diff options
| author | Olivier Wilkinson (reivilibre) <oliverw@matrix.org> | 2022-12-14 13:24:38 +0000 |
|---|---|---|
| committer | Olivier Wilkinson (reivilibre) <oliverw@matrix.org> | 2022-12-14 13:24:38 +0000 |
| commit | bbd59237fcaec9c3cc3a6d98d06012d19f26ef41 (patch) | |
| tree | d5055b172f25d05ae7bf3a603e29942cdba377a2 /synapse | |
| parent | Use uvloop as an asyncio reactor event loop (diff) | |
| parent | Tidy up CHANGES.md (diff) | |
| download | synapse-rei/uvloop.tar.xz | |
Merge tag 'v1.74.0rc1' into rei/uvloop github/rei/uvloop rei/uvloop
Synapse 1.74.0rc1 (2022-12-13) ============================== Features -------- - Improve user search for international display names. ([\#14464](https://github.com/matrix-org/synapse/issues/14464)) - Stop using deprecated `keyIds` parameter when calling `/_matrix/key/v2/server`. ([\#14490](https://github.com/matrix-org/synapse/issues/14490), [\#14525](https://github.com/matrix-org/synapse/issues/14525)) - Add new `push.enabled` config option to allow opting out of push notification calculation. ([\#14551](https://github.com/matrix-org/synapse/issues/14551), [\#14619](https://github.com/matrix-org/synapse/issues/14619)) - Advertise support for Matrix 1.5 on `/_matrix/client/versions`. ([\#14576](https://github.com/matrix-org/synapse/issues/14576)) - Improve opentracing and logging for to-device message handling. ([\#14598](https://github.com/matrix-org/synapse/issues/14598)) - Allow selecting "prejoin" events by state keys in addition to event types. ([\#14642](https://github.com/matrix-org/synapse/issues/14642)) Bugfixes -------- - Fix a long-standing bug where a device list update might not be sent to clients in certain circumstances. ([\#14435](https://github.com/matrix-org/synapse/issues/14435), [\#14592](https://github.com/matrix-org/synapse/issues/14592), [\#14604](https://github.com/matrix-org/synapse/issues/14604)) - Suppress a spurious warning when `POST /rooms/<room_id>/<membership>/`, `POST /join/<room_id_or_alias`, or the unspecced `PUT /join/<room_id_or_alias>/<txn_id>` receive an empty HTTP request body. ([\#14600](https://github.com/matrix-org/synapse/issues/14600)) - Return spec-compliant JSON errors when unknown endpoints are requested. ([\#14620](https://github.com/matrix-org/synapse/issues/14620), [\#14621](https://github.com/matrix-org/synapse/issues/14621)) - Update html templates to load images over HTTPS. Contributed by @ashfame. ([\#14625](https://github.com/matrix-org/synapse/issues/14625)) - Fix a long-standing bug where the user directory would return 1 more row than requested. ([\#14631](https://github.com/matrix-org/synapse/issues/14631)) - Reject invalid read receipt requests with empty room or event IDs. Contributed by Nick @ Beeper (@fizzadar). ([\#14632](https://github.com/matrix-org/synapse/issues/14632)) - Fix a bug introduced in Synapse 1.67.0 where not specifying a config file or a server URL would lead to the `register_new_matrix_user` script failing. ([\#14637](https://github.com/matrix-org/synapse/issues/14637)) - Fix a long-standing bug where the user directory and room/user stats might be out of sync. ([\#14639](https://github.com/matrix-org/synapse/issues/14639), [\#14643](https://github.com/matrix-org/synapse/issues/14643)) - Fix a bug introduced in Synapse 1.72.0 where the background updates to add non-thread unique indexes on receipts would fail if they were previously interrupted. ([\#14650](https://github.com/matrix-org/synapse/issues/14650)) - Improve validation of field size limits in events. ([\#14664](https://github.com/matrix-org/synapse/issues/14664)) - Fix bugs introduced in Synapse 1.55.0 and 1.69.0 where application services would not be notified of events in the correct rooms, due to stale caches. ([\#14670](https://github.com/matrix-org/synapse/issues/14670)) Improved Documentation ---------------------- - Update worker settings for `pusher` and `federation_sender` functionality. ([\#14493](https://github.com/matrix-org/synapse/issues/14493)) - Add links to third party package repositories, and point to the bug which highlights Ubuntu's out-of-date packages. ([\#14517](https://github.com/matrix-org/synapse/issues/14517)) - Remove old, incorrect minimum postgres version note and replace with a link to the [Dependency Deprecation Policy](https://matrix-org.github.io/synapse/v1.73/deprecation_policy.html). ([\#14590](https://github.com/matrix-org/synapse/issues/14590)) - Add Single-Sign On setup instructions for Mastodon-based instances. ([\#14594](https://github.com/matrix-org/synapse/issues/14594)) - Change `turn_allow_guests` example value to lowercase `true`. ([\#14634](https://github.com/matrix-org/synapse/issues/14634)) Internal Changes ---------------- - Optimise push badge count calculations. Contributed by Nick @ Beeper (@fizzadar). ([\#14255](https://github.com/matrix-org/synapse/issues/14255)) - Faster remote room joins: stream the un-partial-stating of rooms over replication. ([\#14473](https://github.com/matrix-org/synapse/issues/14473), [\#14474](https://github.com/matrix-org/synapse/issues/14474)) - Share the `ClientRestResource` for both workers and the main process. ([\#14528](https://github.com/matrix-org/synapse/issues/14528)) - Add `--editable` flag to `complement.sh` which uses an editable install of Synapse for faster turn-around times whilst developing iteratively. ([\#14548](https://github.com/matrix-org/synapse/issues/14548)) - Faster joins: use servers list approximation to send read receipts when in partial state instead of waiting for the full state of the room. ([\#14549](https://github.com/matrix-org/synapse/issues/14549)) - Modernize unit tests configuration related to workers. ([\#14568](https://github.com/matrix-org/synapse/issues/14568)) - Bump jsonschema from 4.17.0 to 4.17.3. ([\#14591](https://github.com/matrix-org/synapse/issues/14591)) - Fix Rust lint CI. ([\#14602](https://github.com/matrix-org/synapse/issues/14602)) - Bump JasonEtco/create-an-issue from 2.5.0 to 2.8.1. ([\#14607](https://github.com/matrix-org/synapse/issues/14607)) - Alter some unit test environment parameters to decrease time spent running tests. ([\#14610](https://github.com/matrix-org/synapse/issues/14610)) - Switch to Go recommended installation method for `gotestfmt` template in CI. ([\#14611](https://github.com/matrix-org/synapse/issues/14611)) - Bump phonenumbers from 8.13.0 to 8.13.1. ([\#14612](https://github.com/matrix-org/synapse/issues/14612)) - Bump types-setuptools from 65.5.0.3 to 65.6.0.1. ([\#14613](https://github.com/matrix-org/synapse/issues/14613)) - Bump twine from 4.0.1 to 4.0.2. ([\#14614](https://github.com/matrix-org/synapse/issues/14614)) - Bump types-requests from 2.28.11.2 to 2.28.11.5. ([\#14615](https://github.com/matrix-org/synapse/issues/14615)) - Bump cryptography from 38.0.3 to 38.0.4. ([\#14616](https://github.com/matrix-org/synapse/issues/14616)) - Remove useless cargo install with apt from Dockerfile. ([\#14636](https://github.com/matrix-org/synapse/issues/14636)) - Bump certifi from 2021.10.8 to 2022.12.7. ([\#14645](https://github.com/matrix-org/synapse/issues/14645)) - Bump flake8-bugbear from 22.10.27 to 22.12.6. ([\#14656](https://github.com/matrix-org/synapse/issues/14656)) - Bump packaging from 21.3 to 22.0. ([\#14657](https://github.com/matrix-org/synapse/issues/14657)) - Bump types-pillow from 9.3.0.1 to 9.3.0.4. ([\#14658](https://github.com/matrix-org/synapse/issues/14658)) - Bump serde from 1.0.148 to 1.0.150. ([\#14659](https://github.com/matrix-org/synapse/issues/14659)) - Bump phonenumbers from 8.13.1 to 8.13.2. ([\#14660](https://github.com/matrix-org/synapse/issues/14660)) - Bump authlib from 1.1.0 to 1.2.0. ([\#14661](https://github.com/matrix-org/synapse/issues/14661)) - Move `StateFilter` to `synapse.types`. ([\#14668](https://github.com/matrix-org/synapse/issues/14668)) - Improve type hints. ([\#14597](https://github.com/matrix-org/synapse/issues/14597), [\#14646](https://github.com/matrix-org/synapse/issues/14646), [\#14671](https://github.com/matrix-org/synapse/issues/14671))
Diffstat (limited to '')
250 files changed, 7820 insertions, 4535 deletions
diff --git a/synapse/replication/slave/storage/keys.py b/stubs/icu.pyi index a00b38c512..efeda7938a 100644 --- a/synapse/replication/slave/storage/keys.py +++ b/stubs/icu.pyi @@ -1,4 +1,4 @@ -# Copyright 2015, 2016 OpenMarket Ltd +# Copyright 2022 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. @@ -12,9 +12,14 @@ # See the License for the specific language governing permissions and # limitations under the License. -from synapse.storage.databases.main.keys import KeyStore +# Stub for PyICU. -# KeyStore isn't really safe to use from a worker, but for now we do so and hope that -# the races it creates aren't too bad. +class Locale: + @staticmethod + def getDefault() -> Locale: ... -SlavedKeyStore = KeyStore +class BreakIterator: + @staticmethod + def createWordInstance(locale: Locale) -> BreakIterator: ... + def setText(self, text: str) -> None: ... + def nextBoundary(self) -> int: ... diff --git a/synapse/__init__.py b/synapse/__init__.py index 7bb07cfd05..a6623e2956 100644 --- a/synapse/__init__.py +++ b/synapse/__init__.py @@ -21,6 +21,7 @@ import os import sys from synapse.util.rust import check_rust_lib_up_to_date +from synapse.util.stringutils import strtobool # Check that we're not running on an unsupported Python version. if sys.version_info < (3, 7): @@ -28,31 +29,29 @@ if sys.version_info < (3, 7): sys.exit(1) # Allow using the asyncio reactor via env var. -if bool(os.environ.get("SYNAPSE_ASYNC_IO_REACTOR", False)): - try: - from incremental import Version +if strtobool(os.environ.get("SYNAPSE_ASYNC_IO_REACTOR", "0")): + from incremental import Version - import twisted + import twisted - # We need a bugfix that is included in Twisted 21.2.0: - # https://twistedmatrix.com/trac/ticket/9787 - if twisted.version < Version("Twisted", 21, 2, 0): - print("Using asyncio reactor requires Twisted>=21.2.0") - sys.exit(1) + # We need a bugfix that is included in Twisted 21.2.0: + # https://twistedmatrix.com/trac/ticket/9787 + if twisted.version < Version("Twisted", 21, 2, 0): + print("Using asyncio reactor requires Twisted>=21.2.0") + sys.exit(1) - import asyncio + import asyncio - from twisted.internet import asyncioreactor + from twisted.internet import asyncioreactor - if bool(os.environ.get("SYNAPSE_UVLOOP", False)): - import uvloop + if bool(os.environ.get("SYNAPSE_UVLOOP", False)): + import uvloop - uvloop.install() - print("Using uvloop") + uvloop.install() + print("Using uvloop") + + asyncioreactor.install(asyncio.get_event_loop()) - asyncioreactor.install(asyncio.get_event_loop()) - except ImportError: - pass # Twisted and canonicaljson will fail to import when this file is executed to # get the __version__ during a fresh install. That's OK and subsequent calls to diff --git a/synapse/_scripts/register_new_matrix_user.py b/synapse/_scripts/register_new_matrix_user.py index 0c4504d5d8..2b74a40166 100644 --- a/synapse/_scripts/register_new_matrix_user.py +++ b/synapse/_scripts/register_new_matrix_user.py @@ -222,6 +222,7 @@ def main() -> None: args = parser.parse_args() + config: Optional[Dict[str, Any]] = None if "config" in args and args.config: config = yaml.safe_load(args.config) @@ -229,7 +230,7 @@ def main() -> None: secret = args.shared_secret else: # argparse should check that we have either config or shared secret - assert config + assert config is not None secret = config.get("registration_shared_secret") secret_file = config.get("registration_shared_secret_path") @@ -244,7 +245,7 @@ def main() -> None: if args.server_url: server_url = args.server_url - elif config: + elif config is not None: server_url = _find_client_listener(config) if not server_url: server_url = _DEFAULT_SERVER_URL diff --git a/synapse/_scripts/synapse_port_db.py b/synapse/_scripts/synapse_port_db.py index 5fa599e70e..d850e54e17 100755 --- a/synapse/_scripts/synapse_port_db.py +++ b/synapse/_scripts/synapse_port_db.py @@ -72,6 +72,7 @@ from synapse.storage.databases.main.registration import ( RegistrationBackgroundUpdateStore, find_max_generated_user_id_localpart, ) +from synapse.storage.databases.main.relations import RelationsWorkerStore from synapse.storage.databases.main.room import RoomBackgroundUpdateStore from synapse.storage.databases.main.roommember import RoomMemberBackgroundUpdateStore from synapse.storage.databases.main.search import SearchBackgroundUpdateStore @@ -206,6 +207,7 @@ class Store( PusherWorkerStore, PresenceBackgroundUpdateStore, ReceiptsBackgroundUpdateStore, + RelationsWorkerStore, ): def execute(self, f: Callable[..., R], *args: Any, **kwargs: Any) -> Awaitable[R]: return self.db_pool.runInteraction(f.