diff options
Diffstat (limited to 'synapse')
83 files changed, 2247 insertions, 1088 deletions
diff --git a/synapse/api/constants.py b/synapse/api/constants.py index 46013cde15..592abd844b 100644 --- a/synapse/api/constants.py +++ b/synapse/api/constants.py @@ -155,3 +155,8 @@ class EventContentFields: class RoomEncryptionAlgorithms: MEGOLM_V1_AES_SHA2 = "m.megolm.v1.aes-sha2" DEFAULT = MEGOLM_V1_AES_SHA2 + + +class AccountDataTypes: + DIRECT = "m.direct" + IGNORED_USER_LIST = "m.ignored_user_list" diff --git a/synapse/app/_base.py b/synapse/app/_base.py index fb476ddaf5..f6f7b2bf42 100644 --- a/synapse/app/_base.py +++ b/synapse/app/_base.py @@ -28,6 +28,7 @@ from twisted.protocols.tls import TLSMemoryBIOFactory import synapse from synapse.app import check_bind_error +from synapse.app.phone_stats_home import start_phone_stats_home from synapse.config.server import ListenerConfig from synapse.crypto import context_factory from synapse.logging.context import PreserveLoggingContext @@ -271,9 +272,19 @@ def start(hs: "synapse.server.HomeServer", listeners: Iterable[ListenerConfig]): hs.get_datastore().db_pool.start_profiling() hs.get_pusherpool().start() + # Log when we start the shut down process. + hs.get_reactor().addSystemEventTrigger( + "before", "shutdown", logger.info, "Shutting down..." + ) + setup_sentry(hs) setup_sdnotify(hs) + # If background tasks are running on the main process, start collecting the + # phone home stats. + if hs.config.run_background_tasks: + start_phone_stats_home(hs) + # We now freeze all allocated objects in the hopes that (almost) # everything currently allocated are things that will be used for the # rest of time. Doing so means less work each GC (hopefully). diff --git a/synapse/app/admin_cmd.py b/synapse/app/admin_cmd.py index 7d309b1bb0..f0d65d08d7 100644 --- a/synapse/app/admin_cmd.py +++ b/synapse/app/admin_cmd.py @@ -208,6 +208,7 @@ def start(config_options): # Explicitly disable background processes config.update_user_directory = False + config.run_background_tasks = False config.start_pushers = False config.send_federation = False diff --git a/synapse/app/generic_worker.py b/synapse/app/generic_worker.py index c38413c893..d53181deb1 100644 --- a/synapse/app/generic_worker.py +++ b/synapse/app/generic_worker.py @@ -127,12 +127,16 @@ from synapse.rest.health import HealthResource from synapse.rest.key.v2 import KeyApiV2Resource from synapse.server import HomeServer, cache_in_self from synapse.storage.databases.main.censor_events import CensorEventsStore +from synapse.storage.databases.main.client_ips import ClientIpWorkerStore from synapse.storage.databases.main.media_repository import MediaRepositoryStore +from synapse.storage.databases.main.metrics import ServerMetricsStore from synapse.storage.databases.main.monthly_active_users import ( MonthlyActiveUsersWorkerStore, ) from synapse.storage.databases.main.presence import UserPresenceState from synapse.storage.databases.main.search import SearchWorkerStore +from synapse.storage.databases.main.stats import StatsStore +from synapse.storage.databases.main.transactions import TransactionWorkerStore from synapse.storage.databases.main.ui_auth import UIAuthWorkerStore from synapse.storage.databases.main.user_directory import UserDirectoryStore from synapse.types import ReadReceipt @@ -454,6 +458,7 @@ class GenericWorkerSlavedStore( # FIXME(#3714): We need to add UserDirectoryStore as we write directly # rather than going via the correct worker. UserDirectoryStore, + StatsStore, UIAuthWorkerStore, SlavedDeviceInboxStore, SlavedDeviceStore, @@ -463,6 +468,7 @@ class GenericWorkerSlavedStore( SlavedAccountDataStore, SlavedPusherStore, CensorEventsStore, + ClientIpWorkerStore, SlavedEventStore, SlavedKeyStore, RoomStore, @@ -476,7 +482,9 @@ class GenericWorkerSlavedStore( SlavedFilteringStore, MonthlyActiveUsersWorkerStore, MediaRepositoryStore, + ServerMetricsStore, SearchWorkerStore, + TransactionWorkerStore, BaseSlavedStore, ): pass diff --git a/synapse/app/homeserver.py b/synapse/app/homeserver.py index dff739e106..2b5465417f 100644 --- a/synapse/app/homeserver.py +++ b/synapse/app/homeserver.py @@ -17,14 +17,10 @@ import gc import logging -import math import os -import resource import sys from typing import Iterable -from prometheus_client import Gauge - from twisted.application import service from twisted.internet import defer, reactor from twisted.python.failure import Failure @@ -60,8 +56,6 @@ from synapse.http.server import ( from synapse.http.site import SynapseSite from synapse.logging.context import LoggingContext from synapse.metrics import METRICS_PREFIX, MetricsResource, RegistryProxy -from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.module_api import ModuleApi from synapse.python_dependencies import check_requirements from synapse.replication.http import REPLICATION_PREFIX, ReplicationRestResource from synapse.replication.tcp.resource import ReplicationStreamProtocolFactory @@ -111,7 +105,7 @@ class SynapseHomeServer(HomeServer): additional_resources = listener_config.http_options.additional_resources logger.debug("Configuring additional resources: %r", additional_resources) - module_api = ModuleApi(self, self.get_auth_handler()) + module_api = self.get_module_api() for path, resmodule in additional_resources.items(): handler_cls, config = load_module(resmodule) handler = handler_cls(config, module_api) @@ -334,20 +328,6 @@ class SynapseHomeServer(HomeServer): logger.warning("Unrecognized listener type: %s", listener.type) -# Gauges to expose monthly active user control metrics -current_mau_gauge = Gauge("synapse_admin_mau:current", "Current MAU") -current_mau_by_service_gauge = Gauge( - "synapse_admin_mau_current_mau_by_service", - "Current MAU by service", - ["app_service"], -) -max_mau_gauge = Gauge("synapse_admin_mau:max", "MAU Limit") -registered_reserved_users_mau_gauge = Gauge( - "synapse_admin_mau:registered_reserved_users", - "Registered users with reserved threepids", -) - - def setup(config_options): """ Args: @@ -389,8 +369,6 @@ def setup(config_options): except UpgradeDatabaseException as e: quit_with_error("Failed to upgrade database: %s" % (e,)) - hs.setup_master() - async def do_acme() -> bool: """ Reprovision an ACME certificate, if it's required. @@ -486,92 +464,6 @@ class SynapseService(service.Service): return self._port.stopListening() -# Contains the list of processes we will be monitoring -# currently either 0 or 1 -_stats_process = [] - - -async def phone_stats_home(hs, stats, stats_process=_stats_process): - logger.info("Gathering stats for reporting") - now = int(hs.get_clock().time()) - uptime = int(now - hs.start_time) - if uptime < 0: - uptime = 0 - - # - # Performance statistics. Keep this early in the function to maintain reliability of `test_performance_100` test. - # - old = stats_process[0] - new = (now, resource.getrusage(resource.RUSAGE_SELF)) - stats_process[0] = new - - # Get RSS in bytes - stats["memory_rss"] = new[1].ru_maxrss - - # Get CPU time in % of a single core, not % of all cores - used_cpu_time = (new[1].ru_utime + new[1].ru_stime) - ( - old[1].ru_utime + old[1].ru_stime - ) - if used_cpu_time == 0 or new[0] == old[0]: - stats["cpu_average"] = 0 - else: - stats["cpu_average"] = math.floor(used_cpu_time / (new[0] - old[0]) * 100) - - # - # General statistics - # - - stats["homeserver"] = hs.config.server_name - stats["server_context"] = hs.config.server_context - stats["timestamp"] = now - stats["uptime_seconds"] = uptime - version = sys.version_info - stats["python_version"] = "{}.{}.{}".format( - version.major, version.minor, version.micro - ) - stats["total_users"] = await hs.get_datastore().count_all_users() - - total_nonbridged_users = await hs.get_datastore().count_nonbridged_users() - stats["total_nonbridged_users"] = total_nonbridged_users - - daily_user_type_results = await hs.get_datastore().count_daily_user_type() - for name, count in daily_user_type_results.items(): - stats["daily_user_type_" + name] = count - - room_count = await hs.get_datastore().get_room_count() - stats["total_room_count"] = room_count - - stats["daily_active_users"] = await hs.get_datastore().count_daily_users() - stats["monthly_active_users"] = await hs.get_datastore().count_monthly_users() - stats["daily_active_rooms"] = await hs.get_datastore().count_daily_active_rooms() - stats["daily_messages"] = await hs.get_datastore().count_daily_messages() - - r30_results = await hs.get_datastore().count_r30_users() - for name, count in r30_results.items(): - stats["r30_users_" + name] = count - - daily_sent_messages = await hs.get_datastore().count_daily_sent_messages() - stats["daily_sent_messages"] = daily_sent_messages - stats["cache_factor"] = hs.config.caches.global_factor - stats["event_cache_size"] = hs.config.caches.event_cache_size - - # - # Database version - # - - # This only reports info about the *main* database. - stats["database_engine"] = hs.get_datastore().db_pool.engine.module.__name__ - stats["database_server_version"] = hs.get_datastore().db_pool.engine.server_version - - logger.info("Reporting stats to %s: %s" % (hs.config.report_stats_endpoint, stats)) - try: - await hs.get_proxied_http_client().put_json( - hs.config.report_stats_endpoint, stats - ) - except Exception as e: - logger.warning("Error reporting stats: %s", e) - - def run(hs): PROFILE_SYNAPSE = False if PROFILE_SYNAPSE: @@ -597,81 +489,6 @@ def run(hs): ThreadPool._worker = profile(ThreadPool._worker) reactor.run = profile(reactor.run) - clock = hs.get_clock() - - stats = {} - - def performance_stats_init(): - _stats_process.clear() - _stats_process.append( - (int(hs.get_clock().time()), resource.getrusage(resource.RUSAGE_SELF)) - ) - - def start_phone_stats_home(): - return run_as_background_process( - "phone_stats_home", phone_stats_home, hs, stats - ) - - def generate_user_daily_visit_stats(): - return run_as_background_process( - "generate_user_daily_visits", hs.get_datastore().generate_user_daily_visits - ) - - # Rather than update on per session basis, batch up the requests. - # If you increase the loop period, the accuracy of user_daily_visits - # table will decrease - clock.looping_call(generate_user_daily_visit_stats, 5 * 60 * 1000) - - # monthly active user limiting functionality - def reap_monthly_active_users(): - return run_as_background_process( - "reap_monthly_active_users", hs.get_datastore().reap_monthly_active_users - ) - - clock.looping_call(reap_monthly_active_users, 1000 * 60 * 60) - reap_monthly_active_users() - - async def generate_monthly_active_users(): - current_mau_count = 0 - current_mau_count_by_service = {} - reserved_users = () - store = hs.get_datastore() - if hs.config.limit_usage_by_mau or hs.config.mau_stats_only: - current_mau_count = await store.get_monthly_active_count() - current_mau_count_by_service = ( - await store.get_monthly_active_count_by_service() - ) - reserved_users = await store.get_registered_reserved_users() - current_mau_gauge.set(float(current_mau_count)) - - for app_service, count in current_mau_count_by_service.items(): - current_mau_by_service_gauge.labels(app_service).set(float(count)) - - registered_reserved_users_mau_gauge.set(float(len(reserved_users))) - max_mau_gauge.set(float(hs.config.max_mau_value)) - - def start_generate_monthly_active_users(): - return run_as_background_process( - "generate_monthly_active_users", generate_monthly_active_users - ) - - start_generate_monthly_active_users() - if hs.config.limit_usage_by_mau or hs.config.mau_stats_only: - clock.looping_call(start_generate_monthly_active_users, 5 * 60 * 1000) - # End of monthly active user settings - - if hs.config.report_stats: - logger.info("Scheduling stats reporting for 3 hour intervals") - clock.looping_call(start_phone_stats_home, 3 * 60 * 60 * 1000) - - # We need to defer this init for the cases that we daemonize - # otherwise the process ID we get is that of the non-daemon process - clock.call_later(0, performance_stats_init) - - # We wait 5 minutes to send the first set of stats as the server can - # be quite busy the first few minutes - clock.call_later(5 * 60, start_phone_stats_home) - _base.start_reactor( "synapse-homeserver", soft_file_limit=hs.config.soft_file_limit, diff --git a/synapse/app/phone_stats_home.py b/synapse/app/phone_stats_home.py new file mode 100644 index 0000000000..8a69104a04 --- /dev/null +++ b/synapse/app/phone_stats_home.py @@ -0,0 +1,198 @@ +# Copyright 2020 The Matrix.org Foundation C.I.C. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +import logging +import math +import resource +import sys + +from prometheus_client import Gauge + +from synapse.metrics.background_process_metrics import ( + run_as_background_process, + wrap_as_background_process, +) + +logger = logging.getLogger("synapse.app.homeserver") + +# Contains the list of processes we will be monitoring +# currently either 0 or 1 +_stats_process = [] + +# Gauges to expose monthly active user control metrics +current_mau_gauge = Gauge("synapse_admin_mau:current", "Current MAU") +current_mau_by_service_gauge = Gauge( + "synapse_admin_mau_current_mau_by_service", + "Current MAU by service", + ["app_service"], +) +max_mau_gauge = Gauge("synapse_admin_mau:max", "MAU Limit") +registered_reserved_users_mau_gauge = Gauge( + "synapse_admin_mau:registered_reserved_users", + "Registered users with reserved threepids", +) + + +@wrap_as_background_process("phone_stats_home") +async def phone_stats_home(hs, stats, stats_process=_stats_process): + logger.info("Gathering stats for reporting") + now = int(hs.get_clock().time()) + uptime = int(now - hs.start_time) + if uptime < 0: + uptime = 0 + + # + # Performance statistics. Keep this early in the function to maintain reliability of `test_performance_100` test. + # + old = stats_process[0] + new = (now, resource.getrusage(resource.RUSAGE_SELF)) + stats_process[0] = new + + # Get RSS in bytes + stats["memory_rss"] = new[1].ru_maxrss + + # Get CPU time in % of a single core, not % of all cores + used_cpu_time = (new[1].ru_utime + new[1].ru_stime) - ( + old[1].ru_utime + old[1].ru_stime + ) + if used_cpu_time == 0 or new[0] == old[0]: + stats["cpu_average"] = 0 + else: + stats["cpu_average"] = math.floor(used_cpu_time / (new[0] - old[0]) * 100) + + # + # General statistics + # + + stats["homeserver"] = hs.config.server_name + stats["server_context"] = hs.config.server_context + stats["timestamp"] = now + stats["uptime_seconds"] = uptime + version = sys.version_info + stats["python_version"] = "{}.{}.{}".format( + version.major, version.minor, version.micro + ) + stats["total_users"] = await hs.get_datastore().count_all_users() + + total_nonbridged_users = await hs.get_datastore().count_nonbridged_users() + stats["total_nonbridged_users"] = total_nonbridged_users + + daily_user_type_results = await hs.get_datastore().count_daily_user_type() + for name, count in daily_user_type_results.items(): + stats["daily_user_type_" + name] = count + + room_count = await hs.get_datastore().get_room_count() + stats["total_room_count"] = room_count + + stats["daily_active_users"] = await hs.get_datastore().count_daily_users() + stats["monthly_active_users"] = await hs.get_datastore().count_monthly_users() + stats["daily_active_rooms"] = await hs.get_datastore().count_daily_active_rooms() + stats["daily_messages"] = await hs.get_datastore().count_daily_messages() + + r30_results = await hs.get_datastore().count_r30_users() + for name, count in r30_results.items(): + stats["r30_users_" + name] = count + + daily_sent_messages = await hs.get_datastore().count_daily_sent_messages() + stats["daily_sent_messages"] = daily_sent_messages + stats["cache_factor"] = hs.config.caches.global_factor + stats["event_cache_size"] = hs.config.caches.event_cache_size + + # + # Database version + # + + # This only reports info about the *main* database. + stats["database_engine"] = hs.get_datastore().db_pool.engine.module.__name__ + stats["database_server_version"] = hs.get_datastore().db_pool.engine.server_version + + # + # Logging configuration + # + synapse_logger = logging.getLogger("synapse") + log_level = synapse_logger.getEffectiveLevel() + stats["log_level"] = logging.getLevelName(log_level) + + logger.info("Reporting stats to %s: %s" % (hs.config.report_stats_endpoint, stats)) + try: + await hs.get_proxied_http_client().put_json( + hs.config.report_stats_endpoint, stats + ) + except Exception as e: + logger.warning("Error reporting stats: %s", e) + + +def start_phone_stats_home(hs): + """ + Start the background tasks which report phone home stats. + """ + clock = hs.get_clock() + + stats = {} + + def performance_stats_init(): + _stats_process.clear() + _stats_process.append( + (int(hs.get_clock().time()), resource.getrusage(resource.RUSAGE_SELF)) + ) + + # Rather than update on per session basis, batch up the requests. + # If you increase the loop period, the accuracy of user_daily_visits + # table will decrease + clock.looping_call(hs.get_datastore().generate_user_daily_visits, 5 * 60 * 1000) + + # monthly active user limiting functionality + def reap_monthly_active_users(): + return run_as_background_process( + "reap_monthly_active_users", hs.get_datastore().reap_monthly_active_users + ) + + clock.looping_call(reap_monthly_active_users, 1000 * 60 * 60) + reap_monthly_active_users() + + @wrap_as_background_process("generate_monthly_active_users") + async def generate_monthly_active_users(): + current_mau_count = 0 + current_mau_count_by_service = {} + reserved_users = () + store = hs.get_datastore() + if hs.config.limit_usage_by_mau or hs.config.mau_stats_only: + current_mau_count = await store.get_monthly_active_count() + current_mau_count_by_service = ( + await store.get_monthly_active_count_by_service() + ) + reserved_users = await store.get_registered_reserved_users() + current_mau_gauge.set(float(current_mau_count)) + + for app_service, count in current_mau_count_by_service.items(): + current_mau_by_service_gauge.labels(app_service).set(float(count)) + + registered_reserved_users_mau_gauge.set(float(len(reserved_users))) + max_mau_gauge.set(float(hs.config.max_mau_value)) + + if hs.config.limit_usage_by_mau or hs.config.mau_stats_only: + generate_monthly_active_users() + clock.looping_call(generate_monthly_active_users, 5 * 60 * 1000) + # End of monthly active user settings + + if hs.config.report_stats: + logger.info("Scheduling stats reporting for 3 hour intervals") + clock.looping_call(phone_stats_home, 3 * 60 * 60 * 1000, hs, stats) + + # We need to defer this init for the cases that we daemonize + # otherwise the process ID we get is that of the non-daemon process + clock.call_later(0, performance_stats_init) + + # We wait 5 minutes to send the first set of stats as the server can + # be quite busy the first few minutes + clock.call_later(5 * 60, phone_stats_home, hs, stats) diff --git a/synapse/config/oidc_config.py b/synapse/config/oidc_config.py index f924116819..7597fbc864 100644 --- a/synapse/config/oidc_config.py +++ b/synapse/config/oidc_config.py @@ -56,6 +56,7 @@ class OIDCConfig(Config): self.oidc_userinfo_endpoint = oidc_config.get("userinfo_endpoint") self.oidc_jwks_uri = oidc_config.get("jwks_uri") self.oidc_skip_verification = oidc_config.get("skip_verification", False) + self.oidc_user_profile_method = oidc_config.get("user_profile_method", "auto") self.oidc_allow_existing_users = oidc_config.get("allow_existing_users", False) ump_config = oidc_config.get("user_mapping_provider", {}) @@ -159,6 +160,14 @@ class OIDCConfig(Config): # #skip_verification: true + # Whether to fetch the user profile from the userinfo endpoint. Valid + # values are: "auto" or "userinfo_endpoint". + # + # Defaults to "auto", which fetches the userinfo endpoint if "openid" is included + # in `scopes`. Uncomment the following to always fetch the userinfo endpoint. + # + #user_profile_method: "userinfo_endpoint" + # Uncomment to allow a user logging in via OIDC to match a pre-existing account instead # of failing. This could be used if switching from password logins to OIDC. Defaults to false. # diff --git a/synapse/config/server.py b/synapse/config/server.py index ef6d70e3f8..85aa49c02d 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -39,7 +39,7 @@ logger = logging.Logger(__name__) # in the list. DEFAULT_BIND_ADDRESSES = ["::", "0.0.0.0"] -DEFAULT_ROOM_VERSION = "5" +DEFAULT_ROOM_VERSION = "6" ROOM_COMPLEXITY_TOO_GREAT = ( "Your homeserver is unable to join rooms this large or complex. " diff --git a/synapse/config/tls.py b/synapse/config/tls.py index 9ddb8b546b..ad37b93c02 100644 --- a/synapse/config/tls.py +++ b/synapse/config/tls.py @@ -18,7 +18,7 @@ import os import warnings from datetime import datetime from hashlib import sha256 -from typing import List +from typing import List, Optional from unpaddedbase64 import encode_base64 @@ -177,8 +177,8 @@ class TlsConfig(Config): "use_insecure_ssl_client_just_for_testing_do_not_use" ) - self.tls_certificate = None - self.tls_private_key = None + self.tls_certificate = None # type: Optional[crypto.X509] + self.tls_private_key = None # type: Optional[crypto.PKey] def is_disk_cert_valid(self, allow_self_signed=True): """ @@ -226,12 +226,12 @@ class TlsConfig(Config): days_remaining = (expires_on - now).days return days_remaining - def read_certificate_from_disk(self, require_cert_and_key): + def read_certificate_from_disk(self, require_cert_and_key: bool): """ Read the certificates and private key from disk. Args: - require_cert_and_key (bool): set to True to throw an error if the certificate + require_cert_and_key: set to True to throw an error if the certificate and key file are not given """ if require_cert_and_key: @@ -479,13 +479,13 @@ class TlsConfig(Config): } ) - def read_tls_certificate(self): + def read_tls_certificate(self) -> crypto.X509: """Reads the TLS certificate from the configured file, and returns it Also checks if it is self-signed, and warns if so Returns: - OpenSSL.crypto.X509: the certificate + The certificate """ cert_path = self.tls_certificate_file logger.info("Loading TLS certificate from %s", cert_path) @@ -504,11 +504,11 @@ class TlsConfig(Config): return cert - def read_tls_private_key(self): + def read_tls_private_key(self) -> crypto.PKey: """Reads the TLS private key from the configured file, and returns it Returns: - OpenSSL.crypto.PKey: the private key + The private key """ private_key_path = self.tls_private_key_file logger.info("Loading TLS key from %s", private_key_path) diff --git a/synapse/config/workers.py b/synapse/config/workers.py index f23e42cdf9..57ab097eba 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -132,6 +132,19 @@ class WorkerConfig(Config): self.events_shard_config = ShardedWorkerHandlingConfig(self.writers.events) + # Whether this worker should run background tasks or not. + # + # As a note for developers, the background tasks guarded by this should + # be able to run on only a single instance (meaning that they don't + # depend on any in-memory state of a particular worker). + # + # No effort is made to ensure only a single instance of