diff options
| author | Eric Eastwood <erice@element.io> | 2021-11-11 00:56:50 -0600 |
|---|---|---|
| committer | Eric Eastwood <erice@element.io> | 2021-11-11 00:56:50 -0600 |
| commit | b559e2326b419b71bdbcf88ab35a6f3fbbea8219 (patch) | |
| tree | e3ef5d870af0ee1888a88201f888933dd9e1a2b4 /synapse | |
| parent | Only run the tests package to get streaming Complement output (diff) | |
| parent | Fix typos in the `username_available` admin API documentation. (#11286) (diff) | |
| download | synapse-b559e2326b419b71bdbcf88ab35a6f3fbbea8219.tar.xz | |
Merge branch 'develop' into madlittlemods/return-historical-events-in-order-from-backfill
Conflicts: synapse/rest/client/room_batch.py
Diffstat (limited to 'synapse')
45 files changed, 956 insertions, 387 deletions
diff --git a/synapse/api/filtering.py b/synapse/api/filtering.py index 4b0a9b2974..13dd6ce248 100644 --- a/synapse/api/filtering.py +++ b/synapse/api/filtering.py @@ -1,7 +1,7 @@ # Copyright 2015, 2016 OpenMarket Ltd # Copyright 2017 Vector Creations Ltd # Copyright 2018-2019 New Vector Ltd -# Copyright 2019 The Matrix.org Foundation C.I.C. +# Copyright 2019-2021 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. @@ -86,6 +86,9 @@ ROOM_EVENT_FILTER_SCHEMA = { # cf https://github.com/matrix-org/matrix-doc/pull/2326 "org.matrix.labels": {"type": "array", "items": {"type": "string"}}, "org.matrix.not_labels": {"type": "array", "items": {"type": "string"}}, + # MSC3440, filtering by event relations. + "io.element.relation_senders": {"type": "array", "items": {"type": "string"}}, + "io.element.relation_types": {"type": "array", "items": {"type": "string"}}, }, } @@ -146,14 +149,16 @@ def matrix_user_id_validator(user_id_str: str) -> UserID: class Filtering: def __init__(self, hs: "HomeServer"): - super().__init__() + self._hs = hs self.store = hs.get_datastore() + self.DEFAULT_FILTER_COLLECTION = FilterCollection(hs, {}) + async def get_user_filter( self, user_localpart: str, filter_id: Union[int, str] ) -> "FilterCollection": result = await self.store.get_user_filter(user_localpart, filter_id) - return FilterCollection(result) + return FilterCollection(self._hs, result) def add_user_filter( self, user_localpart: str, user_filter: JsonDict @@ -191,21 +196,22 @@ FilterEvent = TypeVar("FilterEvent", EventBase, UserPresenceState, JsonDict) class FilterCollection: - def __init__(self, filter_json: JsonDict): + def __init__(self, hs: "HomeServer", filter_json: JsonDict): self._filter_json = filter_json room_filter_json = self._filter_json.get("room", {}) self._room_filter = Filter( - {k: v for k, v in room_filter_json.items() if k in ("rooms", "not_rooms")} + hs, + {k: v for k, v in room_filter_json.items() if k in ("rooms", "not_rooms")}, ) - self._room_timeline_filter = Filter(room_filter_json.get("timeline", {})) - self._room_state_filter = Filter(room_filter_json.get("state", {})) - self._room_ephemeral_filter = Filter(room_filter_json.get("ephemeral", {})) - self._room_account_data = Filter(room_filter_json.get("account_data", {})) - self._presence_filter = Filter(filter_json.get("presence", {})) - self._account_data = Filter(filter_json.get("account_data", {})) + self._room_timeline_filter = Filter(hs, room_filter_json.get("timeline", {})) + self._room_state_filter = Filter(hs, room_filter_json.get("state", {})) + self._room_ephemeral_filter = Filter(hs, room_filter_json.get("ephemeral", {})) + self._room_account_data = Filter(hs, room_filter_json.get("account_data", {})) + self._presence_filter = Filter(hs, filter_json.get("presence", {})) + self._account_data = Filter(hs, filter_json.get("account_data", {})) self.include_leave = filter_json.get("room", {}).get("include_leave", False) self.event_fields = filter_json.get("event_fields", []) @@ -232,25 +238,37 @@ class FilterCollection: def include_redundant_members(self) -> bool: return self._room_state_filter.include_redundant_members - def filter_presence( + async def filter_presence( self, events: Iterable[UserPresenceState] ) -> List[UserPresenceState]: - return self._presence_filter.filter(events) + return await self._presence_filter.filter(events) - def filter_account_data(self, events: Iterable[JsonDict]) -> List[JsonDict]: - return self._account_data.filter(events) + async def filter_account_data(self, events: Iterable[JsonDict]) -> List[JsonDict]: + return await self._account_data.filter(events) - def filter_room_state(self, events: Iterable[EventBase]) -> List[EventBase]: - return self._room_state_filter.filter(self._room_filter.filter(events)) + async def filter_room_state(self, events: Iterable[EventBase]) -> List[EventBase]: + return await self._room_state_filter.filter( + await self._room_filter.filter(events) + ) - def filter_room_timeline(self, events: Iterable[EventBase]) -> List[EventBase]: - return self._room_timeline_filter.filter(self._room_filter.filter(events)) + async def filter_room_timeline( + self, events: Iterable[EventBase] + ) -> List[EventBase]: + return await self._room_timeline_filter.filter( + await self._room_filter.filter(events) + ) - def filter_room_ephemeral(self, events: Iterable[JsonDict]) -> List[JsonDict]: - return self._room_ephemeral_filter.filter(self._room_filter.filter(events)) + async def filter_room_ephemeral(self, events: Iterable[JsonDict]) -> List[JsonDict]: + return await self._room_ephemeral_filter.filter( + await self._room_filter.filter(events) + ) - def filter_room_account_data(self, events: Iterable[JsonDict]) -> List[JsonDict]: - return self._room_account_data.filter(self._room_filter.filter(events)) + async def filter_room_account_data( + self, events: Iterable[JsonDict] + ) -> List[JsonDict]: + return await self._room_account_data.filter( + await self._room_filter.filter(events) + ) def blocks_all_presence(self) -> bool: return ( @@ -274,7 +292,9 @@ class FilterCollection: class Filter: - def __init__(self, filter_json: JsonDict): + def __init__(self, hs: "HomeServer", filter_json: JsonDict): + self._hs = hs + self._store = hs.get_datastore() self.filter_json = filter_json self.limit = filter_json.get("limit", 10) @@ -297,6 +317,20 @@ class Filter: self.labels = filter_json.get("org.matrix.labels", None) self.not_labels = filter_json.get("org.matrix.not_labels", []) + # Ideally these would be rejected at the endpoint if they were provided + # and not supported, but that would involve modifying the JSON schema + # based on the homeserver configuration. + if hs.config.experimental.msc3440_enabled: + self.relation_senders = self.filter_json.get( + "io.element.relation_senders", None + ) + self.relation_types = self.filter_json.get( + "io.element.relation_types", None + ) + else: + self.relation_senders = None + self.relation_types = None + def filters_all_types(self) -> bool: return "*" in self.not_types @@ -306,7 +340,7 @@ class Filter: def filters_all_rooms(self) -> bool: return "*" in self.not_rooms - def check(self, event: FilterEvent) -> bool: + def _check(self, event: FilterEvent) -> bool: """Checks whether the filter matches the given event. Args: @@ -420,8 +454,30 @@ class Filter: return room_ids - def filter(self, events: Iterable[FilterEvent]) -> List[FilterEvent]: - return list(filter(self.check, events)) + async def _check_event_relations( + self, events: Iterable[FilterEvent] + ) -> List[FilterEvent]: + # The event IDs to check, mypy doesn't understand the ifinstance check. + event_ids = [event.event_id for event in events if isinstance(event, EventBase)] # type: ignore[attr-defined] + event_ids_to_keep = set( + await self._store.events_have_relations( + event_ids, self.relation_senders, self.relation_types + ) + ) + + return [ + event + for event in events + if not isinstance(event, EventBase) or event.event_id in event_ids_to_keep + ] + + async def filter(self, events: Iterable[FilterEvent]) -> List[FilterEvent]: + result = [event for event in events if self._check(event)] + + if self.relation_senders or self.relation_types: + return await self._check_event_relations(result) + + return result def with_room_ids(self, room_ids: Iterable[str]) -> "Filter": """Returns a new filter with the given room IDs appended. @@ -433,7 +489,7 @@ class Filter: filter: A new filter including the given rooms and the old filter's rooms. """ - newFilter = Filter(self.filter_json) + newFilter = Filter(self._hs, self.filter_json) newFilter.rooms += room_ids return newFilter @@ -444,6 +500,3 @@ def _matches_wildcard(actual_value: Optional[str], filter_value: str) -> bool: return actual_value.startswith(type_prefix) else: return actual_value == filter_value - - -DEFAULT_FILTER_COLLECTION = FilterCollection({}) diff --git a/synapse/api/urls.py b/synapse/api/urls.py index 6e84b1524f..4486b3bc7d 100644 --- a/synapse/api/urls.py +++ b/synapse/api/urls.py @@ -38,9 +38,6 @@ class ConsentURIBuilder: def __init__(self, hs_config: HomeServerConfig): if hs_config.key.form_secret is None: raise ConfigError("form_secret not set in config") - if hs_config.server.public_baseurl is None: - raise ConfigError("public_baseurl not set in config") - self._hmac_secret = hs_config.key.form_secret.encode("utf-8") self._public_baseurl = hs_config.server.public_baseurl diff --git a/synapse/config/account_validity.py b/synapse/config/account_validity.py index b56c2a24df..c533452cab 100644 --- a/synapse/config/account_validity.py +++ b/synapse/config/account_validity.py @@ -75,10 +75,6 @@ class AccountValidityConfig(Config): self.account_validity_period * 10.0 / 100.0 ) - if self.account_validity_renew_by_email_enabled: - if not self.root.server.public_baseurl: - raise ConfigError("Can't send renewal emails without 'public_baseurl'") - # Load account validity templates. account_validity_template_dir = account_validity_config.get("template_dir") if account_validity_template_dir is not None: diff --git a/synapse/config/cas.py b/synapse/config/cas.py index 9b58ecf3d8..3f81814043 100644 --- a/synapse/config/cas.py +++ b/synapse/config/cas.py @@ -16,7 +16,7 @@ from typing import Any, List from synapse.config.sso import SsoAttributeRequirement -from ._base import Config, ConfigError +from ._base import Config from ._util import validate_config @@ -35,14 +35,10 @@ class CasConfig(Config): if self.cas_enabled: self.cas_server_url = cas_config["server_url"] - # The public baseurl is required because it is used by the redirect - # template. - public_baseurl = self.root.server.public_baseurl - if not public_baseurl: - raise ConfigError("cas_config requires a public_baseurl to be set") - # TODO Update this to a _synapse URL. + public_baseurl = self.root.server.public_baseurl self.cas_service_url = public_baseurl + "_matrix/client/r0/login/cas/ticket" + self.cas_displayname_attribute = cas_config.get("displayname_attribute") required_attributes = cas_config.get("required_attributes") or {} self.cas_required_attributes = _parsed_required_attributes_def( diff --git a/synapse/config/emailconfig.py b/synapse/config/emailconfig.py index 8ff59aa2f8..afd65fecd3 100644 --- a/synapse/config/emailconfig.py +++ b/synapse/config/emailconfig.py @@ -186,11 +186,6 @@ class EmailConfig(Config): if