__name__, f, *args, **kwargs) diff --git a/synapse/_scripts/update_synapse_database.py b/synapse/_scripts/update_synapse_database.py index fb1fb83f50..0adf94bba6 100755..100644 --- a/synapse/_scripts/update_synapse_database.py +++ b/synapse/_scripts/update_synapse_database.py @@ -15,7 +15,6 @@ import argparse import logging -import sys from typing import cast import yaml @@ -100,13 +99,6 @@ def main() -> None: # Load, process and sanity-check the config. hs_config = yaml.safe_load(args.database_config) - if "database" not in hs_config and "databases" not in hs_config: - sys.stderr.write( - "The configuration file must have a 'database' or 'databases' section. " - "See https://matrix-org.github.io/synapse/latest/usage/configuration/config_documentation.html#database" - ) - sys.exit(4) - config = HomeServerConfig() config.parse_config_dict(hs_config, "", "") diff --git a/synapse/api/constants.py b/synapse/api/constants.py index 44c5ffc6a5..6a5e7171da 100644 --- a/synapse/api/constants.py +++ b/synapse/api/constants.py @@ -125,6 +125,8 @@ class EventTypes: MSC2716_BATCH: Final = "org.matrix.msc2716.batch" MSC2716_MARKER: Final = "org.matrix.msc2716.marker" + Reaction: Final = "m.reaction" + class ToDeviceEventTypes: RoomKeyRequest: Final = "m.room_key_request" @@ -150,6 +152,7 @@ class EduTypes: class RejectedReason: AUTH_ERROR: Final = "auth_error" + OVERSIZED_EVENT: Final = "oversized_event" class RoomCreationPreset: @@ -228,6 +231,9 @@ class EventContentFields: # The authorising user for joining a restricted room. AUTHORISING_USER: Final = "join_authorised_via_users_server" + # an unspecced field added to to-device messages to identify them uniquely-ish + TO_DEVICE_MSGID: Final = "org.matrix.msgid" + class RoomTypes: """Understood values of the room_type field of m.room.create events.""" diff --git a/synapse/api/errors.py b/synapse/api/errors.py index c606207569..c2c177fd71 100644 --- a/synapse/api/errors.py +++ b/synapse/api/errors.py @@ -155,7 +155,13 @@ class RedirectException(CodeMessageException): class SynapseError(CodeMessageException): """A base exception type for matrix errors which have an errcode and error - message (as well as an HTTP status code). + message (as well as an HTTP status code). These often bubble all the way up to the + client API response so the error code and status often reach the client directly as + defined here. If the error doesn't make sense to present to a client, then it + probably shouldn't be a `SynapseError`. For example, if we contact another + homeserver over federation, we shouldn't automatically ferry response errors back to + the client on our end (a 500 from a remote server does not make sense to a client + when our server did not experience a 500). Attributes: errcode: Matrix error code e.g 'M_FORBIDDEN' @@ -294,10 +300,8 @@ class InteractiveAuthIncompleteError(Exception): class UnrecognizedRequestError(SynapseError): """An error indicating we don't understand the request you're trying to make""" - def __init__( - self, msg: str = "Unrecognized request", errcode: str = Codes.UNRECOGNIZED - ): - super().__init__(400, msg, errcode) + def __init__(self, msg: str = "Unrecognized request", code: int = 400): + super().__init__(code, msg, Codes.UNRECOGNIZED) class NotFoundError(SynapseError): @@ -420,8 +424,17 @@ class ResourceLimitError(SynapseError): class EventSizeError(SynapseError): """An error raised when an event is too big.""" - def __init__(self, msg: str): + def __init__(self, msg: str, unpersistable: bool): + """ + unpersistable: + if True, the PDU must not be persisted, not even as a rejected PDU + when received over federation. + This is notably true when the entire PDU exceeds the size limit for a PDU, + (as opposed to an individual key's size limit being exceeded). + """ + super().__init__(413, msg, Codes.TOO_LARGE) + self.unpersistable = unpersistable class LoginError(SynapseError): @@ -600,8 +613,20 @@ def cs_error(msg: str, code: str = Codes.UNKNOWN, **kwargs: Any) -> "JsonDict": class FederationError(RuntimeError): - """This class is used to inform remote homeservers about erroneous - PDUs they sent us. + """ + Raised when we process an erroneous PDU. + + There are two kinds of scenarios where this exception can be raised: + + 1. We may pull an invalid PDU from a remote homeserver (e.g. during backfill). We + raise this exception to signal an error to the rest of the application. + 2. We may be pushed an invalid PDU as part of a `/send` transaction from a remote + homeserver. We raise so that we can respond to the transaction and include the + error string in the "PDU Processing Result". The message which will likely be + ignored by the remote homeserver and is not machine parse-able since it's just a + string. + + TODO: In the future, we should split these usage scenarios into their own error types. FATAL: The remote server could not interpret the source event. (e.g., it was missing a required field) @@ -640,6 +665,27 @@ class FederationError(RuntimeError): } +class FederationPullAttemptBackoffError(RuntimeError): + """ + Raised to indicate that we are are deliberately not attempting to pull the given + event over federation because we've already done so recently and are backing off. + + Attributes: + event_id: The event_id which we are refusing to pull + message: A custom error message that gives more context + """ + + def __init__(self, event_ids: List[str], message: Optional[str]): + self.event_ids = event_ids + + if message: + error_message = message + else: + error_message = f"Not attempting to pull event_ids={self.event_ids} because we already tried to pull them recently (backing off)." + + super().__init__(error_message) + + class HttpResponseException(CodeMessageException): """ Represents an HTTP-level failure of an outbound request @@ -674,7 +720,7 @@ class HttpResponseException(CodeMessageException): set to the reason code from the HTTP response. Returns: - SynapseError: + The error converted to a SynapseError. """ # try to parse the body as json, to get better errcode/msg, but # default to M_UNKNOWN with the HTTP status as the error text diff --git a/synapse/api/filtering.py b/synapse/api/filtering.py index cc31cf8cc7..a9888381b4 100644 --- a/synapse/api/filtering.py +++ b/synapse/api/filtering.py @@ -36,14 +36,14 @@ from jsonschema import FormatChecker from synapse.api.constants import EduTypes, EventContentFields from synapse.api.errors import SynapseError from synapse.api.presence import UserPresenceState -from synapse.events import EventBase +from synapse.events import EventBase, relation_from_event from synapse.types import JsonDict, RoomID, UserID if TYPE_CHECKING: from synapse.server import HomeServer FILTER_SCHEMA = { - "additionalProperties": False, + "additionalProperties": True, # Allow new fields for forward compatibility "type": "object", "properties": { "limit": {"type": "number"}, @@ -53,11 +53,17 @@ FILTER_SCHEMA = { # check types are valid event types "types": {"type": "array", "items": {"type": "string"}}, "not_types": {"type": "array", "items": {"type": "string"}}, + # MSC3874, filtering /messages. + "org.matrix.msc3874.rel_types": {"type": "array", "items": {"type": "string"}}, + "org.matrix.msc3874.not_rel_types": { + "type": "array", + "items": {"type": "string"}, + }, }, } ROOM_FILTER_SCHEMA = { - "additionalProperties": False, + "additionalProperties": True, # Allow new fields for forward compatibility "type": "object", "properties": { "not_rooms": {"$ref": "#/definitions/room_id_array"}, @@ -71,7 +77,7 @@ ROOM_FILTER_SCHEMA = { } ROOM_EVENT_FILTER_SCHEMA = { - "additionalProperties": False, + "additionalProperties": True, # Allow new fields for forward compatibility "type": "object", "properties": { "limit": {"type": "number"}, @@ -137,7 +143,7 @@ USER_FILTER_SCHEMA = { }, }, }, - "additionalProperties": False, + "additionalProperties": True, # Allow new fields for forward compatibility } @@ -334,8 +340,15 @@ class Filter: self.labels = filter_json.get("org.matrix.labels", None) self.not_labels = filter_json.get("org.matrix.not_labels", []) - self.related_by_senders = self.filter_json.get("related_by_senders", None) - self.related_by_rel_types = self.filter_json.get("related_by_rel_types", None) + self.related_by_senders = filter_json.get("related_by_senders", None) + self.related_by_rel_types = filter_json.get("related_by_rel_types", None) + + # For compatibility with _check_fields. + self.rel_types = None + self.not_rel_types = [] + if hs.config.experimental.msc3874_enabled: + self.rel_types = filter_json.get("org.matrix.msc3874.rel_types", None) + self.not_rel_types = filter_json.get("org.matrix.msc3874.not_rel_types", []) def filters_all_types(self) -> bool: return "*" in self.not_types @@ -386,11 +399,19 @@ class Filter: # check if there is a string url field in the content for filtering purposes labels = content.get(EventContentFields.LABELS, []) + # Check if the event has a relation. + rel_type = None + if isinstance(event, EventBase): + relation = relation_from_event(event) + if relation: + rel_type = relation.rel_type + field_matchers = { "rooms": lambda v: room_id == v, "senders": lambda v: sender == v, "types": lambda v: _matches_wildcard(ev_type, v), "labels": lambda v: v in labels, + "rel_types": lambda v: rel_type == v, } result = self._check_fields(field_matchers) diff --git a/synapse/api/ratelimiting.py b/synapse/api/ratelimiting.py index 044c7d4926..511790c7c5 100644 --- a/synapse/api/ratelimiting.py +++ b/synapse/api/ratelimiting.py @@ -343,6 +343,7 @@ class RequestRatelimiter: requester: Requester, update: bool = True, is_admin_redaction: bool = False, + n_actions: int = 1, ) -> None: """Ratelimits requests. @@ -355,6 +356,8 @@ class RequestRatelimiter: is_admin_redaction: Whether this is a room admin/moderator redacting an event. If so then we may apply different ratelimits depending on config. + n_actions: Multiplier for the number of actions to apply to the + rate limiter at once. Raises: LimitExceededError if the request should be ratelimited @@ -383,7 +386,9 @@ class RequestRatelimiter: if is_admin_redaction and self.admin_redaction_ratelimiter: # If we have separate config for admin redactions, use a separate # ratelimiter as to not have user_ids clash - await self.admin_redaction_ratelimiter.ratelimit(requester, update=update) + await self.admin_redaction_ratelimiter.ratelimit( + requester, update=update, n_actions=n_actions + ) else: # Override rate and burst count per-user await self.request_ratelimiter.ratelimit( @@ -391,4 +396,5 @@ class RequestRatelimiter: rate_hz=messages_per_second, burst_count=burst_count, update=update, + n_actions=n_actions, ) diff --git a/synapse/api/room_versions.py b/synapse/api/room_versions.py index e37acb0f1e..c397920fe5 100644 --- a/synapse/api/room_versions.py +++ b/synapse/api/room_versions.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import Callable, Dict, Optional +from typing import Callable, Dict, Optional, Tuple import attr @@ -51,6 +51,13 @@ class RoomDisposition: UNSTABLE = "unstable" +class PushRuleRoomFlag: + """Enum for listing possible MSC3931 room version feature flags, for push rules""" + + # MSC3932: Room version supports MSC1767 Extensible Events. + EXTENSIBLE_EVENTS = "org.matrix.msc3932.extensible_events" + + @attr.s(slots=True, frozen=True, auto_attribs=True) class RoomVersion: """An object which describes the unique attributes of a room version.""" @@ -91,6 +98,12 @@ class RoomVersion: msc3787_knock_restricted_join_rule: bool # MSC3667: Enforce integer power levels msc3667_int_only_power_levels: bool + # MSC3931: Adds a push rule condition for "room version feature flags", making + # some push rules room version dependent. Note that adding a flag to this list + # is not enough to mark it "supported": the push rule evaluator also needs to + # support the flag. Unknown flags are ignored by the evaluator, making conditions + # fail if used. + msc3931_push_features: Tuple[str, ...] # values from PushRuleRoomFlag class RoomVersions: @@ -111,6 +124,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V2 = RoomVersion( "2", @@ -129,6 +143,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V3 = RoomVersion( "3", @@ -147,6 +162,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V4 = RoomVersion( "4", @@ -165,6 +181,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V5 = RoomVersion( "5", @@ -183,6 +200,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V6 = RoomVersion( "6", @@ -201,6 +219,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) MSC2176 = RoomVersion( "org.matrix.msc2176", @@ -219,6 +238,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V7 = RoomVersion( "7", @@ -237,6 +257,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V8 = RoomVersion( "8", @@ -255,6 +276,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V9 = RoomVersion( "9", @@ -273,6 +295,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) MSC3787 = RoomVersion( "org.matrix.msc3787", @@ -291,6 +314,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=True, msc3667_int_only_power_levels=False, + msc3931_push_features=(), ) V10 = RoomVersion( "10", @@ -309,6 +333,7 @@ class RoomVersions: msc2716_redactions=False, msc3787_knock_restricted_join_rule=True, msc3667_int_only_power_levels=True, + msc3931_push_features=(), ) MSC2716v4 = RoomVersion( "org.matrix.msc2716v4", @@ -327,6 +352,27 @@ class RoomVersions: msc2716_redactions=True, msc3787_knock_restricted_join_rule=False, msc3667_int_only_power_levels=False, + msc3931_push_features=(), + ) + MSC1767v10 = RoomVersion( + # MSC1767 (Extensible Events) based on room version "10" + "org.matrix.msc1767.10", + RoomDisposition.UNSTABLE, + EventFormatVersions.ROOM_V4_PLUS, + StateResolutionVersions.V2, + enforce_key_validity=True, + special_case_aliases_auth=False, + strict_canonicaljson=True, + limit_notifications_power_levels=True, + msc2176_redaction_rules=False, + msc3083_join_rules=True, + msc3375_redaction_rules=True, + msc2403_knocking=True, + msc2716_historical=False, + msc2716_redactions=False, + msc3787_knock_restricted_join_rule=True, + msc3667_int_only_power_levels=True, + msc3931_push_features=(PushRuleRoomFlag.EXTENSIBLE_EVENTS,), ) diff --git a/synapse/api/urls.py b/synapse/api/urls.py index bd49fa6a5f..a918579f50 100644 --- a/synapse/api/urls.py +++ b/synapse/api/urls.py @@ -28,7 +28,7 @@ FEDERATION_V1_PREFIX = FEDERATION_PREFIX + "/v1" FEDERATION_V2_PREFIX = FEDERATION_PREFIX + "/v2" FEDERATION_UNSTABLE_PREFIX = FEDERATION_PREFIX + "/unstable" STATIC_PREFIX = "/_matrix/static" -SERVER_KEY_V2_PREFIX = "/_matrix/key/v2" +SERVER_KEY_PREFIX = "/_matrix/key" MEDIA_R0_PREFIX = "/_matrix/media/r0" MEDIA_V3_PREFIX = "/_matrix/media/v3" LEGACY_MEDIA_PREFIX = "/_matrix/media/v1" diff --git a/synapse/app/_base.py b/synapse/app/_base.py index 000912e86e..a5aa2185a2 100644 --- a/synapse/app/_base.py +++ b/synapse/app/_base.py @@ -47,6 +47,7 @@ from twisted.internet.tcp import Port from twisted.logger import LoggingFile, LogLevel from twisted.protocols.tls import TLSMemoryBIOFactory from twisted.python.threadpool import ThreadPool +from twisted.web.resource import Resource import synapse.util.caches from synapse.api.constants import MAX_PDU_SIZE @@ -55,12 +56,13 @@ from synapse.app.phone_stats_home import start_phone_stats_home from synapse.config import ConfigError from synapse.config._base import format_config_error from synapse.config.homeserver import HomeServerConfig -from synapse.config.server import ManholeConfig +from synapse.config.server