these tasks is + # running. + background_tasks_instance = config.get("run_background_tasks_on") or "master" + self.run_background_tasks = ( + self.worker_name is None and background_tasks_instance == "master" + ) or self.worker_name == background_tasks_instance + def generate_config_section(self, config_dir_path, server_name, **kwargs): return """\ ## Workers ## @@ -167,6 +180,11 @@ class WorkerConfig(Config): #stream_writers: # events: worker1 # typing: worker1 + + # The worker that is used to run background tasks (e.g. cleaning up expired + # data). If not provided this defaults to the main process. + # + #run_background_tasks_on: worker1 """ def read_arguments(self, args): diff --git a/synapse/event_auth.py b/synapse/event_auth.py index 8c907ad596..56f8dc9caf 100644 --- a/synapse/event_auth.py +++ b/synapse/event_auth.py @@ -446,6 +446,8 @@ def check_redaction( if room_version_obj.event_format == EventFormatVersions.V1: redacter_domain = get_domain_from_id(event.event_id) + if not isinstance(event.redacts, str): + return False redactee_domain = get_domain_from_id(event.redacts) if redacter_domain == redactee_domain: return True diff --git a/synapse/events/__init__.py b/synapse/events/__init__.py index dc49df0812..7a51d0a22f 100644 --- a/synapse/events/__init__.py +++ b/synapse/events/__init__.py @@ -97,13 +97,16 @@ class DefaultDictProperty(DictProperty): class _EventInternalMetadata: - __slots__ = ["_dict"] + __slots__ = ["_dict", "stream_ordering"] def __init__(self, internal_metadata_dict: JsonDict): # we have to copy the dict, because it turns out that the same dict is # reused. TODO: fix that self._dict = dict(internal_metadata_dict) + # the stream ordering of this event. None, until it has been persisted. + self.stream_ordering = None # type: Optional[int] + outlier = DictProperty("outlier") # type: bool out_of_band_membership = DictProperty("out_of_band_membership") # type: bool send_on_behalf_of = DictProperty("send_on_behalf_of") # type: str @@ -113,7 +116,6 @@ class _EventInternalMetadata: redacted = DictProperty("redacted") # type: bool txn_id = DictProperty("txn_id") # type: str token_id = DictProperty("token_id") # type: str - stream_ordering = DictProperty("stream_ordering") # type: int # XXX: These are set by StreamWorkerStore._set_before_and_after. # I'm pretty sure that these are never persisted to the database, so shouldn't diff --git a/synapse/events/spamcheck.py b/synapse/events/spamcheck.py index b0fc859a47..bad18f7fdf 100644 --- a/synapse/events/spamcheck.py +++ b/synapse/events/spamcheck.py @@ -17,24 +17,25 @@ import inspect from typing import Any, Dict, List, Optional, Tuple -from synapse.spam_checker_api import RegistrationBehaviour, SpamCheckerApi +from synapse.spam_checker_api import RegistrationBehaviour from synapse.types import Collection MYPY = False if MYPY: + import synapse.events import synapse.server class SpamChecker: def __init__(self, hs: "synapse.server.HomeServer"): self.spam_checkers = [] # type: List[Any] + api = hs.get_module_api() for module, config in hs.config.spam_checkers: # Older spam checkers don't accept the `api` argument, so we # try and detect support. spam_args = inspect.getfullargspec(module) if "api" in spam_args.args: - api = SpamCheckerApi(hs) self.spam_checkers.append(module(config=config, api=api)) else: self.spam_checkers.append(module(config=config)) diff --git a/synapse/events/third_party_rules.py b/synapse/events/third_party_rules.py index 9d5310851c..1535cc5339 100644 --- a/synapse/events/third_party_rules.py +++ b/synapse/events/third_party_rules.py @@ -12,10 +12,11 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +from typing import Callable from synapse.events import EventBase from synapse.events.snapshot import EventContext -from synapse.types import Requester +from synapse.types import Requester, StateMap class ThirdPartyEventRules: @@ -38,7 +39,7 @@ class ThirdPartyEventRules: if module is not None: self.third_party_rules = module( - config=config, http_client=hs.get_simple_http_client() + config=config, module_api=hs.get_module_api(), ) async def check_event_allowed( @@ -59,12 +60,14 @@ class ThirdPartyEventRules: prev_state_ids = await context.get_prev_state_ids() # Retrieve the state events from the database. - state_events = {} - for key, event_id in prev_state_ids.items(): - state_events[key] = await self.store.get_event(event_id, allow_none=True) + events = await self.store.get_events(prev_state_ids.values()) + state_events = {(ev.type, ev.state_key): ev for ev in events.values()} - ret = await self.third_party_rules.check_event_allowed(event, state_events) - return ret + # The module can modify the event slightly if it wants, but caution should be + # exercised, and it's likely to go very wrong if applied to events received over + # federation. + + return await self.third_party_rules.check_event_allowed(event, state_events) async def on_create_room( self, requester: Requester, config: dict, is_requester_admin: bool @@ -106,6 +109,48 @@ class ThirdPartyEventRules: if self.third_party_rules is None: return True + state_events = await self._get_state_map_for_room(room_id) + + ret = await self.third_party_rules.check_threepid_can_be_invited( + medium, address, state_events + ) + return ret + + async def check_visibility_can_be_modified( + self, room_id: str, new_visibility: str + ) -> bool: + """Check if a room is allowed to be published to, or removed from, the public room + list. + + Args: + room_id: The ID of the room. + new_visibility: The new visibility state. Either "public" or "private". + + Returns: + True if the room's visibility can be modified, False if not. + """ + if self.third_party_rules is None: + return True + + check_func = getattr( + self.third_party_rules, "check_visibility_can_be_modified", None + ) + if not check_func or not isinstance(check_func, Callable): + return True + + state_events = await self._get_state_map_for_room(room_id) + + return await check_func(room_id, state_events, new_visibility) + + async def _get_state_map_for_room(self, room_id: str) -> StateMap[EventBase]: + """Given a room ID, return the state events of that room. + + Args: + room_id: The ID of the room. + + Returns: + A dict mapping (event type, state key) to state event. + """ state_ids = await self.store.get_filtered_current_state_ids(room_id) room_state_events = await self.store.get_events(state_ids.values()) @@ -113,7 +158,4 @@ class ThirdPartyEventRules: for key, event_id in state_ids.items(): state_events[key] = room_state_events[event_id] - ret = await self.third_party_rules.check_threepid_can_be_invited( - medium, address, state_events - ) - return ret + return state_events diff --git a/synapse/events/utils.py b/synapse/events/utils.py index 32c73d3413..355cbe05f1 100644 --- a/synapse/events/utils.py +++ b/synapse/events/utils.py @@ -49,6 +49,11 @@ def prune_event(event: EventBase) -> EventBase: pruned_event_dict, event.room_version, event.internal_metadata.get_dict() ) + # copy the internal fields + pruned_event.internal_metadata.stream_ordering = ( + event.internal_metadata.stream_ordering + ) + # Mark the event as redacted pruned_event.internal_metadata.redacted = True diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index 24329dd0e3..02f11e1209 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -22,7 +22,6 @@ from typing import ( Callable, Dict, List, - Match, Optional, Tuple, Union, @@ -825,14 +824,14 @@ def server_matches_acl_event(server_name: str, acl_event: EventBase) -> bool: return False -def _acl_entry_matches(server_name: str, acl_entry: str) -> Match: +def _acl_entry_matches(server_name: str, acl_entry: Any) -> bool: if not isinstance(acl_entry, str): logger.warning( "Ignoring non-str ACL entry '%s' (is %s)", acl_entry, type(acl_entry) ) return False regex = glob_to_regex(acl_entry) - return regex.match(server_name) + return bool(regex.match(server_name)) class FederationHandlerRegistry: diff --git a/synapse/federation/sender/__init__.py b/synapse/federation/sender/__init__.py index 8bb17b3a05..e33b29a42c 100644 --- a/synapse/federation/sender/__init__.py +++ b/synapse/federation/sender/__init__.py @@ -297,6 +297,8 @@ class FederationSender: sent_pdus_destination_dist_total.inc(len(destinations)) sent_pdus_destination_dist_count.inc() + assert pdu.internal_metadata.stream_ordering + # track the fact that we have a PDU for these destinations, # to allow us to perform catch-up later on if the remote is unreachable # for a while. diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py index bc99af3fdd..db8e456fe8 100644 --- a/synapse/federation/sender/per_destination_queue.py +++ b/synapse/federation/sender/per_destination_queue.py @@ -158,6 +158,7 @@ class PerDestinationQueue: # yet know if we have anything to catch up (None) self._pending_pdus.append(pdu) else: + assert pdu.internal_metadata.stream_ordering self._catchup_last_skipped = pdu.internal_metadata.stream_ordering self.attempt_new_transaction() @@ -361,6 +362,7 @@ class PerDestinationQueue: last_successful_stream_ordering = ( final_pdu.internal_metadata.stream_ordering ) + assert last_successful_stream_ordering await self._store.set_destination_last_successful_stream_ordering( self._destination, last_successful_stream_ordering ) diff --git a/synapse/handlers/auth.py b/synapse/handlers/auth.py index 00eae92052..f6d17c53b1 100644 --- a/synapse/handlers/auth.py +++ b/synapse/handlers/auth.py @@ -164,7 +164,14 @@ class AuthHandler(BaseHandler): self.bcrypt_rounds = hs.config.bcrypt_rounds + # we can't use hs.get_module_api() here, because to do so will create an + # import loop. + # + # TODO: refactor this class to separate the lower-level stuff that + # ModuleApi can use from the higher-level stuff that uses ModuleApi, as + # better way to break the loop account_handler = ModuleApi(hs, self) + self.password_providers = [ module(config=config, account_handler=account_handler) for module, config in hs.config.password_providers @@ -212,7 +219,7 @@ class AuthHandler(BaseHandler): self._clock = self.hs.get_clock() # Expire old UI auth sessions after a period of time. - if hs.config.worker_app is None: + if hs.config.run_background_tasks: self._clock.looping_call( run_as_background_process, 5 * 60 * 1000, diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index b9d9098104..debb1b4f29 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -1,7 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2016 OpenMarket Ltd # Copyright 2019 New Vector Ltd -# Copyright 2019 The Matrix.org Foundation C.I.C. +# Copyright 2019,2020 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -15,7 +15,7 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging -from typing import Any, Dict, List, Optional +from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Optional, Set, Tuple from synapse.api import errors from synapse.api.constants import EventTypes @@ -29,7 +29,10 @@ from synapse.api.errors import ( from synapse.logging.opentracing import log_kv, set_tag, trace from synapse.metrics.background_process_metrics import run_as_background_process from synapse.types import ( + Collection, + JsonDict, StreamToken, + UserID, get_domain_from_id, get_verify_key_from_cross_signing_key, ) @@ -41,13 +44,16 @@ from synapse.util.retryutils import NotRetryingDestination from ._base import BaseHandler +if TYPE_CHECKING: + from synapse.app.homeserver import HomeServer + logger = logging.getLogger(__name__) MAX_DEVICE_DISPLAY_NAME_LEN = 100 class DeviceWorkerHandler(BaseHandler): - def __init__(self, hs): + def __init__(self, hs: "HomeServer"): super().__init__(hs) self.hs = hs @@ -105,7 +111,9 @@ class DeviceWorkerHandler(BaseHandler): @trace @measure_func("device.get_user_ids_changed") - async def get_user_ids_changed(self, user_id: str, from_token: StreamToken): + async def get_user_ids_changed( + self, user_id: str, from_token: StreamToken + ) -> JsonDict: """Get list of users that have had the devices updated, or have newly joined a room, that `user_id` may be interested in. """ @@ -221,8 +229,8 @@ class DeviceWorkerHandler(BaseHandler): possibly_joined = possibly_changed & users_who_share_room possibly_left = (possibly_changed | possibly_left) - users_who_share_room else: - possibly_joined = [] - possibly_left = [] + possibly_joined = set() + possibly_left = set() result = {"changed": list(possibly_joined), "left": list(possibly_left)} @@ -230,7 +238,7 @@ class DeviceWorkerHandler(BaseHandler): return result - async def on_federation_query_user_devices(self, user_id): + async def on_federation_query_user_devices(self, user_id: str) -> JsonDict: stream_id, devices = await self.store.get_e2e_device_keys_for_federation_query( user_id ) @@ -249,7 +257,7 @@ class DeviceWorkerHandler(BaseHandler): class DeviceHandler(DeviceWorkerHandler): - def __init__(self, hs): + def __init__(self, hs: "HomeServer"): super().__init__(hs) self.federation_sender = hs.get_federation_sender() @@ -264,7 +272,7 @@ class DeviceHandler(DeviceWorkerHandler): hs.get_distributor().observe("user_left_room", self.user_left_room) - def _check_device_name_length(self, name: str): + def _check_device_name_length(self, name: Optional[str]): """ Checks whether a device name is longer than the maximum allowed length. @@ -283,8 +291,11 @@ class DeviceHandler(DeviceWorkerHandler): ) async def check_device_registered( - self, user_id, device_id, initial_device_display_name=None - ): + self, + user_id: str, + device_id: Optional[str], + initial_device_display_name: Optional[str] = None, + ) -> str: """ If the given device has not been registered, register it with the supplied display name. @@ -292,12 +303,11 @@ class DeviceHandler(DeviceWorkerHandler): If no device_id is supplied, we make one up. Args: - user_id (str): @user:id - device_id (str | None): device id supplied by client - initial_device_display_name (str | None): device display name from - client + user_id: @user:id + device_id: device id supplied by client + initial_device_display_name: device display name from client Returns: - str: device id (generated if none was supplied) + device id (generated if none was supplied) """ self._check_device_name_length(initial_device_display_name) @@ -316,15 +326,15 @@ class DeviceHandler(DeviceWorkerHandler): # times in case of a clash. attempts = 0 while attempts < 5: - device_id = stringutils.random_string(10).upper() + new_device_id = stringutils.random_string(10).upper() new_device = await self.store.store_device( user_id=user_id, - device_id=device_id, + device_id=new_device_id, initial_device_display_name=initial_device_display_name, ) if new_device: - await self.notify_device_update(user_id, [device_id]) - return device_id + await self.notify_device_update(user_id, [new_device_id]) + return new_device_id attempts += 1 raise errors.StoreError(500, "Couldn't generate a device ID.") @@ -433,7 +443,9 @@ class DeviceHandler(DeviceWorkerHandler): @trace @measure_func("notify_device_update") - async def notify_device_update(self, user_id, device_ids): + async def notify_device_update( + self, user_id: str, device_ids: Collection[str] + ) -> None: """Notify that a user's device(s) has changed. Pokes the notifier, and remote servers if the user is local. """ @@ -445,7 +457,7 @@ class DeviceHandler(DeviceWorkerHandler): user_id ) - hosts = set() + hosts = set() # type: Set[str] if self.hs.is_mine_id(user_id): hosts.update(get_domain_from_id(u) for u in users_who_share_room) hosts.discard(self.server_name) @@ -497,7 +509,7 @@ class DeviceHandler(DeviceWorkerHandler): self.notifier.on_new_event("device_list_key", position, users=[from_user_id]) - async def user_left_room(self, user, room_id): + async def user_left_room(self, user: UserID, room_id: str) -> None: user_id = user.to_string() room_ids = await self.store.get_rooms_for_user(user_id) if not room_ids: @@ -505,8 +517,89 @@ class DeviceHandler(DeviceWorkerHandler): # receive device updates. Mark this in DB. await self.store.mark_remote_user_device_list_as_unsubscribed(user_id) + async def store_dehydrated_device( + self, + user_id: str, + device_data: JsonDict, + initial_device_display_name: Optional[str] = None, + ) -> str: + """Store a dehydrated device for a user. If the user had a previous + dehydrated device, it is removed. + + Args: + user_id: the user that we are storing the device for + device_data: the dehydrated device information + initial_device_display_name: The display name to use for the device + Returns: + device id of the dehydrated device + """ + device_id = await self.check_device_registered( + user_id, None, initial_device_display_name, + ) + old_device_id = await self.store.store_dehydrated_device( + user_id, device_id, device_data + ) + if old_device_id is not None: + await self.delete_device(user_id, old_device_id) + return device_id + + async def get_dehydrated_device( + self, user_id: str + ) -> Optional[Tuple[str, JsonDict]]: + """Retrieve the information for a dehydrated device. + + Args: + user_id: the user whose dehydrated device we are looking for + Returns: + a tuple whose first item is the device ID, and the second item is + the dehydrated device information + """ + return await self.store.get_dehydrated_device(user_id) + + async def rehydrate_device( + self, user_id: str, access_token: str, device_id: str + ) -> dict: + """Process a rehydration request from the user. + + Args: + user_id: the user who is rehydrating the device + access_token: the access token used for the request + device_id: the ID of the device that will be rehydrated + Returns: + a dict containing {"success": True} + """ + success = await self.store.remove_dehydrated_device(user_id, device_id) + + if not success: + raise errors.NotFoundError() + + # If the dehydrated device was successfully deleted (the device ID + # matched the stored dehydrated device), then modify the access + # token to use the dehydrated device's ID and copy the old device + # display name to the dehydrated device, and destroy the old device + # ID + old_device_id = await self.store.set_device_for_access_token( + access_token, device_id + ) + old_device = await self.store.get_device(user_id, old_device_id) + await self.store.update_device(user_id, device_id, old_device["display_name"]) + # can't call self.delete_device because that will clobber the + # access token so call the storage layer directly + await self.store.delete_device(user_id, old_device_id) + await self.store.delete_e2e_keys_by_device( + user_id=user_id, device_id=old_device_id + ) -def _update_device_from_client_ips(device, client_ips): + # tell everyone that the old device is gone and that the dehydrated + # device has a new display name + await self.notify_device_update(user_id, [old_device_id, device_id]) + + return {"success": True} + + +def _update_device_from_client_ips( + device: Dict[str, Any], client_ips: Dict[Tuple[str, str], Dict[str, Any]] +) -> None: 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")}) @@ -514,7 +607,7 @@ def _update_device_from_client_ips(device, client_ips): class DeviceListUpdater: "Handles incoming device list updates from federation and updates the DB" - def __init__(self, hs, device_handler): + def __init__(self, hs: "HomeServer", device_handler: DeviceHandler): self.store = hs.get_datastore() self.federation = hs.get_federation_client() self.clock = hs.get_clock() @@ -523,7 +616,9 @@ class DeviceListUpdater: self._remote_edu_linearizer = Linearizer(name="remote_device_list") # user_id -> list of updates waiting to be handled. - self._pending_updates = {} + self._pending_updates = ( + {} + ) # type: Dict[str, List[Tuple[str, str, Iterable[str], JsonDict]]] # 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 @@ -546,7 +641,9 @@ class DeviceListUpdater: ) @trace - async def incoming_device_list_update(self, origin, edu_content): + async def incoming_device_list_update( + self, origin: str, edu_content: JsonDict + ) -> None: """Called on incoming device list update from federation. Responsible for parsing the EDU and adding to pending updates list. """ @@ -607,7 +704,7 @@ class DeviceListUpdater: await self._handle_device_updates(user_id) @measure_func("_incoming_device_list_update") - async def _handle_device_updates(self, user_id): + async def _handle_device_updates(self, user_id: str) -> None: "Actually handle pending updates." with (await self._remote_edu_linearizer.queue(user_id)): @@ -655,7 +752,9 @@ class DeviceListUpdater: stream_id for _, stream_id, _, _ in pending_updates ) - async def _need_to_do_resync(self, user_id, updates): + async def _need_to_do_resync( + self, user_id: str, updates: Iterable[Tuple[str, str, Iterable[str], JsonDict]] + ) -> bool: """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. """ @@ -686,7 +785,7 @@ class DeviceListUpdater: return False @trace - async def _maybe_retry_device_resync(self): + async def _maybe_retry_device_resync(self) -> None: """Retry to resync device lists that are out of sync, except if another retry is in progress. """ @@ -729,7 +828,7 @@ class DeviceListUpdater: async def user_device_resync( self, user_id: str, mark_failed_as_stale: bool = True - ) -> Optional[dict]: + ) -> Optional[JsonDict]: """Fetches all devices for a user and updates the device cache with them. Args: @@ -753,7 +852,7 @@ class DeviceListUpdater: # it later. await self.store.mark_remote_user_device_cache_as_stale(user_id) - return + return None except (RequestSendFailed, HttpResponseException) as e: logger.warning( "Failed to handle device list update for %s: %s", user_id, e, @@ -770,12 +869,12 @@ class DeviceListUpdater: # 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 + return None except FederationDeniedError as e: set_tag("error", True) log_kv({"reason": "FederationDeniedError"}) logger.info(e) - return + return None except Exception as e: set_tag("error", True) log_kv( @@ -788,7 +887,7 @@ class DeviceListUpdater: # it later. await self.store.mark_remote_user_device_cache_as_stale(user_id) - return + return None log_kv({"result": result}) stream_id = result["stream_id"] devices = result["devices"] @@ -849,7 +948,7 @@ class DeviceListUpdater: user_id: str, master_key: Optional[Dict[str, Any]], self_signing_key: Optional[Dict[str, Any]], - ) -> list: + ) -> List[str]: """Process the given new master and self-signing key for the given remote user. Args: diff --git a/synapse/handlers/directory.py b/synapse/handlers/directory.py index 62aa9a2da8..ad5683d251 100644 --- a/synapse/handlers/directory.py +++ b/synapse/handlers/directory.py @@ -46,6 +46,7 @@ class DirectoryHandler(BaseHandler): self.config = hs.config self.enable_room_list_search = hs.config.enable_room_list_search self.require_membership = hs.config.require_membership_for_aliases + self.third_party_event_rules = hs.get_third_party_event_rules() self.federation = hs.get_federation_client() hs.get_federation_registry().register_query_handler( @@ -383,7 +384,7 @@ class DirectoryHandler(BaseHandler): """ creator = await self.store.get_room_alias_creator(alias.to_string()) - if creator is not None and creator == user_id: + if creator == user_id: return True # Resolve the alias