logging from collections import namedtuple -import ujson as json logger = logging.getLogger(__name__) @@ -57,7 +56,7 @@ class TypingHandler(object): self.federation = hs.get_federation_sender() - hs.get_replication_layer().register_edu_handler("m.typing", self._recv_edu) + hs.get_federation_registry().register_edu_handler("m.typing", self._recv_edu) hs.get_distributor().observe("user_left_room", self.user_left_room) @@ -90,7 +89,7 @@ class TypingHandler(object): until = self._member_typing_until.get(member, None) if not until or until <= now: logger.info("Timing out typing for: %s", member.user_id) - preserve_fn(self._stopped_typing)(member) + self._stopped_typing(member) continue # Check if we need to resend a keep alive over federation for this @@ -98,7 +97,8 @@ class TypingHandler(object): if self.hs.is_mine_id(member.user_id): last_fed_poke = self._member_last_federation_poke.get(member, None) if not last_fed_poke or last_fed_poke + FEDERATION_PING_INTERVAL <= now: - preserve_fn(self._push_remote)( + run_in_background( + self._push_remote, member=member, typing=True ) @@ -148,7 +148,7 @@ class TypingHandler(object): # No point sending another notification defer.returnValue(None) - yield self._push_update( + self._push_update( member=member, typing=True, ) @@ -172,7 +172,7 @@ class TypingHandler(object): member = RoomMember(room_id=room_id, user_id=target_user_id) - yield self._stopped_typing(member) + self._stopped_typing(member) @defer.inlineCallbacks def user_left_room(self, user, room_id): @@ -181,7 +181,6 @@ class TypingHandler(object): member = RoomMember(room_id=room_id, user_id=user_id) yield self._stopped_typing(member) - @defer.inlineCallbacks def _stopped_typing(self, member): if member.user_id not in self._room_typing.get(member.room_id, set()): # No point @@ -190,16 +189,15 @@ class TypingHandler(object): self._member_typing_until.pop(member, None) self._member_last_federation_poke.pop(member, None) - yield self._push_update( + self._push_update( member=member, typing=False, ) - @defer.inlineCallbacks def _push_update(self, member, typing): if self.hs.is_mine_id(member.user_id): # Only send updates for changes to our own users. - yield self._push_remote(member, typing) + run_in_background(self._push_remote, member, typing) self._push_update_local( member=member, @@ -208,28 +206,31 @@ class TypingHandler(object): @defer.inlineCallbacks def _push_remote(self, member, typing): - users = yield self.state.get_current_user_in_room(member.room_id) - self._member_last_federation_poke[member] = self.clock.time_msec() + try: + users = yield self.state.get_current_user_in_room(member.room_id) + self._member_last_federation_poke[member] = self.clock.time_msec() - now = self.clock.time_msec() - self.wheel_timer.insert( - now=now, - obj=member, - then=now + FEDERATION_PING_INTERVAL, - ) + now = self.clock.time_msec() + self.wheel_timer.insert( + now=now, + obj=member, + then=now + FEDERATION_PING_INTERVAL, + ) - for domain in set(get_domain_from_id(u) for u in users): - if domain != self.server_name: - self.federation.send_edu( - destination=domain, - edu_type="m.typing", - content={ - "room_id": member.room_id, - "user_id": member.user_id, - "typing": typing, - }, - key=member, - ) + for domain in set(get_domain_from_id(u) for u in users): + if domain != self.server_name: + self.federation.send_edu( + destination=domain, + edu_type="m.typing", + content={ + "room_id": member.room_id, + "user_id": member.user_id, + "typing": typing, + }, + key=member, + ) + except Exception: + logger.exception("Error pushing typing notif to remotes") @defer.inlineCallbacks def _recv_edu(self, origin, content): @@ -288,11 +289,13 @@ class TypingHandler(object): for room_id, serial in self._room_serials.items(): if last_id < serial and serial <= current_id: typing = self._room_typing[room_id] - typing_bytes = json.dumps(list(typing), ensure_ascii=False) - rows.append((serial, room_id, typing_bytes)) + rows.append((serial, room_id, list(typing))) rows.sort() return rows + def get_current_token(self): + return self._latest_room_serial + class TypingNotificationEventSource(object): def __init__(self, hs): diff --git a/synapse/handlers/user_directory.py