import ListenerConfig, ManholeConfig from synapse.crypto import context_factory from synapse.events.presence_router import load_legacy_presence_router from synapse.events.spamcheck import load_legacy_spam_checkers from synapse.events.third_party_rules import load_legacy_third_party_event_rules from synapse.handlers.auth import load_legacy_password_auth_providers +from synapse.http.site import SynapseSite from synapse.logging.context import PreserveLoggingContext from synapse.logging.opentracing import init_tracer from synapse.metrics import install_gc_manager, register_threadpool @@ -264,26 +266,18 @@ def register_start( reactor.callWhenRunning(lambda: defer.ensureDeferred(wrapper())) -def listen_metrics( - bind_addresses: Iterable[str], port: int, enable_legacy_metric_names: bool -) -> None: +def listen_metrics(bind_addresses: Iterable[str], port: int) -> None: """ Start Prometheus metrics server. """ from prometheus_client import start_http_server as start_http_server_prometheus - from synapse.metrics import ( - RegistryProxy, - start_http_server as start_http_server_legacy, - ) + from synapse.metrics import RegistryProxy for host in bind_addresses: logger.info("Starting metrics listener on %s:%d", host, port) - if enable_legacy_metric_names: - start_http_server_legacy(port, addr=host, registry=RegistryProxy) - else: - _set_prometheus_client_use_created_metrics(False) - start_http_server_prometheus(port, addr=host, registry=RegistryProxy) + _set_prometheus_client_use_created_metrics(False) + start_http_server_prometheus(port, addr=host, registry=RegistryProxy) def _set_prometheus_client_use_created_metrics(new_value: bool) -> None: @@ -357,6 +351,55 @@ def listen_tcp( return r # type: ignore[return-value] +def listen_http( + listener_config: ListenerConfig, + root_resource: Resource, + version_string: str, + max_request_body_size: int, + context_factory: Optional[IOpenSSLContextFactory], + reactor: ISynapseReactor = reactor, +) -> List[Port]: + port = listener_config.port + bind_addresses = listener_config.bind_addresses + tls = listener_config.tls + + assert listener_config.http_options is not None + + site_tag = listener_config.http_options.tag + if site_tag is None: + site_tag = str(port) + + site = SynapseSite( + "synapse.access.%s.%s" % ("https" if tls else "http", site_tag), + site_tag, + listener_config, + root_resource, + version_string, + max_request_body_size=max_request_body_size, + reactor=reactor, + ) + if tls: + # refresh_certificate should have been called before this. + assert context_factory is not None + ports = listen_ssl( + bind_addresses, + port, + site, + context_factory, + reactor=reactor, + ) + logger.info("Synapse now listening on TCP port %d (TLS)", port) + else: + ports = listen_tcp( + bind_addresses, + port, + site, + reactor=reactor, + ) + logger.info("Synapse now listening on TCP port %d", port) + return ports + + def listen_ssl( bind_addresses: Collection[str], port: int, @@ -558,7 +601,7 @@ def reload_cache_config(config: HomeServerConfig) -> None: logger.warning(f) else: logger.debug( - "New cache config. Was:\n %s\nNow:\n", + "New cache config. Was:\n %s\nNow:\n %s", previous_cache_config.__dict__, config.caches.__dict__, ) diff --git a/synapse/app/admin_cmd.py b/synapse/app/admin_cmd.py index 3c8c00ea5b..165d1c5db0 100644 --- a/synapse/app/admin_cmd.py +++ b/synapse/app/admin_cmd.py @@ -28,10 +28,6 @@ from synapse.config.homeserver import HomeServerConfig from synapse.config.logger import setup_logging from synapse.events import EventBase from synapse.handlers.admin import ExfiltrationWriter -from synapse.replication.slave.storage.devices import SlavedDeviceStore -from synapse.replication.slave.storage.events import SlavedEventStore -from synapse.replication.slave.storage.filtering import SlavedFilteringStore -from synapse.replication.slave.storage.push_rule import SlavedPushRuleStore from synapse.server import HomeServer from synapse.storage.database import DatabasePool, LoggingDatabaseConnection from synapse.storage.databases.main.account_data import AccountDataWorkerStore @@ -40,10 +36,24 @@ from synapse.storage.databases.main.appservice import ( ApplicationServiceWorkerStore, ) from synapse.storage.databases.main.deviceinbox import DeviceInboxWorkerStore +from synapse.storage.databases.main.devices import DeviceWorkerStore +from synapse.storage.databases.main.event_federation import EventFederationWorkerStore +from synapse.storage.databases.main.event_push_actions import ( + EventPushActionsWorkerStore, +) +from synapse.storage.databases.main.events_worker import EventsWorkerStore +from synapse.storage.databases.main.filtering import FilteringWorkerStore +from synapse.storage.databases.main.push_rule import PushRulesWorkerStore from synapse.storage.databases.main.receipts import ReceiptsWorkerStore from synapse.storage.databases.main.registration import RegistrationWorkerStore +from synapse.storage.databases.main.relations import RelationsWorkerStore from synapse.storage.databases.main.room import RoomWorkerStore +from synapse.storage.databases.main.roommember import RoomMemberWorkerStore +from synapse.storage.databases.main.signatures import SignatureWorkerStore +from synapse.storage.databases.main.state import StateGroupWorkerStore +from synapse.storage.databases.main.stream import StreamWorkerStore from synapse.storage.databases.main.tags import TagsWorkerStore +from synapse.storage.databases.main.user_erasure_store import UserErasureWorkerStore from synapse.types import StateMap from synapse.util import SYNAPSE_VERSION from synapse.util.logcontext import LoggingContext @@ -52,17 +62,25 @@ logger = logging.getLogger("synapse.app.admin_cmd") class AdminCmdSlavedStore( - SlavedFilteringStore, - SlavedPushRuleStore, - SlavedEventStore, - SlavedDeviceStore, + FilteringWorkerStore, + DeviceWorkerStore, TagsWorkerStore, DeviceInboxWorkerStore, AccountDataWorkerStore, + PushRulesWorkerStore, ApplicationServiceTransactionWorkerStore, ApplicationServiceWorkerStore, - RegistrationWorkerStore, + RoomMemberWorkerStore, + RelationsWorkerStore, + EventFederationWorkerStore, + EventPushActionsWorkerStore, + StateGroupWorkerStore, + SignatureWorkerStore, + UserErasureWorkerStore, ReceiptsWorkerStore, + StreamWorkerStore, + EventsWorkerStore, + RegistrationWorkerStore, RoomWorkerStore, ): def __init__( diff --git a/synapse/app/complement_fork_starter.py b/synapse/app/complement_fork_starter.py index b22f315453..8c0f4a57e7 100644 --- a/synapse/app/complement_fork_starter.py +++ b/synapse/app/complement_fork_starter.py @@ -55,13 +55,13 @@ import os import signal import sys from types import FrameType -from typing import Any, Callable, List, Optional +from typing import Any, Callable, Dict, List, Optional from twisted.internet.main import installReactor # a list of the original signal handlers, before we installed our custom ones. # We restore these in our child processes. -_original_signal_handlers: dict[int, Any] = {} +_original_signal_handlers: Dict[int, Any] = {} class ProxiedReactor: diff --git a/synapse/app/generic_worker.py b/synapse/app/generic_worker.py index 5e3825fca6..bcc8abe20c 100644 --- a/synapse/app/generic_worker.py +++ b/synapse/app/generic_worker.py @@ -14,21 +14,19 @@ # limitations under the License. import logging import sys -from typing import Dict, List, Optional, Tuple +from typing import Dict, List -from twisted.internet import address from twisted.web.resource import Resource import synapse import synapse.events -from synapse.api.errors import HttpResponseException, RequestSendFailed, SynapseError from synapse.api.urls import ( CLIENT_API_PREFIX, FEDERATION_PREFIX, LEGACY_MEDIA_PREFIX, MEDIA_R0_PREFIX, MEDIA_V3_PREFIX, - SERVER_KEY_V2_PREFIX, + SERVER_KEY_PREFIX, ) from synapse.app import _base from synapse.app._base import ( @@ -43,52 +41,13 @@ from synapse.config.logger import setup_logging from synapse.config.server import ListenerConfig from synapse.federation.transport.server import TransportLayerServer from synapse.http.server import JsonResource, OptionsResource -from synapse.http.servlet import RestServlet, parse_json_object_from_request -from synapse.http.site import SynapseRequest, SynapseSite from synapse.logging.context import LoggingContext from synapse.metrics import METRICS_PREFIX, MetricsResource, RegistryProxy from synapse.replication.http import REPLICATION_PREFIX, ReplicationRestResource -from synapse.replication.slave.storage.devices import SlavedDeviceStore -from synapse.replication.slave.storage.events import SlavedEventStore -from synapse.replication.slave.storage.filtering import SlavedFilteringStore -from synapse.replication.slave.storage.keys import SlavedKeyStore -from synapse.replication.slave.storage.push_rule import SlavedPushRuleStore -from synapse.replication.slave.storage.pushers import SlavedPusherStore +from synapse.rest import ClientRestResource from synapse.rest.admin import register_servlets_for_media_repo -from synapse.rest.client import ( - account_data, - events, - initial_sync, - login, - presence, - profile, - push_rule, - read_marker, - receipts, - room, - room_batch, - room_keys, - sendtodevice, - sync, - tags, - user_directory, - versions, - voip, -) -from synapse.rest.client._base import client_patterns -from synapse.rest.client.account import ThreepidRestServlet, WhoamiRestServlet -from synapse.rest.client.devices import DevicesRestServlet -from synapse.rest.client.keys import ( - KeyChangesServlet, - KeyQueryServlet, - OneTimeKeyServlet, -) -from synapse.rest.client.register import ( - RegisterRestServlet, - RegistrationTokenValidityRestServlet, -) from synapse.rest.health import HealthResource -from synapse.rest.key.v2 import KeyApiV2Resource +from synapse.rest.key.v2 import KeyResource from synapse.rest.synapse.client import build_synapse_client_resource_tree from synapse.rest.well_known import well_known_resource from synapse.server import HomeServer @@ -100,8 +59,16 @@ from synapse.storage.databases.main.appservice import ( from synapse.storage.databases.main.censor_events import CensorEventsStore from synapse.storage.databases.main.client_ips import ClientIpWorkerStore from synapse.storage.databases.main.deviceinbox import DeviceInboxWorkerStore +from synapse.storage.databases.main.devices import DeviceWorkerStore from synapse.storage.databases.main.directory import DirectoryWorkerStore from synapse.storage.databases.main.e2e_room_keys import EndToEndRoomKeyStore +from synapse.storage.databases.main.event_federation import EventFederationWorkerStore +from synapse.storage.databases.main.event_push_actions import ( + EventPushActionsWorkerStore, +) +from synapse.storage.databases.main.events_worker import EventsWorkerStore +from synapse.storage.databases.main.filtering import FilteringWorkerStore +from synapse.storage.databases.main.keys import KeyStore from synapse.storage.databases.main.lock import LockStore from synapse.storage.databases.main.media_repository import MediaRepositoryStore from synapse.storage.databases.main.metrics import ServerMetricsStore @@ -110,118 +77,31 @@ from synapse.storage.databases.main.monthly_active_users import ( ) from synapse.storage.databases.main.presence import PresenceStore from synapse.storage.databases.main.profile import ProfileWorkerStore +from synapse.storage.databases.main.push_rule import PushRulesWorkerStore +from synapse.storage.databases.main.pusher import PusherWorkerStore from synapse.storage.databases.main.receipts import ReceiptsWorkerStore from synapse.storage.databases.main.registration import RegistrationWorkerStore +from synapse.storage.databases.main.relations import RelationsWorkerStore from synapse.storage.databases.main.room import RoomWorkerStore from synapse.storage.databases.main.room_batch import RoomBatchStore +from synapse.storage.databases.main.roommember import RoomMemberWorkerStore from synapse.storage.databases.main.search import SearchStore from synapse.storage.databases.main.session import SessionStore +from synapse.storage.databases.main.signatures import SignatureWorkerStore +from synapse.storage.databases.main.state import StateGroupWorkerStore from synapse.storage.databases.main.stats import StatsStore +from synapse.storage.databases.main.stream import StreamWorkerStore from synapse.storage.databases.main.tags import TagsWorkerStore 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 JsonDict +from synapse.storage.databases.main.user_erasure_store import UserErasureWorkerStore from synapse.util import SYNAPSE_VERSION from synapse.util.httpresourcetree import create_resource_tree logger = logging.getLogger("synapse.app.generic_worker") -class KeyUploadServlet(RestServlet): - """An implementation of the `KeyUploadServlet` that responds to read only - requests, but otherwise proxies through to the master instance. - """ - - PATTERNS = client_patterns("/keys/upload(/(?P<device_id>[^/]+))?$") - - def __init__(self, hs: HomeServer): - """ - Args: - hs: server - """ - super().__init__() - self.auth = hs.get_auth() - self.store = hs.get_datastores().main - self.http_client = hs.get_simple_http_client() - self.main_uri = hs.config.worker.worker_main_http_uri - - async def on_POST( - self, request: SynapseRequest, device_id: Optional[str] - ) -> Tuple[int, JsonDict]: - requester = await self.auth.get_user_by_req(request, allow_guest=True) - user_id = requester.user.to_string() - 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. - if requester.device_id is not None and device_id != requester.