not self.email_notif_from: missing.append("email.notif_from") - # public_baseurl is required to build password reset and validation links that - # will be emailed to users - if config.get("public_baseurl") is None: - missing.append("public_baseurl") - if missing: raise ConfigError( MISSING_PASSWORD_RESET_CONFIG_ERROR % (", ".join(missing),) @@ -296,9 +291,6 @@ class EmailConfig(Config): if not self.email_notif_from: missing.append("email.notif_from") - if config.get("public_baseurl") is None: - missing.append("public_baseurl") - if missing: raise ConfigError( "email.enable_notifs is True but required keys are missing: %s" diff --git a/synapse/config/oidc.py b/synapse/config/oidc.py index 10f5796330..42f113cd24 100644 --- a/synapse/config/oidc.py +++ b/synapse/config/oidc.py @@ -59,8 +59,6 @@ class OIDCConfig(Config): ) public_baseurl = self.root.server.public_baseurl - if public_baseurl is None: - raise ConfigError("oidc_config requires a public_baseurl to be set") self.oidc_callback_url = public_baseurl + "_synapse/client/oidc/callback" @property diff --git a/synapse/config/registration.py b/synapse/config/registration.py index a3d2a38c4c..5379e80715 100644 --- a/synapse/config/registration.py +++ b/synapse/config/registration.py @@ -45,17 +45,6 @@ class RegistrationConfig(Config): account_threepid_delegates = config.get("account_threepid_delegates") or {} self.account_threepid_delegate_email = account_threepid_delegates.get("email") self.account_threepid_delegate_msisdn = account_threepid_delegates.get("msisdn") - if ( - self.account_threepid_delegate_msisdn - and not self.root.server.public_baseurl - ): - raise ConfigError( - "The configuration option `public_baseurl` is required if " - "`account_threepid_delegate.msisdn` is set, such that " - "clients know where to submit validation tokens to. Please " - "configure `public_baseurl`." - ) - self.default_identity_server = config.get("default_identity_server") self.allow_guest_access = config.get("allow_guest_access", False) @@ -240,7 +229,7 @@ class RegistrationConfig(Config): # in on this server. # # (By default, no suggestion is made, so it is left up to the client. - # This setting is ignored unless public_baseurl is also set.) + # This setting is ignored unless public_baseurl is also explicitly set.) # #default_identity_server: https://matrix.org @@ -265,8 +254,6 @@ class RegistrationConfig(Config): # by the Matrix Identity Service API specification: # https://matrix.org/docs/spec/identity_service/latest # - # If a delegate is specified, the config option public_baseurl must also be filled out. - # account_threepid_delegates: #email: https://example.com # Delegate email sending to example.com #msisdn: http://localhost:8090 # Delegate SMS sending to this local process diff --git a/synapse/config/saml2.py b/synapse/config/saml2.py index 9c51b6a25a..ba2b0905ff 100644 --- a/synapse/config/saml2.py +++ b/synapse/config/saml2.py @@ -199,14 +199,11 @@ class SAML2Config(Config): """ import saml2 - public_baseurl = self.root.server.public_baseurl - if public_baseurl is None: - raise ConfigError("saml2_config requires a public_baseurl to be set") - if self.saml2_grandfathered_mxid_source_attribute: optional_attributes.add(self.saml2_grandfathered_mxid_source_attribute) optional_attributes -= required_attributes + public_baseurl = self.root.server.public_baseurl metadata_url = public_baseurl + "_synapse/client/saml2/metadata.xml" response_url = public_baseurl + "_synapse/client/saml2/authn_response" return { diff --git a/synapse/config/server.py b/synapse/config/server.py index a387fd9310..7bc0030a9e 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -16,6 +16,7 @@ import itertools import logging import os.path import re +import urllib.parse from textwrap import indent from typing import Any, Dict, Iterable, List, Optional, Set, Tuple, Union @@ -264,10 +265,44 @@ class ServerConfig(Config): self.use_frozen_dicts = config.get("use_frozen_dicts", False) self.serve_server_wellknown = config.get("serve_server_wellknown", False) - self.public_baseurl = config.get("public_baseurl") - if self.public_baseurl is not None: - if self.public_baseurl[-1] != "/": - self.public_baseurl += "/" + # Whether we should serve a "client well-known": + # (a) at .well-known/matrix/client on our client HTTP listener + # (b) in the response to /login + # + # ... which together help ensure that clients use our public_baseurl instead of + # whatever they were told by the user. + # + # For the sake of backwards compatibility with existing installations, this is + # True if public_baseurl is specified explicitly, and otherwise False. (The + # reasoning here is that we have no way of knowing that the default + # public_baseurl is actually correct for existing installations - many things + # will not work correctly, but that's (probably?) better than sending clients + # to a completely broken URL. + self.serve_client_wellknown = False + + public_baseurl = config.get("public_baseurl") + if public_baseurl is None: + public_baseurl = f"https://{self.server_name}/" + logger.info("Using default public_baseurl %s", public_baseurl) + else: + self.serve_client_wellknown = True + if public_baseurl[-1] != "/": + public_baseurl += "/" + self.public_baseurl = public_baseurl + + # check that public_baseurl is valid + try: + splits = urllib.parse.urlsplit(self.public_baseurl) + except Exception as e: + raise ConfigError(f"Unable to parse URL: {e}", ("public_baseurl",)) + if splits.scheme not in ("https", "http"): + raise ConfigError( + f"Invalid scheme '{splits.scheme}': only https and http are supported" + ) + if splits.query or splits.fragment: + raise ConfigError( + "public_baseurl cannot contain query parameters or a #-fragment" + ) # Whether to enable user presence. presence_config = config.get("presence") or {} @@ -773,6 +808,8 @@ class ServerConfig(Config): # Otherwise, it should be the URL to reach Synapse's client HTTP listener (see # 'listeners' below). # + # Defaults to 'https://<server_name>/'. + # #public_baseurl: https://example.com/ # Uncomment the following to tell other servers to send federation traffic on diff --git a/synapse/config/sso.py b/synapse/config/sso.py index 11a9b76aa0..60aacb13ea 100644 --- a/synapse/config/sso.py +++ b/synapse/config/sso.py @@ -101,13 +101,10 @@ class SSOConfig(Config): # gracefully to the client). This would make it pointless to ask the user for # confirmation, since the URL the confirmation page would be showing wouldn't be # the client's. - # public_baseurl is an optional setting, so we only add the fallback's URL to the - # list if it's provided (because we can't figure out what that URL is otherwise). - if self.root.server.public_baseurl: - login_fallback_url = ( - self.root.server.public_baseurl + "_matrix/static/client/login" - ) - self.sso_client_whitelist.append(login_fallback_url) + login_fallback_url = ( + self.root.server.public_baseurl + "_matrix/static/client/login" + ) + self.sso_client_whitelist.append(login_fallback_url) def generate_config_section(self, **kwargs): return """\ @@ -128,11 +125,10 @@ class SSOConfig(Config): # phishing attacks from evil.site. To avoid this, include a slash after the # hostname: "https://my.client/". # - # If public_baseurl is set, then the login fallback page (used by clients - # that don't natively support the required login flows) is whitelisted in - # addition to any URLs in this list. + # The login fallback page (used by clients that don't natively support the + # required login flows) is whitelisted in addition to any URLs in this list. # - # By default, this list is empty. + # By default, this list contains only the login fallback page. # #client_whitelist: # - https://riot.im/develop diff --git a/synapse/config/workers.py b/synapse/config/workers.py index 462630201d..4507992031 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -63,7 +63,8 @@ class WriterLocations: Attributes: events: The instances that write to the event and backfill streams. - typing: The instance that writes to the typing stream. + typing: The instances that write to the typing stream. Currently + can only be a single instance. to_device: The instances that write to the to_device stream. Currently can only be a single instance. account_data: The instances that write to the account data streams. Currently @@ -75,9 +76,15 @@ class WriterLocations: """ events = attr.ib( - default=["master"], type=List[str], converter=_instance_to_list_converter + default=["master"], + type=List[str], + converter=_instance_to_list_converter, + ) + typing = attr.ib( + default=["master"], + type=List[str], + converter=_instance_to_list_converter, ) - typing = attr.ib(default="master", type=str) to_device = attr.ib( default=["master"], type=List[str], @@ -217,6 +224,11 @@ class WorkerConfig(Config): % (instance, stream) ) + if len(self.writers.typing) != 1: + raise ConfigError( + "Must only specify one instance to handle `typing` messages." + ) + if len(self.writers.to_device) != 1: raise ConfigError( "Must only specify one instance to handle `to_device` messages." diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py index 670186f548..3b85b135e0 100644 --- a/synapse/federation/federation_client.py +++ b/synapse/federation/federation_client.py @@ -277,6 +277,58 @@ class FederationClient(FederationBase): return pdus + async def get_pdu_from_destination_raw( + self, + destination: str, + event_id: str, + room_version: RoomVersion, + outlier: bool = False, + timeout: Optional[int] = None, + ) -> Optional[EventBase]: + """Requests the PDU with given origin and ID from the remote home + server. Does not have any caching or rate limiting! + + Args: + destination: Which homeserver to query + event_id: event to fetch + room_version: version of the room + outlier: Indicates whether the PDU is an `outlier`, i.e. if + it's from an arbitrary point in the context as opposed to part + of the current block of PDUs. Defaults to `False` + timeout: How long to try (in ms) each destination for before + moving to the next destination. None indicates no timeout. + + Returns: + The requested PDU, or None if we were unable to find it. + + Raises: + SynapseError, NotRetryingDestination, FederationDeniedError + """ + transaction_data = await self.transport_layer.get_event( + destination, event_id, timeout=timeout + ) + + logger.debug( + "retrieved event id %s from %s: %r", + event_id, + destination, + transaction_data, + ) + + pdu_list: List[EventBase] = [ + event_from_pdu_json(p, room_version, outlier=outlier) + for p in transaction_data["pdus"] + ] + + if pdu_list and pdu_list[0]: + pdu = pdu_list[0] + + # Check signatures are correct. + signed_pdu = await self._check_sigs_and_hash(room_version, pdu) + return signed_pdu + + return None + async def get_pdu( self, destinations: Iterable[str], @@ -321,30 +373,14 @@ class FederationClient(FederationBase): continue try: - transaction_data = await self.transport_layer.get_event( - destination, event_id, timeout=timeout - ) - - logger.debug( - "retrieved event id %s from %s: %r", - event_id, - destination, - transaction_data, + signed_pdu = await self.get_pdu_from_destination_raw( + destination=destination, + event_id=event_id, + room_version=room_version, + outlier=outlier, + timeout=timeout, ) - pdu_list: List[EventBase] = [ - event_from_pdu_json(p, room_version, outlier=outlier) - for p in transaction_data["pdus"] - ] - - if pdu_list and pdu_list[0]: - pdu = pdu_list[0] - - # Check signatures are correct. - signed_pdu = await self._check_sigs_and_hash(room_version, pdu) - - break - pdu_attempts[destination] = now except SynapseError as e: diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index 32a75993d9..9a8758e9a6 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -213,6 +213,11 @@ class FederationServer(FederationBase): self._started_handling_of_staged_events = True self._handle_old_staged_events() + # Start a periodic check for old staged events. This is to handle + # the case where locks time out, e.g. if another process gets killed + # without dropping its locks. + self._clock.looping_call(self._handle_old_staged_events, 60 * 1000) + # keep this as early as possible to make the calculated origin ts as # accurate as possible. request_time = self._clock.time_msec() @@ -1232,10 +1237,6 @@ class FederationHandlerRegistry: self.query_handlers[query_type] = handler - def register_instance_for_edu(self, edu_type: str, instance_name: str) -> None: - """Register that the EDU handler is on a different instance than master.""" - self._edu_type_to_instance[edu_type] = [instance_name] - def register_instances_for_edu( self, edu_type: str, instance_names: List[str] ) -> None: diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index 67f8ffcaff..9abdad262b 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -34,6 +34,7 @@ from synapse.metrics.background_process_metrics import ( ) from synapse.storage.databases.main.directory import RoomAliasMapping from synapse.types import JsonDict, RoomAlias, RoomStreamToken, UserID +from synapse.util.async_helpers import Linearizer from synapse.util.metrics import Measure if TYPE_CHECKING: @@ -58,6 +59,10 @@ class ApplicationServicesHandler: self.current_max = 0 self.is_processing = False + self._ephemeral_events_linearizer = Linearizer( + name="appservice_ephemeral_events" + ) + def notify_interested_services(self, max_token: RoomStreamToken) -> None: """Notifies (pushes) all application services interested in this event. @@ -183,7 +188,7 @@ class ApplicationServicesHandler: self, stream_key: str, new_token: Union[int, RoomStreamToken], - users: Optional[Collection[Union[str, UserID]]] = None, + users: Collection[Union[str, UserID]], ) -> None: """ This is called by the notifier in the background when an ephemeral event is handled @@ -198,7 +203,9 @@ class ApplicationServicesHandler: value for `stream_key` will cause this function to return early. Ephemeral events will only be pushed to appservices that have opted into - them. + receiving them by setting `push_ephemeral` to true in their registration + file. Note that while MSC2409 is experimental, this option is called + `de.sorunome.msc2409.push_ephemeral`. Appservices will only receive ephemeral events that fall within their registered user and room namespaces. @@ -209,6 +216,7 @@ class ApplicationServicesHandler: if not self.notify_appservices: return + # Ignore any unsupported streams if stream_key not in ("typing_key", "receipt_key", "presence_key"): return @@ -225,18 +233,25 @@ class ApplicationServicesHandler: # Additional context: https://github.com/matrix-org/synapse/pull/11137 assert isinstance(new_token, int) + # Check whether there are any appservices which have registered to receive + # ephemeral events. + # + # Note that whether these events are actually relevant to these appservices + # is decided later on. services = [ service for service in self.store.get_app_services() if service.supports_ephemeral ] if not services: + # Bail out early if none of the target appservices have explicitly registered + # to receive these ephemeral events. return # We only start a new background process if necessary rather than # optimistically (to cut down on overhead). self._notify_interested_services_ephemeral( - services, stream_key, new_token, users or [] + services, stream_key, new_token, users ) @wrap_as_background_process("notify_interested_services_ephemeral") @@ -247,7 +262,7 @@ class ApplicationServicesHandler: new_token: int, users: Collection[Union[str, UserID]], ) -> None: - logger.debug("Checking interested services for %s" % (stream_key)) + logger.debug("Checking interested services for %s", stream_key) with Measure(self.clock, "notify_interested_services_ephemeral"): for service in services: if stream_key == "typing_key": @@ -260,26 +275,37 @@ class ApplicationServicesHandler: events = await self._handle_typing(service, new_token) if events: self.scheduler.submit_ephemeral_events_for_as(service, events) + continue - elif stream_key == "receipt_key": - events = await self._handle_receipts(service) - if events: - self.scheduler.submit_ephemeral_events_for_as(service, events) - - # Persist the latest handled stream token for this appservice - await self.store.set_type_stream_id_for_appservice( - service, "read_receipt", new_token + # Since we read/update the stream position for this AS/stream + with ( + await self._ephemeral_events_linearizer.queue( + (service.id, stream_key) ) + ): + if stream_key == "receipt_key": + events = await self._handle_receipts(service, new_token) + if events: + self.scheduler.submit_ephemeral_events_for_as( + service, events + ) + + # Persist the latest handled stream token for this appservice + await self.store.set_type_stream_id_for_appservice( + service, "read_receipt", new_token + ) - elif stream_key == "presence_key": - events = await self._handle_presence(service, users) - if events: - self.scheduler.submit_ephemeral_events_for_as(service, events) + elif stream_key == "presence_key": + events = await self._handle_presence(service, users, new_token) + if events: + self.scheduler.submit_ephemeral_events_for_as( + service, events + ) - # Persist the latest handled stream token for this appservice - await self.store.set_type_stream_id_for_appservice( - service, "presence", new_token - ) + # Persist the latest handled stream token for this appservice + await self.store.set_type_stream_id_for_appservice( + service, "presence", new_token + ) async def _handle_typing( self, service: ApplicationService, new_token: int @@ -316,7 +342,9 @@ class ApplicationServicesHandler: ) return typing - async def _handle_receipts(self, service: ApplicationService) -> List[JsonDict]: + async def _handle_receipts( + self, service: ApplicationService, new_token: Optional[int] + ) -> List[JsonDict]: """ Return the latest read receipts that the given application service should receive. @@ -327,6 +355,9 @@ class ApplicationServicesHandler: Args: service: The application service to check for which events it should receive. + new_token: A receipts event stream token. Purely used to double-check that the + from_token we pull from the database isn't greater than or equal to this + token. Prevents accidentally duplicating work. Returns: A list of JSON dictionaries containing data derived from the read receipts that @@ -335,6 +366,12 @@ class ApplicationServicesHandler: from_key = await self.store.get_type_stream_id_for_appservice( service, "read_receipt" ) + if new_token is not None and new_token <= from_key: + logger.debug( + "Rejecting token lower than or equal to stored: %s" % (new_token,) + ) + return [] + receipts_source = self.event_sources.sources.receipt receipts, _ = await receipts_source.get_new_events_as( service=service, from_key=from_key @@ -342,7 +379,10 @@ class ApplicationServicesHandler: return receipts async def _handle_presence( - self, service: ApplicationService, users: Collection[Union[str, UserID]] + self, + service: ApplicationService, + users: Collection[Union[str, UserID]], + new_token: Optional[int], ) -> List[JsonDict]: """ Return the latest presence updates that the given application service should receive. @@ -355,6 +395,9 @@ class ApplicationServicesHandler: Args: service: The application service that ephemeral events are being sent to. users: The users that should receive the presence update. + new_token: A presence update stream token. Purely used to double-check that the + from_token we pull from the database isn't greater than or equal to this + token. Prevents accidentally duplicating work. Returns: A list of json dictionaries containing data derived from the presence events @@ -365,6 +408,12 @@ class ApplicationServicesHandler: from_key = await self.store.get_type_stream_id_for_appservice( service, "presence" ) + if new_token is not None and new_token <= from_key: + logger.debug( + "Rejecting token lower than or equal to stored: %s" % (new_token,) + ) + return [] + for user in users: if isinstance(user, str): user = UserID.from_string(user) diff --git a/synapse/handlers/auth.py b/synapse/handlers/auth.py index d508d7d32a..60e59d11a0 100644 --- a/synapse/handlers/auth.py +++ b/synapse/handlers/auth.py @@ -1989,7 +1989,9 @@ class PasswordAuthProvider: self, check_3pid_auth: Optional[CHECK_3PID_AUTH_CALLBACK] = None, on_logged_out: Optional[ON_LOGGED_OUT_CALLBACK] = None, - auth_checkers: Optional[Dict[Tuple[str, Tuple], CHECK_AUTH_CALLBACK]] = None, + auth_checkers: Optional[ + Dict[Tuple[str, Tuple[str, ...]], CHECK_AUTH_CALLBACK] + ] = None, ) -> None: # Register check_3pid_auth callback if check_3pid_auth is not None: diff --git a/synapse/handlers/devicemessage.py b/synapse/handlers/devicemessage.py index b6a2a34ab7..b582266af9 100644 --- a/synapse/handlers/devicemessage.py +++ b/synapse/handlers/devicemessage.py @@ -89,6 +89,13 @@ class DeviceMessageHandler: ) async def on_direct_to_device_edu(self, origin: str, content: JsonDict) -> None: + """ + Handle receiving to-device messages from remote homeservers. + + Args: + origin: The remote homeserver. + content: The JSON dictionary containing the to-device messages. + """ local_messages = {} sender_user_id = content["sender"] if origin != get_domain_from_id(sender_user_id): @@ -135,12 +142,16 @@ class DeviceMessageHandler: message_type, sender_user_id, by_device ) - stream_id = await self.store.add_messages_from_remote_to_device_inbox( + # Add messages to the database. + # Retrieve the stream id of the last-processed to-device message. + last_stream_id = await self.store.add_messages_from_remote_to_device_inbox( origin, message_id, local_messages ) + # Notify listeners that there are new to-device messages to process, + # handing them the latest stream id. self.notifier.on_new_event( - "to_device_key", stream_id, users=local_messages.keys() + "to_device_key", last_stream_id, users=local_messages.keys() ) async def _check_for_unknown_devices( @@ -195,6 +206,14 @@ class DeviceMessageHandler: message_type: str, messages: Dict[str, Dict[str, JsonDict]], ) -> None: + """ + Handle a request from a user to send to-device message(s). + + Args: + requester: The user that is sending the to-device