diff options
Diffstat (limited to 'synapse')
33 files changed, 397 insertions, 249 deletions
diff --git a/synapse/api/auth.py b/synapse/api/auth.py index ac0a3655a5..f17fda6315 100644 --- a/synapse/api/auth.py +++ b/synapse/api/auth.py @@ -204,8 +204,8 @@ class Auth(object): ip_addr = self.hs.get_ip_from_request(request) user_agent = request.requestHeaders.getRawHeaders( - "User-Agent", - default=[""] + b"User-Agent", + default=[b""] )[0] if user and access_token and ip_addr: self.store.insert_client_ip( @@ -672,7 +672,7 @@ def has_access_token(request): bool: False if no access_token was given, True otherwise. """ query_params = request.args.get("access_token") - auth_headers = request.requestHeaders.getRawHeaders("Authorization") + auth_headers = request.requestHeaders.getRawHeaders(b"Authorization") return bool(query_params) or bool(auth_headers) @@ -692,8 +692,8 @@ def get_access_token_from_request(request, token_not_found_http_status=401): AuthError: If there isn't an access_token in the request. """ - auth_headers = request.requestHeaders.getRawHeaders("Authorization") - query_params = request.args.get("access_token") + auth_headers = request.requestHeaders.getRawHeaders(b"Authorization") + query_params = request.args.get(b"access_token") if auth_headers: # Try the get the access_token from a "Authorization: Bearer" # header diff --git a/synapse/app/frontend_proxy.py b/synapse/app/frontend_proxy.py index de889357c3..b349e3e3ce 100644 --- a/synapse/app/frontend_proxy.py +++ b/synapse/app/frontend_proxy.py @@ -90,7 +90,7 @@ class KeyUploadServlet(RestServlet): # They're actually trying to upload something, proxy to main synapse. # Pass through the auth headers, if any, in case the access token # is there. - auth_headers = request.requestHeaders.getRawHeaders("Authorization", []) + auth_headers = request.requestHeaders.getRawHeaders(b"Authorization", []) headers = { "Authorization": auth_headers, } diff --git a/synapse/appservice/__init__.py b/synapse/appservice/__init__.py index d5a7a5ce2f..5fdb579723 100644 --- a/synapse/appservice/__init__.py +++ b/synapse/appservice/__init__.py @@ -21,6 +21,8 @@ from twisted.internet import defer import logging import re +from six import string_types + logger = logging.getLogger(__name__) @@ -146,7 +148,7 @@ class ApplicationService(object): ) regex = regex_obj.get("regex") - if isinstance(regex, basestring): + if isinstance(regex, string_types): regex_obj["regex"] = re.compile(regex) # Pre-compile regex else: raise ValueError( diff --git a/synapse/appservice/api.py b/synapse/appservice/api.py index 40c433d7ae..11e9c37c63 100644 --- a/synapse/appservice/api.py +++ b/synapse/appservice/api.py @@ -73,7 +73,8 @@ class ApplicationServiceApi(SimpleHttpClient): super(ApplicationServiceApi, self).__init__(hs) self.clock = hs.get_clock() - self.protocol_meta_cache = ResponseCache(hs, timeout_ms=HOUR_IN_MS) + self.protocol_meta_cache = ResponseCache(hs, "as_protocol_meta", + timeout_ms=HOUR_IN_MS) @defer.inlineCallbacks def query_user(self, service, user_id): diff --git a/synapse/config/_base.py b/synapse/config/_base.py index fa105bce72..32b439d20a 100644 --- a/synapse/config/_base.py +++ b/synapse/config/_base.py @@ -19,6 +19,8 @@ import os import yaml from textwrap import dedent +from six import integer_types + class ConfigError(Exception): pass @@ -49,7 +51,7 @@ Missing mandatory `server_name` config option. class Config(object): @staticmethod def parse_size(value): - if isinstance(value, int) or isinstance(value, long): + if isinstance(value, integer_types): return value sizes = {"K": 1024, "M": 1024 * 1024} size = 1 @@ -61,7 +63,7 @@ class Config(object): @staticmethod def parse_duration(value): - if isinstance(value, int) or isinstance(value, long): + if isinstance(value, integer_types): return value second = 1000 minute = 60 * second @@ -288,22 +290,22 @@ class Config(object): ) obj.invoke_all("generate_files", config) config_file.write(config_bytes) - print ( + print(( "A config file has been generated in %r for server name" " %r with corresponding SSL keys and self-signed" " certificates. Please review this file and customise it" " to your needs." - ) % (config_path, server_name) - print ( + ) % (config_path, server_name)) + print( "If this server name is incorrect, you will need to" " regenerate the SSL certificates" ) return else: - print ( + print(( "Config file %r already exists. Generating any missing key" " files." - ) % (config_path,) + ) % (config_path,)) generate_keys = True parser = argparse.ArgumentParser( diff --git a/synapse/config/appservice.py b/synapse/config/appservice.py index aba0aec6e8..9a2359b6fd 100644 --- a/synapse/config/appservice.py +++ b/synapse/config/appservice.py @@ -21,6 +21,8 @@ import urllib import yaml import logging +from six import string_types + logger = logging.getLogger(__name__) @@ -89,14 +91,14 @@ def _load_appservice(hostname, as_info, config_filename): "id", "as_token", "hs_token", "sender_localpart" ] for field in required_string_fields: - if not isinstance(as_info.get(field), basestring): + if not isinstance(as_info.get(field), string_types): raise KeyError("Required string field: '%s' (%s)" % ( field, config_filename, )) # 'url' must either be a string or explicitly null, not missing # to avoid accidentally turning off push for ASes. - if (not isinstance(as_info.get("url"), basestring) and + if (not isinstance(as_info.get("url"), string_types) and as_info.get("url", "") is not None): raise KeyError( "Required string field or explicit null: 'url' (%s)" % (config_filename,) @@ -128,7 +130,7 @@ def _load_appservice(hostname, as_info, config_filename): "Expected namespace entry in %s to be an object," " but got %s", ns, regex_obj ) - if not isinstance(regex_obj.get("regex"), basestring): + if not isinstance(regex_obj.get("regex"), string_types): raise ValueError( "Missing/bad type 'regex' key in %s", regex_obj ) diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index bea7fd0b71..e4ce037acf 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -65,7 +65,7 @@ class FederationServer(FederationBase): # We cache responses to state queries, as they take a while and often # come in waves. - self._state_resp_cache = ResponseCache(hs, timeout_ms=30000) + self._state_resp_cache = ResponseCache(hs, "state_resp", timeout_ms=30000) @defer.inlineCallbacks @log_function diff --git a/synapse/federation/transaction_queue.py b/synapse/federation/transaction_queue.py index a141ec9953..963d938edd 100644 --- a/synapse/federation/transaction_queue.py +++ b/synapse/federation/transaction_queue.py @@ -169,7 +169,7 @@ class TransactionQueue(object): while True: last_token = yield self.store.get_federation_out_pos("events") next_token, events = yield self.store.get_all_new_events_stream( - last_token, self._last_poked_id, limit=20, + last_token, self._last_poked_id, limit=100, ) logger.debug("Handling %s -> %s", last_token, next_token) @@ -177,24 +177,33 @@ class TransactionQueue(object): if not events and next_token >= self._last_poked_id: break - for event in events: + @defer.inlineCallbacks + def handle_event(event): # Only send events for this server. send_on_behalf_of = event.internal_metadata.get_send_on_behalf_of() is_mine = self.is_mine_id(event.event_id) if not is_mine and send_on_behalf_of is None: - continue - - # Get the state from before the event. - # We need to make sure that this is the state from before - # the event and not from after it. - # Otherwise if the last member on a server in a room is - # banned then it won't receive the event because it won't - # be in the room after the ban. - destinations = yield self.state.get_current_hosts_in_room( - event.room_id, latest_event_ids=[ - prev_id for prev_id, _ in event.prev_events - ], - ) + return + + try: + # Get the state from before the event. + # We need to make sure that this is the state from before + # the event and not from after it. + # Otherwise if the last member on a server in a room is + # banned then it won't receive the event because it won't + # be in the room after the ban. + destinations = yield self.state.get_current_hosts_in_room( + event.room_id, latest_event_ids=[ + prev_id for prev_id, _ in event.prev_events + ], + ) + except Exception: + logger.exception( + "Failed to calculate hosts in room for event: %s", + event.event_id, + ) + return + destinations = set(destinations) if send_on_behalf_of is not None: @@ -207,12 +216,44 @@ class TransactionQueue(object): self._send_pdu(event, destinations) - events_processed_counter.inc_by(len(events)) + @defer.inlineCallbacks + def handle_room_events(events): + for event in events: + yield handle_event(event) + + events_by_room = {} + for event in events: + events_by_room.setdefault(event.room_id, []).append(event) + + yield logcontext.make_deferred_yieldable(defer.gatherResults( + [ + logcontext.run_in_background(handle_room_events, evs) + for evs in events_by_room.itervalues() + ], + consumeErrors=True + )) yield self.store.update_federation_out_pos( "events", next_token ) + if events: + now = self.clock.time_msec() + ts = yield self.store.get_received_ts(events[-1].event_id) + + synapse.metrics.event_processing_lag.set( + now - ts, "federation_sender", + ) + synapse.metrics.event_processing_last_ts.set( + ts, "federation_sender", + ) + + events_processed_counter.inc_by(len(events)) + + synapse.metrics.event_processing_positions.set( + next_token, "federation_sender", + ) + finally: self._is_processing = False diff --git a/synapse/federation/transport/server.py b/synapse/federation/transport/server.py index 4c94d5a36c..ff0656df3e 100644 --- a/synapse/federation/transport/server.py +++ b/synapse/federation/transport/server.py @@ -94,12 +94,6 @@ class Authenticator(object): "signatures": {}, } - if ( - self.federation_domain_whitelist is not None and - self.server_name not in self.federation_domain_whitelist - ): - raise FederationDeniedError(self.server_name) - if content is not None: json_request["content"] = content @@ -138,6 +132,12 @@ class Authenticator(object): json_request["origin"] = origin json_request["signatures"].setdefault(origin, {})[key] = sig + if ( + self.federation_domain_whitelist is not None and + origin not in self.federation_domain_whitelist + ): + raise FederationDeniedError(origin) + if not json_request["signatures"]: raise NoAuthenticationError( 401, "Missing Authorization headers", Codes.UNAUTHORIZED, diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index 3dd3fa2a27..0245197c02 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -18,7 +18,9 @@ from twisted.internet import defer import synapse from synapse.api.constants import EventTypes from synapse.util.metrics import Measure -from synapse.util.logcontext import make_deferred_yieldable, preserve_fn +from synapse.util.logcontext import ( + make_deferred_yieldable, preserve_fn, run_in_background, +) import logging @@ -84,11 +86,16 @@ class ApplicationServicesHandler(object): 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. @@ -108,9 +115,33 @@ class ApplicationServicesHandler(object): service, event ) - events_processed_counter.inc_by(len(events)) + @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) + + 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 diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 6de6e13b7b..54cd691f91 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -454,40 +454,39 @@ class EventCreationHandler(object): """ 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.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, - requester=requester, - prev_event_ids=prev_event_ids, - ) + event, context = yield self.create_new_client_event( + builder=builder, + requester=requester, + prev_event_ids=prev_event_ids, + ) defer.returnValue((event, context)) @@ -557,27 +556,34 @@ class EventCreationHandler(object): See self.create_event and self.send_nonmember_event. """ - event, context = yield self.create_event( - requester, - event_dict, - token_id=requester.access_token_id, - txn_id=txn_id - ) - 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 + # 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 ) - yield self.send_nonmember_event( - requester, - event, - context, - ratelimit=ratelimit, - ) + 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 + ) + + yield self.send_nonmember_event( + requester, + event, + context, + ratelimit=ratelimit, + ) defer.returnValue(event) @measure_func("create_new_client_event") diff --git a/synapse/handlers/register.py b/synapse/handlers/register.py index dd03705279..f83c6b3cf8 100644 --- a/synapse/handlers/register.py +++ b/synapse/handlers/register.py @@ -23,7 +23,7 @@ from synapse.api.errors import ( ) from synapse.http.client import CaptchaServerHttpClient from synapse import types -from synapse.types import UserID +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 @@ -205,10 +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 @@ -483,3 +490,28 @@ class RegistrationHandler(BaseHandler): ) 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_list.py b/synapse/handlers/room_list.py index 5d81f59b44..8028d793c2 100644 --- a/synapse/handlers/room_list.py +++ b/synapse/handlers/room_list.py @@ -44,8 +44,9 @@ 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, diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index 9977be8831..c45142d38d 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -852,6 +852,14 @@ class RoomMemberMasterHandler(RoomMemberHandler): 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") diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index 0f713ce038..06d17ab20c 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -169,7 +169,7 @@ 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, diff --git a/synapse/http/endpoint.py b/synapse/http/endpoint.py index 87639b9151..00572c2897 100644 --- a/synapse/http/endpoint.py +++ b/synapse/http/endpoint.py @@ -12,8 +12,6 @@ # 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 socket - from twisted.internet.endpoints import HostnameEndpoint, wrapClientTLS from twisted.internet import defer, reactor from twisted.internet.error import ConnectError @@ -33,7 +31,7 @@ SERVER_CACHE = {} # our record of an individual server which can be tried to reach a destination. # -# "host" is actually a dotted-quad or ipv6 address string. Except when there's +# "host" is the hostname acquired from the SRV record. Except when there's # no SRV record, in which case it is the original hostname. _Server = collections.namedtuple( "_Server", "priority weight host port expires" @@ -297,20 +295,13 @@ def resolve_service(service_name, dns_client=client, cache=SERVER_CACHE, clock=t payload = answer.payload - hosts = yield _get_hosts_for_srv_record( - dns_client, str(payload.target) - ) - - for (ip, ttl) in hosts: - host_ttl = min(answer.ttl, ttl) - - servers.append(_Server( - host=ip, - port=int(payload.port), - priority=int(payload.priority), - weight=int(payload.weight), - expires=int(clock.time()) + host_ttl, - )) + servers.append(_Server( + host=str(payload.target), + port=int(payload.port), + priority=int(payload.priority), + weight=int(payload.weight), + expires=int(clock.time()) + answer.ttl, + )) servers.sort() cache[service_name] = list(servers) @@ -328,81 +319,3 @@ def resolve_service(service_name, dns_client=client, cache=SERVER_CACHE, clock=t raise e defer.returnValue(servers) - - -@defer.inlineCallbacks -def _get_hosts_for_srv_record(dns_client, host): - """Look up each of the hosts in a SRV record - - Args: - dns_client (twisted.names.dns.IResolver): - host (basestring): host to look up - - Returns: - Deferred[list[(str, int)]]: a list of (host, ttl) pairs - - """ - ip4_servers = [] - ip6_servers = [] - - def cb(res): - # lookupAddress and lookupIP6Address return a three-tuple - # giving the answer, authority, and additional sections of the - # response. - # - # we only care about the answers. - - return res[0] - - def eb(res, record_type): - if res.check(DNSNameError): - return [] - logger.warn("Error looking up %s for %s: %s", record_type, host, res) - return res - - # no logcontexts here, so we can safely fire these off and gatherResults - d1 = dns_client.lookupAddress(host).addCallbacks( - cb, eb, errbackArgs=("A", )) - d2 = dns_client.lookupIPV6Address(host).addCallbacks( - cb, eb, errbackArgs=("AAAA", )) - results = yield defer.DeferredList( - [d1, d2], consumeErrors=True) - - # if all of the lookups failed, raise an exception rather than blowing out - # the cache with an empty result. - if results and all(s == defer.FAILURE for (s, _) in results): - defer.returnValue(results[0][1]) - - for (success, result) in results: - if success == defer.FAILURE: - continue - - for answer in result: - if not answer.payload: - continue - - try: - if answer.type == dns.A: - ip = answer.payload.dottedQuad() - ip4_servers.append((ip, answer.ttl)) - elif