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 - - if device_id is None: - raise SynapseError( - 400, "To upload keys, you must pass device_id when authenticating" - ) - - if body: - # They're actually trying to upload something, proxy to main synapse. - - # Proxy headers from the original request, such as the auth headers - # (in case the access token is there) and the original IP / - # User-Agent of the request. - headers = { - header: request.requestHeaders.getRawHeaders(header, []) - for header in (b"Authorization", b"User-Agent") - } - # Add the previous hop to the X-Forwarded-For header. - x_forwarded_for = request.requestHeaders.getRawHeaders( - b"X-Forwarded-For", [] - ) - # we use request.client here, since we want the previous hop, not the - # original client (as returned by request.getClientAddress()). - if isinstance(request.client, (address.IPv4Address, address.IPv6Address)): - previous_host = request.client.host.encode("ascii") - # If the header exists, add to the comma-separated list of the first - # instance of the header. Otherwise, generate a new header. - if x_forwarded_for: - x_forwarded_for = [x_forwarded_for[0] + b", " + previous_host] - x_forwarded_for.extend(x_forwarded_for[1:]) - else: - x_forwarded_for = [previous_host] - headers[b"X-Forwarded-For"] = x_forwarded_for - - # Replicate the original X-Forwarded-Proto header. Note that - # XForwardedForRequest overrides isSecure() to give us the original protocol - # used by the client, as opposed to the protocol used by our upstream proxy - # - which is what we want here. - headers[b"X-Forwarded-Proto"] = [ - b"https" if request.isSecure() else b"http" - ] - - try: - result = await self.http_client.post_json_get_json( - self.main_uri + request.uri.decode("ascii"), body, headers=headers - ) - except HttpResponseException as e: - raise e.to_synapse_error() from e - except RequestSendFailed as e: - raise SynapseError(502, "Failed to talk to master") from e - - return 200, result - else: - # Just interested in counts. - result = await self.store.count_e2e_one_time_keys(user_id, device_id) - return 200, {"one_time_key_counts": result} - - class GenericWorkerSlavedStore( # FIXME(#3714): We need to add UserDirectoryStore as we write directly # rather than going via the correct worker. @@ -231,26 +111,36 @@ class GenericWorkerSlavedStore( EndToEndRoomKeyStore, PresenceStore, DeviceInboxWorkerStore, - SlavedDeviceStore, - SlavedPushRuleStore, + DeviceWorkerStore, TagsWorkerStore, AccountDataWorkerStore, - SlavedPusherStore, CensorEventsStore, ClientIpWorkerStore, - SlavedEventStore, - SlavedKeyStore, + # KeyStore isn't really safe to use from a worker, but for now we do so and hope that + # the races it creates aren't too bad. + KeyStore, RoomWorkerStore, RoomBatchStore, DirectoryWorkerStore, + PushRulesWorkerStore, ApplicationServiceTransactionWorkerStore, ApplicationServiceWorkerStore, ProfileWorkerStore, - SlavedFilteringStore, + FilteringWorkerStore, MonthlyActiveUsersWorkerStore, MediaRepositoryStore, ServerMetricsStore, + PusherWorkerStore, + RoomMemberWorkerStore, + RelationsWorkerStore, + EventFederationWorkerStore, + EventPushActionsWorkerStore, + StateGroupWorkerStore, + SignatureWorkerStore, + UserErasureWorkerStore, ReceiptsWorkerStore, + StreamWorkerStore, + EventsWorkerStore, RegistrationWorkerStore, SearchStore, TransactionWorkerStore, @@ -267,15 +157,9 @@ class GenericWorkerServer(HomeServer): DATASTORE_CLASS = GenericWorkerSlavedStore # type: ignore def _listen_http(self, listener_config: ListenerConfig) -> None: - port = listener_config.port - bind_addresses = listener_config.bind_addresses assert listener_config.http_options is not None - site_tag = listener_config.http_options.tag - if site_tag is None: - site_tag = str(port) - # We always include a health resource. resources: Dict[str, Resource] = {"/health": HealthResource()} @@ -284,52 +168,15 @@ class GenericWorkerServer(HomeServer): if name == "metrics": resources[METRICS_PREFIX] = MetricsResource(RegistryProxy) elif name == "client": - resource = JsonResource(self, canonical_json=False) - - RegisterRestServlet(self).register(resource) - RegistrationTokenValidityRestServlet(self).register(resource) - login.register_servlets(self, resource) - ThreepidRestServlet(self).register(resource) - WhoamiRestServlet(self).register(resource) - DevicesRestServlet(self).register(resource) - - # Read-only - KeyUploadServlet(self).register(resource) - KeyQueryServlet(self).register(resource) - KeyChangesServlet(self).register(resource) - OneTimeKeyServlet(self).register(resource) - - voip.register_servlets(self, resource) - push_rule.register_servlets(self, resource) - versions.register_servlets(self, resource) - - profile.register_servlets(self, resource) + resource: Resource = ClientRestResource(self) - sync.register_servlets(self, resource) - events.register_servlets(self, resource) - room.register_servlets(self, resource, is_worker=True) - room.register_deprecated_servlets(self, resource) - initial_sync.register_servlets(self, resource) - room_batch.register_servlets(self, resource) - room_keys.register_servlets(self, resource) - tags.register_servlets(self, resource) - account_data.register_servlets(self, resource) - receipts.register_servlets(self, resource) - read_marker.register_servlets(self, resource) - - sendtodevice.register_servlets(self, resource) - - user_directory.register_servlets(self, resource) - - presence.register_servlets(self, resource) - - resources.update({CLIENT_API_PREFIX: resource}) + resources[CLIENT_API_PREFIX] = resource resources.update(build_synapse_client_resource_tree(self)) - resources.update({"/.well-known": well_known_resource(self)}) + resources["/.well-known"] = well_known_resource(self) elif name == "federation": - resources.update({FEDERATION_PREFIX: TransportLayerServer(self)}) + resources[FEDERATION_PREFIX] = TransportLayerServer(self) elif name == "media": if self.config.media.can_load_media_repo: media_repo = self.get_media_repository_resource() @@ -357,16 +204,12 @@ class GenericWorkerServer(HomeServer): # Only load the openid resource separately if federation resource # is not specified since federation resource includes openid # resource. - resources.update( - { - FEDERATION_PREFIX: TransportLayerServer( - self, servlet_groups=["openid"] - ) - } + resources[FEDERATION_PREFIX] = TransportLayerServer( + self, servlet_groups=["openid"] ) if name in ["keys", "federation"]: - resources[SERVER_KEY_V2_PREFIX] = KeyApiV2Resource(self) + resources[SERVER_KEY_PREFIX] = KeyResource(self) if name == "replication": resources[REPLICATION_PREFIX] = ReplicationRestResource(self) @@ -377,23 +220,15 @@ class GenericWorkerServer(HomeServer): root_resource = create_resource_tree(resources, OptionsResource()) - _base.listen_tcp( - bind_addresses, - port, - SynapseSite( - "synapse.access.http.%s" % (site_tag,), - site_tag, - listener_config, - root_resource, - self.version_string, - max_request_body_size=max_request_body_size(self.config), - reactor=self.get_reactor(), - ), + _base.listen_http( + listener_config, + root_resource, + self.version_string, + max_request_body_size(self.config), + self.tls_server_context_factory, reactor=self.get_reactor(), ) - logger.info("Synapse worker now listening on port %d", port) - def start_listening(self) -> None: for listener in self.config.worker.worker_listeners: if listener.type == "http": @@ -415,7 +250,6 @@ class GenericWorkerServer(HomeServer): _base.listen_metrics( listener.bind_addresses, listener.port, - enable_legacy_metric_names=self.config.metrics.enable_legacy_metrics, ) else: logger.warning("Unsupported listener type: %s", listener.type) diff --git a/synapse/app/homeserver.py b/synapse/app/homeserver.py index 883f2fd2ec..b9be558c7e 100644 --- a/synapse/app/homeserver.py +++ b/synapse/app/homeserver.py @@ -31,14 +31,13 @@ from synapse.api.urls import ( LEGACY_MEDIA_PREFIX, MEDIA_R0_PREFIX, MEDIA_V3_PREFIX, - SERVER_KEY_V2_PREFIX, + SERVER_KEY_PREFIX, STATIC_PREFIX, ) from synapse.app import _base from synapse.app._base import ( handle_startup_exception, - listen_ssl, - listen_tcp, + listen_http, max_request_body_size, redirect_stdio_to_logs, register_start, @@ -53,14 +52,13 @@ from synapse.http.server import ( RootOptionsRedirectResource, StaticResource, ) -from synapse.http.site import SynapseSite from synapse.logging.context import LoggingContext from synapse.metrics import METRICS_PREFIX, MetricsResource, RegistryProxy from synapse.replication.http import REPLICATION_PREFIX, ReplicationRestResource from synapse.rest import ClientRestResource from synapse.rest.admin import AdminRestResource from synapse.rest.health import HealthResource -from synapse.rest.key.v2 import KeyApiV2Resource +from synapse.rest.key.v2 import KeyResource from synapse.rest.synapse.client import build_synapse_client_resource_tree from synapse.rest.well_known import well_known_resource from synapse.server import HomeServer @@ -83,8 +81,6 @@ class SynapseHomeServer(HomeServer): self, config: HomeServerConfig, listener_config: ListenerConfig ) -> Iterable[Port]: port = listener_config.port - bind_addresses = listener_config.bind_addresses - tls = listener_config.tls # Must exist since this is an HTTP listener. assert listener_config.http_options is not None site_tag = listener_config.http_options.tag @@ -140,37 +136,15 @@ class SynapseHomeServer(HomeServer): else: root_resource = OptionsResource() - site = SynapseSite( - "synapse.access.%s.%s" % ("https" if tls else "http", site_tag), - site_tag, + ports = listen_http( listener_config, create_resource_tree(resources, root_resource), self.version_string, - max_request_body_size=max_request_body_size(self.config), + max_request_body_size(self.config), + self.tls_server_context_factory, reactor=self.get_reactor(), ) - if tls: - # refresh_certificate should have been called before this. - assert self.tls_server_context_factory is not None - ports = listen_ssl( - bind_addresses, - port, - site, - self.tls_server_context_factory, - reactor=self.get_reactor(), - ) - logger.info("Synapse now listening on TCP port %d (TLS)", port) - - else: - ports = listen_tcp( - bind_addresses, - port, - site, - reactor=self.get_reactor(), - ) - logger.info("Synapse now listening on TCP port %d", port) - return ports def _configure_named_resource( @@ -215,30 +189,22 @@ class SynapseHomeServer(HomeServer): consent_resource: Resource = ConsentResource(self) if compress: consent_resource = gz_wrap(consent_resource) - resources.update({"/_matrix/consent": consent_resource}) + resources["/_matrix/consent"] = consent_resource if name == "federation": federation_resource: Resource = TransportLayerServer(self) if compress: federation_resource = gz_wrap(federation_resource) - resources.update({FEDERATION_PREFIX: federation_resource}) + resources[FEDERATION_PREFIX] = federation_resource if name == "openid": - resources.update( - { - FEDERATION_PREFIX: TransportLayerServer( - self, servlet_groups=["openid"] - ) - } + resources[FEDERATION_PREFIX] = TransportLayerServer( + self, servlet_groups=["openid"] ) if name in ["static", "client"]: - resources.update( - { - STATIC_PREFIX: StaticResource( - os.path.join(os.path.dirname(synapse.__file__), "static") - ) - } + resources[STATIC_PREFIX] = StaticResource( + os.path.join(os.path.dirname(synapse.__file__), "static") ) if name in ["media", "federation", "client"]: @@ -257,7 +223,7 @@ class SynapseHomeServer(HomeServer): ) if name in ["keys", "federation"]: - resources[SERVER_KEY_V2_PREFIX] = KeyApiV2Resource(self) + resources[SERVER_KEY_PREFIX] = KeyResource(self) if name == "metrics" and self.config.metrics.enable_metrics: metrics_resource: Resource = MetricsResource(RegistryProxy) @@ -299,7 +265,6 @@ class SynapseHomeServer(HomeServer): _base.listen_metrics( listener.bind_addresses, listener.port, - enable_legacy_metric_names=self.config.metrics.enable_legacy_metrics, ) else: # this shouldn't happen, as the listener type should have been checked diff --git a/synapse/appservice/__init__.py b/synapse/appservice/__init__.py index 0dfa00df44..65615f50b8 100644 --- a/synapse/appservice/__init__.py +++ b/synapse/appservice/__init__.py @@ -32,9 +32,9 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) -# Type for the `device_one_time_key_counts` field in an appservice transaction +# Type for the `device_one_time_keys_count` field in an appservice transaction # user ID -> {device ID -> {algorithm -> count}} -TransactionOneTimeKeyCounts = Dict[str, Dict[str, Dict[str, int]]] +TransactionOneTimeKeysCount = Dict[str, Dict[str, Dict[str, int]]] # Type for the `device_unused_fallback_key_types` field in an appservice transaction # user ID -> {device ID -> [algorithm]} @@ -172,12 +172,24 @@ class ApplicationService: Returns: True if this service would like to know about this room. """ - member_list = await store.get_users_in_room( + # We can use `get_local_users_in_room(...)