to the corresponding room. @@ -454,6 +455,15 @@ class DirectoryHandler(BaseHandler): # per alias creation rule? raise SynapseError(403, "Not allowed to publish room") + # Check if publishing is blocked by a third party module + allowed_by_third_party_rules = await ( + self.third_party_event_rules.check_visibility_can_be_modified( + room_id, visibility + ) + ) + if not allowed_by_third_party_rules: + raise SynapseError(403, "Not allowed to publish room") + await self.store.set_room_is_public(room_id, making_public) async def edit_published_appservice_room_list( diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index dd40fd1299..611742ae72 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -496,6 +496,22 @@ class E2eKeysHandler: log_kv( {"message": "Did not update one_time_keys", "reason": "no keys given"} ) + fallback_keys = keys.get("org.matrix.msc2732.fallback_keys", None) + if fallback_keys and isinstance(fallback_keys, dict): + log_kv( + { + "message": "Updating fallback_keys for device.", + "user_id": user_id, + "device_id": device_id, + } + ) + await self.store.set_e2e_fallback_keys(user_id, device_id, fallback_keys) + elif fallback_keys: + log_kv({"message": "Did not update fallback_keys", "reason": "not a dict"}) + else: + log_kv( + {"message": "Did not update fallback_keys", "reason": "no keys given"} + ) # the device should have been registered already, but it may have been # deleted due to a race with a DELETE request. Or we may be using an diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 1a8144405a..5ac2fc5656 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -3008,6 +3008,9 @@ class FederationHandler(BaseHandler): elif event.internal_metadata.is_outlier(): return + # the event has been persisted so it should have a stream ordering. + assert event.internal_metadata.stream_ordering + event_pos = PersistedEventPosition( self._instance_name, event.internal_metadata.stream_ordering ) diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index ee271e85e5..3e9a22e8f3 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -635,59 +635,6 @@ class EventCreationHandler: msg = self._block_events_without_consent_error % {"consent_uri": consent_uri} raise ConsentNotGivenError(msg=msg, consent_uri=consent_uri) - async def send_nonmember_event( - self, - requester: Requester, - event: EventBase, - context: EventContext, - ratelimit: bool = True, - ignore_shadow_ban: bool = False, - ) -> int: - """ - Persists and notifies local clients and federation of an event. - - Args: - requester: The requester sending the event. - event: The event to send. - context: The context of the event. - ratelimit: Whether to rate limit this send. - ignore_shadow_ban: True if shadow-banned users should be allowed to - send this event. - - Return: - The stream_id of the persisted event. - - Raises: - ShadowBanError if the requester has been shadow-banned. - """ - if event.type == EventTypes.Member: - raise SynapseError( - 500, "Tried to send member event through non-member codepath" - ) - - if not ignore_shadow_ban and requester.shadow_banned: - # We randomly sleep a bit just to annoy the requester. - await self.clock.sleep(random.randint(1, 10)) - raise ShadowBanError() - - user = UserID.from_string(event.sender) - - assert self.hs.is_mine(user), "User must be our own: %s" % (user,) - - if event.is_state(): - prev_event = await self.deduplicate_state_event(event, context) - if prev_event is not None: - logger.info( - "Not bothering to persist state event %s duplicated by %s", - event.event_id, - prev_event.event_id, - ) - return await self.store.get_stream_id_for_event(prev_event.event_id) - - return await self.handle_new_client_event( - requester=requester, event=event, context=context, ratelimit=ratelimit - ) - async def deduplicate_state_event( self, event: EventBase, context: EventContext ) -> Optional[EventBase]: @@ -728,7 +675,7 @@ class EventCreationHandler: """ Creates an event, then sends it. - See self.create_event and self.send_nonmember_event. + See self.create_event and self.handle_new_client_event. Args: requester: The requester sending the event. @@ -738,9 +685,19 @@ class EventCreationHandler: ignore_shadow_ban: True if shadow-banned users should be allowed to send this event. + Returns: + The event, and its stream ordering (if state event deduplication happened, + the previous, duplicate event). + Raises: ShadowBanError if the requester has been shadow-banned. """ + + if event_dict["type"] == EventTypes.Member: + raise SynapseError( + 500, "Tried to send member event through non-member codepath" + ) + if not ignore_shadow_ban and requester.shadow_banned: # We randomly sleep a bit just to annoy the requester. await self.clock.sleep(random.randint(1, 10)) @@ -756,20 +713,27 @@ class EventCreationHandler: requester, event_dict, token_id=requester.access_token_id, txn_id=txn_id ) + assert self.hs.is_mine_id(event.sender), "User must be our own: %s" % ( + event.sender, + ) + spam_error = self.spam_checker.check_event_for_spam(event) if spam_error: if not isinstance(spam_error, str): spam_error = "Spam is not permitted here" raise SynapseError(403, spam_error, Codes.FORBIDDEN) - stream_id = await self.send_nonmember_event( - requester, - event, - context, + ev = await self.handle_new_client_event( + requester=requester, + event=event, + context=context, ratelimit=ratelimit, ignore_shadow_ban=ignore_shadow_ban, ) - return event, stream_id + + # we know it was persisted, so must have a stream ordering + assert ev.internal_metadata.stream_ordering + return ev, ev.internal_metadata.stream_ordering @measure_func("create_new_client_event") async def create_new_client_event( @@ -843,8 +807,11 @@ class EventCreationHandler: context: EventContext, ratelimit: bool = True, extra_users: List[UserID] = [], - ) -> int: - """Processes a new event. This includes checking auth, persisting it, + ignore_shadow_ban: bool = False, + ) -> EventBase: + """Processes a new event. + + This includes deduplicating, checking auth, persisting, notifying users, sending to remote servers, etc. If called from a worker will hit out to the master process for final @@ -857,10 +824,39 @@ class EventCreationHandler: ratelimit extra_users: Any extra users to notify about event + ignore_shadow_ban: True if shadow-banned users should be allowed to + send this event. + Return: - The stream_id of the persisted event. + If the event was deduplicated, the previous, duplicate, event. Otherwise, + `event`. + + Raises: + ShadowBanError if the requester has been shadow-banned. """ + # we don't apply shadow-banning to membership events here. Invites are blocked + # higher up the stack, and we allow shadow-banned users to send join and leave + # events as normal. + if ( + event.type != EventTypes.Member + and not ignore_shadow_ban + and requester.shadow_banned + ): + # We randomly sleep a bit just to annoy the requester. + await self.clock.sleep(random.randint(1, 10)) + raise ShadowBanError() + + if event.is_state(): + prev_event = await self.deduplicate_state_event(event, context) + if prev_event is not None: + logger.info( + "Not bothering to persist state event %s duplicated by %s", + event.event_id, + prev_event.event_id, + ) + return prev_event + if event.is_state() and (event.type, event.state_key) == ( EventTypes.Create, "", @@ -915,13 +911,13 @@ class EventCreationHandler: ) stream_id = result["stream_id"] event.internal_metadata.stream_ordering = stream_id - return stream_id + return event stream_id = await self.persist_and_notify_client_event( requester, event, context, ratelimit=ratelimit, extra_users=extra_users ) - return stream_id + return event except Exception: # Ensure that we actually remove the entries in the push actions # staging area, if we calculated them. @@ -1232,8 +1228,12 @@ class EventCreationHandler: # Since this is a dummy-event it is OK if it is sent by a # shadow-banned user. - await self.send_nonmember_event( - requester, event, context, ratelimit=False, ignore_shadow_ban=True, + await self.handle_new_client_event( + requester=requester, + event=event, + context=context, + ratelimit=False, + ignore_shadow_ban=True, ) return True except ConsentNotGivenError: diff --git a/synapse/handlers/oidc_handler.py b/synapse/handlers/oidc_handler.py index 19cd652675..05ac86e697 100644 --- a/synapse/handlers/oidc_handler.py +++ b/synapse/handlers/oidc_handler.py @@ -96,6 +96,7 @@ class OidcHandler: self.hs = hs self._callback_url = hs.config.oidc_callback_url # type: str self._scopes = hs.config.oidc_scopes # type: List[str] + self._user_profile_method = hs.config.oidc_user_profile_method # type: str self._client_auth = ClientAuth( hs.config.oidc_client_id, hs.config.oidc_client_secret, @@ -196,11 +197,11 @@ class OidcHandler: % (m["response_types_supported"],) ) - # If the openid scope was not requested, we need a userinfo endpoint to fetch user infos + # Ensure there's a userinfo endpoint to fetch from if it is required. if self._uses_userinfo: if m.get("userinfo_endpoint") is None: raise ValueError( - 'provider has no "userinfo_endpoint", even though it is required because the "openid" scope is not requested' + 'provider has no "userinfo_endpoint", even though it is required' ) else: # If we're not using userinfo, we need a valid jwks to validate the ID token @@ -220,8 +221,10 @@ class OidcHandler: ``access_token`` with the ``userinfo_endpoint``. """ - # Maybe that should be user-configurable and not inferred? - return "openid" not in self._scopes + return ( + "openid" not in self._scopes + or self._user_profile_method == "userinfo_endpoint" + ) async def load_metadata(self) -> OpenIDProviderMetadata: """Load and validate the provider metadata. diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index d5f7c78edf..d0530a446c 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -185,6 +185,7 @@ class RoomCreationHandler(BaseHandler): ShadowBanError if the requester is shadow-banned. """ user_id = requester.user.to_string() + assert self.hs.is_mine_id(user_id), "User must be our own: %s" % (user_id,) # start by allocating a new room id r = await self.store.get_room(old_room_id) @@ -229,8 +230,8 @@ class RoomCreationHandler(BaseHandler): ) # now send the tombstone - await self.event_creation_handler.send_nonmember_event( - requester, tombstone_event, tombstone_context + await self.event_creation_handler.handle_new_client_event( + requester=requester, event=tombstone_event, context=tombstone_context, ) old_room_state = await tombstone_context.get_current_state_ids() @@ -681,6 +682,15 @@ class RoomCreationHandler(BaseHandler): creator_id=user_id, is_public=is_public, room_version=room_version, ) + # Check whether this visibility value is blocked by a third party module + allowed_by_third_party_rules = await ( + self.third_party_event_rules.check_visibility_can_be_modified( + room_id, visibility + ) + ) + if not allowed_by_third_party_rules: + raise SynapseError(403, "Room visibility value not allowed.") + directory_handler = self.hs.get_handlers().directory_handler if room_alias: await directory_handler.create_association( @@ -962,8 +972,6 @@ class RoomCreationHandler(BaseHandler): try: random_string = stringutils.random_string(18) gen_room_id = RoomID(random_string, self.hs.hostname).to_string() - if isinstance(gen_room_id, bytes): - gen_room_id = gen_room_id.decode("utf-8") await self.store.store_room( room_id=gen_room_id, room_creator_user_id=creator_id, diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index 8feba8c90a..fd8114a64d 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -22,7 +22,7 @@ from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple, Union from unpaddedbase64 import encode_base64 from synapse import types -from synapse.api.constants import MAX_DEPTH, EventTypes, Membership +from synapse.api.constants import MAX_DEPTH, AccountDataTypes, EventTypes, Membership from synapse.api.errors import ( AuthError, Codes, @@ -188,15 +188,6 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): require_consent=require_consent, ) - # Check if this event matches the previous membership event for the user. - duplicate = await self.event_creation_handler.deduplicate_state_event( - event, context - ) - if duplicate is not None: - # Discard the new event since this membership change is a no-op. - _, stream_id = await self.store.get_event_ordering(duplicate.event_id) - return duplicate.event_id, stream_id - prev_state_ids = await context.get_prev_state_ids() prev_member_event_id = prev_state_ids.get((EventTypes.Member, user_id), None) @@ -221,7 +212,7 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): retry_after_ms=int(1000 * (time_allowed - time_now_s)) ) - stream_id = await self.event_creation_handler.handle_new_client_event( + result_event = await self.event_creation_handler.handle_new_client_event( requester, event, context, extra_users=[target], ratelimit=ratelimit, ) @@ -231,7 +222,9 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): if prev_member_event.membership == Membership.JOIN: await self._user_left_room(target, room_id) - return event.event_id, stream_id + # we know it was persisted, so should have a stream ordering + assert result_event.internal_metadata.stream_ordering + return result_event.event_id, result_event.internal_metadata.stream_ordering async def copy_room_tags_and_direct_to_room( self, old_room_id, new_room_id, user_id @@ -247,7 +240,7 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): user_account_data, _ = await self.store.get_account_data_for_user(user_id) # Copy direct message state if applicable - direct_rooms = user_account_data.get("m.direct", {}) + direct_rooms = user_account_data.get(AccountDataTypes.DIRECT, {}) # Check which key this room is under if isinstance(direct_rooms, dict): @@ -258,7 +251,7 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): # Save back to user's m.direct account data await self.store.add_account_data_for_user( - user_id, "m.direct", direct_rooms + user_id, AccountDataTypes.DIRECT, direct_rooms ) break @@ -441,12 +434,12 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): same_membership = old_membership == effective_membership_state same_sender = requester.user.to_string() == old_state.sender if same_sender and same_membership and same_content: - _, stream_id = await self.store.get_event_ordering( - old_state.event_id - ) + # duplicate event. + # we know it was persisted, so must have a stream ordering. + assert old_state.internal_metadata.stream_ordering return ( old_state.event_id, - stream_id, + old_state.internal_metadata.stream_ordering, ) if old_membership in ["ban", "leave"] and action == "kick": @@ -642,7 +635,7 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): async def send_membership_event( self, - requester: Requester, + requester: Optional[Requester], event: EventBase, context: EventContext, ratelimit: bool = True, @@ -672,12 +665,6 @@ class RoomMemberHandler(metaclass=abc.ABCMeta): else: requester = types.create_requester(target_user) - prev_event = await self.event_creation_handler.deduplicate_state_event( - event, context - ) - if prev_event is not None: - return - prev_state_ids = await context.get_prev_state_ids() if event.membership == Membership.JOIN: if requester.is_guest: @@ -1185,10 +1172,13 @@ class RoomMemberMasterHandler(RoomMemberHandler): context = await self.state_handler.compute_event_context(event) context.app_service = requester.app_service - stream_id = await self.event_creation_handler.handle_new_client_event( + result_event = await self.event_creation_handler.handle_new_client_event( requester, event, context, extra_users=[UserID.from_string(target_user)], ) - return event.event_id, stream_id + # we know it was persisted, so must have a stream ordering + assert result_event.internal_metadata.stream_ordering + + return result_event.event_id, result_event.internal_metadata.stream_ordering async def _user_left_room(self, target: UserID, room_id: str) -> None: """Implements RoomMemberHandler._user_left_room diff --git a/synapse/handlers/stats.py b/synapse/handlers/stats.py index 249ffe2a55..dc62b21c06 100644 --- a/synapse/handlers/stats.py +++ b/synapse/handlers/stats.py @@ -49,7 +49,7 @@ class StatsHandler: # Guard to ensure we only process deltas one at a time self._is_processing = False - if hs.config.stats_enabled: + if self.stats_enabled and hs.config.run_background_tasks: 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 diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index bfe2583002..6fb8332f93 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -21,7 +21,7 @@ from typing import TYPE_CHECKING, Any, Dict, FrozenSet, List, Optional, Set, Tup import attr from prometheus_client import Counter -from synapse.api.constants import EventTypes, Membership +from synapse.api.constants import AccountDataTypes, EventTypes, Membership from synapse.api.filtering import FilterCollection from synapse.events import EventBase from synapse.logging.context import current_context @@ -87,7 +87,7 @@ class SyncConfig: class TimelineBatch: prev_batch = attr.ib(type=StreamToken) events = attr.ib(type=List[EventBase]) - limited = attr.ib(bool) + limited = attr.ib(type=bool) def __bool__(self) -> bool: """Make the result appear empty if there are no updates. This is used @@ -201,6 +201,8 @@ class SyncResult: device_lists: List of user_ids whose devices have changed device_one_time_keys_count: Dict of algorithm to count for one time keys for this device + device_unused_fallback_key_types: List of key types that have an unused fallback + key groups: Group updates, if any """ @@ -213,6 +215,7 @@ class SyncResult: to_device = attr.ib(type=List[JsonDict]) device_lists = attr.ib(type=DeviceLists) device_one_time_keys_count = attr.ib(type=JsonDict) + device_unused_fallback_key_types = attr.ib(type=List[str]) groups = attr.ib(type=Optional[GroupsSyncResult]) def __bool__(self) -> bool: @@ -457,8 +460,13 @@ class SyncHandler: recents = [] if not limited or block_all_timeline: + prev_batch_token = now_token + if recents: + room_key = recents[0].internal_metadata.before + prev_batch_token = now_token.copy_and_replace("room_key", room_key) + return TimelineBatch( - events=recents, prev_batch=now_token, limited=False + events=recents, prev_batch=prev_batch_token, limited=False ) filtering_factor = 2 @@ -1014,10 +1022,14 @@ class SyncHandler: logger.debug("Fetching OTK data") device_id = sync_config.device_id one_time_key_counts = {} # type: JsonDict + unused_fallback_key_types = [] # type: List[str] if device_id: one_time_key_counts = await self.store.count_e2e_one_time_keys( user_id, device_id ) + unused_fallback_key_types = await self.store.get_e2e_unused_fallback_key_types( + user_id, device_id + ) logger.debug("Fetching group data") await self._generate_sync_entry_for_groups(sync_result_builder) @@ -1041,6 +1053,7 @@ class SyncHandler: device_lists=device_lists, groups=sync_result_builder.groups, device_one_time_keys_count=one_time_key_counts, + device_unused_fallback_key_types=unused_fallback_key_types, next_batch=sync_result_builder.now_token, ) @@ -1378,13 +1391,16 @@ class SyncHandler: return set(), set(), set(), set() ignored_account_data = await self.store.get_global_account_data_by_type_for_user( - "m.ignored_user_list", user_id=user_id + AccountDataTypes.IGNORED_USER_LIST, user_id=user_id ) + # If there is ignored users account data and it matches the proper type, + # then use it. + ignored_users = frozenset() # type: FrozenSet[str] if ignored_account_data: - ignored_users = ignored_account_data.get("ignored_users", {}).keys() - else: - ignored_users = frozenset() + ignored_users_data = ignored_account_data.get("ignored_users", {}) + if isinstance(ignored_users_data, dict): + ignored_users = frozenset(ignored_users_data.keys()) if since_token: room_changes = await self._get_rooms_changed( @@ -1478,7 +1494,7 @@ class SyncHandler: return False async def _get_rooms_changed( - self, sync_result_builder: "SyncResultBuilder", ignored_users: Set[str] + self, sync_result_builder: "SyncResultBuilder", ignored_users: FrozenSet[str] ) -> _RoomChanges: """Gets the the changes that have happened since the last sync. """ @@ -1690,7 +1706,7 @@ class SyncHandler: return _RoomChanges(room_entries, invited, newly_joined_rooms, newly_left_rooms) async def _get_all_rooms( - self, sync_result_builder: "SyncResultBuilder", ignored_users: Set[str] + self, sync_result_builder: "SyncResultBuilder", ignored_users: FrozenSet[str] ) -> _RoomChanges: """Returns entries for all rooms for the user. @@ -1764,7 +1780,7 @@ class SyncHandler: async def _generate_room_entry( self, sync_result_builder: "SyncResultBuilder", - ignored_users: Set[str], + ignored_users: FrozenSet[str], room_builder: "RoomSyncResultBuilder", ephemeral: List[JsonDict], tags: Optional[Dict[str, Dict[str, Any]]], diff --git a/synapse/http/server.py b/synapse/http/server.py index 996a31a9ec..00b98af3d4 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -257,7 +257,7 @@ class _AsyncResource(resource.Resource, metaclass=abc.ABCMeta): if isinstance(raw_callback_return, (defer.Deferred, types.CoroutineType)): callback_return = await raw_callback_return else: - callback_return = raw_callback_return + callback_return = raw_callback_return # type: ignore return callback_return @@ -406,7 +406,7 @@ class JsonResource(DirectServeJsonResource): if isinstance(raw_callback_return, (defer.Deferred, types.CoroutineType)): callback_return = await raw_callback_return else: - callback_return = raw_callback_return + callback_return = raw_callback_return # type: ignore return callback_return @@ -651,6 +651,11 @@ def respond_with_json_bytes( Returns: twisted.web.server.NOT_DONE_YET if the request is still active. """ + if request._disconnected: + logger.warning( + "Not sending response to request %s, already disconnected.", request + ) + return request.setResponseCode(code) request.setHeader(b"Content-Type", b"application/json") diff --git a/synapse/logging/_structured.py b/synapse/logging/_structured.py index 144506c8f2..0fc2ea609e 100644 --- a/synapse/logging/_structured.py +++ b/synapse/logging/_structured.py @@ -12,7 +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 logging import os.path import sys @@ -89,14 +88,7 @@ class LogContextObserver: context = current_context() # Copy the context information to the log event. - if context is not None: - context.copy_to_twisted_log_entry(event) - else: - # If there's no logging context, not even the root one, we might be - # starting up or it might be from non-Synapse code. Log it as if it - # came from the root logger. - event["request"] = None - event["scope"] = None + context.copy_to_twisted_log_entry(event) self.observer(event) diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py index fcbd5378c4..b410e3ad9c 100644 --- a/synapse/module_api/__init__.py +++ b/synapse/module_api/__init__.py @@ -14,13 +14,19 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging +from typing import TYPE_CHECKING, Iterable, Optional, Tuple from twisted.internet import defer +from synapse.http.client import SimpleHttpClient from