answer.type == dns.AAAA: - ip = socket.inet_ntop( - socket.AF_INET6, answer.payload.address, - ) - ip6_servers.append((ip, answer.ttl)) - else: - # the most likely candidate here is a CNAME record. - # rfc2782 says srvs may not point to aliases. - logger.warn( - "Ignoring unexpected DNS record type %s for %s", - answer.type, host, - ) - continue - except Exception as e: - logger.warn("Ignoring invalid DNS response for %s: %s", - host, e) - continue - - # keep the ipv4 results before the ipv6 results, mostly to match historical - # behaviour. - defer.returnValue(ip4_servers + ip6_servers) diff --git a/synapse/http/server.py b/synapse/http/server.py index 64e083ebfc..8d632290de 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -329,7 +329,7 @@ class JsonResource(HttpServer, resource.Resource): register_paths, so will return (possibly via Deferred) either None, or a tuple of (http code, response body). """ - if request.method == "OPTIONS": + if request.method == b"OPTIONS": return _options_handler, {} # Loop through all the registered callbacks to check if the method @@ -543,7 +543,7 @@ def finish_request(request): def _request_user_agent_is_curl(request): user_agents = request.requestHeaders.getRawHeaders( - "User-Agent", default=[] + b"User-Agent", default=[] ) for user_agent in user_agents: if "curl" in user_agent: diff --git a/synapse/http/site.py b/synapse/http/site.py index e422c8dfae..c8b46e1af2 100644 --- a/synapse/http/site.py +++ b/synapse/http/site.py @@ -20,7 +20,7 @@ import logging import re import time -ACCESS_TOKEN_RE = re.compile(r'(\?.*access(_|%5[Ff])token=)[^&]*(.*)$') +ACCESS_TOKEN_RE = re.compile(br'(\?.*access(_|%5[Ff])token=)[^&]*(.*)$') class SynapseRequest(Request): @@ -43,12 +43,12 @@ class SynapseRequest(Request): def get_redacted_uri(self): return ACCESS_TOKEN_RE.sub( - r'\1<redacted>\3', + br'\1<redacted>\3', self.uri ) def get_user_agent(self): - return self.requestHeaders.getRawHeaders("User-Agent", [None])[-1] + return self.requestHeaders.getRawHeaders(b"User-Agent", [None])[-1] def started_processing(self): self.site.access_logger.info( diff --git a/synapse/metrics/__init__.py b/synapse/metrics/__init__.py index 50d99d7a5c..e3b831db67 100644 --- a/synapse/metrics/__init__.py +++ b/synapse/metrics/__init__.py @@ -17,12 +17,13 @@ import logging import functools import time import gc +import platform from twisted.internet import reactor from .metric import ( CounterMetric, CallbackMetric, DistributionMetric, CacheMetric, - MemoryUsageMetric, + MemoryUsageMetric, GaugeMetric, ) from .process_collector import register_process_collector @@ -30,6 +31,7 @@ from .process_collector import register_process_collector logger = logging.getLogger(__name__) +running_on_pypy = platform.python_implementation() == 'PyPy' all_metrics = [] all_collectors = [] @@ -63,6 +65,13 @@ class Metrics(object): """ return self._register(CounterMetric, *args, **kwargs) + def register_gauge(self, *args, **kwargs): + """ + Returns: + GaugeMetric + """ + return self._register(GaugeMetric, *args, **kwargs) + def register_callback(self, *args, **kwargs): """ Returns: @@ -142,6 +151,32 @@ reactor_metrics = get_metrics_for("python.twisted.reactor") tick_time = reactor_metrics.register_distribution("tick_time") pending_calls_metric = reactor_metrics.register_distribution("pending_calls") +synapse_metrics = get_metrics_for("synapse") + +# Used to track where various components have processed in the event stream, +# e.g. federation sending, appservice sending, etc. +event_processing_positions = synapse_metrics.register_gauge( + "event_processing_positions", labels=["name"], +) + +# Used to track the current max events stream position +event_persisted_position = synapse_metrics.register_gauge( + "event_persisted_position", +) + +# Used to track the received_ts of the last event processed by various +# components +event_processing_last_ts = synapse_metrics.register_gauge( + "event_processing_last_ts", labels=["name"], +) + +# Used to track the lag processing events. This is the time difference +# between the last processed event's received_ts and the time it was +# finished being processed. +event_processing_lag = synapse_metrics.register_gauge( + "event_processing_lag", labels=["name"], +) + def runUntilCurrentTimer(func): @@ -174,6 +209,9 @@ def runUntilCurrentTimer(func): tick_time.inc_by(end - start) pending_calls_metric.inc_by(num_pending) + if running_on_pypy: + return ret + # Check if we need to do a manual GC (since its been disabled), and do # one if necessary. threshold = gc.get_threshold() @@ -206,6 +244,7 @@ try: # We manually run the GC each reactor tick so that we can get some metrics # about time spent doing GC, - gc.disable() + if not running_on_pypy: + gc.disable() except AttributeError: pass diff --git a/synapse/metrics/metric.py b/synapse/metrics/metric.py index ff5aa8c0e1..89bd47c3f7 100644 --- a/synapse/metrics/metric.py +++ b/synapse/metrics/metric.py @@ -115,7 +115,7 @@ class CounterMetric(BaseMetric): # dict[list[str]]: value for each set of label values. the keys are the # label values, in the same order as the labels in self.labels. # - # (if the metric is a scalar, the (single) key is the empty list). + # (if the metric is a scalar, the (single) key is the empty tuple). self.counts = {} # Scalar metrics are never empty @@ -145,6 +145,36 @@ class CounterMetric(BaseMetric): ) +class GaugeMetric(BaseMetric): + """A metric that can go up or down + """ + + def __init__(self, *args, **kwargs): + super(GaugeMetric, self).__init__(*args, **kwargs) + + # dict[list[str]]: value for each set of label values. the keys are the + # label values, in the same order as the labels in self.labels. + # + # (if the metric is a scalar, the (single) key is the empty tuple). + self.guages = {} + + def set(self, v, *values): + if len(values) != self.dimension(): + raise ValueError( + "Expected as many values to inc() as labels (%d)" % (self.dimension()) + ) + + # TODO: should assert that the tag values are all strings + + self.guages[values] = v + + def render(self): + return flatten( + self._render_for_labels(k, self.guages[k]) + for k in sorted(self.guages.keys()) + ) + + class CallbackMetric(BaseMetric): """A metric that returns the numeric value returned by a callback whenever it is rendered. Typically this is used to implement gauges that yield the diff --git a/synapse/replication/http/send_event.py b/synapse/replication/http/send_event.py index bbe2f967b7..c6a6551d24 100644 --- a/synapse/replication/http/send_event.py +++ b/synapse/replication/http/send_event.py @@ -115,7 +115,7 @@ class ReplicationSendEventRestServlet(RestServlet): self.clock = hs.get_clock() # The responses are tiny, so we may as well cache them for a while - self.response_cache = ResponseCache(hs, timeout_ms=30 * 60 * 1000) + self.response_cache = ResponseCache(hs, "send_event", timeout_ms=30 * 60 * 1000) def on_PUT(self, request, event_id): result = self.response_cache.get(event_id) diff --git a/synapse/rest/client/v1/logout.py b/synapse/rest/client/v1/logout.py index ca49955935..e092158cb7 100644 --- a/synapse/rest/client/v1/logout.py +++ b/synapse/rest/client/v1/logout.py @@ -44,7 +44,10 @@ class LogoutRestServlet(ClientV1RestServlet): requester = yield self.auth.get_user_by_req(request) except AuthError: # this implies the access token has already been deleted. - pass + defer.returnValue((401, { + "errcode": "M_UNKNOWN_TOKEN", + "error": "Access Token unknown or expired" + })) else: if requester.device_id is None: # the acccess token wasn't associated with a device. diff --git a/synapse/rest/client/v1/register.py b/synapse/rest/client/v1/register.py index 5c5fa8f7ab..8a82097178 100644 --- a/synapse/rest/client/v1/register.py +++ b/synapse/rest/client/v1/register.py @@ -348,9 +348,9 @@ class RegisterRestServlet(ClientV1RestServlet): admin = register_json.get("admin", None) # Its important to check as we use null bytes as HMAC field separators - if "\x00" in user: + if b"\x00" in user: raise SynapseError(400, "Invalid user") - if "\x00" in password: + if b"\x00" in password: raise SynapseError(400, "Invalid password") # str() because otherwise hmac complains that 'unicode' does not diff --git a/synapse/rest/client/v1/room.py b/synapse/rest/client/v1/room.py index d06cbdc35e..2ad0e5943b 100644 --- a/synapse/rest/client/v1/room.py +++ b/synapse/rest/client/v1/room.py @@ -165,17 +165,12 @@ class