` here because an application service + # can only be interested in local users of the server it's on (ignore any remote + # users that might match the user namespace regex). + # + # In the future, we can consider re-using + # `store.get_app_service_users_in_room` which is very similar to this + # function but has a slightly worse performance than this because we + # have an early escape-hatch if we find a single user that the + # appservice is interested in. The juice would be worth the squeeze if + # `store.get_app_service_users_in_room` was used in more places besides + # an experimental MSC. But for now we can avoid doing more work and + # barely using it later. + local_user_ids = await store.get_local_users_in_room( room_id, on_invalidate=cache_context.invalidate ) # check joined member events - for user_id in member_list: + for user_id in local_user_ids: if self.is_interested_in_user(user_id): return True return False @@ -233,7 +245,9 @@ class ApplicationService: return True # likewise with the room's aliases (if it has any) - alias_list = await store.get_aliases_for_room(room_id) + alias_list = await store.get_aliases_for_room( + room_id, on_invalidate=cache_context.invalidate + ) for alias in alias_list: if self.is_room_alias_in_namespace(alias): return True @@ -299,7 +313,9 @@ class ApplicationService: # Find all the rooms the sender is in if self.is_interested_in_user(user_id.to_string()): return True - room_ids = await store.get_rooms_for_user(user_id.to_string()) + room_ids = await store.get_rooms_for_user( + user_id.to_string(), on_invalidate=cache_context.invalidate + ) # Then find out if the appservice is interested in any of those rooms for room_id in room_ids: @@ -364,7 +380,7 @@ class AppServiceTransaction: events: List[EventBase], ephemeral: List[JsonDict], to_device_messages: List[JsonDict], - one_time_key_counts: TransactionOneTimeKeyCounts, + one_time_keys_count: TransactionOneTimeKeysCount, unused_fallback_keys: TransactionUnusedFallbackKeys, device_list_summary: DeviceListUpdates, ): @@ -373,7 +389,7 @@ class AppServiceTransaction: self.events = events self.ephemeral = ephemeral self.to_device_messages = to_device_messages - self.one_time_key_counts = one_time_key_counts + self.one_time_keys_count = one_time_keys_count self.unused_fallback_keys = unused_fallback_keys self.device_list_summary = device_list_summary @@ -390,7 +406,7 @@ class AppServiceTransaction: events=self.events, ephemeral=self.ephemeral, to_device_messages=self.to_device_messages, - one_time_key_counts=self.one_time_key_counts, + one_time_keys_count=self.one_time_keys_count, unused_fallback_keys=self.unused_fallback_keys, device_list_summary=self.device_list_summary, txn_id=self.id, diff --git a/synapse/appservice/api.py b/synapse/appservice/api.py index fbac4375b0..edafd433cd 100644 --- a/synapse/appservice/api.py +++ b/synapse/appservice/api.py @@ -23,7 +23,7 @@ from synapse.api.constants import EventTypes, Membership, ThirdPartyEntityKind from synapse.api.errors import CodeMessageException from synapse.appservice import ( ApplicationService, - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.events import EventBase @@ -123,7 +123,7 @@ class ApplicationServiceApi(SimpleHttpClient): response = await self.get_json( uri, {"access_token": service.hs_token}, - headers={"Authorization": f"Bearer {service.hs_token}"}, + headers={"Authorization": [f"Bearer {service.hs_token}"]}, ) if response is not None: # just an empty json object return True @@ -147,7 +147,7 @@ class ApplicationServiceApi(SimpleHttpClient): response = await self.get_json( uri, {"access_token": service.hs_token}, - headers={"Authorization": f"Bearer {service.hs_token}"}, + headers={"Authorization": [f"Bearer {service.hs_token}"]}, ) if response is not None: # just an empty json object return True @@ -190,7 +190,9 @@ class ApplicationServiceApi(SimpleHttpClient): b"access_token": service.hs_token, } response = await self.get_json( - uri, args=args, headers={"Authorization": f"Bearer {service.hs_token}"} + uri, + args=args, + headers={"Authorization": [f"Bearer {service.hs_token}"]}, ) if not isinstance(response, list): logger.warning( @@ -230,7 +232,7 @@ class ApplicationServiceApi(SimpleHttpClient): info = await self.get_json( uri, {"access_token": service.hs_token}, - headers={"Authorization": f"Bearer {service.hs_token}"}, + headers={"Authorization": [f"Bearer {service.hs_token}"]}, ) if not _is_valid_3pe_metadata(info): @@ -260,7 +262,7 @@ class ApplicationServiceApi(SimpleHttpClient): events: List[EventBase], ephemeral: List[JsonDict], to_device_messages: List[JsonDict], - one_time_key_counts: TransactionOneTimeKeyCounts, + one_time_keys_count: TransactionOneTimeKeysCount, unused_fallback_keys: TransactionUnusedFallbackKeys, device_list_summary: DeviceListUpdates, txn_id: Optional[int] = None, @@ -308,10 +310,13 @@ class ApplicationServiceApi(SimpleHttpClient): # TODO: Update to stable prefixes once MSC3202 completes FCP merge if service.msc3202_transaction_extensions: - if one_time_key_counts: + if one_time_keys_count: body[ "org.matrix.msc3202.device_one_time_key_counts" - ] = one_time_key_counts + ] = one_time_keys_count + body[ + "org.matrix.msc3202.device_one_time_keys_count" + ] = one_time_keys_count if unused_fallback_keys: body[ "org.matrix.msc3202.device_unused_fallback_key_types" @@ -327,7 +332,7 @@ class ApplicationServiceApi(SimpleHttpClient): uri=uri, json_body=body, args={"access_token": service.hs_token}, - headers={"Authorization": f"Bearer {service.hs_token}"}, + headers={"Authorization": [f"Bearer {service.hs_token}"]}, ) if logger.isEnabledFor(logging.DEBUG): logger.debug( diff --git a/synapse/appservice/scheduler.py b/synapse/appservice/scheduler.py index 430ffbcd1f..7b562795a3 100644 --- a/synapse/appservice/scheduler.py +++ b/synapse/appservice/scheduler.py @@ -64,7 +64,7 @@ from typing import ( from synapse.appservice import ( ApplicationService, ApplicationServiceState, - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.appservice.api import ApplicationServiceApi @@ -258,7 +258,7 @@ class _ServiceQueuer: ): return - one_time_key_counts: Optional[TransactionOneTimeKeyCounts] = None + one_time_keys_count: Optional[TransactionOneTimeKeysCount] = None unused_fallback_keys: Optional[TransactionUnusedFallbackKeys] = None if ( @@ -269,7 +269,7 @@ class _ServiceQueuer: # for the users which are mentioned in this transaction, # as well as the appservice's sender. ( - one_time_key_counts, + one_time_keys_count, unused_fallback_keys, ) = await self._compute_msc3202_otk_counts_and_fallback_keys( service, events, ephemeral, to_device_messages_to_send @@ -281,7 +281,7 @@ class _ServiceQueuer: events, ephemeral, to_device_messages_to_send, - one_time_key_counts, + one_time_keys_count, unused_fallback_keys, device_list_summary, ) @@ -296,7 +296,7 @@ class _ServiceQueuer: events: Iterable[EventBase], ephemerals: Iterable[JsonDict], to_device_messages: Iterable[JsonDict], - ) -> Tuple[TransactionOneTimeKeyCounts, TransactionUnusedFallbackKeys]: + ) -> Tuple[TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys]: """ Given a list of the events, ephemeral messages and to-device messages, - first computes a list of application services users that may have @@ -367,7 +367,7 @@ class _TransactionController: events: List[EventBase], ephemeral: Optional[List[JsonDict]] = None, to_device_messages: Optional[List[JsonDict]] = None, - one_time_key_counts: Optional[TransactionOneTimeKeyCounts] = None, + one_time_keys_count: Optional[TransactionOneTimeKeysCount] = None, unused_fallback_keys: Optional[TransactionUnusedFallbackKeys] = None, device_list_summary: Optional[DeviceListUpdates] = None, ) -> None: @@ -380,7 +380,7 @@ class _TransactionController: events: The persistent events to include in the transaction. ephemeral: The ephemeral events to include in the transaction. to_device_messages: The to-device messages to include in the transaction. - one_time_key_counts: Counts of remaining one-time keys for relevant + one_time_keys_count: Counts of remaining one-time keys for relevant appservice devices in the transaction. unused_fallback_keys: Lists of unused fallback keys for relevant appservice devices in the transaction. @@ -397,7 +397,7 @@ class _TransactionController: events=events, ephemeral=ephemeral or [], to_device_messages=to_device_messages or [], - one_time_key_counts=one_time_key_counts or {}, + one_time_keys_count=one_time_keys_count or {}, unused_fallback_keys=unused_fallback_keys or {}, device_list_summary=device_list_summary or DeviceListUpdates(), ) diff --git a/synapse/config/_util.py b/synapse/config/_util.py index 3edb4b7106..d3a4b484ab 100644 --- a/synapse/config/_util.py +++ b/synapse/config/_util.py @@ -33,6 +33,9 @@ def validate_config( config: the configuration value to be validated config_path: the path within the config file. This will be used as a basis for the error message. + + Raises: + ConfigError, if validation fails. """ try: jsonschema.validate(config, json_schema) diff --git a/synapse/config/api.py b/synapse/config/api.py index e46728e73f..27d50d118f 100644 --- a/synapse/config/api.py +++ b/synapse/config/api.py @@ -13,12 +13,13 @@ # limitations under the License. import logging -from typing import Any, Iterable +from typing import Any, Iterable, Optional, Tuple from synapse.api.constants import EventTypes from synapse.config._base import Config, ConfigError from synapse.config._util import validate_config from synapse.types import JsonDict +from synapse.types.state import StateFilter logger = logging.getLogger(__name__) @@ -26,16 +27,20 @@ logger = logging.getLogger(__name__) class ApiConfig(Config): section = "api" + room_prejoin_state: StateFilter + track_puppetted_users_ips: bool + def read_config(self, config: JsonDict, **kwargs: Any) -> None: validate_config(_MAIN_SCHEMA, config, ()) - self.room_prejoin_state = list(self._get_prejoin_state_types(config)) + self.room_prejoin_state = StateFilter.from_types( + self._get_prejoin_state_entries(config) + ) self.track_puppeted_user_ips = config.get("track_puppeted_user_ips", False) - def _get_prejoin_state_types(self, config: JsonDict) -> Iterable[str]: - """Get the event types to include in the prejoin state - - Parses the config and returns an iterable of the event types to be included. - """ + def _get_prejoin_state_entries( + self, config: JsonDict + ) -> Iterable[Tuple[str, Optional[str]]]: + """Get the event types and state keys to include in the prejoin state.""" room_prejoin_state_config = config.get("room_prejoin_state") or {} # backwards-compatibility support for room_invite_state_types @@ -50,33 +55,39 @@ class ApiConfig(Config): logger.warning(_ROOM_INVITE_STATE_TYPES_WARNING) - yield from config["room_invite_state_types"] + for event_type in config["room_invite_state_types"]: + yield event_type, None return if not room_prejoin_state_config.get("disable_default_event_types"): - yield from _DEFAULT_PREJOIN_STATE_TYPES + yield from _DEFAULT_PREJOIN_STATE_TYPES_AND_STATE_KEYS - yield from room_prejoin_state_config.get("additional_event_types", []) + for entry in room_prejoin_state_config.get("additional_event_types", []): + if isinstance(entry, str): + yield entry, None + else: + yield entry _ROOM_INVITE_STATE_TYPES_WARNING = """\ WARNING: The 'room_invite_state_types' configuration setting is now deprecated, and replaced with 'room_prejoin_state'. New features may not work correctly -unless 'room_invite_state_types' is removed. See the sample configuration file for -details of 'room_prejoin_state'. +unless 'room_invite_state_types' is removed. See the config documentation at + https://matrix-org.github.io/synapse/latest/usage/configuration/config_documentation.html#room_prejoin_state +for details of 'room_prejoin_state'. -------------------------------------------------------------------------------- """ -_DEFAULT_PREJOIN_STATE_TYPES = [ - EventTypes.JoinRules, - EventTypes.CanonicalAlias, - EventTypes.RoomAvatar, - EventTypes.RoomEncryption, - EventTypes.Name, +_DEFAULT_PREJOIN_STATE_TYPES_AND_STATE_KEYS = [ + (EventTypes.JoinRules, ""), + (EventTypes.CanonicalAlias, ""), + (EventTypes.RoomAvatar, ""), + (EventTypes.RoomEncryption, ""), + (EventTypes.Name, ""), # Per MSC1772. - EventTypes.Create, + (EventTypes.Create, ""), # Per MSC3173. - EventTypes.Topic, + (EventTypes.Topic, ""), ] @@ -90,7 +101,17 @@ _ROOM_PREJOIN_STATE_CONFIG_SCHEMA = { "disable_default_event_types": {"type": "boolean"}, "additional_event_types": { "type": "array", - "items": {"type": "string"}, + "items": { + "oneOf": [ + {"type": "string"}, + { + "type": "array", + "items": {"type": "string"}, + "minItems": 2, + "maxItems": 2, + }, + ], + }, }, }, }, diff --git a/synapse/config/cache.py b/synapse/config/cache.py index 2db8cfb005..eb4194a5a9 100644 --- a/synapse/config/cache.py +++ b/synapse/config/cache.py @@ -159,7 +159,7 @@ class CacheConfig(Config): self.track_memory_usage = cache_config.get("track_memory_usage", False) if self.track_memory_usage: - check_requirements("cache_memory") + check_requirements("cache-memory") expire_caches = cache_config.get("expire_caches", True) cache_entry_ttl = cache_config.get("cache_entry_ttl", "30m") diff --git a/synapse/config/experimental.py b/synapse/config/experimental.py index c35301207a..573fa0386f 100644 --- a/synapse/config/experimental.py +++ b/synapse/config/experimental.py @@ -12,10 +12,11 @@ # See the License for the specific language governing permissions and # limitations under the License. -from typing import Any +from typing import Any, Optional import attr +from synapse.api.room_versions import KNOWN_ROOM_VERSIONS, RoomVersions from synapse.config._base import Config from synapse.types import JsonDict @@ -53,9 +54,6 @@ class ExperimentalConfig(Config): # MSC3266 (room summary api) self.msc3266_enabled: bool = experimental.get("msc3266_enabled", False) - # MSC3030 (Jump to date API endpoint) - self.msc3030_enabled: bool = experimental.get("msc3030_enabled", False) - # MSC2409 (this setting only relates to optionally sending to-device messages). # Presence, typing and read receipt EDUs are already sent to application services that # have opted in to receive them. If enabled, this adds to-device messages to that list. @@ -95,13 +93,11 @@ class ExperimentalConfig(Config): # MSC2815 (allow room moderators to view redacted event content) self.msc2815_enabled: bool = experimental.get("msc2815_enabled", False) - # MSC3772: A push rule for mutual relations. - self.msc3772_enabled: bool = experimental.get("msc3772_enabled", False) # MSC3773: Thread notifications self.msc3773_enabled: bool = experimental.get("msc3773_enabled", False) - # MSC3715: dir param on /relations. - self.msc3715_enabled: bool = experimental.get("msc3715_enabled", False) + # MSC3664: Pushrules to match on related events + self.msc3664_enabled: bool = experimental.get("msc3664_enabled", False) # MSC3848: Introduce errcodes for specific event sending failures self.msc3848_enabled: bool = experimental.get("msc3848_enabled", False) @@ -122,3 +118,21 @@ class ExperimentalConfig(Config): self.msc3882_token_timeout = self.parse_duration( experimental.get("msc3882_token_timeout", "5m") ) + + # MSC3874: Filtering /messages with rel_types / not_rel_types. + self.msc3874_enabled: bool = experimental.get("msc3874_enabled", False) + + # MSC3886: Simple client rendezvous capability + self.msc3886_endpoint: Optional[str] = experimental.get( + "msc3886_endpoint", None + ) + + # MSC3912: Relation-based redactions. + self.msc3912_enabled: bool = experimental.get("msc3912_enabled", False) + + # MSC1767 and friends: Extensible Events + self.msc1767_enabled: bool = experimental.get("msc1767_enabled", False) + if self.msc1767_enabled: + # Enable room version (and thus applicable push rules from MSC3931/3932) + version_id = RoomVersions.MSC1767v10.identifier + KNOWN_ROOM_VERSIONS[version_id] = RoomVersions.MSC1767v10 diff --git a/synapse/config/groups.py b/synapse/config/groups.py deleted file mode 100644 index baa051fdd4..0000000000 --- a/synapse/config/groups.py +++ /dev/null @@ -1,27 +0,0 @@ -# Copyright 2017 New Vector Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from typing import Any - -from synapse.types import JsonDict - -from ._base import Config - - -class GroupsConfig(Config): - section = "groups" - - def read_config(self, config: JsonDict, **kwargs: Any) -> None: - self.enable_group_creation = config.get("enable_group_creation", False) - self.group_creation_prefix = config.get("group_creation_prefix", "") diff --git a/synapse/config/logger.py b/synapse/config/logger.py index 6c1f78f8df..5468b963a2 100644 --- a/synapse/config/logger.py +++ b/synapse/config/logger.py @@ -53,7 +53,7 @@ DEFAULT_LOG_CONFIG = Template( # Synapse also supports structured logging for machine readable logs which can # be ingested by ELK stacks. See [2] for details. # -# [1]: https://docs.python.org/3.7/library/logging.config.html#configuration-dictionary-schema +# [1]: https://docs.python.org/3/library/logging.config.html#configuration-dictionary-schema # [2]: https://matrix-org.github.io/synapse/latest/structured_logging.html version: 1 @@ -317,15 +317,16 @@ def setup_logging( Set up the logging subsystem. Args: - config (LoggingConfig | synapse.config.worker.WorkerConfig): - configuration data + config: configuration data - use_worker_options (bool): True to use the 'worker_log_config' option + use_worker_options: True to use the 'worker_log_config' option instead of 'log_config'. logBeginner: The Twisted logBeginner to use. """ + from twisted.internet import reactor + log_config_path = ( config.worker.worker_log_config if use_worker_options @@ -348,3 +349,4 @@ def setup_logging( ) logging.info("Server hostname: %s", config.server.server_name) logging.info("Instance name: %s", hs.get_instance_name()) + logging.info("Twisted reactor: %s", type(reactor).