synapse.http.site import SynapseRequest from synapse.logging.context import make_deferred_yieldable, run_in_background +from synapse.storage.state import StateFilter from synapse.types import UserID +if TYPE_CHECKING: + from synapse.server import HomeServer + """ This package defines the 'stable' API which can be used by extension modules which are loaded into Synapse. @@ -43,6 +49,27 @@ class ModuleApi: self._auth = hs.get_auth() self._auth_handler = auth_handler + # We expose these as properties below in order to attach a helpful docstring. + self._http_client = hs.get_simple_http_client() # type: SimpleHttpClient + self._public_room_list_manager = PublicRoomListManager(hs) + + @property + def http_client(self): + """Allows making outbound HTTP requests to remote resources. + + An instance of synapse.http.client.SimpleHttpClient + """ + return self._http_client + + @property + def public_room_list_manager(self): + """Allows adding to, removing from and checking the status of rooms in the + public room list. + + An instance of synapse.module_api.PublicRoomListManager + """ + return self._public_room_list_manager + def get_user_by_req(self, req, allow_guest=False): """Check the access_token provided for a request @@ -266,3 +293,70 @@ class ModuleApi: await self._auth_handler.complete_sso_login( registered_user_id, request, client_redirect_url, ) + + @defer.inlineCallbacks + def get_state_events_in_room( + self, room_id: str, types: Iterable[Tuple[str, Optional[str]]] + ) -> defer.Deferred: + """Gets current state events for the given room. + + (This is exposed for compatibility with the old SpamCheckerApi. We should + probably deprecate it and replace it with an async method in a subclass.) + + Args: + room_id: The room ID to get state events in. + types: The event type and state key (using None + to represent 'any') of the room state to acquire. + + Returns: + twisted.internet.defer.Deferred[list(synapse.events.FrozenEvent)]: + The filtered state events in the room. + """ + state_ids = yield defer.ensureDeferred( + self._store.get_filtered_current_state_ids( + room_id=room_id, state_filter=StateFilter.from_types(types) + ) + ) + state = yield defer.ensureDeferred(self._store.get_events(state_ids.values())) + return state.values() + + +class PublicRoomListManager: + """Contains methods for adding to, removing from and querying whether a room + is in the public room list. + """ + + def __init__(self, hs: "HomeServer"): + self._store = hs.get_datastore() + + async def room_is_in_public_room_list(self, room_id: str) -> bool: + """Checks whether a room is in the public room list. + + Args: + room_id: The ID of the room. + + Returns: + Whether the room is in the public room list. Returns False if the room does + not exist. + """ + room = await self._store.get_room(room_id) + if not room: + return False + + return room.get("is_public", False) + + async def add_room_to_public_room_list(self, room_id: str) -> None: + """Publishes a room to the public room list. + + Args: + room_id: The ID of the room. + """ + await self._store.set_room_is_public(room_id, True) + + async def remove_room_from_public_room_list(self, room_id: str) -> None: + """Removes a room from the public room list. + + Args: + room_id: The ID of the room. + """ + await self._store.set_room_is_public(room_id, False) diff --git a/synapse/push/push_rule_evaluator.py b/synapse/push/push_rule_evaluator.py index 709ace01e5..3a68ce636f 100644 --- a/synapse/push/push_rule_evaluator.py +++ b/synapse/push/push_rule_evaluator.py @@ -16,7 +16,7 @@ import logging import re -from typing import Any, Dict, List, Pattern, Union +from typing import Any, Dict, List, Optional, Pattern, Union from synapse.events import EventBase from synapse.types import UserID @@ -181,7 +181,7 @@ class PushRuleEvaluatorForEvent: return r.search(body) - def _get_value(self, dotted_key: str) -> str: + def _get_value(self, dotted_key: str) -> Optional[str]: return self._value_cache.get(dotted_key, None) diff --git a/synapse/replication/tcp/handler.py b/synapse/replication/tcp/handler.py index b323841f73..e92da7b263 100644 --- a/synapse/replication/tcp/handler.py +++ b/synapse/replication/tcp/handler.py @@ -251,10 +251,9 @@ class ReplicationCommandHandler: using TCP. """ if hs.config.redis.redis_enabled: - import txredisapi - from synapse.replication.tcp.redis import ( RedisDirectTcpReplicationClientFactory, + lazyConnection, ) logger.info( @@ -271,7 +270,8 @@ class ReplicationCommandHandler: # connection after SUBSCRIBE is called). # First create the connection for sending commands. - outbound_redis_connection = txredisapi.lazyConnection( + outbound_redis_connection = lazyConnection( + reactor=hs.get_reactor(), host=hs.config.redis_host, port=hs.config.redis_port, password=hs.config.redis.redis_password, diff --git a/synapse/replication/tcp/protocol.py b/synapse/replication/tcp/protocol.py index 0b0d204e64..a509e599c2 100644 --- a/synapse/replication/tcp/protocol.py +++ b/synapse/replication/tcp/protocol.py @@ -51,10 +51,11 @@ import fcntl import logging import struct from inspect import isawaitable -from typing import TYPE_CHECKING, List +from typing import TYPE_CHECKING, List, Optional from prometheus_client import Counter +from twisted.internet import task from twisted.protocols.basic import LineOnlyReceiver from twisted.python.failure import Failure @@ -152,9 +153,10 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver): self.last_received_command = self.clock.time_msec() self.last_sent_command = 0 - self.time_we_closed = None # When we requested the connection be closed + # When we requested the connection be closed + self.time_we_closed = None # type: Optional[int] - self.received_ping = False # Have we reecived a ping from the other side + self.received_ping = False # Have we received a ping from the other side self.state = ConnectionStates.CONNECTING @@ -165,7 +167,7 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver): self.pending_commands = [] # type: List[Command] # The LoopingCall for sending pings. - self._send_ping_loop = None + self._send_ping_loop = None # type: Optional[task.LoopingCall] # a logcontext which we use for processing incoming commands. We declare it as a # background process so that the CPU stats get reported to prometheus. diff --git a/synapse/replication/tcp/redis.py b/synapse/replication/tcp/redis.py index f225e533de..de19705c1f 100644 --- a/synapse/replication/tcp/redis.py +++ b/synapse/replication/tcp/redis.py @@ -15,7 +15,7 @@ import logging from inspect import isawaitable -from typing import TYPE_CHECKING +from typing import TYPE_CHECKING, Optional import txredisapi @@ -228,3 +228,41 @@ class RedisDirectTcpReplicationClientFactory(txredisapi.SubscriberFactory): p.password = self.password return p + + +def lazyConnection( + reactor, + host: str = "localhost", + port: int = 6379, + dbid: Optional[int] = None, + reconnect: bool = True, + charset: str = "utf-8", + password: Optional[str] = None, + connectTimeout: Optional[int] = None, + replyTimeout: Optional[int] = None, + convertNumbers: bool = True, +) -> txredisapi.RedisProtocol: + """Equivalent to `txredisapi.lazyConnection`, except allows specifying a + reactor. + """ + + isLazy = True + poolsize = 1 + + uuid = "%s:%d" % (host, port) + factory = txredisapi.RedisFactory( + uuid, + dbid, + poolsize, + isLazy, + txredisapi.ConnectionHandler, + charset, + password, + replyTimeout, + convertNumbers, + ) + factory.continueTrying = reconnect + for x in range(poolsize): + reactor.connectTCP(host, port, factory, connectTimeout) + + return factory.handler diff --git a/synapse/rest/admin/__init__.py b/synapse/rest/admin/__init__.py index 57cac22252..789431ef25 100644 --- a/synapse/rest/admin/__init__.py +++ b/synapse/rest/admin/__init__.py @@ -57,6 +57,7 @@ from synapse.rest.admin.users import ( UsersRestServletV2, WhoisRestServlet, ) +from synapse.types import RoomStreamToken from synapse.util.versionstring import get_version_string logger = logging.getLogger(__name__) @@ -109,7 +110,9 @@ class PurgeHistoryRestServlet(RestServlet): if event.room_id != room_id: raise SynapseError(400, "Event is for wrong room.") - room_token = await self.store.get_topological_token_for_event(event_id) + room_token = RoomStreamToken( + event.depth, event.internal_metadata.stream_ordering + ) token = await room_token.to_string(self.store) logger.info("[purge] purging up to token %s (event_id %s)", token, event_id) diff --git a/synapse/rest/client/v2_alpha/devices.py b/synapse/rest/client/v2_alpha/devices.py index 7e174de692..af117cb27c 100644 --- a/synapse/rest/client/v2_alpha/devices.py +++ b/synapse/rest/client/v2_alpha/devices.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2015, 2016 OpenMarket Ltd +# Copyright 2020 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -21,6 +22,7 @@ from synapse.http.servlet import ( assert_params_in_dict, parse_json_object_from_request, ) +from synapse.http.site import SynapseRequest from ._base import client_patterns, interactive_auth_handler @@ -151,7 +153,139 @@ class DeviceRestServlet(RestServlet): return 200, {} +class DehydratedDeviceServlet(RestServlet): + """Retrieve or store a dehydrated device. + + GET /org.matrix.msc2697.v2/dehydrated_device + + HTTP/1.1 200 OK + Content-Type: application/json + + { + "device_id": "dehydrated_device_id", + "device_data": { + "algorithm": "org.matrix.msc2697.v1.dehydration.v1.olm", + "account": "dehydrated_device" + } + } + + PUT /org.matrix.msc2697/dehydrated_device + Content-Type: application/json + + { + "device_data": { + "algorithm": "org.matrix.msc2697.v1.dehydration.v1.olm", + "account": "dehydrated_device" + } + } + + HTTP/1.1 200 OK + Content-Type: application/json + + { + "device_id": "dehydrated_device_id" + } + + """ + + PATTERNS = client_patterns("/org.matrix.msc2697.v2/dehydrated_device", releases=()) + + def __init__(self, hs): + super().__init__() + self.hs = hs + self.auth = hs.get_auth() + self.device_handler = hs.get_device_handler() + + async def on_GET(self, request: SynapseRequest): + requester = await self.auth.get_user_by_req(request) + dehydrated_device = await self.device_handler.get_dehydrated_device( + requester.user.to_string() + ) + if dehydrated_device is not None: + (device_id, device_data) = dehydrated_device + result = {"device_id": device_id, "device_data": device_data} + return (200, result) + else: + raise errors.NotFoundError("No dehydrated device available") + + async def on_PUT(self, request: SynapseRequest): + submission = parse_json_object_from_request(request) + requester = await self.auth.get_user_by_req(request) + + if "device_data" not in submission: + raise errors.SynapseError( + 400, "device_data missing", errcode=errors.Codes.MISSING_PARAM, + ) + elif not isinstance(submission["device_data"], dict): + raise errors.SynapseError( + 400, + "device_data must be an object", + errcode=errors.Codes.INVALID_PARAM, + ) + + device_id = await self.device_handler.store_dehydrated_device( + requester.user.to_string(), + submission["device_data"], + submission.get("initial_device_display_name", None), + ) + return 200, {"device_id": device_id} + + +class ClaimDehydratedDeviceServlet(RestServlet): + """Claim a dehydrated device. + + POST /org.matrix.msc2697.v2/dehydrated_device/claim + Content-Type: application/json + + { + "device_id": "dehydrated_device_id" + } + + HTTP/1.1 200 OK + Content-Type: application/json + + { + "success": true, + } + + """ + + PATTERNS = client_patterns( + "/org.matrix.msc2697.v2/dehydrated_device/claim", releases=() + ) + + def __init__(self, hs): + super().__init__() + self.hs = hs + self.auth = hs.get_auth() + self.device_handler = hs.get_device_handler() + + async def on_POST(self, request: SynapseRequest): + requester = await self.auth.get_user_by_req(request) + + submission = parse_json_object_from_request(request) + + if "device_id" not in submission: + raise errors.SynapseError( + 400, "device_id missing", errcode=errors.Codes.MISSING_PARAM, + ) + elif not isinstance(submission["device_id"], str): + raise errors.SynapseError( + 400, "device_id must be a string", errcode=errors.Codes.INVALID_PARAM, + ) + + result = await self.device_handler.rehydrate_device( + requester.user.to_string(), + self.auth.get_access_token_from_request(request), + submission["device_id"], + ) + + return (200, result) + + def register_servlets(hs, http_server): DeleteDevicesRestServlet(hs).register(http_server) DevicesRestServlet(hs).register(http_server) DeviceRestServlet(hs).register(http_server) + DehydratedDeviceServlet(hs).register(http_server) + ClaimDehydratedDeviceServlet(hs).register(http_server) diff --git a/synapse/rest/client/v2_alpha/keys.py b/synapse/rest/client/v2_alpha/keys.py index 55c4606569..b91996c738 100644 --- a/synapse/rest/client/v2_alpha/keys.py +++ b/synapse/rest/client/v2_alpha/keys.py @@ -1,6 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2015, 2016 OpenMarket Ltd # Copyright 2019 New Vector Ltd +# Copyright 2020 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -67,6 +68,7 @@ class KeyUploadServlet(RestServlet): super().__init__() self.auth = hs.get_auth() self.e2e_keys_handler = hs.get_e2e_keys_handler() + self.device_handler = hs.get_device_handler() @trace(opname="upload_keys") async def on_POST(self, request, device_id): @@ -75,23 +77,28 @@ class KeyUploadServlet(RestServlet): body = parse_json_object_from_request(request) if device_id is not None: - # passing the device_id here is deprecated; however, we allow it - # for now for compatibility with older clients. + # Providing the device_id should only be done for setting keys + # for dehydrated devices; however, we allow it for any device for + # compatibility with older clients. if requester.device_id is not None and device_id != requester.device_id: - set_tag("error", True) - log_kv( - { - "message": "Client uploading keys for a different device", - "logged_in_id": requester.device_id, - "key_being_uploaded": device_id, - } - ) - logger.warning( - "Client uploading keys for a different device " - "(logged in as %s, uploading for %s)", - requester.device_id, - device_id, + dehydrated_device = await self.device_handler.get_dehydrated_device( + user_id ) + if dehydrated_device is not None and device_id != dehydrated_device[0]: + set_tag("error", True) + log_kv( + { + "message": "Client uploading keys for a different device", + "logged_in_id": requester.device_id, + "key_being_uploaded": device_id, + } + ) + logger.warning( + "Client uploading keys for a different device " + "(logged in as %s, uploading for %s)", + requester.device_id, + device_id, + ) else: device_id = requester.device_id diff --git a/synapse/rest/client/v2_alpha/sync.py b/synapse/rest/client/v2_alpha/sync.py index 6779df952f..2b84eb89c0 100644 --- a/synapse/rest/client/v2_alpha/sync.py +++ b/synapse/rest/client/v2_alpha/sync.py @@ -236,6 +236,7 @@ class SyncRestServlet(RestServlet): "leave": sync_result.groups.leave, }, "device_one_time_keys_count": sync_result.device_one_time_keys_count, + "org.matrix.msc2732.device_unused_fallback_key_types": sync_result.device_unused_fallback_key_types, "next_batch": await sync_result.next_batch.to_string(self.store), } diff --git a/synapse/rest/media/v1/_base.py b/synapse/rest/media/v1/_base.py index 6568e61829..67aa993f19 100644 --- a/synapse/rest/media/v1/_base.py +++ b/synapse/rest/media/v1/_base.py @@ -213,6 +213,12 @@ async def respond_with_responder( file_size (int|None): Size in bytes of the media. If not known it should be None upload_name (str|None): The name of the requested file, if any. """ + if request._disconnected: + logger.warning( + "Not sending response to request %s, already disconnected.", request + ) + return + if not responder: respond_404(request) return diff --git a/synapse/server.py b/synapse/server.py index 5e3752c333..f83dd6148c 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -91,6 +91,7 @@ from synapse.handlers.typing import FollowerTypingHandler, TypingWriterHandler from synapse.handlers.user_directory import UserDirectoryHandler from synapse.http.client import InsecureInterceptableContextFactory, SimpleHttpClient from synapse.http.matrixfederationclient import MatrixFederationHttpClient +from synapse.module_api import ModuleApi from synapse.notifier import Notifier from synapse.push.action_generator import ActionGenerator from synapse.push.pusherpool import PusherPool @@ -185,7 +186,10 @@ class HomeServer(metaclass=abc.ABCMeta): we are listening on to provide HTTP services. """ - REQUIRED_ON_MASTER_STARTUP = ["user_directory_handler", "stats_handler"] + REQUIRED_ON_BACKGROUND_TASK_STARTUP = [ + "auth", + "stats", + ] # This is overridden in derived application classes # (such as synapse.app.homeserver.SynapseHomeServer) and gives the class to be @@ -251,14 +255,20 @@ class HomeServer(metaclass=abc.ABCMeta): self.datastores = Databases(self.DATASTORE_CLASS, self) logger.info("Finished setting up.") - def setup_master(self) -> None: + # Register background tasks required by this server. This must be done + # somewhat manually due to the background tasks not being registered + # unless handlers are instantiated. + if self.config.run_background_tasks: + self.setup_background_tasks() + + def setup_background_tasks(self) -> None: """ Some handlers have side effects on instantiation (like registering background updates). This function causes them to be fetched, and therefore instantiated, to run those side effects. """ - for i in self.REQUIRED_ON_MASTER_STARTUP: - getattr(self, "get_" + i)() + for i in self.REQUIRED_ON_BACKGROUND_TASK_STARTUP: + getattr(self, "get_" + i + "_handler")() def get_reactor(self) -> twisted.internet.base.ReactorBase: """ @@ -647,6 +657,10 @@ class HomeServer(metaclass=abc.ABCMeta): def get_federation_ratelimiter(self) -> FederationRateLimiter: return FederationRateLimiter(self.clock, config=self.config.rc_federation) + @cache_in_self + def get_module_api(self) -> ModuleApi: + return ModuleApi(self, self.get_auth_handler()) + async def remove_pusher(self, app_id: str, push_key: str, user_id: str): return await self.get_pusherpool().remove_pusher(app_id, push_key, user_id) diff --git a/synapse/spam_checker_api/__init__.py b/synapse/spam_checker_api/__init__.py index 395ac5ab02..3ce25bb012 100644 --- a/synapse/spam_checker_api/__init__.py +++ b/synapse/spam_checker_api/__init__.py @@ -12,19 +12,8 @@ # 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 enum import Enum -from twisted.internet import defer - -from synapse.storage.state import StateFilter - -MYPY = False -if MYPY: - import synapse.server - -logger = logging.getLogger(__name__) - class RegistrationBehaviour(Enum): """ @@ -34,35 +23,3 @@ class RegistrationBehaviour(Enum): ALLOW = "allow" SHADOW_BAN = "shadow_ban" DENY = "deny" - - -class SpamCheckerApi: - """A proxy object that gets passed to spam checkers so they can get - access to rooms and other relevant information. - """ - - def __init__(self, hs: "synapse.server.HomeServer"): - self.hs = hs - - self._store = hs.get_datastore() - - @defer.inlineCallbacks - def get_state_events_in_room(self, room_id: str, types: tuple) -> defer.Deferred: - """Gets state events for the given room. - - Args: - room_id: The room ID to get state events in. - types: The event type and state key (using None - to represent 'any') of the room state to acquire. - - Returns: - twisted.internet.defer.Deferred[list(synapse.events.FrozenEvent)]: - The filtered state events in the room. - """ - state_ids = yield defer.ensureDeferred( - self._store.get_filtered_current_state_ids( - room_id=room_id, state_filter=StateFilter.from_types(types) - ) - ) - state = yield defer.ensureDeferred(self._store.get_events(state_ids.values())) - return state.values() diff --git a/synapse/state/__init__.py b/synapse/state/__init__.py index 31082bb16a..5b0900aa3c 100644 --- a/synapse/state/__init__.py +++ b/synapse/state/__init__.py @@ -738,7 +738,7 @@ def _make_state_cache_entry( # failing that, look for the closest match. prev_group = None - delta_ids = None + delta_ids = None # type: Optional[StateMap[str]] for old_group, old_state in state_groups_ids.items(): n_delta_ids = {k: v for k, v in new_state.items() if old_state.get(k) != v} diff --git a/synapse/storage/database.py b/synapse/storage/database.py index 6116191b16..0ba3a025cf 100644 --- a/synapse/storage/database.py +++ b/synapse/storage/database.py @@ -32,6 +32,7 @@ from typing import ( overload, ) +import attr from prometheus_client import Histogram from typing_extensions import Literal @@ -90,13 +91,17 @@ def make_pool( return adbapi.ConnectionPool( db_config.config["name"], cp_reactor=reactor, - cp_openfun=engine.on_new_connection, + cp_openfun=lambda conn: engine.on_new_connection( + LoggingDatabaseConnection(conn, engine, "on_new_connection") + ), **db_config.config.get("args", {}) ) def make_conn( - db_config: DatabaseConnectionConfig, engine: BaseDatabaseEngine + db_config: DatabaseConnectionConfig, + engine: BaseDatabaseEngine, + default_txn_name: str, ) -> Connection: """Make a new connection to the database and return it. @@ -109,11 +114,60 @@ def make_conn( for k, v in db_config.config.get("args", {}).items() if not k.startswith("cp_") } - db_conn = engine.module.connect(**db_params) + native_db_conn = engine.module.connect(**db_params) + db_conn = LoggingDatabaseConnection(native_db_conn, engine, default_txn_name) + engine.on_new_connection(db_conn) return db_conn +@attr.s(slots=True) +class LoggingDatabaseConnection: + """A wrapper around a database connection that returns `LoggingTransaction` + as its cursor class. + + This is mainly used on startup to ensure that queries get logged correctly + """ + + conn = attr.ib(type=Connection) + engine = attr.ib(type=BaseDatabaseEngine) + default_txn_name = attr.ib(type=str) + + def cursor( + self, *, txn_name=None, after_callbacks=None, exception_callbacks=None + ) -> "LoggingTransaction": + if not txn_name: + txn_name = self.default_txn_name + + return LoggingTransaction( + self.conn.cursor(), + name=txn_name, + database_engine=self.engine, + after_callbacks=after_callbacks, + exception_callbacks=exception_callbacks, + ) + + def close(self) -> None: + self.conn.close() + + def commit(self) -> None: + self.conn.commit() + + def rollback(self, *args, **kwargs) -> None: + self.conn.rollback(*args, **kwargs) + + def __enter__(self) -> "Connection": + self.conn.__enter__() + return self + + def __exit__(self, exc_type, exc_value, traceback) -> bool: + return self.conn.