RoomStateEventRestServlet(ClientV1RestServlet): content=content, ) else: - event, context = yield self.event_creation_hander.create_event( + event = yield self.event_creation_hander.create_and_send_nonmember_event( requester, event_dict, - token_id=requester.access_token_id, txn_id=txn_id, ) - yield self.event_creation_hander.send_nonmember_event( - requester, event, context, - ) - ret = {} if event: ret = {"event_id": event.event_id} diff --git a/synapse/rest/client/v2_alpha/register.py b/synapse/rest/client/v2_alpha/register.py index 0ba62bddc1..f317c919dc 100644 --- a/synapse/rest/client/v2_alpha/register.py +++ b/synapse/rest/client/v2_alpha/register.py @@ -20,7 +20,6 @@ import synapse import synapse.types from synapse.api.auth import get_access_token_from_request, has_access_token from synapse.api.constants import LoginType -from synapse.types import RoomID, RoomAlias from synapse.api.errors import SynapseError, Codes, UnrecognizedRequestError from synapse.http.servlet import ( RestServlet, parse_json_object_from_request, assert_params_in_request, parse_string @@ -405,14 +404,6 @@ class RegisterRestServlet(RestServlet): generate_token=False, ) - # auto-join the user to any rooms we're supposed to dump them into - fake_requester = synapse.types.create_requester(registered_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) - # remember that we've now registered that user account, and with # what user ID (since the user may not have specified) self.auth_handler.set_session_data( @@ -446,29 +437,6 @@ class RegisterRestServlet(RestServlet): return 200, {} @defer.inlineCallbacks - def _join_user_to_room(self, requester, room_identifier): - room_id = None - 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 self.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 self.room_member_handler.update_membership( - requester=requester, - target=requester.user, - room_id=room_id, - action="join", - ) - - @defer.inlineCallbacks def _do_appservice_registration(self, username, as_token, body): user_id = yield self.registration_handler.appservice_register( username, as_token diff --git a/synapse/rest/media/v1/media_storage.py b/synapse/rest/media/v1/media_storage.py index 83471b3173..7f263db239 100644 --- a/synapse/rest/media/v1/media_storage.py +++ b/synapse/rest/media/v1/media_storage.py @@ -16,6 +16,8 @@ from twisted.internet import defer, threads from twisted.protocols.basic import FileSender +import six + from ._base import Responder from synapse.util.file_consumer import BackgroundFileConsumer @@ -119,7 +121,7 @@ class MediaStorage(object): os.remove(fname) except Exception: pass - raise t, v, tb + six.reraise(t, v, tb) if not finished_called: raise Exception("Finished callback not called") diff --git a/synapse/storage/__init__.py b/synapse/storage/__init__.py index eacd49d6a5..8cdfd50f90 100644 --- a/synapse/storage/__init__.py +++ b/synapse/storage/__init__.py @@ -266,9 +266,9 @@ class DataStore(RoomMemberStore, RoomStore, def count_r30_users(self): """ Counts the number of 30 day retained users, defined as:- - * Users who have created their accounts more than 30 days + * Users who have created their accounts more than 30 days ago * Where last seen at most 30 days ago - * Where account creation and last_seen are > 30 days + * Where account creation and last_seen are > 30 days apart Returns counts globaly for a given user as well as breaking by platform diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py index 2fbebd4907..2262776ab2 100644 --- a/synapse/storage/_base.py +++ b/synapse/storage/_base.py @@ -376,7 +376,7 @@ class SQLBaseStore(object): Returns: A list of dicts where the key is the column header. """ - col_headers = list(intern(column[0]) for column in cursor.description) + col_headers = list(intern(str(column[0])) for column in cursor.description) results = list( dict(zip(col_headers, row)) for row in cursor ) diff --git a/synapse/storage/engines/__init__.py b/synapse/storage/engines/__init__.py index 338b495611..8c868ece75 100644 --- a/synapse/storage/engines/__init__.py +++ b/synapse/storage/engines/__init__.py @@ -18,6 +18,7 @@ from .postgres import PostgresEngine from .sqlite3 import Sqlite3Engine import importlib +import platform