__name__) diff --git a/synapse/config/metrics.py b/synapse/config/metrics.py index bb065f9f2f..8c1c9bd12d 100644 --- a/synapse/config/metrics.py +++ b/synapse/config/metrics.py @@ -43,8 +43,6 @@ class MetricsConfig(Config): def read_config(self, config: JsonDict, **kwargs: Any) -> None: self.enable_metrics = config.get("enable_metrics", False) - self.enable_legacy_metrics = config.get("enable_legacy_metrics", True) - self.report_stats = config.get("report_stats", None) self.report_stats_endpoint = config.get( "report_stats_endpoint", "https://matrix.org/report-usage-stats/push" diff --git a/synapse/config/oidc.py b/synapse/config/oidc.py index 5418a332da..0bd83f4010 100644 --- a/synapse/config/oidc.py +++ b/synapse/config/oidc.py @@ -123,6 +123,8 @@ OIDC_PROVIDER_CONFIG_SCHEMA = { "userinfo_endpoint": {"type": "string"}, "jwks_uri": {"type": "string"}, "skip_verification": {"type": "boolean"}, + "backchannel_logout_enabled": {"type": "boolean"}, + "backchannel_logout_ignore_sub": {"type": "boolean"}, "user_profile_method": { "type": "string", "enum": ["auto", "userinfo_endpoint"], @@ -292,6 +294,10 @@ def _parse_oidc_config_dict( token_endpoint=oidc_config.get("token_endpoint"), userinfo_endpoint=oidc_config.get("userinfo_endpoint"), jwks_uri=oidc_config.get("jwks_uri"), + backchannel_logout_enabled=oidc_config.get("backchannel_logout_enabled", False), + backchannel_logout_ignore_sub=oidc_config.get( + "backchannel_logout_ignore_sub", False + ), skip_verification=oidc_config.get("skip_verification", False), user_profile_method=oidc_config.get("user_profile_method", "auto"), allow_existing_users=oidc_config.get("allow_existing_users", False), @@ -368,6 +374,12 @@ class OidcProviderConfig: # "openid" scope is used. jwks_uri: Optional[str] + # Whether Synapse should react to backchannel logouts + backchannel_logout_enabled: bool + + # Whether Synapse should ignore the `sub` claim in backchannel logouts or not. + backchannel_logout_ignore_sub: bool + # Whether to skip metadata verification skip_verification: bool diff --git a/synapse/config/push.py b/synapse/config/push.py index 979b128eae..3b5378e6ea 100644 --- a/synapse/config/push.py +++ b/synapse/config/push.py @@ -26,6 +26,7 @@ class PushConfig(Config): def read_config(self, config: JsonDict, **kwargs: Any) -> None: push_config = config.get("push") or {} self.push_include_content = push_config.get("include_content", True) + self.enable_push = push_config.get("enabled", True) self.push_group_unread_count_by_room = push_config.get( "group_unread_count_by_room", True ) diff --git a/synapse/config/ratelimiting.py b/synapse/config/ratelimiting.py index 1ed001e105..5c13fe428a 100644 --- a/synapse/config/ratelimiting.py +++ b/synapse/config/ratelimiting.py @@ -150,8 +150,5 @@ class RatelimitConfig(Config): self.rc_third_party_invite = RatelimitSettings( config.get("rc_third_party_invite", {}), - defaults={ - "per_second": self.rc_message.per_second, - "burst_count": self.rc_message.burst_count, - }, + defaults={"per_second": 0.0025, "burst_count": 5}, ) diff --git a/synapse/config/repository.py b/synapse/config/repository.py index 1033496bb4..e4759711ed 100644 --- a/synapse/config/repository.py +++ b/synapse/config/repository.py @@ -205,7 +205,7 @@ class ContentRepositoryConfig(Config): ) self.url_preview_enabled = config.get("url_preview_enabled", False) if self.url_preview_enabled: - check_requirements("url_preview") + check_requirements("url-preview") proxy_env = getproxies_environment() if "url_preview_ip_range_blacklist" not in config: diff --git a/synapse/config/server.py b/synapse/config/server.py index f2353ce5fb..ec46ca63ad 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -207,6 +207,9 @@ class HttpListenerConfig: additional_resources: Dict[str, dict] = attr.Factory(dict) tag: Optional[str] = None request_id_header: Optional[str] = None + # If true, the listener will return CORS response headers compatible with MSC3886: + # https://github.com/matrix-org/matrix-spec-proposals/pull/3886 + experimental_cors_msc3886: bool = False @attr.s(slots=True, frozen=True, auto_attribs=True) @@ -935,6 +938,7 @@ def parse_listener_def(num: int, listener: Any) -> ListenerConfig: additional_resources=listener.get("additional_resources", {}), tag=listener.get("tag"), request_id_header=listener.get("request_id_header"), + experimental_cors_msc3886=listener.get("experimental_cors_msc3886", False), ) return ListenerConfig(port, bind_addresses, listener_type, tls, http_config) diff --git a/synapse/config/workers.py b/synapse/config/workers.py index 0fb725dd8f..2580660b6c 100644 --- a/synapse/config/workers.py +++ b/synapse/config/workers.py @@ -29,20 +29,6 @@ from ._base import ( ) from .server import DIRECT_TCP_ERROR, ListenerConfig, parse_listener_def -_FEDERATION_SENDER_WITH_SEND_FEDERATION_ENABLED_ERROR = """ -The send_federation config option must be disabled in the main -synapse process before they can be run in a separate worker. - -Please add ``send_federation: false`` to the main config -""" - -_PUSHER_WITH_START_PUSHERS_ENABLED_ERROR = """ -The start_pushers config option must be disabled in the main -synapse process before they can be run in a separate worker. - -Please add ``start_pushers: false`` to the main config -""" - _DEPRECATED_WORKER_DUTY_OPTION_USED = """ The '%s' configuration option is deprecated and will be removed in a future Synapse version. Please use ``%s: name_of_worker`` instead. @@ -67,6 +53,7 @@ class InstanceLocationConfig: host: str port: int + tls: bool = False @attr.s @@ -149,13 +136,25 @@ class WorkerConfig(Config): # The port on the main synapse for HTTP replication endpoint self.worker_replication_http_port = config.get("worker_replication_http_port") + # The tls mode on the main synapse for HTTP replication endpoint. + # For backward compatibility this defaults to False. + self.worker_replication_http_tls = config.get( + "worker_replication_http_tls", False + ) + # The shared secret used for authentication when connecting to the main synapse. self.worker_replication_secret = config.get("worker_replication_secret", None) self.worker_name = config.get("worker_name", self.worker_app) self.instance_name = self.worker_name or "master" + # FIXME: Remove this check after a suitable amount of time. self.worker_main_http_uri = config.get("worker_main_http_uri", None) + if self.worker_main_http_uri is not None: + logger.warning( + "The config option worker_main_http_uri is unused since Synapse 1.73. " + "It can be safely removed from your configuration." + ) # This option is really only here to support `--manhole` command line # argument. @@ -169,40 +168,12 @@ class WorkerConfig(Config): ) ) - # Handle federation sender configuration. - # - # There are two ways of configuring which instances handle federation - # sending: - # 1. The old way where "send_federation" is set to false and running a - # `synapse.app.federation_sender` worker app. - # 2. Specifying the workers sending federation in - # `federation_sender_instances`. - # - - send_federation = config.get("send_federation", True) - - federation_sender_instances = config.get("federation_sender_instances") - if federation_sender_instances is None: - # Default to an empty list, which means "another, unknown, worker is - # responsible for it". - federation_sender_instances = [] - - # If no federation sender instances are set we check if - # `send_federation` is set, which means use master - if send_federation: - federation_sender_instances = ["master"] - - if self.worker_app == "synapse.app.federation_sender": - if send_federation: - # If we're running federation senders, and not using - # `federation_sender_instances`, then we should have - # explicitly set `send_federation` to false. - raise ConfigError( - _FEDERATION_SENDER_WITH_SEND_FEDERATION_ENABLED_ERROR - ) - - federation_sender_instances = [self.worker_name] - + federation_sender_instances = self._worker_names_performing_this_duty( + config, + "send_federation", + "synapse.app.federation_sender", + "federation_sender_instances", + ) self.send_federation = self.instance_name in federation_sender_instances self.federation_shard_config = ShardedWorkerHandlingConfig( federation_sender_instances @@ -269,27 +240,12 @@ class WorkerConfig(Config): ) # Handle sharded push - start_pushers = config.get("start_pushers", True) - pusher_instances = config.get("pusher_instances") - if pusher_instances is None: - # Default to an empty list, which means "another, unknown, worker is - # responsible for it". - pusher_instances = [] - - # If no pushers instances are set we check if `start_pushers` is - # set, which means use master - if start_pushers: - pusher_instances = ["master"] - - if self.worker_app == "synapse.app.pusher": - if start_pushers: - # If we're running pushers, and not using - # `pusher_instances`, then we should have explicitly set - # `start_pushers` to false. - raise ConfigError(_PUSHER_WITH_START_PUSHERS_ENABLED_ERROR) - - pusher_instances = [self.instance_name] - + pusher_instances = self._worker_names_performing_this_duty( + config, + "start_pushers", + "synapse.app.pusher", + "pusher_instances", + ) self.start_pushers = self.instance_name in pusher_instances self.pusher_shard_config = ShardedWorkerHandlingConfig(pusher_instances) @@ -412,6 +368,64 @@ class WorkerConfig(Config): # (By this point, these are either the same value or only one is not None.) return bool(new_option_should_run_here or legacy_option_should_run_here) + def _worker_names_performing_this_duty( + self, + config: Dict[str, Any], + legacy_option_name: str, + legacy_app_name: str, + modern_instance_list_name: str, + ) -> List[str]: + """ + Retrieves the names of the workers handling a given duty, by either legacy + option or instance list. + + There are two ways of configuring which instances handle a given duty, e.g. + for configuring pushers: + + 1. The old way where "start_pushers" is set to false and running a + `synapse.app.pusher'` worker app. + 2. Specifying the workers sending federation in `pusher_instances`. + + Args: + config: settings read from yaml. + legacy_option_name: the old way of enabling options. e.g. 'start_pushers' + legacy_app_name: The historical app name. e.g. 'synapse.app.pusher' + modern_instance_list_name: the string name of the new instance_list. e.g. + 'pusher_instances' + + Returns: + A list of worker instance names handling the given duty. + """ + + legacy_option = config.get(legacy_option_name, True) + + worker_instances = config.get(modern_instance_list_name) + if worker_instances is None: + # Default to an empty list, which means "another, unknown, worker is + # responsible for it". + worker_instances = [] + + # If no worker instances are set we check if the legacy option + # is set, which means use the main process. + if legacy_option: + worker_instances = ["master"] + + if self.worker_app == legacy_app_name: + if legacy_option: + # If we're using `legacy_app_name`, and not using + # `modern_instance_list_name`, then we should have + # explicitly set `legacy_option_name` to false. + raise ConfigError( + f"The '{legacy_option_name}' config option must be disabled in " + "the main synapse process before they can be run in a separate " + "worker.\n" + f"Please add `{legacy_option_name}: false` to the main config.