messages. + message_type: The type of to-device messages that are being sent. + messages: A dictionary containing recipients mapped to messages intended for them. + """ sender_user_id = requester.user.to_string() message_id = random_string(16) @@ -257,12 +276,16 @@ class DeviceMessageHandler: "org.matrix.opentracing_context": json_encoder.encode(context), } - stream_id = await self.store.add_messages_to_device_inbox( + # Add messages to the database. + # Retrieve the stream id of the last-processed to-device message. + last_stream_id = await self.store.add_messages_to_device_inbox( local_messages, remote_edu_contents ) + # Notify listeners that there are new to-device messages to process, + # handing them the latest stream id. self.notifier.on_new_event( - "to_device_key", stream_id, users=local_messages.keys() + "to_device_key", last_stream_id, users=local_messages.keys() ) if self.federation_sender: diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index d0fb2fc7dc..60c11e3d21 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -201,95 +201,19 @@ class E2eKeysHandler: r[user_id] = remote_queries[user_id] # Now fetch any devices that we don't have in our cache - @trace - async def do_remote_query(destination: str) -> None: - """This is called when we are querying the device list of a user on - a remote homeserver and their device list is not in the device list - cache. If we share a room with this user and we're not querying for - specific user we will update the cache with their device list. - """ - - destination_query = remote_queries_not_in_cache[destination] - - # We first consider whether we wish to update the device list cache with - # the users device list. We want to track a user's devices when the - # authenticated user shares a room with the queried user and the query - # has not specified a particular device. - # If we update the cache for the queried user we remove them from further - # queries. We use the more efficient batched query_client_keys for all - # remaining users - user_ids_updated = [] - for (user_id, device_list) in destination_query.items(): - if user_id in user_ids_updated: - continue - - if device_list: - continue - - room_ids = await self.store.get_rooms_for_user(user_id) - if not room_ids: - continue - - # We've decided we're sharing a room with this user and should - # probably be tracking their device lists. However, we haven't - # done an initial sync on the device list so we do it now. - try: - if self._is_master: - user_devices = await self.device_handler.device_list_updater.user_device_resync( - user_id - ) - else: - user_devices = await self._user_device_resync_client( - user_id=user_id - ) - - user_devices = user_devices["devices"] - user_results = results.setdefault(user_id, {}) - for device in user_devices: - user_results[device["device_id"]] = device["keys"] - user_ids_updated.append(user_id) - except Exception as e: - failures[destination] = _exception_to_failure(e) - - if len(destination_query) == len(user_ids_updated): - # We've updated all the users in the query and we do not need to - # make any further remote calls. - return - - # Remove all the users from the query which we have updated - for user_id in user_ids_updated: - destination_query.pop(user_id) - - try: - remote_result = await self.federation.query_client_keys( - destination, {"device_keys": destination_query}, timeout=timeout - ) - - for user_id, keys in remote_result["device_keys"].items(): - if user_id in destination_query: - results[user_id] = keys - - if "master_keys" in remote_result: - for user_id, key in remote_result["master_keys"].items(): - if user_id in destination_query: - cross_signing_keys["master_keys"][user_id] = key - - if "self_signing_keys" in remote_result: - for user_id, key in remote_result["self_signing_keys"].items(): - if user_id in destination_query: - cross_signing_keys["self_signing_keys"][user_id] = key - - except Exception as e: - failure = _exception_to_failure(e) - failures[destination] = failure - set_tag("error", True) - set_tag("reason", failure) - await make_deferred_yieldable( defer.gatherResults( [ - run_in_background(do_remote_query, destination) - for destination in remote_queries_not_in_cache + run_in_background( + self._query_devices_for_destination, + results, + cross_signing_keys, + failures, + destination, + queries, + timeout, + ) + for destination, queries in remote_queries_not_in_cache.items() ], consumeErrors=True, ).addErrback(unwrapFirstError) @@ -301,6 +225,121 @@ class E2eKeysHandler: return ret + @trace + async def _query_devices_for_destination( + self, + results: JsonDict, + cross_signing_keys: JsonDict, + failures: Dict[str, JsonDict], + destination: str, + destination_query: Dict[str, Iterable[str]], + timeout: int, + ) -> None: + """This is called when we are querying the device list of a user on + a remote homeserver and their device list is not in the device list + cache. If we share a room with this user and we're not querying for + specific user we will update the cache with their device list. + + Args: + results: A map from user ID to their device keys, which gets + updated with the newly fetched keys. + cross_signing_keys: Map from user ID to their cross signing keys, + which gets updated with the newly fetched keys. + failures: Map of destinations to failures that have occurred while + attempting to fetch keys. + destination: The remote server to query + destination_query: The query dict of devices to query the remote + server for. + timeout: The timeout for remote HTTP requests. + """ + + # We first consider whether we wish to update the device list cache with + # the users device list. We want to track a user's devices when the + # authenticated user shares a room with the queried user and the query + # has not specified a particular device. + # If we update the cache for the queried user we remove them from further + # queries. We use the more efficient batched query_client_keys for all + # remaining users + user_ids_updated = [] + for (user_id, device_list) in destination_query.items(): + if user_id in user_ids_updated: + continue + + if device_list: + continue + + room_ids = await self.store.get_rooms_for_user(user_id) + if not room_ids: + continue + + # We've decided we're sharing a room with this user and should + # probably be tracking their device lists. However, we haven't + # done an initial sync on the device list so we do it now. + try: + if self._is_master: + resync_results = await self.device_handler.device_list_updater.user_device_resync( + user_id + ) + else: + resync_results = await self._user_device_resync_client( + user_id=user_id + ) + + # Add the device keys to the results. + user_devices = resync_results["devices"] + user_results = results.setdefault(user_id, {}) + for device in user_devices: + user_results[device["device_id"]] = device["keys"] + user_ids_updated.append(user_id) + + # Add any cross signing keys to the results. + master_key = resync_results.get("master_key") + self_signing_key = resync_results.get("self_signing_key") + + if master_key: + cross_signing_keys["master_keys"][user_id] = master_key + + if self_signing_key: + cross_signing_keys["self_signing_keys"][user_id] = self_signing_key + except Exception as e: + failures[destination] = _exception_to_failure(e) + + if len(destination_query) == len(user_ids_updated): + # We've updated all the users in the query and we do not need to + # make any further remote calls. + return + + # Remove all the users from the query which we have updated + for user_id in user_ids_updated: + destination_query.pop(user_id) + + try: + remote_result = await self.federation.query_client_keys( + destination, {"device_keys": destination_query}, timeout=timeout + ) + + for user_id, keys in remote_result["device_keys"].items(): + if user_id in destination_query: + results[user_id] = keys + + if "master_keys" in remote_result: + for user_id, key in remote_result["master_keys"].items(): + if user_id in destination_query: + cross_signing_keys["master_keys"][user_id] = key + + if "self_signing_keys" in remote_result: + for user_id, key in remote_result["self_signing_keys"].items(): + if user_id in destination_query: + cross_signing_keys["self_signing_keys"][user_id] = key + + except Exception as e: + failure = _exception_to_failure(e) + failures[destination] = failure + set_tag("error", True) + set_tag("reason", failure) + + return + async def get_cross_signing_keys_from_cache( self, query: Iterable[str], from_user_id: Optional[str] ) -> Dict[str, Dict[str, dict]]: diff --git a/synapse/handlers/identity.py b/synapse/handlers/identity.py index 6a315117ba..3dbe611f95 100644 --- a/synapse/handlers/identity.py +++ b/synapse/handlers/identity.py @@ -537,10 +537,6 @@ class IdentityHandler: except RequestTimedOutError: raise SynapseError(500, "Timed out contacting identity server") - # It is already checked that public_baseurl is configured since this code - # should only be used if account_threepid_delegate_msisdn is true. - assert self.hs.config.server.public_baseurl - # we need to tell the client to send the token back to us, since it doesn't # otherwise know where to send it, so add submit_url response parameter # (see also MSC2078) diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index abfe7be0e3..aa26911aed 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -424,7 +424,7 @@ class PaginationHandler: if events: if event_filter: - events = event_filter.filter(events) + events = await event_filter.filter(events) events = await filter_events_for_client( self.storage, user_id, events, is_peeking=(member_event_id is None) diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 969eb3b9b0..11af30eee7 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -12,8 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -"""Contains functions for performing events on rooms.""" - +"""Contains functions for performing actions on rooms.""" import itertools import logging import math @@ -31,6 +30,8 @@ from typing import ( Tuple, ) +from typing_extensions import TypedDict + from synapse.api.constants import ( EventContentFields, EventTypes, @@ -1158,8 +1159,10 @@ class RoomContextHandler: ) if event_filter: - results["events_before"] = event_filter.filter(results["events_before"]) - results["events_after"] = event_filter.filter(results["events_after"]) + results["events_before"] = await event_filter.filter( + results["events_before"] + ) + results["events_after"] = await event_filter.filter(results["events_after"]) results["events_before"] = await filter_evts(results["events_before"]) results["events_after"] = await filter_evts(results["events_after"]) @@ -1195,7 +1198,7 @@ class RoomContextHandler: state_events = list(state[last_event_id].values()) if event_filter: - state_events = event_filter.filter(state_events) + state_events = await event_filter.filter(state_events) results["state"] = await filter_evts(state_events) @@ -1275,6 +1278,13 @@ class RoomEventSource(EventSource[RoomStreamToken, EventBase]): return self.store.get_room_events_max_id(room_id) +class ShutdownRoomResponse(TypedDict): + kicked_users: List[str] + failed_to_kick_users: List[str] + local_aliases: List[str] + new_room_id: Optional[str] + + class RoomShutdownHandler: DEFAULT_MESSAGE = ( @@ -1300,7 +1310,7 @@ class RoomShutdownHandler: new_room_name: Optional[str] = None, message: Optional[str] = None, block: bool = False, - ) -> dict: + ) -> ShutdownRoomResponse: """ Shuts down a room. Moves all local users and room aliases automatically to a new room if `new_room_user_id` is set. Otherwise local users only @@ -1334,8 +1344,13 @@ class RoomShutdownHandler: Defaults to `Sharing illegal content on this server is not permitted and rooms in violation will be blocked.