__exit__(exc_type, exc_value, traceback) + + # Proxy through any unknown lookups to the DB conn class. + def __getattr__(self, name): + return getattr(self.conn, name) + + # The type of entry which goes on our after_callbacks and exception_callbacks lists. # # Python 3.5.2 doesn't support Callable with an ellipsis, so we wrap it in quotes so @@ -247,6 +301,12 @@ class LoggingTransaction: def close(self) -> None: self.txn.close() + def __enter__(self) -> "LoggingTransaction": + return self + + def __exit__(self, exc_type, exc_value, traceback): + self.close() + class PerformanceCounters: def __init__(self): @@ -395,7 +455,7 @@ class DatabasePool: def new_transaction( self, - conn: Connection, + conn: LoggingDatabaseConnection, desc: str, after_callbacks: List[_CallbackListEntry], exception_callbacks: List[_CallbackListEntry], @@ -436,12 +496,10 @@ class DatabasePool: i = 0 N = 5 while True: - cursor = LoggingTransaction( - conn.cursor(), - name, - self.engine, - after_callbacks, - exception_callbacks, + cursor = conn.cursor( + txn_name=name, + after_callbacks=after_callbacks, + exception_callbacks=exception_callbacks, ) try: r = func(cursor, *args, **kwargs) @@ -638,7 +696,10 @@ class DatabasePool: if db_autocommit: self.engine.attempt_to_set_autocommit(conn, True) - return func(conn, *args, **kwargs) + db_conn = LoggingDatabaseConnection( + conn, self.engine, "runWithConnection" + ) + return func(db_conn, *args, **kwargs) finally: if db_autocommit: self.engine.attempt_to_set_autocommit(conn, False) @@ -1678,7 +1739,7 @@ class DatabasePool: def get_cache_dict( self, - db_conn: Connection, + db_conn: LoggingDatabaseConnection, table: str, entity_column: str, stream_column: str, @@ -1699,9 +1760,7 @@ class DatabasePool: "limit": limit, } - sql = self.engine.convert_param_style(sql) - - txn = db_conn.cursor() + txn = db_conn.cursor(txn_name="get_cache_dict") txn.execute(sql, (int(max_value),)) cache = {row[0]: int(row[1]) for row in txn} diff --git a/synapse/storage/databases/__init__.py b/synapse/storage/databases/__init__.py index aa5d490624..0c24325011 100644 --- a/synapse/storage/databases/__init__.py +++ b/synapse/storage/databases/__init__.py @@ -46,7 +46,7 @@ class Databases: db_name = database_config.name engine = create_engine(database_config.config) - with make_conn(database_config, engine) as db_conn: + with make_conn(database_config, engine, "startup") as db_conn: logger.info("[database config %r]: Checking database server", db_name) engine.check_database(db_conn) diff --git a/synapse/storage/databases/main/__init__.py b/synapse/storage/databases/main/__init__.py index 0cb12f4c61..9b16f45f3e 100644 --- a/synapse/storage/databases/main/__init__.py +++ b/synapse/storage/databases/main/__init__.py @@ -15,9 +15,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -import calendar import logging -import time from typing import Any, Dict, List, Optional, Tuple from synapse.api.constants import PresenceState @@ -268,9 +266,6 @@ class DataStore( self._stream_order_on_start = self.get_room_max_stream_ordering() self._min_stream_order_on_start = self.get_room_min_stream_ordering() - # Used in _generate_user_daily_visits to keep track of progress - self._last_user_visit_update = self._get_start_of_day() - def get_device_stream_token(self) -> int: return self._device_list_id_gen.get_current_token() @@ -289,7 +284,6 @@ class DataStore( " last_user_sync_ts, status_msg, currently_active FROM presence_stream" " WHERE state != ?" ) - sql = self.database_engine.convert_param_style(sql) txn = db_conn.cursor() txn.execute(sql, (PresenceState.OFFLINE,)) @@ -301,192 +295,6 @@ class DataStore( return [UserPresenceState(**row) for row in rows] - async def count_daily_users(self) -> int: - """ - Counts the number of users who used this homeserver in the last 24 hours. - """ - yesterday = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24) - return await self.db_pool.runInteraction( - "count_daily_users", self._count_users, yesterday - ) - - async def count_monthly_users(self) -> int: - """ - Counts the number of users who used this homeserver in the last 30 days. - Note this method is intended for phonehome metrics only and is different - from the mau figure in synapse.storage.monthly_active_users which, - amongst other things, includes a 3 day grace period before a user counts. - """ - thirty_days_ago = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24 * 30) - return await self.db_pool.runInteraction( - "count_monthly_users", self._count_users, thirty_days_ago - ) - - def _count_users(self, txn, time_from): - """ - Returns number of users seen in the past time_from period - """ - sql = """ - SELECT COALESCE(count(*), 0) FROM ( - SELECT user_id FROM user_ips - WHERE last_seen > ? - GROUP BY user_id - ) u - """ - txn.execute(sql, (time_from,)) - (count,) = txn.fetchone() - return count - - async def count_r30_users(self) -> Dict[str, int]: - """ - Counts the number of 30 day retained users, defined as:- - * 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 apart - - Returns: - A mapping of counts globally as well as broken out by platform. - """ - - def _count_r30_users(txn): - thirty_days_in_secs = 86400 * 30 - now = int(self._clock.time()) - thirty_days_ago_in_secs = now - thirty_days_in_secs - - sql = """ - SELECT platform, COALESCE(count(*), 0) FROM ( - SELECT - users.name, platform, users.creation_ts * 1000, - MAX(uip.last_seen) - FROM users - INNER JOIN ( - SELECT - user_id, - last_seen, - CASE - WHEN user_agent LIKE '%%Android%%' THEN 'android' - WHEN user_agent LIKE '%%iOS%%' THEN 'ios' - WHEN user_agent LIKE '%%Electron%%' THEN 'electron' - WHEN user_agent LIKE '%%Mozilla%%' THEN 'web' - WHEN user_agent LIKE '%%Gecko%%' THEN 'web' - ELSE 'unknown' - END - AS platform - FROM user_ips - ) uip - ON users.name = uip.user_id - AND users.appservice_id is NULL - AND users.creation_ts < ? - AND uip.last_seen/1000 > ? - AND (uip.last_seen/1000) - users.creation_ts > 86400 * 30 - GROUP BY users.name, platform, users.creation_ts - ) u GROUP BY platform - """ - - results = {} - txn.execute(sql, (thirty_days_ago_in_secs, thirty_days_ago_in_secs)) - - for row in txn: - if row[0] == "unknown": - pass - results[row[0]] = row[1] - - sql = """ - SELECT COALESCE(count(*), 0) FROM ( - SELECT users.name, users.creation_ts * 1000, - MAX(uip.last_seen) - FROM users - INNER JOIN ( - SELECT - user_id, - last_seen - FROM user_ips - ) uip - ON users.name = uip.user_id - AND appservice_id is NULL - AND users.creation_ts < ? - AND uip.last_seen/1000 > ? - AND (uip.last_seen/1000) - users.creation_ts > 86400 * 30 - GROUP BY users.name, users.creation_ts - ) u - """ - - txn.execute(sql, (thirty_days_ago_in_secs, thirty_days_ago_in_secs)) - - (count,) = txn.fetchone() - results["all"] = count - - return results - - return await self.db_pool.runInteraction("count_r30_users", _count_r30_users) - - def _get_start_of_day(self): - """ - Returns millisecond unixtime for start of UTC day. - """ - now = time.gmtime() - today_start = calendar.timegm((now.tm_year, now.tm_mon, now.tm_mday, 0, 0, 0)) - return today_start * 1000 - - async def generate_user_daily_visits(self) -> None: - """ - Generates daily visit data for use in cohort/ retention analysis - """ - - def _generate_user_daily_visits(txn): - logger.info("Calling _generate_user_daily_visits") - today_start = self._get_start_of_day() - a_day_in_milliseconds = 24 * 60 * 60 * 1000 - now = self.clock.time_msec() - - sql = """ - INSERT INTO user_daily_visits (user_id, device_id, timestamp) - SELECT u.user_id, u.device_id, ? - FROM user_ips AS u - LEFT JOIN ( - SELECT user_id, device_id, timestamp FROM user_daily_visits - WHERE timestamp = ? - ) udv - ON u.user_id = udv.user_id AND u.device_id=udv.device_id - INNER JOIN users ON users.name=u.user_id - WHERE last_seen > ? AND last_seen <= ? - AND udv.timestamp IS NULL AND users.is_guest=0 - AND users.appservice_id IS NULL - GROUP BY u.user_id, u.device_id - """ - - # This means that the day has rolled over but there could still - # be entries from the previous day. There is an edge case - # where if the user logs in at 23:59 and overwrites their - # last_seen at 00:01 then they will not be counted in the - # previous day's stats - it is important that the query is run - # often to minimise this case. - if today_start > self._last_user_visit_update: - yesterday_start = today_start - a_day_in_milliseconds - txn.execute( - sql, - ( - yesterday_start, - yesterday_start, - self._last_user_visit_update, - today_start, - ), - ) - self._last_user_visit_update = today_start - - txn.execute( - sql, (today_start, today_start, self._last_user_visit_update, now) - ) - # Update _last_user_visit_update to now. The reason to do this - # rather just clamping to the beginning of the day is to limit - # the size of the join - meaning that the query can be run more - # frequently - self._last_user_visit_update = now - - await self.db_pool.runInteraction( - "generate_user_daily_visits", _generate_user_daily_visits - ) - async def get_users(self) -> List[Dict[str, Any]]: """Function to retrieve a list of users in users table. diff --git a/synapse/storage/databases/main/account_data.py b/synapse/storage/databases/main/account_data.py index ef81d73573..49ee23470d 100644 --- a/synapse/storage/databases/main/account_data.py +++ b/synapse/storage/databases/main/account_data.py @@ -18,6 +18,7 @@ import abc import logging from typing import Dict, List, Optional, Tuple +from synapse.api.constants import AccountDataTypes from synapse.storage._base import SQLBaseStore, db_to_json from synapse.storage.database import DatabasePool from synapse.storage.util.id_generators import StreamIdGenerator @@ -291,14 +292,18 @@ class AccountDataWorkerStore(SQLBaseStore, metaclass=abc.ABCMeta): self, ignored_user_id: str, ignorer_user_id: str, cache_context: _CacheContext ) -> bool: ignored_account_data = await self.get_global_account_data_by_type_for_user( - "m.ignored_user_list", + AccountDataTypes.IGNORED_USER_LIST, ignorer_user_id, on_invalidate=cache_context.invalidate, ) if not ignored_account_data: return False - return ignored_user_id in ignored_account_data.get("ignored_users", {}) + try: + return ignored_user_id in ignored_account_data.get("ignored_users", {}) + except TypeError: + # The type of the ignored_users field is invalid. + return False class AccountDataStore(AccountDataWorkerStore): diff --git a/synapse/storage/databases/main/censor_events.py b/synapse/storage/databases/main/censor_events.py index f211ddbaf8..4bb2b9c28c 100644 --- a/synapse/storage/databases/main/censor_events.py +++ b/synapse/storage/databases/main/censor_events.py @@ -21,8 +21,8 @@ from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage._base import SQLBaseStore from synapse.storage.database import DatabasePool from synapse.storage.databases.main.cache import CacheInvalidationWorkerStore -from synapse.storage.databases.main.events import encode_json from synapse.storage.databases.main.events_worker import EventsWorkerStore +from synapse.util.frozenutils import frozendict_json_encoder if TYPE_CHECKING: from synapse.server import HomeServer @@ -105,7 +105,7 @@ class CensorEventsStore(EventsWorkerStore, CacheInvalidationWorkerStore, SQLBase and original_event.internal_metadata.is_redacted() ): # Redaction was allowed - pruned_json = encode_json( + pruned_json = frozendict_json_encoder.encode( prune_event_dict( original_event.room_version, original_event.get_dict() ) @@ -171,7 +171,7 @@ class CensorEventsStore(EventsWorkerStore, CacheInvalidationWorkerStore, SQLBase return # Prune the event's dict then convert it to JSON. - pruned_json = encode_json( + pruned_json = frozendict_json_encoder.encode( prune_event_dict(event.room_version, event.get_dict()) ) diff --git a/synapse/storage/databases/main/client_ips.py b/synapse/storage/databases/main/client_ips.py index 239c7a949c..a25a888443 100644 --- a/synapse/storage/databases/main/client_ips.py +++ b/synapse/storage/databases/main/client_ips.py @@ -351,7 +351,63 @@ class ClientIpBackgroundUpdateStore(SQLBaseStore): return updated -class ClientIpStore(ClientIpBackgroundUpdateStore): +class ClientIpWorkerStore(ClientIpBackgroundUpdateStore): + def __init__(self, database: DatabasePool, db_conn, hs): + super().__init__(database, db_conn, hs) + + self.user_ips_max_age = hs.config.user_ips_max_age + + if hs.config.run_background_tasks and self.user_ips_max_age: + self._clock.looping_call(self._prune_old_user_ips, 5 * 1000) + + @wrap_as_background_process("prune_old_user_ips") + async def _prune_old_user_ips(self): + """Removes entries in user IPs older than the configured period. + """ + + if self.user_ips_max_age is None: + # Nothing to do + return + + if not await self.db_pool.updates.has_completed_background_update( + "devices_last_seen" + ): + # Only start pruning if we have finished populating the devices + # last seen info. + return + + # We do a slightly funky SQL delete to ensure we don't try and delete + # too much at once (as the table may be very large from before we + # started pruning). + # + # This works by finding the max last_seen that is less than the given + # time, but has no more than N rows before it, deleting all rows with + # a lesser last_seen time. (We COALESCE so that the sub-SELECT always + # returns exactly one row). + sql = """ + DELETE FROM user_ips + WHERE last_seen <= ( + SELECT COALESCE(MAX(last_seen), -1) + FROM ( + SELECT last_seen FROM user_ips + WHERE last_seen <= ? + ORDER BY last_seen ASC + LIMIT 5000 + ) AS u + ) + """ + + timestamp = self.clock.time_msec() - self.user_ips_max_age + + def _prune_old_user_ips_txn(txn): + txn.execute(sql, (timestamp,)) + + await self.db_pool.runInteraction( + "_prune_old_user_ips", _prune_old_user_ips_txn + ) + + +class ClientIpStore(ClientIpWorkerStore): def __init__(self, database: DatabasePool, db_conn, hs): self.client_ip_last_seen = Cache( @@ -360,8 +416,6 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): super().__init__(database, db_conn, hs) - self.user_ips_max_age = hs.config.user_ips_max_age - # (user_id, access_token, ip,) -> (user_agent, device_id, last_seen) self._batch_row_update = {} @@ -372,9 +426,6 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): "before", "shutdown", self._update_client_ips_batch ) - if self.user_ips_max_age: - self._clock.looping_call(self._prune_old_user_ips, 5 * 1000) - async def insert_client_ip( self, user_id, access_token, ip, user_agent, device_id, now=None ): @@ -525,49 +576,3 @@ class ClientIpStore(ClientIpBackgroundUpdateStore): } for (access_token, ip), (user_agent, last_seen) in results.items() ] - - @wrap_as_background_process("prune_old_user_ips") - async def _prune_old_user_ips(self): - """Removes entries in user IPs older than the configured period. - """ - - if self.user_ips_max_age is None: - # Nothing to do - return - - if not await self.db_pool.updates.has_completed_background_update( - "devices_last_seen" - ): - # Only start pruning if we have finished populating the devices - # last seen info. - return - - # We do a slightly funky SQL delete to ensure we don't try and delete - # too much at once (as the table may be very large from before we - # started pruning). - # - # This works by finding the max last_seen that is less than the given - # time, but has no more than N rows before it, deleting all rows with - # a lesser last_seen time. (We COALESCE so that the sub-SELECT always - # returns exactly one row). - sql = """ - DELETE FROM user_ips - WHERE last_seen <= ( - SELECT COALESCE(MAX(last_seen), -1) - FROM ( - SELECT last_seen FROM user_ips - WHERE last_seen <= ? - ORDER BY last_seen ASC - LIMIT 5000 - ) AS u - ) - """ - - timestamp = self.clock.time_msec() - self.user_ips_max_age - - def _prune_old_user_ips_txn(txn): - txn.execute(sql, (timestamp,)) - - await self.db_pool.runInteraction( - "_prune_old_user_ips", _prune_old_user_ips_txn - ) diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index fdf394c612..2d0a6408b5 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.py @@ -1,7 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2016 OpenMarket Ltd # Copyright 2019 New Vector Ltd -# Copyright 2019 The Matrix.org Foundation C.I.C. +# Copyright 2019,2020 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -33,7 +33,7 @@ from synapse.storage.database import ( make_tuple_comparison_clause, ) from synapse.types import Collection, JsonDict, get_verify_key_from_cross_signing_key -from synapse.util import json_encoder +from synapse.util import json_decoder, json_encoder from synapse.util.caches.descriptors import Cache, cached, cachedList from synapse.util.iterutils import batch_iter from synapse.util.stringutils import shortstr @@ -698,6 +698,80 @@ class DeviceWorkerStore(SQLBaseStore): _mark_remote_user_device_list_as_unsubscribed_txn, ) + async def get_dehydrated_device( + self, user_id: str + ) -> Optional[Tuple[str, JsonDict]]: + """Retrieve the information for a dehydrated device. + + Args: + user_id: the user whose dehydrated device we are looking for + Returns: + a tuple whose first item is the device ID, and the second item is + the dehydrated device information + """ + # FIXME: make sure device ID still exists in devices table + row = await self.db_pool.simple_select_one( + table="dehydrated_devices", + keyvalues={"user_id": user_id}, + retcols=["device_id", "device_data"], + allow_none=True, + ) + return ( + (row["device_id"], json_decoder.decode(row["device_data"])) if row else None + ) + + def _store_dehydrated_device_txn( + self, txn, user_id: str, device_id: str, device_data: str + ) -> Optional[str]: + old_device_id = self.db_pool.simple_select_one_onecol_txn( + txn, + table="dehydrated_devices", + keyvalues={"user_id": user_id}, + retcol="device_id", + allow_none=True, + ) + self.db_pool.simple_upsert_txn( + txn, + table="dehydrated_devices", + keyvalues={"user_id": user_id}, + values={"device_id": device_id, "device_data": device_data}, + ) + return old_device_id + + async def store_dehydrated_device( + self, user_id: str, device_id: str, device_data: JsonDict + ) -> Optional[str]: + """Store a dehydrated device for a user. + + Args: + user_id: the user that we are storing the device for + device_id: the ID of the dehydrated device + device_data: the dehydrated device information + Returns: + device id of the user's previous dehydrated device, if any + """ + return await self.db_pool.runInteraction( + "store_dehydrated_device_txn", + self._store_dehydrated_device_txn, + user_id, + device_id, + json_encoder.encode(device_data), + ) + + async def remove_dehydrated_device(self, user_id: str, device_id: str) -> bool: + """Remove a dehydrated device. + + Args: + user_id: the user that the dehydrated device belongs to + device_id: the ID of the dehydrated device + """ + count = await self.db_pool.simple_delete( + "dehydrated_devices", + {"user_id": user_id, "device_id": device_id}, + desc="remove_dehydrated_device", + ) + return count >= 1 + class DeviceBackgroundUpdateStore(SQLBaseStore): def __init__(self, database: DatabasePool, db_conn, hs): @@ -837,7 +911,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): self._clock.looping_call(self._prune_old_outbound_device_pokes, 60 * 60 * 1000) async def store_device( - self, user_id: str, device_id: str, initial_device_display_name: str + self, user_id: str, device_id: str, initial_device_display_name: Optional[str] ) -> bool: """Ensure the given device is known; add it to the store if not @@ -955,7 +1029,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): ) async def update_remote_device_list_cache_entry( - self, user_id: str, device_id: str, content: JsonDict, stream_id: int + self, user_id: str, device_id: str, content: JsonDict, stream_id: str ) -> None: """Updates a single device in the cache of a remote user's devicelist. @@ -983,7 +1057,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): user_id: str, device_id: str, content: JsonDict, - stream_id: int, + stream_id: str, ) -> None: if content.get("deleted"): self.db_pool.simple_delete_txn( diff --git a/synapse/storage/databases/main/end_to_end_keys.py b/synapse/storage/databases/main/end_to_end_keys.py index 22e1ed15d0..359dc6e968 100644 --- a/synapse/storage/databases/main/end_to_end_keys.py +++ b/synapse/storage/databases/main/end_to_end_keys.py @@ -1,7 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2015, 2016 OpenMarket Ltd # Copyright 2019 New Vector Ltd -# Copyright 2019 The Matrix.org Foundation C.I.C. +# Copyright 2019,2020 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -367,6 +367,57 @@ class EndToEndKeyWorkerStore(SQLBaseStore): "count_e2e_one_time_keys", _count_e2e_one_time_keys ) + async def set_e2e_fallback_keys( + self, user_id: str, device_id: str, fallback_keys: JsonDict + ) -> None: + """Set the user's e2e fallback keys. + + Args: + user_id: the user whose keys are being set + device_id: the device whose keys are being set + fallback_keys: the keys to set. This is a map from key ID (which is + of the form "algorithm:id") to key data. + """ + # fallback_keys will usually only have one item in it, so using a for + # loop (as opposed to calling simple_upsert_many_txn) won't be too bad + # FIXME: make sure that only one key per algorithm is uploaded + for key_id, fallback_key in fallback_keys.items(): + algorithm, key_id = key_id.split(":", 1) + await self.db_pool.simple_upsert( + "e2e_fallback_keys_json", + keyvalues={ + "user_id": user_id, + "device_id": device_id, + "algorithm": algorithm, + }, + values={ + "key_id": key_id, + "key_json": json_encoder.encode(fallback_key), + "used": False, + }, + desc="set_e2e_fallback_key", + ) + + @cached(max_entries=10000) + async def get_e2e_unused_fallback_key_types( + self, user_id: str, device_id: str + ) -> List[str]: + """Returns the fallback key types that have an unused key. + + Args: + user_id: the user whose keys are being queried + device_id: the device whose keys are being queried + + Returns: + a list of key types + """ + return await self.db_pool.simple_select_onecol( + "e2e_fallback_keys_json", + keyvalues={"user_id": user_id, "device_id": device_id, "used": False}, + retcol="algorithm", + desc="get_e2e_unused_fallback_key_types", + ) + async def get_e2e_cross_signing_key( self, user_id: str, key_type: str, from_user_id: Optional[str] = None ) -> Optional[dict]: @@ -701,15 +752,37 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): " WHERE user_id = ? AND device_id = ? AND algorithm = ?" " LIMIT 1" ) + fallback_sql = ( + "SELECT key_id, key_json, used FROM e2e_fallback_keys_json" + " WHERE user_id = ? AND device_id = ? AND algorithm = ?" + " LIMIT 1" + ) result = {} delete = [] + used_fallbacks = [] for user_id, device_id, algorithm in query_list: user_result = result.setdefault(user_id, {}) device_result = user_result.setdefault(device_id, {}) txn.execute(sql, (user_id, device_id, algorithm)) - for key_id, key_json in txn: + otk_row = txn.fetchone() + if otk_row is not None: + key_id, key_json = otk_row device_result[algorithm + ":" + key_id] = key_json delete.append((user_id, device_id, algorithm, key_id)) + else: + # no one-time key available, so see if there's a fallback + # key + txn.execute(fallback_sql, (user_id, device_id, algorithm)) + fallback_row = txn.fetchone() + if fallback_row is not None: + key_id, key_json, used = fallback_row + device_result[algorithm + ":" + key_id] = key_json + if not used: + used_fallbacks.append( + (user_id, device_id, algorithm, key_id) + ) + + # drop any one-time keys that were claimed sql = ( "DELETE FROM e2e_one_time_keys_json" " WHERE user_id = ? AND device_id = ? AND algorithm = ?" @@ -726,6 +799,23 