\n", + ) + + worker_instances = [self.worker_name] + + return worker_instances + def read_arguments(self, args: argparse.Namespace) -> None: # We support a bunch of command line arguments that override options in # the config. A lot of these options have a worker_* prefix when running diff --git a/synapse/crypto/keyring.py b/synapse/crypto/keyring.py index c88afb2986..69310d9035 100644 --- a/synapse/crypto/keyring.py +++ b/synapse/crypto/keyring.py @@ -14,7 +14,6 @@ import abc import logging -import urllib from typing import TYPE_CHECKING, Callable, Dict, Iterable, List, Optional, Tuple import attr @@ -213,7 +212,7 @@ class Keyring: def verify_json_objects_for_server( self, server_and_json: Iterable[Tuple[str, dict, int]] - ) -> List[defer.Deferred]: + ) -> List["defer.Deferred[None]"]: """Bulk verifies signatures of json objects, bulk fetching keys as necessary. @@ -226,10 +225,9 @@ class Keyring: valid. Returns: - List<Deferred[None]>: for each input triplet, a deferred indicating success - or failure to verify each json object's signature for the given - server_name. The deferreds run their callbacks in the sentinel - logcontext. + For each input triplet, a deferred indicating success or failure to + verify each json object's signature for the given server_name. The + deferreds run their callbacks in the sentinel logcontext. """ return [ run_in_background( @@ -814,31 +812,27 @@ class ServerKeyFetcher(BaseV2KeyFetcher): results = {} - async def get_key(key_to_fetch_item: _FetchKeyRequest) -> None: + async def get_keys(key_to_fetch_item: _FetchKeyRequest) -> None: server_name = key_to_fetch_item.server_name - key_ids = key_to_fetch_item.key_ids try: - keys = await self.get_server_verify_key_v2_direct(server_name, key_ids) + keys = await self.get_server_verify_keys_v2_direct(server_name) results[server_name] = keys except KeyLookupError as e: - logger.warning( - "Error looking up keys %s from %s: %s", key_ids, server_name, e - ) + logger.warning("Error looking up keys from %s: %s", server_name, e) except Exception: - logger.exception("Error getting keys %s from %s", key_ids, server_name) + logger.exception("Error getting keys from %s", server_name) - await yieldable_gather_results(get_key, keys_to_fetch) + await yieldable_gather_results(get_keys, keys_to_fetch) return results - async def get_server_verify_key_v2_direct( - self, server_name: str, key_ids: Iterable[str] + async def get_server_verify_keys_v2_direct( + self, server_name: str ) -> Dict[str, FetchKeyResult]: """ Args: - server_name: - key_ids: + server_name: Server to request keys from Returns: Map from key ID to lookup result @@ -846,57 +840,41 @@ class ServerKeyFetcher(BaseV2KeyFetcher): Raises: KeyLookupError if there was a problem making the lookup """ - keys: Dict[str, FetchKeyResult] = {} - - for requested_key_id in key_ids: - # we may have found this key as a side-effect of asking for another. - if requested_key_id in keys: - continue - - time_now_ms = self.clock.time_msec() - try: - response = await self.client.get_json( - destination=server_name, - path="/_matrix/key/v2/server/" - + urllib.parse.quote(requested_key_id), - ignore_backoff=True, - # we only give the remote server 10s to respond. It should be an - # easy request to handle, so if it doesn't reply within 10s, it's - # probably not going to. - # - # Furthermore, when we are acting as a notary server, we cannot - # wait all day for all of the origin servers, as the requesting - # server will otherwise time out before we can respond. - # - # (Note that get_json may make 4 attempts, so this can still take - # almost 45 seconds to fetch the headers, plus up to another 60s to - # read the response). - timeout=10000, - ) - except (NotRetryingDestination, RequestSendFailed) as e: - # these both have str() representations which we can't really improve - # upon - raise KeyLookupError(str(e)) - except HttpResponseException as e: - raise KeyLookupError("Remote server returned an error: %s" % (e,)) - - assert isinstance(response, dict) - if response["server_name"] != server_name: - raise KeyLookupError( - "Expected a response for server %r not %r" - % (server_name, response["server_name"]) - ) - - response_keys = await self.process_v2_response( - from_server=server_name, - response_json=response, - time_added_ms=time_now_ms, + time_now_ms = self.clock.time_msec() + try: + response = await self.client.get_json( + destination=server_name, + path="/_matrix/key/v2/server", + ignore_backoff=True, + # we only give the remote server 10s to respond. It should be an + # easy request to handle, so if it doesn't reply within 10s, it's + # probably not going to. + # + # Furthermore, when we are acting as a notary server, we cannot + # wait all day for all of the origin servers, as the requesting + # server will otherwise time out before we can respond. + # + # (Note that get_json may make 4 attempts, so this can still take + # almost 45 seconds to fetch the headers, plus up to another 60s to + # read the response). + timeout=10000, ) - await self.store.store_server_verify_keys( - server_name, - time_now_ms, - ((server_name, key_id, key) for key_id, key in response_keys.items()), + except (NotRetryingDestination, RequestSendFailed) as e: + # these both have str() representations which we can't really improve + # upon + raise KeyLookupError(str(e)) + except HttpResponseException as e: + raise KeyLookupError("Remote server returned an error: %s" % (e,)) + + assert isinstance(response, dict) + if response["server_name"] != server_name: + raise KeyLookupError( + "Expected a response for server %r not %r" + % (server_name, response["server_name"]) ) - keys.update(response_keys) - return keys + return await self.process_v2_response( + from_server=server_name, + response_json=response, + time_added_ms=time_now_ms, + ) diff --git a/synapse/event_auth.py b/synapse/event_auth.py index c7d5ef92fc..d437b7e5d1 100644 --- a/synapse/event_auth.py +++ b/synapse/event_auth.py @@ -15,7 +15,18 @@ import logging import typing -from typing import Any, Collection, Dict, Iterable, List, Optional, Set, Tuple, Union +from typing import ( + Any, + Collection, + Dict, + Iterable, + List, + Mapping, + Optional, + Set, + Tuple, + Union, +) from canonicaljson import encode_canonical_json from signedjson.key import decode_verify_key_bytes @@ -41,6 +52,7 @@ from synapse.api.room_versions import ( KNOWN_ROOM_VERSIONS, EventFormatVersions, RoomVersion, + RoomVersions, ) from synapse.storage.databases.main.events_worker import EventRedactBehaviour from synapse.types import MutableStateMap, StateMap, UserID, get_domain_from_id @@ -134,6 +146,7 @@ def validate_event_for_room_version(event: "EventBase") -> None: async def check_state_independent_auth_rules( store: _EventSourceStore, event: "EventBase", + batched_auth_events: Optional[Mapping[str, "EventBase"]] = None, ) -> None: """Check that an event complies with auth rules that are independent of room state @@ -143,6 +156,8 @@ async def check_state_independent_auth_rules( Args: store: the datastore; used to fetch the auth events for validation event: the event being checked. + batched_auth_events: if the event being authed is part of a batch, any events + from the same batch that may be necessary to auth the current event Raises: AuthError if the checks fail @@ -162,6 +177,9 @@ async def check_state_independent_auth_rules( redact_behaviour=EventRedactBehaviour.as_is, allow_rejected=True, ) + if batched_auth_events: + auth_events.update(batched_auth_events) + room_id = event.room_id auth_dict: MutableStateMap[str] = {} expected_auth_types = auth_types_for_event(event.room_version, event) @@ -324,19 +342,80 @@ def check_state_dependent_auth_rules( logger.debug("Allowing! %s", event) +# Set of room versions where Synapse did not apply event key size limits +# in bytes, but rather in codepoints. +# In these room versions, we are more lenient with event size validation. +LENIENT_EVENT_BYTE_LIMITS_ROOM_VERSIONS = { + RoomVersions.V1, + RoomVersions.V2, + RoomVersions.V3, + RoomVersions.V4, + RoomVersions.V5, + RoomVersions.V6, + RoomVersions.MSC2176, + RoomVersions.V7, + RoomVersions.V8, + RoomVersions.V9, + RoomVersions.MSC3787, + RoomVersions.V10, + RoomVersions.MSC2716v4, + RoomVersions.MSC1767v10, +} + + def _check_size_limits(event: "EventBase") -> None: + """ + Checks the size limits in a PDU. + + The entire size limit of the PDU is checked first. + Then the size of fields is checked, first in codepoints and then in bytes. + + The codepoint size limits are only for Synapse compatibility. + + Raises: + EventSizeError: + when a size limit has been violated. + + unpersistable=True if Synapse never would have accepted the event and + the PDU must NOT be persisted. + + unpersistable=False if a prior version of Synapse would have accepted the + event and so the PDU must be persisted as rejected to avoid + breaking the room. + """ + + # Whole PDU check + if len(encode_canonical_json(event.get_pdu_json())) > MAX_PDU_SIZE: + raise EventSizeError("event too large", unpersistable=True) + + # Codepoint size check: Synapse always enforced these limits, so apply + # them strictly. if len(event.user_id) > 255: - raise EventSizeError("'user_id' too large") + raise EventSizeError("'user_id' too large", unpersistable=True) if len(event.room_id) > 255: - raise EventSizeError("'room_id' too large") + raise EventSizeError("'room_id' too large", unpersistable=True) if event.is_state() and len(event.state_key) > 255: - raise EventSizeError("'state_key' too large") + raise EventSizeError("'state_key' too large", unpersistable=True) if len(event.type) > 255: - raise EventSizeError("'type' too large") + raise EventSizeError("'type' too large", unpersistable=True) if len(event.event_id) > 255: - raise EventSizeError("'event_id' too large") - if len(encode_canonical_json(event.get_pdu_json())) > MAX_PDU_SIZE: - raise EventSizeError("event too large") + raise EventSizeError("'event_id' too large", unpersistable=True) + + strict_byte_limits = ( + event.room_version not in LENIENT_EVENT_BYTE_LIMITS_ROOM_VERSIONS + ) + + # Byte size check: if these fail, then be lenient to avoid breaking rooms. + if len(event.user_id.encode("utf-8")) > 255: + raise EventSizeError("'user_id' too large", unpersistable=strict_byte_limits) + if len(event.room_id.encode("utf-8")) > 255: + raise EventSizeError("'room_id' too large", unpersistable=strict_byte_limits) + if event.is_state() and len(event.state_key.encode("utf-8")) > 255: + raise EventSizeError("'state_key' too large", unpersistable=strict_byte_limits) + if len(event.type.encode("utf-8")) > 255: + raise EventSizeError("'type' too large", unpersistable=strict_byte_limits) + if len(event.event_id.encode("utf-8")) > 255: + raise EventSizeError("'event_id' too large", unpersistable=strict_byte_limits) def _check_create(event: "EventBase") -> None: diff --git a/synapse/events/__init__.py b/synapse/events/__init__.py index 030c3ca408..8aca9a3ab9 100644 --- a/synapse/events/__init__.py +++ b/synapse/events/__init__.py @@ -597,8 +597,7 @@ def _event_type_from_format_version( format_version: The event format version Returns: - type: A type that can be initialized as per the initializer of - `FrozenEvent` + A type that can be initialized as per the initializer of `FrozenEvent` """ if format_version == EventFormatVersions.ROOM_V1_V2: diff --git a/synapse/events/builder.py b/synapse/events/builder.py index e2ee10dd3d..94dd1298e1 100644 --- a/synapse/events/builder.py +++ b/synapse/events/builder.py @@ -28,8 +28,8 @@ from synapse.event_auth import auth_types_for_event from synapse.events import EventBase, _EventInternalMetadata, make_event_from_dict from synapse.state import StateHandler from synapse.storage.databases.main import DataStore -from synapse.storage.state import StateFilter from synapse.types import EventID, JsonDict +from synapse.types.state import StateFilter from synapse.util import Clock from synapse.util.stringutils import random_string @@ -128,6 +128,7 @@ class EventBuilder: state_filter=StateFilter.from_types( auth_types_for_event(self.room_version, self) ), + await_full_state=False, ) auth_event_ids = self._event_auth_handler.compute_auth_events( self, state_ids diff --git a/synapse/events/snapshot.py b/synapse/events/snapshot.py index d3c8083e4a..6eaef8b57a 100644 --- a/synapse/events/snapshot.py +++ b/synapse/events/snapshot.py @@ -23,7 +23,7 @@ from synapse.types import JsonDict, StateMap if TYPE_CHECKING: from synapse.storage.controllers import StorageControllers from synapse.storage.databases.main import DataStore - from synapse.storage.state import StateFilter + from synapse.types.state import StateFilter @attr.s(slots=True, auto_attribs=True) @@ -65,7 +65,8 @@ class EventContext: None does not necessarily mean that ``state_group`` does not have a prev_group! - If the event is a state event, this is normally the same as ``prev_group``. + If the event is a state event, this is normally the same as + ``state_group_before_event``. If ``state_group`` is None (ie, the event is an outlier), ``prev_group`` will always also be ``None``. diff --git a/synapse/events/utils.py b/synapse/events/utils.py index 71853caad8..13fa93afb8 100644 --- a/synapse/events/utils.py +++ b/synapse/events/utils.py @@ -28,8 +28,14 @@ from typing import ( ) import attr +from canonicaljson import encode_canonical_json -from synapse.api.constants import EventContentFields, EventTypes, RelationTypes +from synapse.api.constants import ( + MAX_PDU_SIZE, + EventContentFields, + EventTypes, + RelationTypes, +) from synapse.api.errors import Codes, SynapseError from synapse.api.room_versions import RoomVersion from synapse.types import JsonDict @@ -674,3 +680,27 @@ def validate_canonicaljson(value: Any) -> None: elif not isinstance(value, (bool, str)) and value is not None: # Other potential JSON values (bool, None, str) are safe. raise SynapseError(400, "Unknown JSON value", Codes.BAD_JSON) + + +def maybe_upsert_event_field( + event: EventBase, container: JsonDict, key: str, value: object +) -> bool: + """Upsert an event field, but only if this doesn't make the event too large. + + Returns true iff the upsert took place. + """ + if key in container: + old_value: object = container[key] + container[key] = value + # NB: here and below, we assume that passing a non-None `time_now` argument to + # get_pdu_json doesn't increase the size of the encoded result. + upsert_okay = len(encode_canonical_json(event.get_pdu_json())) <= MAX_PDU_SIZE + if not upsert_okay: + container[key] = old_value + else: + container[key] = value + upsert_okay = len(encode_canonical_json(event.get_pdu_json())) <= MAX_PDU_SIZE + if not upsert_okay: + del container[key] + + return upsert_okay diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py index 4dca711cd2..137cfb3346 100644 --- a/synapse/federation/federation_client.py +++ b/synapse/federation/federation_client.py @@ -80,6 +80,18 @@ PDU_RETRY_TIME_MS = 1 * 60 * 1000 T = TypeVar("T") +@attr.s(frozen=True, slots=True, auto_attribs=True) +class PulledPduInfo: + """ + A result object that