` block: - If set to `true`, this room will be added to a blocking list, - preventing future attempts to join the room. Defaults to `false`. + If set to `True`, users will be prevented from joining the old + room. This option can also be used to pre-emptively block a room, + even if it's unknown to this homeserver. In this case, the room + will be blocked, and no further action will be taken. If `False`, + attempting to delete an unknown room is invalid. + + Defaults to `False`. Returns: a dict containing the following keys: kicked_users: An array of users (`user_id`) that were kicked. @@ -1344,7 +1359,9 @@ class RoomShutdownHandler: local_aliases: An array of strings representing the local aliases that were migrated from the old room to the new. - new_room_id: A string representing the room ID of the new room. + new_room_id: + A string representing the room ID of the new room, or None if + no such room was created. """ if not new_room_name: @@ -1355,14 +1372,28 @@ class RoomShutdownHandler: if not RoomID.is_valid(room_id): raise SynapseError(400, "%s is not a legal room ID" % (room_id,)) - if not await self.store.get_room(room_id): - raise NotFoundError("Unknown room id %s" % (room_id,)) - - # This will work even if the room is already blocked, but that is - # desirable in case the first attempt at blocking the room failed below. + # Action the block first (even if the room doesn't exist yet) if block: + # This will work even if the room is already blocked, but that is + # desirable in case the first attempt at blocking the room failed below. await self.store.block_room(room_id, requester_user_id) + if not await self.store.get_room(room_id): + if block: + # We allow you to block an unknown room. + return { + "kicked_users": [], + "failed_to_kick_users": [], + "local_aliases": [], + "new_room_id": None, + } + else: + # But if you don't want to preventatively block another room, + # this function can't do anything useful. + raise NotFoundError( + "Cannot shut down room: unknown room id %s" % (room_id,) + ) + if new_room_user_id is not None: if not self.hs.is_mine_id(new_room_user_id): raise SynapseError( diff --git a/synapse/handlers/search.py b/synapse/handlers/search.py index 6e4dff8056..ab7eaab2fb 100644 --- a/synapse/handlers/search.py +++ b/synapse/handlers/search.py @@ -180,7 +180,7 @@ class SearchHandler: % (set(group_keys) - {"room_id", "sender"},), ) - search_filter = Filter(filter_dict) + search_filter = Filter(self.hs, filter_dict) # TODO: Search through left rooms too rooms = await self.store.get_rooms_for_local_user_where_membership_is( @@ -242,7 +242,7 @@ class SearchHandler: rank_map.update({r["event"].event_id: r["rank"] for r in results}) - filtered_events = search_filter.filter([r["event"] for r in results]) + filtered_events = await search_filter.filter([r["event"] for r in results]) events = await filter_events_for_client( self.storage, user.to_string(), filtered_events @@ -292,7 +292,9 @@ class SearchHandler: rank_map.update({r["event"].event_id: r["rank"] for r in results}) - filtered_events = search_filter.filter([r["event"] for r in results]) + filtered_events = await search_filter.filter( + [r["event"] for r in results] + ) events = await filter_events_for_client( self.storage, user.to_string(), filtered_events diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index 2c7c6d63a9..891435c14d 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -510,7 +510,7 @@ class SyncHandler: log_kv({"limited": limited}) if potential_recents: - recents = sync_config.filter_collection.filter_room_timeline( + recents = await sync_config.filter_collection.filter_room_timeline( potential_recents ) log_kv({"recents_after_sync_filtering": len(recents)}) @@ -575,8 +575,8 @@ class SyncHandler: log_kv({"loaded_recents": len(events)}) - loaded_recents = sync_config.filter_collection.filter_room_timeline( - events + loaded_recents = ( + await sync_config.filter_collection.filter_room_timeline(events) ) log_kv({"loaded_recents_after_sync_filtering": len(loaded_recents)}) @@ -1015,7 +1015,7 @@ class SyncHandler: return { (e.type, e.state_key): e - for e in sync_config.filter_collection.filter_room_state( + for e in await sync_config.filter_collection.filter_room_state( list(state.values()) ) if e.type != EventTypes.Aliases # until MSC2261 or alternative solution @@ -1383,7 +1383,7 @@ class SyncHandler: sync_config.user ) - account_data_for_user = sync_config.filter_collection.filter_account_data( + account_data_for_user = await sync_config.filter_collection.filter_account_data( [ {"type": account_data_type, "content": content} for account_data_type, content in account_data.items() @@ -1448,7 +1448,7 @@ class SyncHandler: # Deduplicate the presence entries so that there's at most one per user presence = list({p.user_id: p for p in presence}.values()) - presence = sync_config.filter_collection.filter_presence(presence) + presence = await sync_config.filter_collection.filter_presence(presence) sync_result_builder.presence = presence @@ -2021,12 +2021,14 @@ class SyncHandler: ) account_data_events = ( - sync_config.filter_collection.filter_room_account_data( + await sync_config.filter_collection.filter_room_account_data( account_data_events ) ) - ephemeral = sync_config.filter_collection.filter_room_ephemeral(ephemeral) + ephemeral = await sync_config.filter_collection.filter_room_ephemeral( + ephemeral + ) if not ( always_include diff --git a/synapse/handlers/typing.py b/synapse/handlers/typing.py index c411d69924..22c6174821 100644 --- a/synapse/handlers/typing.py +++ b/synapse/handlers/typing.py @@ -62,8 +62,8 @@ class FollowerTypingHandler: if hs.should_send_federation(): self.federation = hs.get_federation_sender() - if hs.config.worker.writers.typing != hs.get_instance_name(): - hs.get_federation_registry().register_instance_for_edu( + if hs.get_instance_name() not in hs.config.worker.writers.typing: + hs.get_federation_registry().register_instances_for_edu( "m.typing", hs.config.worker.writers.typing, ) @@ -205,7 +205,7 @@ class TypingWriterHandler(FollowerTypingHandler): def __init__(self, hs: "HomeServer"): super().__init__(hs) - assert hs.config.worker.writers.typing == hs.get_instance_name() + assert hs.get_instance_name() in hs.config.worker.writers.typing self.auth = hs.get_auth() self.notifier = hs.get_notifier() diff --git a/synapse/replication/tcp/handler.py b/synapse/replication/tcp/handler.py index 06fd06fdf3..21293038ef 100644 --- a/synapse/replication/tcp/handler.py +++ b/synapse/replication/tcp/handler.py @@ -138,7 +138,7 @@ class ReplicationCommandHandler: if isinstance(stream, TypingStream): # Only add TypingStream as a source on the instance in charge of # typing. - if hs.config.worker.writers.typing == hs.get_instance_name(): + if hs.get_instance_name() in hs.config.worker.writers.typing: self._streams_to_replicate.append(stream) continue diff --git a/synapse/replication/tcp/streams/_base.py b/synapse/replication/tcp/streams/_base.py index c8b188ae4e..743a01da08 100644 --- a/synapse/replication/tcp/streams/_base.py +++ b/synapse/replication/tcp/streams/_base.py @@ -328,8 +328,7 @@ class TypingStream(Stream): ROW_TYPE = TypingStreamRow def __init__(self, hs: "HomeServer"): - writer_instance = hs.config.worker.writers.typing - if writer_instance == hs.get_instance_name(): + if hs.get_instance_name() in hs.config.worker.writers.typing: # On the writer, query the typing handler typing_writer_handler = hs.get_typing_writer_handler() update_function: Callable[ diff --git a/synapse/rest/admin/__init__.py b/synapse/rest/admin/__init__.py index 70514e814f..81e98f81d6 100644 --- a/synapse/rest/admin/__init__.py +++ b/synapse/rest/admin/__init__.py @@ -25,6 +25,10 @@ from synapse.http.server import HttpServer, JsonResource from synapse.http.servlet import RestServlet, parse_json_object_from_request from synapse.http.site import SynapseRequest from synapse.rest.admin._base import admin_patterns, assert_requester_is_admin +from synapse.rest.admin.background_updates import ( + BackgroundUpdateEnabledRestServlet, + BackgroundUpdateRestServlet, +) from synapse.rest.admin.devices import ( DeleteDevicesRestServlet, DeviceRestServlet, @@ -247,6 +251,8 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: # Some servlets only get registered for the main process. if hs.config.worker.worker_app is None: SendServerNoticeServlet(hs).register(http_server) + BackgroundUpdateEnabledRestServlet(hs).register(http_server) + BackgroundUpdateRestServlet(hs).register(http_server) def register_servlets_for_client_rest_resource( diff --git a/synapse/rest/admin/background_updates.py b/synapse/rest/admin/background_updates.py new file mode 100644 index 0000000000..0d0183bf20 --- /dev/null +++ b/synapse/rest/admin/background_updates.py @@ -0,0 +1,107 @@ +# Copyright 2021 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 +from typing import TYPE_CHECKING, Tuple + +from synapse.api.errors import SynapseError +from synapse.http.servlet import RestServlet, parse_json_object_from_request +from synapse.http.site import SynapseRequest +from synapse.rest.admin._base import admin_patterns, assert_user_is_admin +from synapse.types import JsonDict + +if TYPE_CHECKING: + from synapse.server import HomeServer + +logger = logging.getLogger(__name__) + + +class BackgroundUpdateEnabledRestServlet(RestServlet): + """Allows temporarily disabling background updates""" + + PATTERNS = admin_patterns("/background_updates/enabled") + + def __init__(self, hs: "HomeServer"): + self.group_server = hs.get_groups_server_handler() + self.is_mine_id = hs.is_mine_id + self.auth = hs.get_auth() + + self.data_stores = hs.get_datastores() + + async def on_GET(self, request: SynapseRequest) -> Tuple[int, JsonDict]: + requester = await self.auth.get_user_by_req(request) + await assert_user_is_admin(self.auth, requester.user) + + # We need to check that all configured databases have updates enabled. + # (They *should* all be in sync.) + enabled = all(db.updates.enabled for db in self.data_stores.databases) + + return 200, {"enabled": enabled} + + async def on_POST(self, request: SynapseRequest) -> Tuple[int, JsonDict]: + requester = await self.auth.get_user_by_req(request) + await assert_user_is_admin(self.auth, requester.user) + + body = parse_json_object_from_request(request) + + enabled = body.get("enabled", True) + + if not isinstance(enabled, bool): + raise SynapseError(400, "'enabled' parameter