b/synapse/handlers/user_directory.py new file mode 100644 index 0000000000..714f0195c8 --- /dev/null +++ b/synapse/handlers/user_directory.py @@ -0,0 +1,681 @@ +# -*- coding: utf-8 -*- +# Copyright 2017 Vector Creations 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 synapse.api.constants import EventTypes, JoinRules, Membership +from synapse.storage.roommember import ProfileInfo +from synapse.util.metrics import Measure +from synapse.util.async import sleep +from synapse.types import get_localpart_from_id + + +logger = logging.getLogger(__name__) + + +class UserDirectoryHandler(object): + """Handles querying of and keeping updated the user_directory. + + N.B.: ASSUMES IT IS THE ONLY THING THAT MODIFIES THE USER DIRECTORY + + The user directory is filled with users who this server can see are joined to a + world_readable or publically joinable room. We keep a database table up to date + by streaming changes of the current state and recalculating whether users should + be in the directory or not when necessary. + + For each user in the directory we also store a room_id which is public and that the + user is joined to. This allows us to ignore history_visibility and join_rules changes + for that user in all other public rooms, as we know they'll still be in at least + one public room. + """ + + INITIAL_ROOM_SLEEP_MS = 50 + INITIAL_ROOM_SLEEP_COUNT = 100 + INITIAL_ROOM_BATCH_SIZE = 100 + INITIAL_USER_SLEEP_MS = 10 + + def __init__(self, hs): + self.store = hs.get_datastore() + self.state = hs.get_state_handler() + self.server_name = hs.hostname + self.clock = hs.get_clock() + self.notifier = hs.get_notifier() + self.is_mine_id = hs.is_mine_id + self.update_user_directory = hs.config.update_user_directory + self.search_all_users = hs.config.user_directory_search_all_users + + # When start up for the first time we need to populate the user_directory. + # This is a set of user_id's we've inserted already + self.initially_handled_users = set() + self.initially_handled_users_in_public = set() + + self.initially_handled_users_share = set() + self.initially_handled_users_share_private_room = set() + + # The current position in the current_state_delta stream + self.pos = None + + # Guard to ensure we only process deltas one at a time + self._is_processing = False + + if self.update_user_directory: + self.notifier.add_replication_callback(self.notify_new_event) + + # We kick this off so that we don't have to wait for a change before + # we start populating the user directory + self.clock.call_later(0, self.notify_new_event) + + def search_users(self, user_id, search_term, limit): + """Searches for users in directory + + Returns: + dict of the form:: + + { + "limited": <bool>, # whether there were more results or not + "results": [ # Ordered by best match first + { + "user_id": <user_id>, + "display_name": <display_name>, + "avatar_url": <avatar_url> + } + ] + } + """ + return self.store.search_user_dir(user_id, search_term, limit) + + @defer.inlineCallbacks + def notify_new_event(self): + """Called when there may be more deltas to process + """ + if not self.update_user_directory: + return + + if self._is_processing: + return + + self._is_processing = True + try: + yield self._unsafe_process() + finally: + self._is_processing = False + + @defer.inlineCallbacks + def handle_local_profile_change(self, user_id, profile): + """Called to update index of our local user profiles when they change + irrespective of any rooms the user may be in. + """ + yield self.store.update_profile_in_user_dir( + user_id, profile.display_name, profile.avatar_url, None, + ) + + @defer.inlineCallbacks + def _unsafe_process(self): + # If self.pos is None then means we haven't fetched it from DB + if self.pos is None: + self.pos = yield self.store.get_user_directory_stream_pos() + + # If still None then we need to do the initial fill of directory + if self.pos is None: + yield self._do_initial_spam() + self.pos = yield self.store.get_user_directory_stream_pos() + + # Loop round handling deltas until we're up to date + while True: + with Measure(self.clock, "user_dir_delta"): + deltas = yield self.store.get_current_state_deltas(self.pos) + if not deltas: + return + + logger.info("Handling %d state deltas", len(deltas)) + yield self._handle_deltas(deltas) + + self.pos = deltas[-1]["stream_id"] + yield self.store.update_user_directory_stream_pos(self.pos) + + @defer.inlineCallbacks + def _do_initial_spam(self): + """Populates the user_directory from the current state of the DB, used + when synapse first starts with user_directory support + """ + new_pos = yield self.store.get_max_stream_id_in_current_state_deltas() + + # Delete any existing entries just in case there are any + yield self.store.delete_all_from_user_dir() + + # We process by going through each existing room at a time. + room_ids = yield self.store.get_all_rooms() + + logger.info("Doing initial update of user directory. %d rooms", len(room_ids)) + num_processed_rooms = 0 + + for room_id in room_ids: + logger.info("Handling room %d/%d", num_processed_rooms + 1, len(room_ids)) + yield self._handle_initial_room(room_id) + num_processed_rooms += 1 + yield sleep(self.INITIAL_ROOM_SLEEP_MS / 1000.) + + logger.info("Processed all rooms.") + + if self.search_all_users: + num_processed_users = 0 + user_ids = yield self.store.get_all_local_users() + logger.info("Doing initial update of user directory. %d users", len(user_ids)) + for user_id in user_ids: + # We add profiles for all users even if they don't match the + # include pattern, just in case we want to change it in future + logger.info("Handling user %d/%d", num_processed_users + 1, len(user_ids)) + yield self._handle_local_user(user_id) + num_processed_users += 1 + yield sleep(self.INITIAL_USER_SLEEP_MS / 1000.) + + logger.info("Processed all users") + + self.initially_handled_users = None + self.initially_handled_users_in_public = None + self.initially_handled_users_share = None + self.initially_handled_users_share_private_room = None + + yield self.store.update_user_directory_stream_pos(new_pos) + + @defer.inlineCallbacks + def _handle_initial_room(self, room_id): + """Called when we initially fill out user_directory one room at a time + """ + is_in_room = yield self.store.is_host_joined(room_id, self.server_name) + if not is_in_room: + return + + is_public = yield self.store.is_room_world_readable_or_publicly_joinable(room_id) + + users_with_profile = yield self.state.get_current_user_in_room(room_id) + user_ids = set(users_with_profile) + unhandled_users = user_ids - self.initially_handled_users + + yield self.store.add_profiles_to_user_dir( + room_id, { + user_id: users_with_profile[user_id] for user_id in unhandled_users + } + ) + + self.initially_handled_users |= unhandled_users + + if is_public: + yield self.store.add_users_to_public_room( + room_id, + user_ids=user_ids - self.initially_handled_users_in_public + ) + self.initially_handled_users_in_public |= user_ids + + # We now go and figure out the new users who share rooms with user entries + # We sleep aggressively here as otherwise it can starve resources. + # We also batch up inserts/updates, but try to avoid too many at once. + to_insert = set() + to_update = set() + count = 0 + for user_id in user_ids: + if count % self.INITIAL_ROOM_SLEEP_COUNT == 0: + yield sleep(self.INITIAL_ROOM_SLEEP_MS / 1000.) + + if not self.is_mine_id(user_id): + count += 1 + continue + + if self.store.get_if_app_services_interested_in_user(user_id): + count += 1 + continue + + for other_user_id in user_ids: + if user_id == other_user_id: + continue + + if count % self.INITIAL_ROOM_SLEEP_COUNT == 0: + yield sleep(self.INITIAL_ROOM_SLEEP_MS / 1000.) + count += 1 + + user_set = (user_id, other_user_id) + + if user_set in self.initially_handled_users_share_private_room: + continue + + if user_set in self.initially_handled_users_share: + if is_public: + continue + to_update.add(user_set) + else: + to_insert.add(user_set) + + if is_public: + self.initially_handled_users_share.add(user_set) + else: + self.initially_handled_users_share_private_room.add(user_set) + + if len(to_insert) > self.INITIAL_ROOM_BATCH_SIZE: + yield self.store.add_users_who_share_room( + room_id, not is_public, to_insert, + ) + to_insert.clear() + + if len(to_update) > self.INITIAL_ROOM_BATCH_SIZE: + yield self.store.update_users_who_share_room( + room_id, not is_public, to_update, + ) + to_update.clear() + + if