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): self._invalidate_cache_and_stream( txn, self.count_e2e_one_time_keys, (user_id, device_id) ) + # mark fallback keys as used + for user_id, device_id, algorithm, key_id in used_fallbacks: + self.db_pool.simple_update_txn( + txn, + "e2e_fallback_keys_json", + { + "user_id": user_id, + "device_id": device_id, + "algorithm": algorithm, + "key_id": key_id, + }, + {"used": True}, + ) + self._invalidate_cache_and_stream( + txn, self.get_e2e_unused_fallback_key_types, (user_id, device_id) + ) + return result return await self.db_pool.runInteraction( @@ -754,6 +844,19 @@ class EndToEndKeyStore(EndToEndKeyWorkerStore, SQLBaseStore): self._invalidate_cache_and_stream( txn, self.count_e2e_one_time_keys, (user_id, device_id) ) + self.db_pool.simple_delete_txn( + txn, + table="dehydrated_devices", + keyvalues={"user_id": user_id, "device_id": device_id}, + ) + self.db_pool.simple_delete_txn( + txn, + table="e2e_fallback_keys_json", + keyvalues={"user_id": user_id, "device_id": device_id}, + ) + self._invalidate_cache_and_stream( + txn, self.get_e2e_unused_fallback_key_types, (user_id, device_id) + ) await self.db_pool.runInteraction( "delete_e2e_keys_by_device", delete_e2e_keys_by_device_txn diff --git a/synapse/storage/databases/main/event_push_actions.py b/synapse/storage/databases/main/event_push_actions.py index 62f1738732..80f3b4d740 100644 --- a/synapse/storage/databases/main/event_push_actions.py +++ b/synapse/storage/databases/main/event_push_actions.py @@ -20,7 +20,7 @@ from typing import Dict, List, Optional, Tuple, Union import attr from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.storage._base import LoggingTransaction, SQLBaseStore, db_to_json +from synapse.storage._base import SQLBaseStore, db_to_json from synapse.storage.database import DatabasePool from synapse.util import json_encoder from synapse.util.caches.descriptors import cached @@ -74,11 +74,7 @@ class EventPushActionsWorkerStore(SQLBaseStore): self.stream_ordering_month_ago = None self.stream_ordering_day_ago = None - cur = LoggingTransaction( - db_conn.cursor(), - name="_find_stream_orderings_for_times_txn", - database_engine=self.database_engine, - ) + cur = db_conn.cursor(txn_name="_find_stream_orderings_for_times_txn") self._find_stream_orderings_for_times_txn(cur) cur.close() diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py index 18def01f50..b4abd961b9 100644 --- a/synapse/storage/databases/main/events.py +++ b/synapse/storage/databases/main/events.py @@ -52,16 +52,6 @@ event_counter = Counter( ) -def encode_json(json_object): - """ - Encode a Python object as JSON and return it in a Unicode string. - """ - out = frozendict_json_encoder.encode(json_object) - if isinstance(out, bytes): - out = out.decode("utf8") - return out - - _EventCacheEntry = namedtuple("_EventCacheEntry", ("event", "redacted_event")) @@ -341,6 +331,10 @@ class PersistEventsStore: min_stream_order = events_and_contexts[0][0].internal_metadata.stream_ordering max_stream_order = events_and_contexts[-1][0].internal_metadata.stream_ordering + # stream orderings should have been assigned by now + assert min_stream_order + assert max_stream_order + self._update_forward_extremities_txn( txn, new_forward_extremities=new_forward_extremeties, @@ -743,7 +737,9 @@ class PersistEventsStore: logger.exception("") raise - metadata_json = encode_json(event.internal_metadata.get_dict()) + metadata_json = frozendict_json_encoder.encode( + event.internal_metadata.get_dict() + ) sql = "UPDATE event_json SET internal_metadata = ? WHERE event_id = ?" txn.execute(sql, (metadata_json, event.event_id)) @@ -797,10 +793,10 @@ class PersistEventsStore: { "event_id": event.event_id, "room_id": event.room_id, - "internal_metadata": encode_json( + "internal_metadata": frozendict_json_encoder.encode( event.internal_metadata.get_dict() ), - "json": encode_json(event_dict(event)), + "json": frozendict_json_encoder.encode(event_dict(event)), "format_version": event.format_version, } for event, _ in events_and_contexts diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index f95679ebc4..b7ed8ca6ab 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -74,6 +74,13 @@ class EventRedactBehaviour(Names): class EventsWorkerStore(SQLBaseStore): + # Whether to use dedicated DB threads for event fetching. This is only used + # if there are multiple DB threads available. When used will lock the DB + # thread for periods of time (so unit tests want to disable this when they + # run DB transactions on the main thread). See EVENT_QUEUE_* for more + # options controlling this. + USE_DEDICATED_DB_THREADS_FOR_EVENT_FETCHING = True + def __init__(self, database: DatabasePool, db_conn, hs): super().__init__(database, db_conn, hs) @@ -522,7 +529,11 @@ class EventsWorkerStore(SQLBaseStore): if not event_list: single_threaded = self.database_engine.single_threaded - if single_threaded or i > EVENT_QUEUE_ITERATIONS: + if ( + not self.USE_DEDICATED_DB_THREADS_FOR_EVENT_FETCHING + or single_threaded + or i > EVENT_QUEUE_ITERATIONS + ): self._event_fetch_ongoing -= 1 return else: @@ -712,6 +723,7 @@ class EventsWorkerStore(SQLBaseStore): internal_metadata_dict=internal_metadata, rejected_reason=rejected_reason, ) + original_ev.internal_metadata.stream_ordering = row["stream_ordering"] event_map[event_id] = original_ev @@ -779,6 +791,8 @@ class EventsWorkerStore(SQLBaseStore): * event_id (str) + * stream_ordering (int): stream ordering for this event + * json (str): json-encoded event structure * internal_metadata (str): json-encoded internal metadata dict @@ -811,13 +825,15 @@ class EventsWorkerStore(SQLBaseStore): sql = """\ SELECT e.event_id, - e.internal_metadata, - e.json, - e.format_version, + e.stream_ordering, + ej.internal_metadata, + ej.json, + ej.format_version, r.room_version, rej.reason - FROM event_json as e - LEFT JOIN rooms r USING (room_id) + FROM events AS e + JOIN event_json AS ej USING (event_id) + LEFT JOIN rooms r ON r.room_id = e.room_id LEFT JOIN rejections as rej USING (event_id) WHERE """ @@ -831,11 +847,12 @@ class EventsWorkerStore(SQLBaseStore): event_id = row[0] event_dict[event_id] = { "event_id": event_id, - "internal_metadata": row[1], - "json": row[2], - "format_version": row[3], - "room_version_id": row[4], - "rejected_reason": row[5], + "stream_ordering": row[1], + "internal_metadata": row[2], + "json": row[3], + "format_version": row[4], + "room_version_id": row[5], + "rejected_reason": row[6], "redactions": [], } diff --git a/synapse/storage/databases/main/metrics.py b/synapse/storage/databases/main/metrics.py index 92099f95ce..0acf0617ca 100644 --- a/synapse/storage/databases/main/metrics.py +++ b/synapse/storage/databases/main/metrics.py @@ -12,15 +12,21 @@ # 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 calendar +import logging +import time +from typing import Dict from synapse.metrics import GaugeBucketCollector -from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.metrics.background_process_metrics import wrap_as_background_process from synapse.storage._base import SQLBaseStore from synapse.storage.database import DatabasePool from synapse.storage.databases.main.event_push_actions import ( EventPushActionsWorkerStore, ) +logger = logging.getLogger(__name__) + # Collect metrics on the number of forward extremities that exist. _extremities_collecter = GaugeBucketCollector( "synapse_forward_extremities", @@ -51,15 +57,13 @@ class ServerMetricsStore(EventPushActionsWorkerStore, SQLBaseStore): super().__init__(database, db_conn, hs) # Read the extrems every 60 minutes - def read_forward_extremities(): - # run as a background process to make sure that the database transactions - # have a logcontext to report to - return run_as_background_process( - "read_forward_extremities", self._read_forward_extremities - ) + if hs.config.run_background_tasks: + self._clock.looping_call(self._read_forward_extremities, 60 * 60 * 1000) - hs.get_clock().looping_call(read_forward_extremities, 60 * 60 * 1000) + # Used in _generate_user_daily_visits to keep track of progress + self._last_user_visit_update = self._get_start_of_day() + @wrap_as_background_process("read_forward_extremities") async def _read_forward_extremities(self): def fetch(txn): txn.execute( @@ -137,3 +141,190 @@ class ServerMetricsStore(EventPushActionsWorkerStore, SQLBaseStore): return count return await self.db_pool.runInteraction("count_daily_active_rooms", _count) + + async def count_daily_users(self) -> int: + """ + Counts the number of users who used this homeserver in the last 24 hours. + """ + yesterday = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24) + return await self.db_pool.runInteraction( + "count_daily_users", self._count_users, yesterday + ) + + async def count_monthly_users(self) -> int: + """ + Counts the number of users who used this homeserver in the last 30 days. + Note this method is intended for phonehome metrics only and is different + from the mau figure in synapse.storage.monthly_active_users which, + amongst other things, includes a 3 day grace period before a user counts. + """ + thirty_days_ago = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24 * 30) + return await self.db_pool.runInteraction( + "count_monthly_users", self._count_users, thirty_days_ago + ) + + def _count_users(self, txn, time_from): + """ + Returns number of users seen in the past time_from period + """ + sql = """ + SELECT COALESCE(count(*), 0) FROM ( + SELECT user_id FROM user_ips + WHERE last_seen > ? + GROUP BY user_id + ) u + """ + txn.execute(sql, (time_from,)) + (count,) = txn.fetchone() + return count + + async def count_r30_users(self) -> Dict[str, int]: + """ + Counts the number of 30 day retained users, defined as:- + * 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 apart + + Returns: + A mapping of counts globally as well as broken out by platform. + """ + + def _count_r30_users(txn): + thirty_days_in_secs = 86400 * 30 + now = int(self._clock.time()) + thirty_days_ago_in_secs = now - thirty_days_in_secs + + sql = """ + SELECT platform, COALESCE(count(*), 0) FROM ( + SELECT + users.name, platform, users.creation_ts * 1000, + MAX(uip.last_seen) + FROM users + INNER JOIN ( + SELECT + user_id, + last_seen, + CASE + WHEN user_agent LIKE '%%Android%%' THEN 'android' + WHEN user_agent LIKE '%%iOS%%' THEN 'ios' + WHEN user_agent LIKE '%%Electron%%' THEN 'electron' + WHEN user_agent LIKE '%%Mozilla%%' THEN 'web' + WHEN user_agent LIKE '%%Gecko%%' THEN 'web' + ELSE 'unknown' + END + AS platform + FROM user_ips + ) uip + ON users.name = uip.user_id + AND users.appservice_id is NULL + AND users.creation_ts < ? + AND uip.last_seen/1000 > ? + AND (uip.last_seen/1000) - users.creation_ts > 86400 * 30 + GROUP BY users.name, platform, users.creation_ts + ) u GROUP BY platform + """ + + results = {} + txn.execute(sql, (thirty_days_ago_in_secs, thirty_days_ago_in_secs)) + + for row in txn: + if row[0] == "unknown": + pass + results[row[0]] = row[1] + + sql = """ + SELECT COALESCE(count(*), 0) FROM ( + SELECT users.name, users.creation_ts * 1000, + MAX(uip.last_seen) + FROM users + INNER JOIN ( + SELECT + user_id, + last_seen + FROM user_ips + ) uip + ON users.name = uip.user_id + AND appservice_id is NULL + AND users.creation_ts < ? + AND uip.last_seen/1000 > ? + AND (uip.last_seen/1000) - users.creation_ts > 86400 * 30 + GROUP BY users.name, users.creation_ts + ) u + """ + + txn.execute(sql, (thirty_days_ago_in_secs, thirty_days_ago_in_secs)) + + (count,) = txn.fetchone() + results["all"] = count + + return results + + return await self.db_pool.runInteraction("count_r30_users", _count_r30_users) + + def _get_start_of_day(self): + """ + Returns millisecond unixtime for start of UTC day. + """ + now = time.gmtime() + today_start = calendar.timegm((now.tm_year, now.tm_mon, now.tm_mday, 0, 0, 0)) + return today_start * 1000 + + @wrap_as_background_process("generate_user_daily_visits") + async def generate_user_daily_visits(self) -> None: + """ + Generates daily visit data for use in cohort/ retention analysis + """ + + def _generate_user_daily_visits(txn): + logger.info("Calling _generate_user_daily_visits") + today_start = self._get_start_of_day() + a_day_in_milliseconds = 24 * 60 * 60 * 1000 + now = self._clock.time_msec() + + sql = """ + INSERT INTO user_daily_visits (user_id, device_id, timestamp) + SELECT u.user_id, u.device_id, ? + FROM user_ips AS u + LEFT JOIN ( + SELECT user_id, device_id, timestamp FROM user_daily_visits + WHERE timestamp = ? + ) udv + ON u.user_id = udv.user_id AND u.device_id=udv.device_id + INNER JOIN users ON users.name=u.user_id + WHERE last_seen > ? AND last_seen <= ? + AND udv.timestamp IS NULL AND users.is_guest=0 + AND users.appservice_id IS NULL + GROUP BY u.user_id, u.device_id + """ + + # This means that the day has rolled over but there could still + # be entries from the previous day. There is an edge case + # where if the user logs in at 23:59 and overwrites their + # last_seen at 00:01 then they will not be counted in the + # previous day's stats - it is important that the query is run + # often to minimise this case. + if today_start > self._last_user_visit_update: + yesterday_start = today_start - a_day_in_milliseconds + txn.execute( + sql, + ( + yesterday_start, + yesterday_start, + self._last_user_visit_update, + today_start, + ), + ) + self._last_user_visit_update = today_start + + txn.execute( + sql, (today_start, today_start, self._last_user_visit_update, now) + ) + # Update _last_user_visit_update to now. The reason to do this + # rather just clamping to the beginning of the day is to limit + # the size of the join - meaning that the query can be run more + # frequently + self._last_user_visit_update = now + + await self.db_pool.runInteraction( + "generate_user_daily_visits", _generate_user_daily_visits + ) diff --git a/synapse/storage/databases/main/monthly_active_users.py b/synapse/storage/databases/main/monthly_active_users.py index e93aad33cd..c66f558567 100644 --- a/synapse/storage/databases/main/monthly_active_users.py +++ b/synapse/storage/databases/main/monthly_active_users.py @@ -32,6 +32,9 @@ class MonthlyActiveUsersWorkerStore(SQLBaseStore): self._clock = hs.get_clock() self.hs = hs + self._limit_usage_by_mau = hs.config.limit_usage_by_mau + self._max_mau_value = hs.config.max_mau_value + @cached(num_args=0) async def get_monthly_active_count(self) -> int: """Generates current count of monthly active users @@ -124,60 +127,6 @@ class MonthlyActiveUsersWorkerStore(SQLBaseStore): desc="user_last_seen_monthly_active", ) - -class MonthlyActiveUsersStore(MonthlyActiveUsersWorkerStore): - def __init__(self, database: DatabasePool, db_conn, hs): - super().__init__(database, db_conn, hs) - - self._limit_usage_by_mau = hs.config.limit_usage_by_mau - self._mau_stats_only = hs.config.mau_stats_only - self._max_mau_value = hs.config.max_mau_value - - # Do not add more reserved users than the total allowable number - # cur = LoggingTransaction( - self.db_pool.new_transaction( - db_conn, - "initialise_mau_threepids", - [], - [], - self._initialise_reserved_users, - hs.config.mau_limits_reserved_threepids[: self._max_mau_value], - ) - - def _initialise_reserved_users(self, txn, threepids): - """Ensures that reserved threepids are accounted for in the MAU table, should - be called on start up. - - Args: - txn (cursor): - threepids (list[dict]): List of threepid dicts to reserve - """ - - # XXX what is this function trying to achieve? It upserts into - # monthly_active_users for each *registered* reserved mau user, but why? - # - # - shouldn't there already be an entry for each reserved user (at least - # if they have been active recently)? - # - # - if it's important that the timestamp is kept up to date, why do we only - # run this at startup? - - for tp in threepids: - user_id = self.get_user_id_by_threepid_txn(txn, tp["medium"], tp["address"]) - - if user_id: - is_support = self.is_support_user_txn(txn, user_id) - if not is_support: - # We do this manually here to avoid hitting #6791 - self.db_pool.simple_upsert_txn( - txn, - table="monthly_active_users", - keyvalues={"user_id": user_id}, - values={"timestamp": int(self._clock.time_msec())}, - ) - else: - logger.warning("mau limit reserved threepid %s not found in db" % tp) - async def reap_monthly_active_users(self): """Cleans out monthly active user table to ensure that no stale entries exist. @@ -257,6 +206,57 @@ class MonthlyActiveUsersStore(MonthlyActiveUsersWorkerStore): "reap_monthly_active_users", _reap_users, reserved_users ) + +class MonthlyActiveUsersStore(MonthlyActiveUsersWorkerStore): + def __init__(self, database: DatabasePool, db_conn, hs): + super().__init__(database, db_conn, hs) + + self._mau_stats_only = hs.config.mau_stats_only + + # Do not add more reserved users than the total allowable number + self.db_pool.new_transaction( + db_conn, + "initialise_mau_threepids", + [], + [], + self._initialise_reserved_users, + hs.config.mau_limits_reserved_threepids[: self._max_mau_value], + ) + + def _initialise_reserved_users(self, txn, threepids): + """Ensures that reserved threepids are accounted for in the MAU table, should + be called on start up. + + Args: + txn (cursor): + threepids (list[dict]): List of threepid dicts to reserve + """ + + # XXX what is this function trying to achieve? It upserts into + # monthly_active_users for each *registered* reserved mau user, but why? + # + # - shouldn't there already be an entry for each reserved user (at least + # if they have been active recently)? + # + # - if it's important that the timestamp is kept up to date, why do we only + # run this at startup? + + for tp in threepids: + user_id = self.get_user_id_by_threepid_txn(txn, tp["medium"], tp["address"]) + + if user_id: + is_support = self.is_support_user_txn(txn, user_id) + if not is_support: + # We do this manually here to avoid hitting #6791 + self.db_pool.simple_upsert_txn( + txn, + table="monthly_active_users", + keyvalues={"user_id": user_id}, + values={"timestamp": int(self._clock.time_msec())}, + ) + else: + logger.warning("mau limit reserved threepid %s not found in db" % tp) + async def upsert_monthly_active_user(self, user_id: str) -> None: """Updates or inserts the user into the monthly active user table, which is used to track the current MAU usage of the server diff --git a/synapse/storage/databases/main/registration.py b/synapse/storage/databases/main/registration.py index a83df7759d..a85867936f 100644 --- a/synapse/storage/databases/main/registration.py +++ b/synapse/storage/databases/main/registration.py @@ -1,7 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 OpenMarket Ltd # Copyright 2017-2018 New Vector Ltd -# Copyright 2019 The Matrix.org Foundation C.I.C. +# Copyright 2019,2020 The Matrix.org Foundation C.I.C. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -14,14 +14,16 @@ # 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 re from typing import Any, Dict, List, Optional, Tuple from synapse.api.constants import UserTypes from synapse.api.errors import Codes, StoreError, SynapseError, ThreepidValidationError -from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.metrics.background_process_metrics import ( + run_as_background_process, + wrap_as_background_process, +) from synapse.storage._base import SQLBaseStore from synapse.storage.database import DatabasePool from synapse.storage.types import Cursor @@ -48,6 +50,21 @@ class RegistrationWorkerStore(SQLBaseStore): database.engine, find_max_generated_user_id_localpart, "user_id_seq", ) + self._account_validity = hs.config.account_validity + if hs.config.run_background_tasks and self._account_validity.enabled: + self._clock.call_later( + 0.0, + run_as_background_process, + "account_validity_set_expiration_dates", + self._set_expiration_date_when_missing, + ) + + # Create a background job for culling expired 3PID validity tokens + if hs.config.run_background_tasks: + self.clock.looping_call( + self.cull_expired_threepid_validation_tokens, THIRTY_MINUTES_IN_MS + ) + @cached() async def get_user_by_id(self, user_id: str) -> Optional[Dict[str, Any]]: return await self.db_pool.simple_select_one( @@ -778,6 +795,78 @@ class RegistrationWorkerStore(SQLBaseStore): "delete_threepid_session", delete_threepid_session_txn ) + @wrap_as_background_process("cull_expired_threepid_validation_tokens") + async def cull_expired_threepid_validation_tokens(self) -> None: + """Remove threepid validation tokens with expiry dates that have passed""" + + def cull_expired_threepid_validation_tokens_txn(txn, ts): + sql = """ + DELETE FROM threepid_validation_token WHERE + expires < ? + """ + txn.execute(sql, (ts,)) + + await self.db_pool.runInteraction( + "cull_expired_threepid_validation_tokens", + cull_expired_threepid_validation_tokens_txn, + self.clock.time_msec(), + ) + + async def _set_expiration_date_when_missing(self): + """ + Retrieves the list of registered users that don't have an expiration date, and + adds an expiration date for each of them. + """ + + def select_users_with_no_expiration_date_txn(txn): + """Retrieves the list of registered users with no expiration date from the + database, filtering out deactivated users. + """ + sql = ( + "SELECT users.name FROM users" + " LEFT JOIN account_validity ON (users.name = account_validity.user_id)" + " WHERE account_validity.user_id is NULL AND users.deactivated = 0;" + ) + txn.execute(sql, []) + + res = self.db_pool.cursor_to_dict(txn) + if res: + for user in res: + self.set_expiration_date_for_user_txn( + txn, user["name"], use_delta=True + ) + + await self.db_pool.runInteraction( + "get_users_with_no_expiration_date", + select_users_with_no_expiration_date_txn, + ) + + def set_expiration_date_for_user_txn(self, txn, user_id, use_delta=False): + """Sets an expiration date to the account with the given user ID. + + Args: + user_id (str): User ID to set an expiration date for. + use_delta (bool): If set to False, the expiration date for the user will be + now + validity period. If set to True, this expiration date will be a + random value in the [now + period - d ; now + period] range, d being a + delta equal to 10% of the validity period. + """ + now_ms = self._clock.time_msec() + expiration_ts = now_ms + self._account_validity.period + + if use_delta: + expiration_ts = self.rand.randrange( + expiration_ts - self._account_validity.startup_job_max_delta, + expiration_ts, + ) + + self.db_pool.simple_upsert_txn( + txn, + "account_validity", + keyvalues={"user_id": user_id}, + values={"expiration_ts_ms": expiration_ts, "email_sent": False}, + ) + class RegistrationBackgroundUpdateStore(RegistrationWorkerStore): def __init__(self, database: DatabasePool, db_conn, hs): @@ -911,28 +1000,8 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): def __init__(self, database: DatabasePool, db_conn, hs): super().