must be a boolean") + + for db in self.data_stores.databases: + db.updates.enabled = enabled + + # If we're re-enabling them ensure that we start the background + # process again. + if enabled: + db.updates.start_doing_background_updates() + + return 200, {"enabled": enabled} + + +class BackgroundUpdateRestServlet(RestServlet): + """Fetch information about background updates""" + + PATTERNS = admin_patterns("/background_updates/status") + + def __init__(self, hs: "HomeServer"): + self.group_server = hs.get_groups_server_handler() + self.is_mine_id = hs.is_mine_id + self.auth = hs.get_auth() + + self.data_stores = hs.get_datastores() + + async def on_GET(self, request: SynapseRequest) -> Tuple[int, JsonDict]: + requester = await self.auth.get_user_by_req(request) + await assert_user_is_admin(self.auth, requester.user) + + # We need to check that all configured databases have updates enabled. + # (They *should* all be in sync.) + enabled = all(db.updates.enabled for db in self.data_stores.databases) + + current_updates = {} + + for db in self.data_stores.databases: + update = db.updates.get_current_update() + if not update: + continue + + current_updates[db.name()] = { + "name": update.name, + "total_item_count": update.total_item_count, + "total_duration_ms": update.total_duration_ms, + "average_items_per_ms": update.average_items_per_ms(), + } + + return 200, {"enabled": enabled, "current_updates": current_updates} diff --git a/synapse/rest/admin/rooms.py b/synapse/rest/admin/rooms.py index 05c5b4bf0c..a2f4edebb8 100644 --- a/synapse/rest/admin/rooms.py +++ b/synapse/rest/admin/rooms.py @@ -13,7 +13,7 @@ # limitations under the License. import logging from http import HTTPStatus -from typing import TYPE_CHECKING, List, Optional, Tuple +from typing import TYPE_CHECKING, List, Optional, Tuple, cast from urllib import parse as urlparse from synapse.api.constants import EventTypes, JoinRules, Membership @@ -239,9 +239,22 @@ class RoomRestServlet(RestServlet): # Purge room if purge: - await pagination_handler.purge_room(room_id, force=force_purge) - - return 200, ret + try: + await pagination_handler.purge_room(room_id, force=force_purge) + except NotFoundError: + if block: + # We can block unknown rooms with this endpoint, in which case + # a failed purge is expected. + pass + else: + # But otherwise, we expect this purge to have succeeded. + raise + + # Cast safety: cast away the knowledge that this is a TypedDict. + # See https://github.com/python/mypy/issues/4976#issuecomment-579883622 + # for some discussion on why this is necessary. Either way, + # `ret` is an opaque dictionary blob as far as the rest of the app cares. + return 200, cast(JsonDict, ret) class RoomMembersRestServlet(RestServlet): @@ -583,6 +596,7 @@ class RoomEventContextServlet(RestServlet): def __init__(self, hs: "HomeServer"): super().__init__() + self._hs = hs self.clock = hs.get_clock() self.room_context_handler = hs.get_room_context_handler() self._event_serializer = hs.get_event_client_serializer() @@ -600,7 +614,9 @@ class RoomEventContextServlet(RestServlet): filter_str = parse_string(request, "filter", encoding="utf-8") if filter_str: filter_json = urlparse.unquote(filter_str) - event_filter: Optional[Filter] = Filter(json_decoder.decode(filter_json)) + event_filter: Optional[Filter] = Filter( + self._hs, json_decoder.decode(filter_json) + ) else: event_filter = None diff --git a/synapse/rest/client/receipts.py b/synapse/rest/client/receipts.py index 9770413c61..2b25b9aad6 100644 --- a/synapse/rest/client/receipts.py +++ b/synapse/rest/client/receipts.py @@ -13,10 +13,12 @@ # limitations under the License. import logging +import re from typing import TYPE_CHECKING, Tuple from synapse.api.constants import ReadReceiptEventFields from synapse.api.errors import Codes, SynapseError +from synapse.http import get_request_user_agent from synapse.http.server import HttpServer from synapse.http.servlet import RestServlet, parse_json_object_from_request from synapse.http.site import SynapseRequest @@ -24,6 +26,8 @@ from synapse.types import JsonDict from ._base import client_patterns +pattern = re.compile(r"(?:Element|SchildiChat)/1\.[012]\.") + if TYPE_CHECKING: from synapse.server import HomeServer @@ -52,7 +56,13 @@ class ReceiptRestServlet(RestServlet): if receipt_type != "m.read": raise SynapseError(400, "Receipt type must be 'm.read'") - body = parse_json_object_from_request(request, allow_empty_body=True) + # Do not allow older SchildiChat and Element Android clients (prior to Element/1.[012].x) to send an empty body. + user_agent = get_request_user_agent(request) + allow_empty_body = False + if "Android" in user_agent: + if pattern.match(user_agent) or "Riot" in user_agent: + allow_empty_body = True + body = parse_json_object_from_request(request, allow_empty_body) hidden = body.get(ReadReceiptEventFields.MSC2285_HIDDEN, False) if not isinstance(hidden, bool): diff --git a/synapse/rest/client/relations.py b/synapse/rest/client/relations.py index 58f6699073..184cfbe196 100644 --- a/synapse/rest/client/relations.py +++ b/synapse/rest/client/relations.py @@ -298,7 +298,9 @@ class RelationAggregationPaginationServlet(RestServlet): raise SynapseError(404, "Unknown parent event.") if relation_type not in (RelationTypes.ANNOTATION, None): - raise SynapseError(400, "Relation type must be 'annotation'") + raise SynapseError( + 400, f"Relation type must be '{RelationTypes.ANNOTATION}'" + ) limit = parse_integer(request, "limit", default=5) from_token_str = parse_string(request, "from") diff --git a/synapse/rest/client/room.py b/synapse/rest/client/room.py index ed95189b6d..03a353d53c 100644 --- a/synapse/rest/client/room.py +++ b/synapse/rest/client/room.py @@ -550,6 +550,7 @@ class RoomMessageListRestServlet(RestServlet): def __init__(self, hs: "HomeServer"): super().__init__() + self._hs = hs self.pagination_handler = hs.get_pagination_handler() self.auth = hs.get_auth() self.store = hs.get_datastore() @@ -567,7 +568,9 @@ class RoomMessageListRestServlet(RestServlet): filter_str = parse_string(request, "filter", encoding="utf-8") if filter_str: filter_json = urlparse.unquote(filter_str) - event_filter: Optional[Filter] = Filter(json_decoder.decode(filter_json)) + event_filter: Optional[Filter] = Filter( + self._hs, json_decoder.decode(filter_json) + ) if ( event_filter and event_filter.filter_json.get("event_format", "client") @@ -672,6 +675,7 @@ class RoomEventContextServlet(RestServlet): def __init__(self, hs: "HomeServer"): super().__init__() + self._hs = hs self.clock = hs.get_clock() self.room_context_handler = hs.get_room_context_handler() self._event_serializer = hs.get_event_client_serializer() @@ -688,7 +692,9 @@ class RoomEventContextServlet(RestServlet): filter_str = parse_string(request, "filter", encoding="utf-8") if filter_str: filter_json = urlparse.unquote(filter_str) - event_filter: Optional[Filter] = Filter(json_decoder.decode(filter_json)) + event_filter: Optional[Filter] = Filter( + self._hs, json_decoder.decode(filter_json) + ) else: event_filter = None @@ -914,7 +920,7 @@ class RoomTypingRestServlet(RestServlet): # If we're not on the typing writer instance we should scream if we get # requests. self._is_typing_writer = ( - hs.config.worker.writers.typing == hs.get_instance_name() + hs.get_instance_name() in hs.config.worker.writers.typing ) async def on_PUT( diff --git a/synapse/rest/client/room_batch.py b/synapse/rest/client/room_batch.py index 982763f2d8..c9509d2ae3 100644 --- a/synapse/rest/client/room_batch.py +++ b/synapse/rest/client/room_batch.py @@ -139,20 +139,22 @@ class RoomBatchSendEventRestServlet(RestServlet): errcode=Codes.INVALID_PARAM, ) + state_event_ids_at_start = [] # Create and persist all of the state events that float off on their own # before the batch. These will most likely be all of the invite/member # state events used to auth the upcoming historical messages. - state_event_ids_at_start = ( - await self.room_batch_handler.persist_state_events_at_start( - state_events_at_start=body["state_events_at_start"], - room_id=room_id, - initial_auth_event_ids=auth_event_ids, - app_service_requester=requester, + if body["state_events_at_start"]: + state_event_ids_at_start = ( + await self.room_batch_handler.persist_state_events_at_start( + state_events_at_start=body["state_events_at_start"], + room_id=room_id, + initial_auth_event_ids=auth_event_ids, + app_service_requester=requester, + ) ) - ) - # Update our ongoing auth event ID list with all of the new state we - # just created - auth_event_ids.extend(state_event_ids_at_start) + # Update our ongoing auth event ID list with all of the new state we + # just created + auth_event_ids.extend(state_event_ids_at_start) inherited_depth = await self.room_batch_handler.inherit_depth_from_prev_ids( prev_event_ids_from_query @@ -205,8 +207,11 @@ class RoomBatchSendEventRestServlet(RestServlet): # Also connect the historical event chain to the end of the floating # state chain, which causes the HS to ask for the state at the start of - # the batch later. - prev_event_ids = [state_event_ids_at_start[-1]] + # the batch later. If there is no state chain to connect to, just make + # the insertion event float itself. + prev_event_ids = [] + if len(state_event_ids_at_start): + prev_event_ids = [state_event_ids_at_start[-1]] # Create and persist all of the historical events as well as insertion # and batch meta events to make the batch navigable in the DAG. diff --git a/synapse/rest/client/sync.py b/synapse/rest/client/sync.py index 913216a7c4..8c0fdb1940 100644 --- a/synapse/rest/client/sync.py +++ b/synapse/rest/client/sync.py @@ -29,7 +29,7 @@ from typing import ( from synapse.api.constants import Membership, PresenceState from synapse.api.errors import Codes, StoreError, SynapseError -from synapse.api.filtering import DEFAULT_FILTER_COLLECTION, FilterCollection +from synapse.api.filtering import FilterCollection from synapse.api.presence import UserPresenceState from synapse.events import EventBase from synapse.events.utils import ( @@ -150,7 +150,7 @@ class SyncRestServlet(RestServlet): request_key = (user, timeout, since, filter_id, full_state, device_id) if filter_id is None: - filter_collection = DEFAULT_FILTER_COLLECTION + filter_collection = self.filtering.DEFAULT_FILTER_COLLECTION elif filter_id.startswith("{"): try: filter_object = json_decoder.decode(filter_id) @@ -160,7 +160,7 @@ class SyncRestServlet(RestServlet): except Exception: raise SynapseError(400, "Invalid filter JSON") self.filtering.check_valid_filter(filter_object) - filter_collection = FilterCollection(filter_object) + filter_collection = FilterCollection(self.hs, filter_object) else: try: filter_collection = await self.filtering.get_user_filter( diff --git a/synapse/rest/well_known.py b/synapse/rest/well_known.py index edbf5ce5d0..04b035a1b1 100644 --- a/synapse/rest/well_known.py +++ b/synapse/rest/well_known.py @@ -34,8 +34,7 @@ class WellKnownBuilder: self._config = hs.config def get_well_known(self) -> Optional[JsonDict]: - # if we don't have a