SUPPORTED_MODULE = { @@ -31,6 +32,10 @@ def create_engine(database_config): engine_class = SUPPORTED_MODULE.get(name, None) if engine_class: + # pypy requires psycopg2cffi rather than psycopg2 + if (name == "psycopg2" and + platform.python_implementation() == "PyPy"): + name = "psycopg2cffi" module = importlib.import_module(name) return engine_class(module, database_config) diff --git a/synapse/storage/events.py b/synapse/storage/events.py index ece5e6c41f..da44b52fd6 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -444,6 +444,9 @@ class EventsStore(EventsWorkerStore): new_forward_extremeties=new_forward_extremeties, ) persist_event_counter.inc_by(len(chunk)) + synapse.metrics.event_persisted_position.set( + chunk[-1][0].internal_metadata.stream_ordering, + ) for event, context in chunk: if context.app_service: origin_type = "local" diff --git a/synapse/storage/events_worker.py b/synapse/storage/events_worker.py index 2e23dd78ba..a937b9bceb 100644 --- a/synapse/storage/events_worker.py +++ b/synapse/storage/events_worker.py @@ -51,6 +51,26 @@ _EventCacheEntry = namedtuple("_EventCacheEntry", ("event", "redacted_event")) class EventsWorkerStore(SQLBaseStore): + def get_received_ts(self, event_id): + """Get received_ts (when it was persisted) for the event. + + Raises an exception for unknown events. + + Args: + event_id (str) + + Returns: + Deferred[int|None]: Timestamp in milliseconds, or None for events + that were persisted before received_ts was implemented. + """ + return self._simple_select_one_onecol( + table="events", + keyvalues={ + "event_id": event_id, + }, + retcol="received_ts", + desc="get_received_ts", + ) @defer.inlineCallbacks def get_event(self, event_id, check_redacted=True, diff --git a/synapse/types.py b/synapse/types.py index 7cb24cecb2..cc7c182a78 100644 --- a/synapse/types.py +++ b/synapse/types.py @@ -169,7 +169,7 @@ class DomainSpecificString( except Exception: return False - __str__ = to_string + __repr__ = to_string class UserID(DomainSpecificString): diff --git a/synapse/util/caches/response_cache.py b/synapse/util/caches/response_cache.py index 00af539880..066fa423fd 100644 --- a/synapse/util/caches/response_cache.py +++ b/synapse/util/caches/response_cache.py @@ -14,6 +14,7 @@ # limitations under the License. from synapse.util.async import ObservableDeferred +from synapse.util.caches import metrics as cache_metrics class ResponseCache(object): @@ -24,20 +25,63 @@ class ResponseCache(object): used rather than trying to compute a new response. """ - def __init__(self, hs, timeout_ms=0): + def __init__(self, hs, name, timeout_ms=0): self.pending_result_cache = {} # Requests that haven't finished yet. self.clock = hs.get_clock() self.timeout_sec = timeout_ms / 1000. + self._metrics = cache_metrics.register_cache( + "response_cache", + size_callback=lambda: self.size(), + cache_name=name, + ) + + def size(self): + return len(self.pending_result_cache) + def get(self, key): + """Look up the given key. + + Returns a deferred which doesn't follow the synapse logcontext rules, + so you'll probably want to make_deferred_yieldable it. + + Args: + key (str): + + Returns: + twisted.internet.defer.Deferred|None: None if there is no entry + for this key; otherwise a deferred result. + """ result = self.pending_result_cache.get(key) if result is not None: + self._metrics.inc_hits() return result.observe() else: + self._metrics.inc_misses() return None def set(self, key, deferred): + """Set the entry for the given key to the given deferred. + + *deferred* should run its callbacks in the sentinel logcontext (ie, + you should wrap normal synapse deferreds with + logcontext.run_in_background). + + Returns a new Deferred which also doesn't follow the synapse logcontext + rules, so you will want to make_deferred_yieldable it + + (TODO: before using this more widely, it might make sense to refactor + it and get() so that they do the necessary wrapping rather than having + to do it everywhere ResponseCache is used.) + + Args: + key (str): + deferred (twisted.internet.defer.Deferred): + + Returns: + twisted.internet.defer.Deferred + """ result = ObservableDeferred(deferred, consumeErrors=True) self.pending_result_cache[key] = result |