__init__(database, db_conn, hs) - self._account_validity = hs.config.account_validity self._ignore_unknown_session_error = hs.config.request_token_inhibit_3pid_errors - if self._account_validity.enabled: - self._clock.call_later( - 0.0, - run_as_background_process, - "account_validity_set_expiration_dates", - self._set_expiration_date_when_missing, - ) - - # Create a background job for culling expired 3PID validity tokens - def start_cull(): - # run as a background process to make sure that the database transactions - # have a logcontext to report to - return run_as_background_process( - "cull_expired_threepid_validation_tokens", - self.cull_expired_threepid_validation_tokens, - ) - - hs.get_clock().looping_call(start_cull, THIRTY_MINUTES_IN_MS) - async def add_access_token_to_user( self, user_id: str, @@ -964,6 +1033,36 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): desc="add_access_token_to_user", ) + def _set_device_for_access_token_txn(self, txn, token: str, device_id: str) -> str: + old_device_id = self.db_pool.simple_select_one_onecol_txn( + txn, "access_tokens", {"token": token}, "device_id" + ) + + self.db_pool.simple_update_txn( + txn, "access_tokens", {"token": token}, {"device_id": device_id} + ) + + self._invalidate_cache_and_stream(txn, self.get_user_by_access_token, (token,)) + + return old_device_id + + async def set_device_for_access_token(self, token: str, device_id: str) -> str: + """Sets the device ID associated with an access token. + + Args: + token: The access token to modify. + device_id: The new device ID. + Returns: + The old device ID associated with the access token. + """ + + return await self.db_pool.runInteraction( + "set_device_for_access_token", + self._set_device_for_access_token_txn, + token, + device_id, + ) + async def register_user( self, user_id: str, @@ -1447,22 +1546,6 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): start_or_continue_validation_session_txn, ) - async def cull_expired_threepid_validation_tokens(self) -> None: - """Remove threepid validation tokens with expiry dates that have passed""" - - def cull_expired_threepid_validation_tokens_txn(txn, ts): - sql = """ - DELETE FROM threepid_validation_token WHERE - expires < ? - """ - txn.execute(sql, (ts,)) - - await self.db_pool.runInteraction( - "cull_expired_threepid_validation_tokens", - cull_expired_threepid_validation_tokens_txn, - self.clock.time_msec(), - ) - async def set_user_deactivated_status( self, user_id: str, deactivated: bool ) -> None: @@ -1492,61 +1575,6 @@ class RegistrationStore(RegistrationBackgroundUpdateStore): ) txn.call_after(self.is_guest.invalidate, (user_id,)) - async def _set_expiration_date_when_missing(self): - """ - Retrieves the list of registered users that don't have an expiration date, and - adds an expiration date for each of them. - """ - - def select_users_with_no_expiration_date_txn(txn): - """Retrieves the list of registered users with no expiration date from the - database, filtering out deactivated users. - """ - sql = ( - "SELECT users.name FROM users" - " LEFT JOIN account_validity ON (users.name = account_validity.user_id)" - " WHERE account_validity.user_id is NULL AND users.deactivated = 0;" - ) - txn.execute(sql, []) - - res = self.db_pool.cursor_to_dict(txn) - if res: - for user in res: - self.set_expiration_date_for_user_txn( - txn, user["name"], use_delta=True - ) - - await self.db_pool.runInteraction( - "get_users_with_no_expiration_date", - select_users_with_no_expiration_date_txn, - ) - - def set_expiration_date_for_user_txn(self, txn, user_id, use_delta=False): - """Sets an expiration date to the account with the given user ID. - - Args: - user_id (str): User ID to set an expiration date for. - use_delta (bool): If set to False, the expiration date for the user will be - now + validity period. If set to True, this expiration date will be a - random value in the [now + period - d ; now + period] range, d being a - delta equal to 10% of the validity period. - """ - now_ms = self._clock.time_msec() - expiration_ts = now_ms + self._account_validity.period - - if use_delta: - expiration_ts = self.rand.randrange( - expiration_ts - self._account_validity.startup_job_max_delta, - expiration_ts, - ) - - self.db_pool.simple_upsert_txn( - txn, - "account_validity", - keyvalues={"user_id": user_id}, - values={"expiration_ts_ms": expiration_ts, "email_sent": False}, - ) - def find_max_generated_user_id_localpart(cur: Cursor) -> int: """ diff --git a/synapse/storage/databases/main/room.py b/synapse/storage/databases/main/room.py index 3c7630857f..c0f2af0785 100644 --- a/synapse/storage/databases/main/room.py +++ b/synapse/storage/databases/main/room.py @@ -192,6 +192,18 @@ class RoomWorkerStore(SQLBaseStore): "count_public_rooms", _count_public_rooms_txn ) + async def get_room_count(self) -> int: + """Retrieve the total number of rooms. + """ + + def f(txn): + sql = "SELECT count(*) FROM rooms" + txn.execute(sql) + row = txn.fetchone() + return row[0] or 0 + + return await self.db_pool.runInteraction("get_rooms", f) + async def get_largest_public_rooms( self, network_tuple: Optional[ThirdPartyInstanceID], @@ -1292,18 +1304,6 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore, SearchStore): ) self.hs.get_notifier().on_new_replication_data() - async def get_room_count(self) -> int: - """Retrieve the total number of rooms. - """ - - def f(txn): - sql = "SELECT count(*) FROM rooms" - txn.execute(sql) - row = txn.fetchone() - return row[0] or 0 - - return await self.db_pool.runInteraction("get_rooms", f) - async def add_event_report( self, room_id: str, diff --git a/synapse/storage/databases/main/roommember.py b/synapse/storage/databases/main/roommember.py index 86ffe2479e..20fcdaa529 100644 --- a/synapse/storage/databases/main/roommember.py +++ b/synapse/storage/databases/main/roommember.py @@ -21,12 +21,7 @@ from synapse.events import EventBase from synapse.events.snapshot import EventContext from synapse.metrics import LaterGauge from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.storage._base import ( - LoggingTransaction, - SQLBaseStore, - db_to_json, - make_in_list_sql_clause, -) +from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause from synapse.storage.database import DatabasePool from synapse.storage.databases.main.events_worker import EventsWorkerStore from synapse.storage.engines import Sqlite3Engine @@ -60,15 +55,16 @@ class RoomMemberWorkerStore(EventsWorkerStore): # background update still running? self._current_state_events_membership_up_to_date = False - txn = LoggingTransaction( - db_conn.cursor(), - name="_check_safe_current_state_events_membership_updated", - database_engine=self.database_engine, + txn = db_conn.cursor( + txn_name="_check_safe_current_state_events_membership_updated" ) self._check_safe_current_state_events_membership_updated_txn(txn) txn.close() - if self.hs.config.metrics_flags.known_servers: + if ( + self.hs.config.run_background_tasks + and self.hs.config.metrics_flags.known_servers + ): self._known_servers_count = 1 self.hs.get_clock().looping_call( run_as_background_process, diff --git a/synapse/storage/databases/main/schema/delta/20/pushers.py b/synapse/storage/databases/main/schema/delta/20/pushers.py index 3edfcfd783..45b846e6a7 100644 --- a/synapse/storage/databases/main/schema/delta/20/pushers.py +++ b/synapse/storage/databases/main/schema/delta/20/pushers.py @@ -66,16 +66,15 @@ def run_create(cur, database_engine, *args, **kwargs): row[8] = bytes(row[8]).decode("utf-8") row[11] = bytes(row[11]).decode("utf-8") cur.execute( - database_engine.convert_param_style( - """ - INSERT into pushers2 ( - id, user_name, access_token, profile_tag, kind, - app_id, app_display_name, device_display_name, - pushkey, ts, lang, data, last_token, last_success, - failing_since - ) values (%s)""" - % (",".join(["?" for _ in range(len(row))])) - ), + """ + INSERT into pushers2 ( + id, user_name, access_token, profile_tag, kind, + app_id, app_display_name, device_display_name, + pushkey, ts, lang, data, last_token, last_success, + failing_since + ) values (%s) + """ + % (",".join(["?" for _ in range(len(row))])), row, ) count += 1 diff --git a/synapse/storage/databases/main/schema/delta/25/fts.py b/synapse/storage/databases/main/schema/delta/25/fts.py index ee675e71ff..21f57825d4 100644 --- a/synapse/storage/databases/main/schema/delta/25/fts.py +++ b/synapse/storage/databases/main/schema/delta/25/fts.py @@ -71,8 +71,6 @@ def run_create(cur, database_engine, *args, **kwargs): " VALUES (?, ?)" ) - sql = database_engine.convert_param_style(sql) - cur.execute(sql, ("event_search", progress_json)) diff --git a/synapse/storage/databases/main/schema/delta/27/ts.py b/synapse/storage/databases/main/schema/delta/27/ts.py index b7972cfa8e..1c6058063f 100644 --- a/synapse/storage/databases/main/schema/delta/27/ts.py +++ b/synapse/storage/databases/main/schema/delta/27/ts.py @@ -50,8 +50,6 @@ def run_create(cur, database_engine, *args, **kwargs): " VALUES (?, ?)" ) - sql = database_engine.convert_param_style(sql) - cur.execute(sql, ("event_origin_server_ts", progress_json)) diff --git a/synapse/storage/databases/main/schema/delta/30/as_users.py b/synapse/storage/databases/main/schema/delta/30/as_users.py index b42c02710a..7f08fabe9f 100644 --- a/synapse/storage/databases/main/schema/delta/30/as_users.py +++ b/synapse/storage/databases/main/schema/delta/30/as_users.py @@ -59,9 +59,7 @@ def run_upgrade(cur, database_engine, config, *args, **kwargs): user_chunks = (user_ids[i : i + 100] for i in range(0, len(user_ids), n)) for chunk in user_chunks: cur.execute( - database_engine.convert_param_style( - "UPDATE users SET appservice_id = ? WHERE name IN (%s)" - % (",".join("?" for _ in chunk),) - ), + "UPDATE users SET appservice_id = ? WHERE name IN (%s)" + % (",".join("?" for _ in chunk),), [as_id] + chunk, ) diff --git a/synapse/storage/databases/main/schema/delta/31/pushers.py b/synapse/storage/databases/main/schema/delta/31/pushers.py index 9bb504aad5..5be81c806a 100644 --- a/synapse/storage/databases/main/schema/delta/31/pushers.py +++ b/synapse/storage/databases/main/schema/delta/31/pushers.py @@ -65,16 +65,15 @@ def run_create(cur, database_engine, *args, **kwargs): row = list(row) row[12] = token_to_stream_ordering(row[12]) cur.execute( - database_engine.convert_param_style( - """ - INSERT into pushers2 ( - id, user_name, access_token, profile_tag, kind, - app_id, app_display_name, device_display_name, - pushkey, ts, lang, data, last_stream_ordering, last_success, - failing_since - ) values (%s)""" - % (",".join(["?" for _ in range(len(row))])) - ), + """ + INSERT into pushers2 ( + id, user_name, access_token, profile_tag, kind, + app_id, app_display_name, device_display_name, + pushkey, ts, lang, data, last_stream_ordering, last_success, + failing_since + ) values (%s) + """ + % (",".join(["?" for _ in range(len(row))])), row, ) count += 1 diff --git a/synapse/storage/databases/main/schema/delta/31/search_update.py b/synapse/storage/databases/main/schema/delta/31/search_update.py index 63b757ade6..b84c844e3a 100644 --- a/synapse/storage/databases/main/schema/delta/31/search_update.py +++ b/synapse/storage/databases/main/schema/delta/31/search_update.py @@ -55,8 +55,6 @@ def run_create(cur, database_engine, *args, **kwargs): " VALUES (?, ?)" ) - sql = database_engine.convert_param_style(sql) - cur.execute(sql, ("event_search_order", progress_json)) diff --git a/synapse/storage/databases/main/schema/delta/33/event_fields.py b/synapse/storage/databases/main/schema/delta/33/event_fields.py index a3e81eeac7..e928c66a8f 100644 --- a/synapse/storage/databases/main/schema/delta/33/event_fields.py +++ b/synapse/storage/databases/main/schema/delta/33/event_fields.py @@ -50,8 +50,6 @@ def run_create(cur, database_engine, *args, **kwargs): " VALUES (?, ?)" ) - sql = database_engine.convert_param_style(sql) - cur.execute(sql, ("event_fields_sender_url", progress_json)) diff --git a/synapse/storage/databases/main/schema/delta/33/remote_media_ts.py b/synapse/storage/databases/main/schema/delta/33/remote_media_ts.py index a26057dfb6..ad875c733a 100644 --- a/synapse/storage/databases/main/schema/delta/33/remote_media_ts.py +++ b/synapse/storage/databases/main/schema/delta/33/remote_media_ts.py @@ -23,8 +23,5 @@ def run_create(cur, database_engine, *args, **kwargs): def run_upgrade(cur, database_engine, *args, **kwargs): cur.execute( - database_engine.convert_param_style( - "UPDATE remote_media_cache SET last_access_ts = ?" - ), - (int(time.time() * 1000),), + "UPDATE remote_media_cache SET last_access_ts = ?", (int(time.time() * 1000),), ) diff --git a/synapse/storage/databases/main/schema/delta/56/unique_user_filter_index.py b/synapse/storage/databases/main/schema/delta/56/unique_user_filter_index.py index 1de8b54961..bb7296852a 100644 --- a/synapse/storage/databases/main/schema/delta/56/unique_user_filter_index.py +++ b/synapse/storage/databases/main/schema/delta/56/unique_user_filter_index.py @@ -1,6 +1,8 @@ import logging +from io import StringIO from synapse.storage.engines import PostgresEngine +from synapse.storage.prepare_database import execute_statements_from_stream logger = logging.getLogger(__name__) @@ -46,7 +48,4 @@ def run_create(cur, database_engine, *args, **kwargs): select_clause, ) - if isinstance(database_engine, PostgresEngine): - cur.execute(sql) - else: - cur.executescript(sql) + execute_statements_from_stream(cur, StringIO(sql)) diff --git a/synapse/storage/databases/main/schema/delta/57/local_current_membership.py b/synapse/storage/databases/main/schema/delta/57/local_current_membership.py index 63b5acdcf7..44917f0a2e 100644 --- a/synapse/storage/databases/main/schema/delta/57/local_current_membership.py +++ b/synapse/storage/databases/main/schema/delta/57/local_current_membership.py @@ -68,7 +68,6 @@ def run_upgrade(cur, database_engine, config, *args, **kwargs): INNER JOIN room_memberships AS r USING (event_id) WHERE type = 'm.room.member' AND state_key LIKE ? """ - sql = database_engine.convert_param_style(sql) cur.execute(sql, ("%:" + config.server_name,)) cur.execute( diff --git a/synapse/storage/databases/main/schema/delta/58/11dehydration.sql b/synapse/storage/databases/main/schema/delta/58/11dehydration.sql new file mode 100644 index 0000000000..7851a0a825 --- /dev/null +++ b/synapse/storage/databases/main/schema/delta/58/11dehydration.sql @@ -0,0 +1,20 @@ +/* Copyright 2020 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +CREATE TABLE IF NOT EXISTS dehydrated_devices( + user_id TEXT NOT NULL PRIMARY KEY, + device_id TEXT NOT NULL, + device_data TEXT NOT NULL -- JSON-encoded client-defined data +); diff --git a/synapse/storage/databases/main/schema/delta/58/11fallback.sql b/synapse/storage/databases/main/schema/delta/58/11fallback.sql new file mode 100644 index 0000000000..4ed981dbf8 --- /dev/null +++ b/synapse/storage/databases/main/schema/delta/58/11fallback.sql @@ -0,0 +1,24 @@ +/* Copyright 2020 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +CREATE TABLE IF NOT EXISTS e2e_fallback_keys_json ( + user_id TEXT NOT NULL, -- The user this fallback key is for. + device_id TEXT NOT NULL, -- The device this fallback key is for. + algorithm TEXT NOT NULL, -- Which algorithm this fallback key is for. + key_id TEXT NOT NULL, -- An id for suppressing duplicate uploads. + key_json TEXT NOT NULL, -- The key as a JSON blob. + used BOOLEAN NOT NULL DEFAULT FALSE, -- Whether the key has been used or not. + CONSTRAINT e2e_fallback_keys_json_uniqueness UNIQUE (user_id, device_id, algorithm) +); diff --git a/synapse/storage/databases/main/schema/delta/58/19instance_map.sql.postgres b/synapse/storage/databases/main/schema/delta/58/19instance_map.sql.postgres new file mode 100644 index 0000000000..841186b826 --- /dev/null +++ b/synapse/storage/databases/main/schema/delta/58/19instance_map.sql.postgres @@ -0,0 +1,25 @@ +/* Copyright 2020 The Matrix.org Foundation C.I.C + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + + +-- A unique and immutable mapping between instance name and an integer ID. This +-- lets us refer to instances via a small ID in e.g. stream tokens, without +-- having to encode the full name. +CREATE TABLE IF NOT EXISTS instance_map ( + instance_id SERIAL PRIMARY KEY, + instance_name TEXT NOT NULL +); + +CREATE UNIQUE INDEX IF NOT EXISTS instance_map_idx ON instance_map(instance_name); diff --git a/synapse/storage/databases/main/stream.py b/synapse/storage/databases/main/stream.py index 37249f1e3f..e3b9ff5ca6 100644 --- a/synapse/storage/databases/main/stream.py +++ b/synapse/storage/databases/main/stream.py @@ -53,7 +53,9 @@ from synapse.storage.database import ( ) from synapse.storage.databases.main.events_worker import EventsWorkerStore from synapse.storage.engines import BaseDatabaseEngine, PostgresEngine +from synapse.storage.util.id_generators import MultiWriterIdGenerator from synapse.types import Collection, PersistedEventPosition, RoomStreamToken +from synapse.util.caches.descriptors import cached from synapse.util.caches.stream_change_cache import StreamChangeCache if TYPE_CHECKING: @@ -208,6 +210,55 @@ def _make_generic_sql_bound( ) +def _filter_results( + lower_token: Optional[RoomStreamToken], + upper_token: Optional[RoomStreamToken], + instance_name: str, + topological_ordering: int, + stream_ordering: int, +) -> bool: + """Returns True if the event persisted by the given instance at the given + topological/stream_ordering falls between the two tokens (taking a None + token to mean unbounded). + + Used to filter results from fetching events in the DB against the given + tokens. This is necessary to handle the case where the tokens include + position maps, which we handle by fetching more than necessary from the DB + and then filtering (rather than attempting to construct a complicated SQL + query). + """ + + event_historical_tuple = ( + topological_ordering, + stream_ordering, + ) + + if lower_token: + if lower_token.topological is not None: + # If these are historical tokens we compare the `(topological, stream)` + # tuples. + if event_historical_tuple <= lower_token.as_historical_tuple(): + return False + + else: + # If these are live tokens we compare the stream ordering against the + # writers stream position. + if stream_ordering <= lower_token.get_stream_pos_for_instance( + instance_name + ): + return False + + if upper_token: + if upper_token.topological is not None: + if upper_token.as_historical_tuple() < event_historical_tuple: + return False + else: + if upper_token.get_stream_pos_for_instance(instance_name) < stream_ordering: + return False + + return True + + def filter_to_clause(event_filter: Optional[Filter]) -> Tuple[str, List[str]]: # NB: This may create SQL clauses that don't optimise well (and we don't # have indices on all possible clauses). E.g. it may create @@ -305,7 +356,31 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): raise NotImplementedError() def get_room_max_token(self) -> RoomStreamToken: - return RoomStreamToken(None, self.get_room_max_stream_ordering()) + """Get a `RoomStreamToken` that marks the current maximum persisted + position of the events stream. Useful to get a token that represents + "now". + + The token returned is a "live" token that may have an instance_map + component. + """ + + min_pos = self._stream_id_gen.get_current_token() + + positions = {} + if isinstance(self._stream_id_gen, MultiWriterIdGenerator): + # The `min_pos` is the minimum position that we know all instances + # have finished persisting to, so we only care about instances whose + # positions are ahead of that. (Instance positions can be behind the + # min position as there are times we can work out that the minimum + # position is ahead of the naive minimum across all current + # positions. See MultiWriterIdGenerator for details) + positions = { + i: p + for i, p in self._stream_id_gen.get_positions().items() + if p > min_pos + } + + return RoomStreamToken(None, min_pos, positions) async def get_room_events_stream_for_rooms( self, @@ -404,25 +479,43 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): if from_key == to_key: return [], from_key - from_id = from_key.stream - to_id = to_key.stream - - has_changed = self._events_stream_cache.has_entity_changed(room_id, from_id) + has_changed = self._events_stream_cache.has_entity_changed( + room_id, from_key.stream + ) if not has_changed: return [], from_key def f(txn): - sql = ( - "SELECT event_id, stream_ordering FROM events WHERE" - " room_id = ?" - " AND not outlier" - " AND stream_ordering > ? AND stream_ordering <= ?" - " ORDER BY stream_ordering %s LIMIT ?" - ) % (order,) - txn.execute(sql, (room_id, from_id, to_id, limit)) - - rows = [_EventDictReturn(row[0], None, row[1]) for row in txn] + # To handle tokens with a non-empty instance_map we fetch more + # results than necessary and then filter down + min_from_id = from_key.stream + max_to_id = to_key.get_max_stream_pos() + + sql = """ + SELECT event_id, instance_name, topological_ordering, stream_ordering + FROM events + WHERE + room_id = ? + AND not outlier + AND stream_ordering > ? AND stream_ordering <= ? + ORDER BY stream_ordering %s LIMIT ? + """ % ( + order, + ) + txn.execute(sql, (room_id, min_from_id, max_to_id, 2 * limit)) + + rows = [ + _EventDictReturn(event_id, None, stream_ordering) + for event_id, instance_name, topological_ordering, stream_ordering in txn + if _filter_results( + from_key, + to_key, + instance_name, + topological_ordering, + stream_ordering, + ) + ][:limit] return rows rows = await self.db_pool.runInteraction("get_room_events_stream_for_room", f) @@ -431,7 +524,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): [r.event_id for r in rows], get_prev_content=True ) - self._set_before_and_after(ret, rows, topo_order=from_id is None) + self._set_before_and_after(ret, rows, topo_order=False) if order.lower() == "desc": ret.reverse() @@ -448,31 +541,43 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): async def get_membership_changes_for_user( self, user_id: str, from_key: RoomStreamToken, to_key: RoomStreamToken ) -> List[EventBase]: - from_id = from_key.stream - to_id = to_key.stream - if from_key == to_key: return [] - if from_id: + if from_key: has_changed = self._membership_stream_cache.has_entity_changed( - user_id, int(from_id) + user_id, int(from_key.stream) ) if not has_changed: return [] def f(txn): - sql = ( - "SELECT m.event_id, stream_ordering FROM events AS e," - " room_memberships AS m" - " WHERE e.event_id = m.event_id" - " AND m.user_id = ?" - " AND e.stream_ordering > ? AND e.stream_ordering <= ?" - " ORDER BY e.stream_ordering ASC" - ) - txn.execute(sql, (user_id, from_id, to_id)) - - rows = [_EventDictReturn(row[0], None, row[1]) for row in txn] + # To handle tokens with a non-empty instance_map we fetch more + # results than necessary and then filter down + min_from_id = from_key.stream + max_to_id = to_key.get_max_stream_pos() + + sql = """ + SELECT m.event_id, instance_name, topological_ordering, stream_ordering + FROM events AS e, room_memberships AS m + WHERE e.event_id = m.event_id + AND m.user_id = ? + AND e.stream_ordering > ? AND e.stream_ordering <= ? + ORDER BY e.stream_ordering ASC + """ + txn.execute(sql, (user_id, min_from_id, max_to_id,)) + + rows = [ + _EventDictReturn(event_id, None, stream_ordering) + for