public_baseurl, we can't help much here. - if self._config.server.public_baseurl is None: + if not self._config.server.serve_client_wellknown: return None result = {"m.homeserver": {"base_url": self._config.server.public_baseurl}} diff --git a/synapse/server.py b/synapse/server.py index 0fbf36ba99..013a7bacaa 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -463,7 +463,7 @@ class HomeServer(metaclass=abc.ABCMeta): @cache_in_self def get_typing_writer_handler(self) -> TypingWriterHandler: - if self.config.worker.writers.typing == self.get_instance_name(): + if self.get_instance_name() in self.config.worker.writers.typing: return TypingWriterHandler(self) else: raise Exception("Workers cannot write typing") @@ -474,7 +474,7 @@ class HomeServer(metaclass=abc.ABCMeta): @cache_in_self def get_typing_handler(self) -> FollowerTypingHandler: - if self.config.worker.writers.typing == self.get_instance_name(): + if self.get_instance_name() in self.config.worker.writers.typing: # Use get_typing_writer_handler to ensure that we use the same # cached version. return self.get_typing_writer_handler() diff --git a/synapse/storage/background_updates.py b/synapse/storage/background_updates.py index 82b31d24f1..b9a8ca997e 100644 --- a/synapse/storage/background_updates.py +++ b/synapse/storage/background_updates.py @@ -100,29 +100,58 @@ class BackgroundUpdater: ] = {} self._all_done = False + # Whether we're currently running updates + self._running = False + + # Whether background updates are enabled. This allows us to + # enable/disable background updates via the admin API. + self.enabled = True + + def get_current_update(self) -> Optional[BackgroundUpdatePerformance]: + """Returns the current background update, if any.""" + + update_name = self._current_background_update + if not update_name: + return None + + perf = self._background_update_performance.get(update_name) + if not perf: + perf = BackgroundUpdatePerformance(update_name) + + return perf + def start_doing_background_updates(self) -> None: - run_as_background_process("background_updates", self.run_background_updates) + if self.enabled: + run_as_background_process("background_updates", self.run_background_updates) async def run_background_updates(self, sleep: bool = True) -> None: - logger.info("Starting background schema updates") - while True: - if sleep: - await self._clock.sleep(self.BACKGROUND_UPDATE_INTERVAL_MS / 1000.0) + if self._running or not self.enabled: + return - try: - result = await self.do_next_background_update( - self.BACKGROUND_UPDATE_DURATION_MS - ) - except Exception: - logger.exception("Error doing update") - else: - if result: - logger.info( - "No more background updates to do." - " Unscheduling background update task." + self._running = True + + try: + logger.info("Starting background schema updates") + while self.enabled: + if sleep: + await self._clock.sleep(self.BACKGROUND_UPDATE_INTERVAL_MS / 1000.0) + + try: + result = await self.do_next_background_update( + self.BACKGROUND_UPDATE_DURATION_MS ) - self._all_done = True - return None + except Exception: + logger.exception("Error doing update") + else: + if result: + logger.info( + "No more background updates to do." + " Unscheduling background update task." + ) + self._all_done = True + return None + finally: + self._running = False async def has_completed_background_updates(self) -> bool: """Check if all the background updates have completed diff --git a/synapse/storage/database.py b/synapse/storage/database.py index 5c71e27518..d4cab69ebf 100644 --- a/synapse/storage/database.py +++ b/synapse/storage/database.py @@ -446,6 +446,10 @@ class DatabasePool: self._check_safe_to_upsert, ) + def name(self) -> str: + "Return the name of this database" + return self._database_config.name + def is_running(self) -> bool: """Is the database pool currently running""" return self._db_pool.running diff --git a/synapse/storage/databases/main/appservice.py b/synapse/storage/databases/main/appservice.py index 2da2659f41..baec35ee27 100644 --- a/synapse/storage/databases/main/appservice.py +++ b/synapse/storage/databases/main/appservice.py @@ -412,16 +412,16 @@ class ApplicationServiceTransactionWorkerStore( ) async def set_type_stream_id_for_appservice( - self, service: ApplicationService, type: str, pos: Optional[int] + self, service: ApplicationService, stream_type: str, pos: Optional[int] ) -> None: - if type not in ("read_receipt", "presence"): + if stream_type not in ("read_receipt", "presence"): raise ValueError( "Expected type to be a valid application stream id type, got %s" - % (type,) + % (stream_type,) ) def set_type_stream_id_for_appservice_txn(txn): - stream_id_type = "%s_stream_id" % type + stream_id_type = "%s_stream_id" % stream_type txn.execute( "UPDATE application_services_state SET %s = ? WHERE as_id=?" % stream_id_type, diff --git a/synapse/storage/databases/main/deviceinbox.py b/synapse/storage/databases/main/deviceinbox.py index 264e625bd7..ae3afdd5d2 100644 --- a/synapse/storage/databases/main/deviceinbox.py +++ b/synapse/storage/databases/main/deviceinbox.py @@ -134,7 +134,10 @@ class DeviceInboxWorkerStore(SQLBaseStore): limit: The maximum number of messages to retrieve. Returns: - A list of messages for the device and where in the stream the messages got to. + A tuple containing: + * A list of messages for the device. + * The max stream token of these messages. There may be more to retrieve + if the given limit was reached. """ has_changed = self._device_inbox_stream_cache.has_entity_changed( user_id, last_stream_id @@ -153,12 +156,19 @@ class DeviceInboxWorkerStore(SQLBaseStore): txn.execute( sql, (user_id, device_id, last_stream_id, current_stream_id, limit) ) + messages = [] + stream_pos = current_stream_id + for row in txn: stream_pos = row[0] messages.append(db_to_json(row[1])) + + # If the limit was not reached we know that there's no more data for this + # user/device pair up to current_stream_id. if len(messages) < limit: stream_pos = current_stream_id + return messages, stream_pos return await self.db_pool.runInteraction( @@ -260,13 +270,20 @@ class DeviceInboxWorkerStore(SQLBaseStore): " LIMIT ?" ) txn.execute(sql, (destination, last_stream_id, current_stream_id, limit)) + messages = [] + stream_pos = current_stream_id + for row in txn: stream_pos = row[0] messages.append(db_to_json(row[1])) + + # If the limit was not reached we know that there's no more data for this + # user/device pair up to current_stream_id. if len(messages) < limit: log_kv({"message": "Set stream position to current position"}) stream_pos = current_stream_id + return messages, stream_pos return await self.db_pool.runInteraction( @@ -372,8 +389,8 @@ class DeviceInboxWorkerStore(SQLBaseStore): """Used to send messages from this server. Args: - local_messages_by_user_and_device: - Dictionary of user_id to device_id to message. + local_messages_by_user_then_device: + Dictionary of recipient user_id to recipient device_id to message. remote_messages_by_destination: Dictionary of destination server_name to the EDU JSON to send. diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index ae37901be9..c6bf316d5b 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -28,6 +28,7 @@ from typing import ( import attr from constantly import NamedConstant, Names +from prometheus_client import Gauge from typing_extensions import Literal from twisted.internet import defer @@ -81,6 +82,12 @@ EVENT_QUEUE_ITERATIONS = 3 # No. times we block waiting for requests for events EVENT_QUEUE_TIMEOUT_S = 0.1 # Timeout when waiting for requests for events +event_fetch_ongoing_gauge = Gauge( + "synapse_event_fetch_ongoing", + "The number of event fetchers that are running", +) + + @attr.s(slots=True, auto_attribs=True) class _EventCacheEntry: event: EventBase @@ -222,6 +229,7 @@ class EventsWorkerStore(SQLBaseStore): self._event_fetch_lock = threading.Condition() self._event_fetch_list = [] self._event_fetch_ongoing = 0 + event_fetch_ongoing_gauge.set(self._event_fetch_ongoing) # We define this sequence here so that it can be referenced from both # the DataStore and PersistEventStore. @@ -732,28 +740,31 @@ class EventsWorkerStore(SQLBaseStore): """Takes a database connection and waits for requests for events from the _event_fetch_list queue. """ - i = 0 - while True: - with self._event_fetch_lock: - event_list = self._event_fetch_list - self._event_fetch_list = [] - - if not event_list: - single_threaded = self.database_engine.single_threaded - 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: - self._event_fetch_lock.wait(EVENT_QUEUE_TIMEOUT_S) - i += 1 - continue - i = 0 - - self._fetch_event_list(conn, event_list) + try: + i = 0 + while True: + with self._event_fetch_lock: + event_list = self._event_fetch_list + self._event_fetch_list = [] + + if not event_list: + single_threaded = self.database_engine.single_threaded + if ( + not self.USE_DEDICATED_DB_THREADS_FOR_EVENT_FETCHING + or single_threaded + or i > EVENT_QUEUE_ITERATIONS + ): + break + else: + self._event_fetch_lock.wait(EVENT_QUEUE_TIMEOUT_S) + i += 1 + continue + i = 0 + + self._fetch_event_list(conn, event_list) + finally: + self._event_fetch_ongoing -= 1 + event_fetch_ongoing_gauge.set(self._event_fetch_ongoing) def _fetch_event_list( self, conn: Connection, event_list: List[Tuple[List[str], defer.Deferred]] @@ -977,6 +988,7 @@ class EventsWorkerStore(SQLBaseStore): if self._event_fetch_ongoing < EVENT_QUEUE_THREADS: self._event_fetch_ongoing += 1 + event_fetch_ongoing_gauge.set(self._event_fetch_ongoing) should_start = True else: should_start = False diff --git a/synapse/storage/databases/main/lock.py b/synapse/storage/databases/main/lock.py index 3d1dff660b..3d0df0cbd4 100644 --- a/synapse/storage/databases/main/lock.py +++ b/synapse/storage/databases/main/lock.py @@ -14,6 +14,7 @@ import logging from types import TracebackType from typing import TYPE_CHECKING, Dict, Optional, Tuple, Type +from weakref import WeakValueDictionary from twisted.internet.interfaces import IReactorCore @@ -61,7 +62,7 @@ class LockStore(SQLBaseStore): # A map from `(lock_name, lock_key)` to the token of any locks that we # think we currently hold. - self._live_tokens: Dict[Tuple[str, str], str] = {} + self._live_tokens: Dict[Tuple[str, str], Lock] = WeakValueDictionary() # When we shut down we want to remove the locks. Technically this can # lead to a race, as we may drop the lock while we are still processing. @@ -80,10 +81,10 @@ class LockStore(SQLBaseStore): # We need to take a copy of the tokens dict as dropping the locks will # cause the dictionary to change. - tokens = dict(self._live_tokens) + locks = dict(self._live_tokens) - for (lock_name, lock_key), token in tokens.items(): - await self._drop_lock(lock_name, lock_key, token) + for lock in locks.values(): + await lock.release() logger.info("Dropped locks due to shutdown") @@ -93,6 +94,11 @@ class LockStore(SQLBaseStore): used (otherwise the lock will leak). """ + # Check if this process has taken out a lock and if it's still valid. + lock = self._live_tokens.get((lock_name, lock_key)) + if lock and