to_insert: + yield self.store.add_users_who_share_room( + room_id, not is_public, to_insert, + ) + to_insert.clear() + + if to_update: + yield self.store.update_users_who_share_room( + room_id, not is_public, to_update, + ) + to_update.clear() + + @defer.inlineCallbacks + def _handle_deltas(self, deltas): + """Called with the state deltas to process + """ + for delta in deltas: + typ = delta["type"] + state_key = delta["state_key"] + room_id = delta["room_id"] + event_id = delta["event_id"] + prev_event_id = delta["prev_event_id"] + + logger.debug("Handling: %r %r, %s", typ, state_key, event_id) + + # For join rule and visibility changes we need to check if the room + # may have become public or not and add/remove the users in said room + if typ in (EventTypes.RoomHistoryVisibility, EventTypes.JoinRules): + yield self._handle_room_publicity_change( + room_id, prev_event_id, event_id, typ, + ) + elif typ == EventTypes.Member: + change = yield self._get_key_change( + prev_event_id, event_id, + key_name="membership", + public_value=Membership.JOIN, + ) + + if change is None: + # Handle any profile changes + yield self._handle_profile_change( + state_key, room_id, prev_event_id, event_id, + ) + continue + + if not change: + # Need to check if the server left the room entirely, if so + # we might need to remove all the users in that room + is_in_room = yield self.store.is_host_joined( + room_id, self.server_name, + ) + if not is_in_room: + logger.info("Server left room: %r", room_id) + # Fetch all the users that we marked as being in user + # directory due to being in the room and then check if + # need to remove those users or not + user_ids = yield self.store.get_users_in_dir_due_to_room(room_id) + for user_id in user_ids: + yield self._handle_remove_user(room_id, user_id) + return + else: + logger.debug("Server is still in room: %r", room_id) + + if change: # The user joined + event = yield self.store.get_event(event_id, allow_none=True) + profile = ProfileInfo( + avatar_url=event.content.get("avatar_url"), + display_name=event.content.get("displayname"), + ) + + yield self._handle_new_user(room_id, state_key, profile) + else: # The user left + yield self._handle_remove_user(room_id, state_key) + else: + logger.debug("Ignoring irrelevant type: %r", typ) + + @defer.inlineCallbacks + def _handle_room_publicity_change(self, room_id, prev_event_id, event_id, typ): + """Handle a room having potentially changed from/to world_readable/publically + joinable. + + Args: + room_id (str) + prev_event_id (str|None): The previous event before the state change + event_id (str|None): The new event after the state change + typ (str): Type of the event + """ + logger.debug("Handling change for %s: %s", typ, room_id) + + if typ == EventTypes.RoomHistoryVisibility: + change = yield self._get_key_change( + prev_event_id, event_id, + key_name="history_visibility", + public_value="world_readable", + ) + elif typ == EventTypes.JoinRules: + change = yield self._get_key_change( + prev_event_id, event_id, + key_name="join_rule", + public_value=JoinRules.PUBLIC, + ) + else: + raise Exception("Invalid event type") + # If change is None, no change. True => become world_readable/public, + # False => was world_readable/public + if change is None: + logger.debug("No change") + return + + # There's been a change to or from being world readable. + + is_public = yield self.store.is_room_world_readable_or_publicly_joinable( + room_id + ) + + logger.debug("Change: %r, is_public: %r", change, is_public) + + if change and not is_public: + # If we became world readable but room isn't currently public then + # we ignore the change + return + elif not change and is_public: + # If we stopped being world readable but are still public, + # ignore the change + return + + if change: + users_with_profile = yield self.state.get_current_user_in_room(room_id) + for user_id, profile in users_with_profile.iteritems(): + yield self._handle_new_user(room_id, user_id, profile) + else: + users = yield self.store.get_users_in_public_due_to_room(room_id) + for user_id in users: + yield self._handle_remove_user(room_id, user_id) + + @defer.inlineCallbacks + def _handle_local_user(self, user_id): + """Adds