event_id, instance_name, topological_ordering, stream_ordering in txn + if _filter_results( + from_key, + to_key, + instance_name, + topological_ordering, + stream_ordering, + ) + ] return rows @@ -546,7 +651,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): async def get_room_event_before_stream_ordering( self, room_id: str, stream_ordering: int - ) -> Tuple[int, int, str]: + ) -> Optional[Tuple[int, int, str]]: """Gets details of the first event in a room at or before a stream ordering Args: @@ -589,19 +694,6 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): ) return "t%d-%d" % (topo, token) - async def get_stream_id_for_event(self, event_id: str) -> int: - """The stream ID for an event - Args: - event_id: The id of the event to look up a stream token for. - Raises: - StoreError if the event wasn't in the database. - Returns: - A stream ID. - """ - return await self.db_pool.runInteraction( - "get_stream_id_for_event", self.get_stream_id_for_event_txn, event_id, - ) - def get_stream_id_for_event_txn( self, txn: LoggingTransaction, event_id: str, allow_none=False, ) -> int: @@ -979,11 +1071,46 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): else: order = "ASC" + # The bounds for the stream tokens are complicated by the fact + # that we need to handle the instance_map part of the tokens. We do this + # by fetching all events between the min stream token and the maximum + # stream token (as returned by `RoomStreamToken.get_max_stream_pos`) and + # then filtering the results. + if from_token.topological is not None: + from_bound = ( + from_token.as_historical_tuple() + ) # type: Tuple[Optional[int], int] + elif direction == "b": + from_bound = ( + None, + from_token.get_max_stream_pos(), + ) + else: + from_bound = ( + None, + from_token.stream, + ) + + to_bound = None # type: Optional[Tuple[Optional[int], int]] + if to_token: + if to_token.topological is not None: + to_bound = to_token.as_historical_tuple() + elif direction == "b": + to_bound = ( + None, + to_token.stream, + ) + else: + to_bound = ( + None, + to_token.get_max_stream_pos(), + ) + bounds = generate_pagination_where_clause( direction=direction, column_names=("topological_ordering", "stream_ordering"), - from_token=from_token.as_tuple(), - to_token=to_token.as_tuple() if to_token else None, + from_token=from_bound, + to_token=to_bound, engine=self.database_engine, ) @@ -993,7 +1120,8 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): bounds += " AND " + filter_clause args.extend(filter_args) - args.append(int(limit)) + # We fetch more events as we'll filter the result set + args.append(int(limit) * 2) select_keywords = "SELECT" join_clause = "" @@ -1015,7 +1143,9 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): select_keywords += "DISTINCT" sql = """ - %(select_keywords)s event_id, topological_ordering, stream_ordering + %(select_keywords)s + event_id, instance_name, + topological_ordering, stream_ordering FROM events %(join_clause)s WHERE outlier = ? AND room_id = ? AND %(bounds)s @@ -1030,7 +1160,18 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): txn.execute(sql, args) - rows = [_EventDictReturn(row[0], row[1], row[2]) for row in txn] + # Filter the result set. + rows = [ + _EventDictReturn(event_id, topological_ordering, stream_ordering) + for event_id, instance_name, topological_ordering, stream_ordering in txn + if _filter_results( + lower_token=to_token if direction == "b" else from_token, + upper_token=from_token if direction == "b" else to_token, + instance_name=instance_name, + topological_ordering=topological_ordering, + stream_ordering=stream_ordering, + ) + ][:limit] if rows: topo = rows[-1].topological_ordering @@ -1095,6 +1236,58 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): return (events, token) + @cached() + async def get_id_for_instance(self, instance_name: str) -> int: + """Get a unique, immutable ID that corresponds to the given Synapse worker instance. + """ + + def _get_id_for_instance_txn(txn): + instance_id = self.db_pool.simple_select_one_onecol_txn( + txn, + table="instance_map", + keyvalues={"instance_name": instance_name}, + retcol="instance_id", + allow_none=True, + ) + if instance_id is not None: + return instance_id + + # If we don't have an entry upsert one. + # + # We could do this before the first check, and rely on the cache for + # efficiency, but each UPSERT causes the next ID to increment which + # can quickly bloat the size of the generated IDs for new instances. + self.db_pool.simple_upsert_txn( + txn, + table="instance_map", + keyvalues={"instance_name": instance_name}, + values={}, + ) + + return self.db_pool.simple_select_one_onecol_txn( + txn, + table="instance_map", + keyvalues={"instance_name": instance_name}, + retcol="instance_id", + ) + + return await self.db_pool.runInteraction( + "get_id_for_instance", _get_id_for_instance_txn + ) + + @cached() + async def get_name_from_instance_id(self, instance_id: int) -> str: + """Get the instance name from an ID previously returned by + `get_id_for_instance`. + """ + + return await self.db_pool.simple_select_one_onecol( + table="instance_map", + keyvalues={"instance_id": instance_id}, + retcol="instance_name", + desc="get_name_from_instance_id", + ) + class StreamStore(StreamWorkerStore): def get_room_max_stream_ordering(self) -> int: diff --git a/synapse/storage/databases/main/transactions.py b/synapse/storage/databases/main/transactions.py index 97aed1500e..7d46090267 100644 --- a/synapse/storage/databases/main/transactions.py +++ b/synapse/storage/databases/main/transactions.py @@ -19,7 +19,7 @@ from typing import Iterable, List, Optional, Tuple from canonicaljson import encode_canonical_json -from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.metrics.background_process_metrics import wrap_as_background_process from synapse.storage._base import SQLBaseStore, db_to_json from synapse.storage.database import DatabasePool, LoggingTransaction from synapse.storage.engines import PostgresEngine, Sqlite3Engine @@ -43,15 +43,33 @@ _UpdateTransactionRow = namedtuple( SENTINEL = object() -class TransactionStore(SQLBaseStore): +class TransactionWorkerStore(SQLBaseStore): + def __init__(self, database: DatabasePool, db_conn, hs): + super().__init__(database, db_conn, hs) + + if hs.config.run_background_tasks: + self._clock.looping_call(self._cleanup_transactions, 30 * 60 * 1000) + + @wrap_as_background_process("cleanup_transactions") + async def _cleanup_transactions(self) -> None: + now = self._clock.time_msec() + month_ago = now - 30 * 24 * 60 * 60 * 1000 + + def _cleanup_transactions_txn(txn): + txn.execute("DELETE FROM received_transactions WHERE ts < ?", (month_ago,)) + + await self.db_pool.runInteraction( + "_cleanup_transactions", _cleanup_transactions_txn + ) + + +class TransactionStore(TransactionWorkerStore): """A collection of queries for handling PDUs. """ def __init__(self, database: DatabasePool, db_conn, hs): super().__init__(database, db_conn, hs) - self._clock.looping_call(self._start_cleanup_transactions, 30 * 60 * 1000) - self._destination_retry_cache = ExpiringCache( cache_name="get_destination_retry_timings", clock=self._clock, @@ -266,22 +284,6 @@ class TransactionStore(SQLBaseStore): }, ) - def _start_cleanup_transactions(self): - return run_as_background_process( - "cleanup_transactions", self._cleanup_transactions - ) - - async def _cleanup_transactions(self) -> None: - now = self._clock.time_msec() - month_ago = now - 30 * 24 * 60 * 60 * 1000 - - def _cleanup_transactions_txn(txn): - txn.execute("DELETE FROM received_transactions WHERE ts < ?", (month_ago,)) - - await self.db_pool.runInteraction( - "_cleanup_transactions", _cleanup_transactions_txn - ) - async def store_destination_rooms_entries( self, destinations: Iterable[str], room_id: str, stream_ordering: int, ) -> None: diff --git a/synapse/storage/databases/main/ui_auth.py b/synapse/storage/databases/main/ui_auth.py index 3b9211a6d2..79b7ece330 100644 --- a/synapse/storage/databases/main/ui_auth.py +++ b/synapse/storage/databases/main/ui_auth.py @@ -288,8 +288,6 @@ class UIAuthWorkerStore(SQLBaseStore): ) return [(row["user_agent"], row["ip"]) for row in rows] - -class UIAuthStore(UIAuthWorkerStore): async def delete_old_ui_auth_sessions(self, expiration_time: int) -> None: """ Remove sessions which were last used earlier than the expiration time. @@ -339,3 +337,7 @@ class UIAuthStore(UIAuthWorkerStore): iterable=session_ids, keyvalues={}, ) + + +class UIAuthStore(UIAuthWorkerStore): + pass diff --git a/synapse/storage/persist_events.py b/synapse/storage/persist_events.py index 72939f3984..4d2d88d1f0 100644 --- a/synapse/storage/persist_events.py +++ b/synapse/storage/persist_events.py @@ -248,6 +248,8 @@ class EventsPersistenceStorage: await make_deferred_yieldable(deferred) event_stream_id = event.internal_metadata.stream_ordering + # stream ordering should have been assigned by now + assert event_stream_id pos = PersistedEventPosition(self._instance_name, event_stream_id) return pos, self.main_store.get_room_max_token() diff --git a/synapse/storage/prepare_database.py b/synapse/storage/prepare_database.py index 4957e77f4c..459754feab 100644 --- a/synapse/storage/prepare_database.py +++ b/synapse/storage/prepare_database.py @@ -13,7 +13,6 @@ # 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 imp import logging import os @@ -24,9 +23,10 @@ from typing import Optional, TextIO import attr from synapse.config.homeserver import HomeServerConfig +from synapse.storage.database import LoggingDatabaseConnection from synapse.storage.engines import BaseDatabaseEngine from synapse.storage.engines.postgres import PostgresEngine -from synapse.storage.types import Connection, Cursor +from synapse.storage.types import Cursor from synapse.types import Collection logger = logging.getLogger(__name__) @@ -67,7 +67,7 @@ UNAPPLIED_DELTA_ON_WORKER_ERROR = ( def prepare_database( - db_conn: Connection, + db_conn: LoggingDatabaseConnection, database_engine: BaseDatabaseEngine, config: Optional[HomeServerConfig], databases: Collection[str] = ["main", "state"], @@ -89,7 +89,7 @@ def prepare_database( """ try: - cur = db_conn.cursor() + cur = db_conn.cursor(txn_name="prepare_database") # sqlite does not automatically start transactions for DDL / SELECT statements, # so we start one before running anything. This ensures that any upgrades @@ -258,9 +258,7 @@ def _setup_new_database(cur, database_engine, databases): executescript(cur, entry.absolute_path) cur.execute( - database_engine.convert_param_style( - "INSERT INTO schema_version (version, upgraded) VALUES (?,?)" - ), + "INSERT INTO schema_version (version, upgraded) VALUES (?,?)", (max_current_ver, False), ) @@ -486,17 +484,13 @@ def _upgrade_existing_database( # Mark as done. cur.execute( - database_engine.convert_param_style( - "INSERT INTO applied_schema_deltas (version, file) VALUES (?,?)" - ), + "INSERT INTO applied_schema_deltas (version, file) VALUES (?,?)", (v, relative_path), ) cur.execute("DELETE FROM schema_version") cur.execute( - database_engine.convert_param_style( - "INSERT INTO schema_version (version, upgraded) VALUES (?,?)" - ), + "INSERT INTO schema_version (version, upgraded) VALUES (?,?)", (v, True), ) @@ -532,10 +526,7 @@ def _apply_module_schema_files(cur, database_engine, modname, names_and_streams) schemas to be applied """ cur.execute( - database_engine.convert_param_style( - "SELECT file FROM applied_module_schemas WHERE module_name = ?" - ), - (modname,), + "SELECT file FROM applied_module_schemas WHERE module_name = ?", (modname,), ) applied_deltas = {d for d, in cur} for (name, stream) in names_and_streams: @@ -553,9 +544,7 @@ def _apply_module_schema_files(cur, database_engine, modname, names_and_streams) # Mark as done. cur.execute( - database_engine.convert_param_style( - "INSERT INTO applied_module_schemas (module_name, file) VALUES (?,?)" - ), + "INSERT INTO applied_module_schemas (module_name, file) VALUES (?,?)", (modname, name), ) @@ -627,9 +616,7 @@ def _get_or_create_schema_state(txn, database_engine): if current_version: txn.execute( - database_engine.convert_param_style( - "SELECT file FROM applied_schema_deltas WHERE version >= ?" - ), + "SELECT file FROM applied_schema_deltas WHERE version >= ?", (current_version,), ) applied_deltas = [d for d, in txn] diff --git a/synapse/storage/types.py b/synapse/storage/types.py index 2d2b560e74..970bb1b9da 100644 --- a/synapse/storage/types.py +++ b/synapse/storage/types.py @@ -61,3 +61,9 @@ class Connection(Protocol): def rollback(self, *args, **kwargs) -> None: ... + + def __enter__(self) -> "Connection": + ... + + def __exit__(self, exc_type, exc_value, traceback) -> bool: + ... diff --git a/synapse/storage/util/id_generators.py b/synapse/storage/util/id_generators.py index ad017207aa..d7e40aaa8b 100644 --- a/synapse/storage/util/id_generators.py +++ b/synapse/storage/util/id_generators.py @@ -55,7 +55,7 @@ def _load_current_id(db_conn, table, column, step=1): """ # debug logging for https://github.com/matrix-org/synapse/issues/7968 logger.info("initialising stream generator for %s(%s)", table, column) - cur = db_conn.cursor() + cur = db_conn.cursor(txn_name="_load_current_id") if step == 1: cur.execute("SELECT MAX(%s) FROM %s" % (column, table)) else: @@ -270,7 +270,7 @@ class MultiWriterIdGenerator: def _load_current_ids( self, db_conn, table: str, instance_column: str, id_column: str ): - cur = db_conn.cursor() + cur = db_conn.cursor(txn_name="_load_current_ids") # Load the current positions of all writers for the stream. if self._writers: @@ -284,15 +284,12 @@ class MultiWriterIdGenerator: stream_name = ? AND instance_name != ALL(?) """ - sql = self._db.engine.convert_param_style(sql) cur.execute(sql, (self._stream_name, self._writers)) sql = """ SELECT instance_name, stream_id FROM stream_positions WHERE stream_name = ? """ - sql = self._db.engine.convert_param_style(sql) - cur.execute(sql, (self._stream_name,)) self._current_positions = { @@ -341,7 +338,6 @@ class MultiWriterIdGenerator: "instance": instance_column, "cmp": "<=" if self._positive else ">=", } - sql = self._db.engine.convert_param_style(sql) cur.execute(sql, (min_stream_id * self._return_factor,)) self._persisted_upto_position = min_stream_id @@ -422,7 +418,7 @@ class MultiWriterIdGenerator: self._unfinished_ids.discard(next_id) self._finished_ids.add(next_id) - new_cur = None + new_cur = None # type: Optional[int] if self._unfinished_ids: # If there are unfinished IDs then the new position will be the diff --git a/synapse/storage/util/sequence.py b/synapse/storage/util/sequence.py index 2dd95e2709..ff2d038ad2 100644 --- a/synapse/storage/util/sequence.py +++ b/synapse/storage/util/sequence.py @@ -17,6 +17,7 @@ import logging import threading from typing import Callable, List, Optional +from synapse.storage.database import LoggingDatabaseConnection from synapse.storage.engines import ( BaseDatabaseEngine, IncorrectDatabaseSetup, @@ -53,7 +54,11 @@ class SequenceGenerator(metaclass=abc.ABCMeta): @abc.abstractmethod def check_consistency( - self, db_conn: Connection, table: str, id_column: str, positive: bool = True + self, + db_conn: LoggingDatabaseConnection, + table: str, + id_column: str, + positive: bool = True, ): """Should be called during start up to test that the current value of the sequence is greater than or equal to the maximum ID in the table. @@ -82,9 +87,13 @@ class PostgresSequenceGenerator(SequenceGenerator): return [i for (i,) in txn] def check_consistency( - self, db_conn: Connection, table: str, id_column: str, positive: bool = True + self, + db_conn: LoggingDatabaseConnection, + table: str, + id_column: str, + positive: bool = True, ): - txn = db_conn.cursor() + txn = db_conn.cursor(txn_name="sequence.check_consistency") # First we get the current max ID from the table. table_sql = "SELECT GREATEST(%(agg)s(%(id)s), 0) FROM %(table)s" % { diff --git a/synapse/types.py b/synapse/types.py index bd271f9f16..5bde67cc07 100644 --- a/synapse/types.py +++ b/synapse/types.py @@ -22,6 +22,7 @@ from typing import ( TYPE_CHECKING, Any, Dict, + Iterable, Mapping, MutableMapping, Optional, @@ -43,7 +44,7 @@ if TYPE_CHECKING: if sys.version_info[:3] >= (3, 6, 0): from typing import Collection else: - from typing import Container, Iterable, Sized + from typing import Container, Sized T_co = TypeVar("T_co", covariant=True) @@ -375,7 +376,7 @@ def map_username_to_mxid_localpart(username, case_sensitive=False): return username.decode("ascii") -@attr.s(frozen=True, slots=True) +@attr.s(frozen=True, slots=True, cmp=False) class RoomStreamToken: """Tokens are positions between events. The token "s1" comes after event 1. @@ -397,6 +398,31 @@ class RoomStreamToken: event it comes after. Historic tokens start with a "t" followed by the "topological_ordering" id of the event it comes after, followed by "-", followed by the "stream_ordering" id of the event it comes after. + + There is also a third mode for live tokens where the token starts with "m", + which is sometimes used when using sharded event persisters. In this case + the events stream is considered to be a set of streams (one for each writer) + and the token encodes the vector clock of positions of each writer in their + respective streams. + + The format of the token in such case is an initial integer min position, + followed by the mapping of instance ID to position separated by '.' and '~': + + m{min_pos}~{writer1}.{pos1}~{writer2}.{pos2}. ... + + The `min_pos` corresponds to the minimum position all writers have persisted + up to, and then only writers that are ahead of that position need to be + encoded. An example token is: + + m56~2.58~3.59 + + Which corresponds to a set of three (or more writers) where instances 2 and + 3 (these are instance IDs that can be looked up in the DB to fetch the more + commonly used instance names) are at positions 58 and 59 respectively, and + all other instances are at position 56. + + Note: The `RoomStreamToken` cannot have both a topological part and an + instance map. """ topological = attr.ib( @@ -405,6 +431,25 @@ class RoomStreamToken: ) stream = attr.ib(type=int, validator=attr.validators.instance_of(int)) + instance_map = attr.ib( + type=Dict[str, int], + factory=dict, + validator=attr.validators.deep_mapping( + key_validator=attr.validators.instance_of(str), + value_validator=attr.validators.instance_of(int), + mapping_validator=attr.validators.instance_of(dict), + ), + ) + + def __attrs_post_init__(self): + """Validates that both `topological` and `instance_map` aren't set. + """ + + if self.instance_map and self.topological: + raise ValueError( + "Cannot set both 'topological' and 'instance_map' on 'RoomStreamToken'." + ) + @classmethod async def parse(cls, store: "DataStore", string: str) -> "RoomStreamToken": try: @@ -413,6 +458,20 @@ class RoomStreamToken: if string[0] == "t": parts = string[1:].split("-", 1) return cls(topological=int(parts[0]), stream=int(parts[1])) + if string[0] == "m": + parts = string[1:].split("~") + stream = int(parts[0]) + + instance_map = {} + for part in parts[1:]: + key, value = part.split(".") + instance_id = int(key) + pos = int(value) + + instance_name = await store.get_name_from_instance_id(instance_id) + instance_map[instance_name] = pos + + return cls(topological=None, stream=stream, instance_map=instance_map,) except Exception: pass raise SynapseError(400, "Invalid token %r" % (string,)) @@ -436,14 +495,61 @@ class RoomStreamToken: max_stream = max(self.stream, other.stream) - return RoomStreamToken(None, max_stream) + instance_map = { + instance: max( + self.instance_map.get(instance, self.stream), + other.instance_map.get(instance, other.stream), + ) + for instance in set(self.instance_map).union(other.instance_map) + } + + return RoomStreamToken(None, max_stream, instance_map) + + def as_historical_tuple(self) -> Tuple[int, int]: + """Returns a tuple of `(topological, stream)` for historical tokens. + + Raises if not an historical token (i.e. doesn't have a topological part). + """ + if self.topological is None: + raise Exception( + "Cannot call `RoomStreamToken.as_historical_tuple` on live token" + ) - def as_tuple(self) -> Tuple[Optional[int], int]: return (self.topological, self.stream) + def get_stream_pos_for_instance(self, instance_name: str) -> int: + """Get the stream position that the given writer was at at this token. + + This only makes sense for "live" tokens that may have a vector clock + component, and so asserts that this is a "live" token. + """ + assert self.topological is None + + # If we don't have an entry for the instance we can assume that it was + # at `self.stream`. + return self.instance_map.get(instance_name, self.stream) + + def get_max_stream_pos(self) -> int: + """Get the maximum stream position referenced in this token. + + The corresponding "min" position is, by definition just `self.stream`. + + This is used to handle tokens that have non-empty `instance_map`, and so + reference stream positions after the `self.stream` position. + """ + return max(self.instance_map.values(), default=self.stream) + async def to_string(self, store: "DataStore") -> str: if self.topological is not None: return "t%d-%d" % (self.topological, self.stream) + elif self.instance_map: + entries = [] + for name, pos in self.instance_map.items(): + instance_id = await store.get_id_for_instance(name) + entries.append("{}.{}".format(instance_id, pos)) + + encoded_map = "~".join(entries) + return "m{}~{}".format(self.stream, encoded_map) else: return "s%d" % (self.stream,) @@ -535,7 +641,7 @@ class PersistedEventPosition: stream = attr.ib(type=int) def persisted_after(self, token: RoomStreamToken) -> bool: - return token.stream < self.stream + return token.get_stream_pos_for_instance(self.instance_name) < self.stream def to_room_stream_token(self) -> RoomStreamToken: """Converts the position to a room stream token such that events diff --git a/synapse/visibility.py b/synapse/visibility.py index e3da7744d2..527365498e 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -16,7 +16,7 @@ import logging import operator -from synapse.api.constants import EventTypes, Membership +from synapse.api.constants import AccountDataTypes, EventTypes, Membership from synapse.events.utils import prune_event from synapse.storage import Storage from synapse.storage.state import StateFilter @@ -77,15 +77,14 @@ async def filter_events_for_client( ) ignore_dict_content = await storage.main.get_global_account_data_by_type_for_user( - "m.ignored_user_list", user_id + AccountDataTypes.IGNORED_USER_LIST, user_id ) - # FIXME: This will explode if people upload something incorrect. - ignore_list = frozenset( - ignore_dict_content.get("ignored_users", {}).keys() - if ignore_dict_content - else [] - ) + ignore_list = frozenset() + if ignore_dict_content: + ignored_users_dict = ignore_dict_content.get("ignored_users", {}) + if isinstance(ignored_users_dict, dict): + ignore_list = frozenset(ignored_users_dict.keys()) erased_senders = await storage.main.are_users_erased((e.sender for e in events)) |