await lock.is_still_valid(): + return None + now = self._clock.time_msec() token = random_string(6) @@ -100,7 +106,9 @@ class LockStore(SQLBaseStore): def _try_acquire_lock_txn(txn: LoggingTransaction) -> bool: # We take out the lock if either a) there is no row for the lock - # already or b) the existing row has timed out. + # already, b) the existing row has timed out, or c) the row is + # for this instance (which means the process got killed and + # restarted) sql = """ INSERT INTO worker_locks (lock_name, lock_key, instance_name, token, last_renewed_ts) VALUES (?, ?, ?, ?, ?) @@ -112,6 +120,7 @@ class LockStore(SQLBaseStore): last_renewed_ts = EXCLUDED.last_renewed_ts WHERE worker_locks.last_renewed_ts < ? + OR worker_locks.instance_name = EXCLUDED.instance_name """ txn.execute( sql, @@ -148,11 +157,11 @@ class LockStore(SQLBaseStore): WHERE lock_name = ? AND lock_key = ? - AND last_renewed_ts < ? + AND (last_renewed_ts < ? OR instance_name = ?) """ txn.execute( sql, - (lock_name, lock_key, now - _LOCK_TIMEOUT_MS), + (lock_name, lock_key, now - _LOCK_TIMEOUT_MS, self._instance_name), ) inserted = self.db_pool.simple_upsert_txn_emulated( @@ -179,9 +188,7 @@ class LockStore(SQLBaseStore): if not did_lock: return None - self._live_tokens[(lock_name, lock_key)] = token - - return Lock( + lock = Lock( self._reactor, self._clock, self, @@ -190,6 +197,10 @@ class LockStore(SQLBaseStore): token=token, ) + self._live_tokens[(lock_name, lock_key)] = lock + + return lock + async def _is_lock_still_valid( self, lock_name: str, lock_key: str, token: str ) -> bool: diff --git a/synapse/storage/databases/main/relations.py b/synapse/storage/databases/main/relations.py index 53576ad52f..907af10995 100644 --- a/synapse/storage/databases/main/relations.py +++ b/synapse/storage/databases/main/relations.py @@ -20,7 +20,7 @@ import attr from synapse.api.constants import RelationTypes from synapse.events import EventBase from synapse.storage._base import SQLBaseStore -from synapse.storage.database import LoggingTransaction +from synapse.storage.database import LoggingTransaction, make_in_list_sql_clause from synapse.storage.databases.main.stream import generate_pagination_where_clause from synapse.storage.relations import ( AggregationPaginationToken, @@ -334,6 +334,62 @@ class RelationsWorkerStore(SQLBaseStore): return count, latest_event + async def events_have_relations( + self, + parent_ids: List[str], + relation_senders: Optional[List[str]], + relation_types: Optional[List[str]], + ) -> List[str]: + """Check which events have a relationship from the given senders of the + given types. + + Args: + parent_ids: The events being annotated + relation_senders: The relation senders to check. + relation_types: The relation types to check. + + Returns: + True if the event has at least one relationship from one of the given senders of the given type. + """ + # If no restrictions are given then the event has the required relations. + if not relation_senders and not relation_types: + return parent_ids + + sql = """ + SELECT relates_to_id FROM event_relations + INNER JOIN events USING (event_id) + WHERE + %s; + """ + + def _get_if_event_has_relations(txn) -> List[str]: + clauses: List[str] = [] + clause, args = make_in_list_sql_clause( + txn.database_engine, "relates_to_id", parent_ids + ) + clauses.append(clause) + + if relation_senders: + clause, temp_args = make_in_list_sql_clause( + txn.database_engine, "sender", relation_senders + ) + clauses.append(clause) + args.extend(temp_args) + if relation_types: + clause, temp_args = make_in_list_sql_clause( + txn.database_engine, "relation_type", relation_types + ) + clauses.append(clause) + args.extend(temp_args) + + txn.execute(sql % " AND ".join(clauses), args) + + return [row[0] for row in txn] + + return await self.db_pool.runInteraction( + "get_if_event_has_relations", _get_if_event_has_relations + ) + async def has_user_annotated_event( self, parent_id: str, event_type: str, aggregation_key: str, sender: str ) -> bool: diff --git a/synapse/storage/databases/main/room.py b/synapse/storage/databases/main/room.py index cefc77fa0f..17b398bb69 100644 --- a/synapse/storage/databases/main/room.py +++ b/synapse/storage/databases/main/room.py @@ -1751,7 +1751,12 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore, SearchStore): ) async def block_room(self, room_id: str, user_id: str) -> None: - """Marks the room as blocked. Can be called multiple times. + """Marks the room as blocked. + + Can be called multiple times (though we'll only track the last user to + block this room). + + Can be called on a room unknown to this homeserver. Args: room_id: Room to block diff --git a/synapse/storage/databases/main/stream.py b/synapse/storage/databases/main/stream.py index dc7884b1c0..42dc807d17 100644 --- a/synapse/storage/databases/main/stream.py +++ b/synapse/storage/databases/main/stream.py @@ -272,31 +272,37 @@ def filter_to_clause(event_filter: Optional[Filter]) -> Tuple[str, List[str]]: args = [] if event_filter.types: - clauses.append("(%s)" % " OR ".join("type = ?" for _ in event_filter.types)) + clauses.append( + "(%s)" % " OR ".join("event.type = ?" for _ in event_filter.types) + ) args.extend(event_filter.types) for typ in event_filter.not_types: - clauses.append("type != ?") + clauses.append("event.type != ?") args.append(typ) if event_filter.senders: - clauses.append("(%s)" % " OR ".join("sender = ?" for _ in event_filter.senders)) + clauses.append( + "(%s)" % " OR ".join("event.sender = ?" for _ in event_filter.senders) + ) args.extend(event_filter.senders) for sender in event_filter.not_senders: - clauses.append("sender != ?") + clauses.append("event.sender != ?") args.append(sender) if event_filter.rooms: - clauses.append("(%s)" % " OR ".join("room_id = ?" for _ in event_filter.rooms)) + clauses.append( + "(%s)" % " OR ".join("event.room_id = ?" for _ in event_filter.rooms) + ) args.extend(event_filter.rooms) for room_id in event_filter.not_rooms: - clauses.append("room_id != ?") + clauses.append("event.room_id != ?") args.append(room_id) if event_filter.contains_url: - clauses.append("contains_url = ?") + clauses.append("event.contains_url = ?") args.append(event_filter.contains_url) # We're only applying the "labels" filter on the database query, because applying the @@ -307,6 +313,23 @@ def filter_to_clause(event_filter: Optional[Filter]) -> Tuple[str, List[str]]: clauses.append("(%s)" % " OR ".join("label = ?" for _ in event_filter.labels)) args.extend(event_filter.labels) + # Filter on relation_senders / relation types from the joined tables. + if event_filter.relation_senders: + clauses.append( + "(%s)" + % " OR ".join( + "related_event.sender = ?" for _ in event_filter.relation_senders + ) + ) + args.extend(event_filter.relation_senders) + + if event_filter.relation_types: + clauses.append( + "(%s)" + % " OR ".join("relation_type = ?" for _ in event_filter.relation_types) + ) + args.extend(event_filter.relation_types) + return " AND ".join(clauses), args @@ -1116,7 +1139,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): bounds = generate_pagination_where_clause( direction=direction, - column_names=("topological_ordering", "stream_ordering"), + column_names=("event.topological_ordering", "event.stream_ordering"), from_token=from_bound, to_token=to_bound, engine=self.database_engine, @@ -1133,32 +1156,51 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore, metaclass=abc.ABCMeta): select_keywords = "SELECT" join_clause = "" + # Using DISTINCT in this SELECT query is quite expensive, because it + # requires the engine to sort on the entire (not limited) result set, + # i.e. the entire events table. Only use it in scenarios that could result + # in the same event ID occurring multiple times in the results. + needs_distinct = False if event_filter and event_filter.labels: # If we're not filtering on a label, then joining on event_labels will # return as many row for a single event as the number of labels it has. To # avoid this, only join if we're filtering on at least one label. - join_clause = """ + join_clause += """ LEFT JOIN event_labels USING (event_id, room_id, topological_ordering) """ if len(event_filter.labels) > 1: - # Using DISTINCT in this SELECT query is quite expensive, because it - # requires the engine to sort on the entire (not limited) result set, - # i.e. the entire events table. We only need to use it when we're - # filtering on more than two labels, because that's the only scenario - # in which we can possibly to get multiple times the same event ID in - # the results. - select_keywords += "DISTINCT" + # Multiple labels could cause the same event to appear multiple times. + needs_distinct = True + + # If there is a filter on relation_senders and relation_types join to the + # relations table. + if event_filter and ( + event_filter.relation_senders or event_filter.relation_types + ): + # Filtering by relations could cause the same event to appear multiple + # times (since there's no limit on the number of relations to an event). + needs_distinct = True + join_clause += """ + LEFT JOIN event_relations AS relation ON (event.event_id = relation.relates_to_id) + """ + if event_filter.relation_senders: + join_clause += """ + LEFT JOIN events AS related_event ON (relation.event_id = related_event.event_id) + """ + + if needs_distinct: + select_keywords += " DISTINCT" sql = """ %(select_keywords)s - event_id, instance_name, - topological_ordering, stream_ordering - FROM events + event.event_id, event.instance_name, + event.topological_ordering, event.stream_ordering + FROM events AS event %(join_clause)s - WHERE outlier = ? AND room_id = ? AND %(bounds)s - ORDER BY topological_ordering %(order)s, - stream_ordering %(order)s LIMIT ? + WHERE event.outlier = ? AND event.room_id = ? AND %(bounds)s + ORDER BY event.topological_ordering %(order)s, + event.stream_ordering %(order)s LIMIT ? """ % { "select_keywords": select_keywords, "join_clause": join_clause, diff --git a/synapse/storage/prepare_database.py b/synapse/storage/prepare_database.py index 1629d2a53c..8b9c6adae2 100644 --- a/synapse/storage/prepare_database.py +++ b/synapse/storage/prepare_database.py @@ -133,22 +133,23 @@ def prepare_database( # if it's a worker app, refuse to upgrade the database, to avoid multiple # workers doing it at once. - if ( - config.worker.worker_app is not None - and version_info.current_version != SCHEMA_VERSION - ): + if config.worker.worker_app is None: + _upgrade_existing_database( + cur, + version_info, + database_engine, + config, + databases=databases, + ) + elif version_info.current_version < SCHEMA_VERSION: + # If the DB is on an older version than we expect then we refuse + # to start the worker (as the main process needs to run first to + # update the schema). raise UpgradeDatabaseException( OUTDATED_SCHEMA_ON_WORKER_ERROR % (SCHEMA_VERSION, version_info.current_version) ) - _upgrade_existing_database( - cur, - version_info, - database_engine, - config, - databases=databases, - ) else: logger.info("%r: Initialising new database", databases) |