a new local roomless user into the user_directory_search table. + Used to populate up the user index when we have an + user_directory_search_all_users specified. + """ + logger.debug("Adding new local user to dir, %r", user_id) + + profile = yield self.store.get_profileinfo(get_localpart_from_id(user_id)) + + row = yield self.store.get_user_in_directory(user_id) + if not row: + yield self.store.add_profiles_to_user_dir(None, {user_id: profile}) + + @defer.inlineCallbacks + def _handle_new_user(self, room_id, user_id, profile): + """Called when we might need to add user to directory + + Args: + room_id (str): room_id that user joined or started being public + user_id (str) + """ + logger.debug("Adding new user to dir, %r", user_id) + + row = yield self.store.get_user_in_directory(user_id) + if not row: + yield self.store.add_profiles_to_user_dir(room_id, {user_id: profile}) + + is_public = yield self.store.is_room_world_readable_or_publicly_joinable( + room_id + ) + + if is_public: + row = yield self.store.get_user_in_public_room(user_id) + if not row: + yield self.store.add_users_to_public_room(room_id, [user_id]) + else: + logger.debug("Not adding new user to public dir, %r", user_id) + + # Now we update users who share rooms with users. We do this by getting + # all the current users in the room and seeing which aren't already + # marked in the database as sharing with `user_id` + + users_with_profile = yield self.state.get_current_user_in_room(room_id) + + to_insert = set() + to_update = set() + + is_appservice = self.store.get_if_app_services_interested_in_user(user_id) + + # First, if they're our user then we need to update for every user + if self.is_mine_id(user_id) and not is_appservice: + # Returns a map of other_user_id -> shared_private. We only need + # to update mappings if for users that either don't share a room + # already (aren't in the map) or, if the room is private, those that + # only share a public room. + user_ids_shared = yield self.store.get_users_who_share_room_from_dir( + user_id + ) + + for other_user_id in users_with_profile: + if user_id == other_user_id: + continue + + shared_is_private = user_ids_shared.get(other_user_id) + if shared_is_private is True: + # We've already marked in the database they share a private room + continue + elif shared_is_private is False: + # They already share a public room, so only update if this is + # a private room + if not is_public: + to_update.add((user_id, other_user_id)) + elif shared_is_private is None: + # This is the first time they both share a room + to_insert.add((user_id, other_user_id)) + + # Next we need to update for every local user in the room + for other_user_id in users_with_profile: + if user_id == other_user_id: + continue + + is_appservice = self.store.get_if_app_services_interested_in_user( + other_user_id + ) + if self.is_mine_id(other_user_id) and not is_appservice: + shared_is_private = yield self.store.get_if_users_share_a_room( + other_user_id, user_id, + ) + if shared_is_private is True: + # We've already marked in the database they share a private room + continue + elif shared_is_private is False: + # They already share a public room, so only update if this is + # a private room + if not is_public: + to_update.add((other_user_id, user_id)) + elif shared_is_private is None: + # This is the first time they both share a room + to_insert.add((other_user_id, user_id)) + + if to_insert: + yield self.store.add_users_who_share_room( + room_id, not is_public, to_insert, + ) + + if to_update: + yield self.store.update_users_who_share_room( + room_id, not is_public, to_update, + ) + + @defer.inlineCallbacks + def _handle_remove_user(self, room_id, user_id): + """Called when we might need to remove user to directory + + Args: + room_id (str): room_id that user left or stopped being public that + user_id (str) + """ + logger.debug("Maybe removing user %r", user_id) + + row = yield self.store.get_user_in_directory(user_id) + update_user_dir = row and row["room_id"] == room_id + + row = yield self.store.get_user_in_public_room(user_id) + update_user_in_public = row and row["room_id"] == room_id + + if (update_user_in_public or update_user_dir): + # XXX: Make this faster? + rooms = yield self.store.get_rooms_for_user(user_id) + for j_room_id in rooms: + if (not update_user_in_public and not update_user_dir): + break + + is_in_room = yield self.store.is_host_joined( + j_room_id, self.server_name, + ) + + if not is_in_room: + continue + + if update_user_dir: + update_user_dir = False + yield self.store.update_user_in_user_dir(user_id, j_room_id) + + is_public = yield self.store.is_room_world_readable_or_publicly_joinable( + j_room_id + ) + + if update_user_in_public and is_public: + yield self.store.update_user_in_public_user_list(user_id, j_room_id) + update_user_in_public = False + + if update_user_dir: + yield self.store.remove_from_user_dir(user_id) + elif update_user_in_public: + yield self.store.remove_from_user_in_public_room(user_id) + + # Now handle users_who_share_rooms. + + # Get a list of user tuples that were in the DB due to this room and + # users (this includes tuples where the other user matches `user_id`) + user_tuples = yield self.store.get_users_in_share_dir_with_room_id( + user_id, room_id, + ) + + for user_id, other_user_id in user_tuples: + # For each user tuple get a list of rooms that they still share, + # trying to find a private room, and update the entry in the DB + rooms = yield self.store.get_rooms_in_common_for_users(user_id, other_user_id) + + # If they dont share a room anymore, remove the mapping + if not rooms: + yield self.store.remove_user_who_share_room( + user_id, other_user_id, + ) + continue + + found_public_share = None + for j_room_id in rooms: + is_public = yield self.store.is_room_world_readable_or_publicly_joinable( + j_room_id + ) + + if is_public: + found_public_share = j_room_id + else: + found_public_share = None + yield self.store.update_users_who_share_room( + room_id, not is_public, [(user_id, other_user_id)], + ) + break + + if found_public_share: + yield self.store.update_users_who_share_room( + room_id, not is_public, [(user_id, other_user_id)], + ) + + @defer.inlineCallbacks + def _handle_profile_change(self, user_id, room_id, prev_event_id, event_id): + """Check member event changes for any profile changes and update the + database if there are. + """ + if not prev_event_id or not event_id: + return + + prev_event = yield self.store.get_event(prev_event_id, allow_none=True) + event = yield self.store.get_event(event_id, allow_none=True) + + if not prev_event or not event: + return + + if event.membership != Membership.JOIN: + return + + prev_name = prev_event.content.get("displayname") + new_name = event.content.get("displayname") + + prev_avatar = prev_event.content.get("avatar_url") + new_avatar = event.content.get("avatar_url") + + if prev_name != new_name or prev_avatar != new_avatar: + yield self.store.update_profile_in_user_dir( + user_id, new_name, new_avatar, room_id, + ) + + @defer.inlineCallbacks + def _get_key_change(self, prev_event_id, event_id, key_name, public_value): + """Given two events check if the `key_name` field in content changed + from not matching `public_value` to doing so. + + For example, check if `history_visibility` (`key_name`) changed from + `shared` to `world_readable` (`public_value`). + + Returns: + None if the field in the events either both match `public_value` + or if neither do, i.e. there has been no change. + True if it didnt match `public_value` but now does + False if it did match `public_value` but now doesn't + """ + prev_event = None + event = None + if prev_event_id: + prev_event = yield self.store.get_event(prev_event_id, allow_none=True) + + if event_id: + event = yield self.store.get_event(event_id, allow_none=True) + + if not event and not prev_event: + logger.debug("Neither event exists: %r %r", prev_event_id, event_id) + defer.returnValue(None) + + prev_value = None + value = None + + if prev_event: + prev_value = prev_event.content.get(key_name) + + if event: + value = event.content.get(key_name) + + logger.debug("prev_value: %r -> value: %r", prev_value, value) + + if value == public_value and prev_value != public_value: + defer.returnValue(True) + elif value != public_value and prev_value == public_value: + defer.returnValue(False) + else: + defer.returnValue(None) |