stores the PDU and info about it like which homeserver we + pulled it from (`pull_origin`) + """ + + pdu: EventBase + # Which homeserver we pulled the PDU from + pull_origin: str + + class InvalidResponseError(RuntimeError): """Helper for _try_destination_list: indicates that the server returned a response we couldn't parse @@ -114,7 +126,9 @@ class FederationClient(FederationBase): self.hostname = hs.hostname self.signing_key = hs.signing_key - self._get_pdu_cache: ExpiringCache[str, EventBase] = ExpiringCache( + # Cache mapping `event_id` to a tuple of the event itself and the `pull_origin` + # (which server we pulled the event from) + self._get_pdu_cache: ExpiringCache[str, Tuple[EventBase, str]] = ExpiringCache( cache_name="get_pdu_cache", clock=self._clock, max_len=1000, @@ -352,11 +366,11 @@ class FederationClient(FederationBase): @tag_args async def get_pdu( self, - destinations: Iterable[str], + destinations: Collection[str], event_id: str, room_version: RoomVersion, timeout: Optional[int] = None, - ) -> Optional[EventBase]: + ) -> Optional[PulledPduInfo]: """Requests the PDU with given origin and ID from the remote home servers. @@ -371,11 +385,11 @@ class FederationClient(FederationBase): moving to the next destination. None indicates no timeout. Returns: - The requested PDU, or None if we were unable to find it. + The requested PDU wrapped in `PulledPduInfo`, or None if we were unable to find it. """ logger.debug( - "get_pdu: event_id=%s from destinations=%s", event_id, destinations + "get_pdu(event_id=%s): from destinations=%s", event_id, destinations ) # TODO: Rate limit the number of times we try and get the same event. @@ -384,19 +398,25 @@ class FederationClient(FederationBase): # it gets persisted to the database), so we cache the results of the lookup. # Note that this is separate to the regular get_event cache which caches # events once they have been persisted. - event = self._get_pdu_cache.get(event_id) + get_pdu_cache_entry = self._get_pdu_cache.get(event_id) + event = None + pull_origin = None + if get_pdu_cache_entry: + event, pull_origin = get_pdu_cache_entry # If we don't see the event in the cache, go try to fetch it from the # provided remote federated destinations - if not event: + else: pdu_attempts = self.pdu_destination_tried.setdefault(event_id, {}) + # TODO: We can probably refactor this to use `_try_destination_list` for destination in destinations: now = self._clock.time_msec() last_attempt = pdu_attempts.get(destination, 0) if last_attempt + PDU_RETRY_TIME_MS > now: logger.debug( - "get_pdu: skipping destination=%s because we tried it recently last_attempt=%s and we only check every %s (now=%s)", + "get_pdu(event_id=%s): skipping destination=%s because we tried it recently last_attempt=%s and we only check every %s (now=%s)", + event_id, destination, last_attempt, PDU_RETRY_TIME_MS, @@ -411,43 +431,48 @@ class FederationClient(FederationBase): room_version=room_version, timeout=timeout, ) + pull_origin = destination pdu_attempts[destination] = now if event: # Prime the cache - self._get_pdu_cache[event.event_id] = event + self._get_pdu_cache[event.event_id] = (event, pull_origin) # Now that we have an event, we can break out of this # loop and stop asking other destinations. break + except NotRetryingDestination as e: + logger.info("get_pdu(event_id=%s): %s", event_id, e) + continue + except FederationDeniedError: + logger.info( + "get_pdu(event_id=%s): Not attempting to fetch PDU from %s because the homeserver is not on our federation whitelist", + event_id, + destination, + ) + continue except SynapseError as e: logger.info( - "Failed to get PDU %s from %s because %s", + "get_pdu(event_id=%s): Failed to get PDU from %s because %s", event_id, destination, e, ) continue - except NotRetryingDestination as e: - logger.info(str(e)) - continue - except FederationDeniedError as e: - logger.info(str(e)) - continue except Exception as e: pdu_attempts[destination] = now logger.info( - "Failed to get PDU %s from %s because %s", + "get_pdu(event_id=%s): Failed to get PDU from %s because %s", event_id, destination, e, ) continue - if not event: + if not event or not pull_origin: return None # `event` now refers to an object stored in `get_pdu_cache`. Our @@ -459,7 +484,7 @@ class FederationClient(FederationBase): event.room_version, ) - return event_copy + return PulledPduInfo(event_copy, pull_origin) @trace @tag_args @@ -699,12 +724,14 @@ class FederationClient(FederationBase): pdu_origin = get_domain_from_id(pdu.sender) if not res and pdu_origin != origin: try: - res = await self.get_pdu( + pulled_pdu_info = await self.get_pdu( destinations=[pdu_origin], event_id=pdu.event_id, room_version=room_version, timeout=10000, ) + if pulled_pdu_info is not None: + res = pulled_pdu_info.pdu except SynapseError: pass @@ -744,17 +771,28 @@ class FederationClient(FederationBase): """ if synapse_error is None: synapse_error = e.to_synapse_error() - # There is no good way to detect an "unknown" endpoint. - # - # Dendrite returns a 404 (with a body of "404 page not found"); - # Conduit returns a 404 (with no body); and Synapse returns a 400 - # with M_UNRECOGNIZED. + # MSC3743 specifies that servers should return a 404 or 405 with an errcode + # of M_UNRECOGNIZED when they receive a request to an unknown endpoint or + # to an unknown method, respectively. # - # This needs to be rather specific as some endpoints truly do return 404 - # errors. + # Older versions of servers don't properly handle this. This needs to be + # rather specific as some endpoints truly do return 404 errors. return ( - e.code == 404 and (not e.response or e.response == b"404 page not found") - ) or (e.code == 400 and synapse_error.errcode == Codes.UNRECOGNIZED) + # 404 is an unknown endpoint, 405 is a known endpoint, but unknown method. + (e.code == 404 or e.code == 405) + and ( + # Older Dendrites returned a text or empty body. + # Older Conduit returned an empty body. + not e.response + or e.response == b"404 page not found" + # The proper response JSON with M_UNRECOGNIZED errcode. + or synapse_error.errcode == Codes.UNRECOGNIZED + ) + ) or ( + # Older Synapses returned a 400 error. + e.code == 400 + and synapse_error.errcode == Codes.UNRECOGNIZED + ) async def _try_destination_list( self, @@ -806,6 +844,7 @@ class FederationClient(FederationBase): ) for destination in destinations: + # We don't want to ask our own server for information we don't have if destination == self.server_name: continue @@ -814,9 +853,21 @@ class FederationClient(FederationBase): except ( RequestSendFailed, InvalidResponseError, - NotRetryingDestination, ) as e: logger.warning("Failed to %s via %s: %s", description, destination, e) + # Skip to the next homeserver in the list to try. + continue + except NotRetryingDestination as e: + logger.info("%s: %s", description, e) + continue + except FederationDeniedError: + logger.info( + "%s: Not attempting to %s from %s because the homeserver is not on our federation whitelist", + description, + description, + destination, + ) + continue except UnsupportedRoomVersionError: raise except HttpResponseException as e: @@ -1294,7 +1345,7 @@ class FederationClient(FederationBase): return resp[1] async def send_knock(self, destinations: List[str], pdu: EventBase) -> JsonDict: - """Attempts to send a knock event to given a list of servers. Iterates + """Attempts to send a knock event to a given list of servers. Iterates through the list until one attempt succeeds. Doing so will cause the remote server to add the event to the graph, @@ -1609,6 +1660,64 @@ class FederationClient(FederationBase): return result async def timestamp_to_event( + self, *, destinations: List[str], room_id: str, timestamp: int, direction: str + ) -> Optional["TimestampToEventResponse"]: + """ + Calls each remote federating server from `destinations` asking for their closest + event to the given timestamp in the given direction until we get a response. + Also validates the response to always return the expected keys or raises an + error. + + Args: + destinations: The domains of homeservers to try fetching from + room_id: Room to fetch the event from + timestamp: The point in time (inclusive) we should navigate from in + the given direction to find the closest event. + direction: ["f"|"b"] to indicate whether we should navigate forward + or backward from the given timestamp to find the closest event. + + Returns: + A parsed TimestampToEventResponse including the closest event_id + and origin_server_ts or None if no destination has a response. + """ + + async def _timestamp_to_event_from_destination( + destination: str, + ) -> TimestampToEventResponse: + return await self._timestamp_to_event_from_destination( + destination, room_id, timestamp, direction + ) + + try: + # Loop through each homeserver candidate until we get a succesful response + timestamp_to_event_response = await self._try_destination_list( + "timestamp_to_event", + destinations, + # TODO: The requested timestamp may lie in a part of the + # event graph that the remote server *also* didn't have, + # in which case they will have returned another event + # which may be nowhere near the requested timestamp. In + # the future, we may need to reconcile that gap and ask + # other homeservers, and/or extend `/timestamp_to_event` + # to return events on *both* sides of the timestamp to + # help reconcile the gap faster. + _timestamp_to_event_from_destination, + # Since this endpoint is new, we should try other servers before giving up. + # We can safely remove this in a year (remove after 2023-11-16). + failover_on_unknown_endpoint=True, + ) + return timestamp_to_event_response + except SynapseError as e: + logger.warn( + "timestamp_to_event(room_id=%s, timestamp=%s, direction=%s): encountered error when trying to fetch from destinations: %s", + room_id, + timestamp, + direction, + e, + ) + return None + + async def _timestamp_to_event_from_destination( self, destination: str, room_id: str, timestamp: int, direction: str ) -> "TimestampToEventResponse": """ diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index 907940e19e..bb20af6e91 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -74,6 +74,8 @@ from synapse.replication.http.federation import ( ) from synapse.storage.databases.main.events import PartialStateConflictError from synapse.storage.databases.main.lock import Lock +from synapse.storage.databases.main.roommember import extract_heroes_from_room_summary +from synapse.storage.roommember import MemberSummary from synapse.types import JsonDict, StateMap, get_domain_from_id from synapse.util import json_decoder, unwrapFirstError from synapse.util.async_helpers import Linearizer, concurrently_execute, gather_results @@ -481,6 +483,14 @@ class FederationServer(FederationBase): pdu_results[pdu.event_id] = await process_pdu(pdu) async def process_pdu(pdu: EventBase) -> JsonDict: + """ + Processes a pushed PDU sent to us via a `/send` transaction + + Returns: + JsonDict representing a "PDU Processing Result" that will be bundled up + with the other processed PDU's in the `/send` transaction and sent back + to remote homeserver. + """ event_id = pdu.event_id with nested_logging_context(event_id): try: @@ -683,8 +693,9 @@ class FederationServer(FederationBase): state_event_ids: Collection[str] servers_in_room: Optional[Collection[str]] if caller_supports_partial_state: + summary = await self.store.get_room_summary(room_id) state_event_ids = _get_event_ids_for_partial_state_join( - event, prev_state_ids + event, prev_state_ids, summary ) servers_in_room = await self.state.get_hosts_in_room_at_events( room_id, event_ids=event.prev_event_ids() @@ -824,7 +835,14 @@ class FederationServer(FederationBase): context, self._room_prejoin_state_types ) ) - return {"knock_state_events": stripped_room_state} + return { + "knock_room_state": stripped_room_state, + # Since v1.37, Synapse incorrectly used "knock_state_events" for this field. + # Thus, we also populate a 'knock_state_events' with the same content to + # support old instances. + # See https://github.com/matrix-org/synapse/issues/14088. + "knock_state_events": stripped_room_state, + } async def _on_send_membership_event( self, origin: str, content: JsonDict, membership_type: str, room_id: str @@ -1480,6 +1498,7 @@ class FederationHandlerRegistry: def _get_event_ids_for_partial_state_join( join_event: EventBase, prev_state_ids: StateMap[str], + summary: Dict[str, MemberSummary], ) -> Collection[str]: """Calculate state to be retuned in a partial_state send_join @@ -1506,8 +1525,19 @@ def _get_event_ids_for_partial_state_join( if current_membership_event_id is not None: state_event_ids.add(current_membership_event_id) - # TODO: return a few more members: - # - those with invites - # - those that are kicked? / banned + name_id = prev_state_ids.get((EventTypes.Name, "")) + canonical_alias_id = prev_state_ids.get((EventTypes.CanonicalAlias, "")) + if not name_id and not canonical_alias_id: + # Also include the hero members of the room (for DM rooms without a title). + # To do this properly, we should select the correct subset of membership events + # from `prev_state_ids`. Instead, we are lazier and use the (cached) + # `get_room_summary` function, which is based on the current state of the room. + # This introduces races; we choose to ignore them because a) they should be rare + # and b) even if it's wrong, joining servers will get the full state eventually. + heroes = extract_heroes_from_room_summary(summary, join_event.state_key) + for hero in heroes: + membership_event_id = prev_state_ids.get((EventTypes.Member, hero)) + if membership_event_id: + state_event_ids.add(membership_event_id) return state_event_ids diff --git a/synapse/federation/sender/__init__.py b/synapse/federation/sender/__init__.py index a6cb3ba58f..30ebd62883 100644 --- a/synapse/federation/sender/__init__.py +++ b/synapse/federation/sender/__init__.py @@ -353,21 +353,25 @@ class FederationSender(AbstractFederationSender): last_token = await self.store.get_federation_out_pos("events") ( next_token, - events, event_to_received_ts, - ) = await self.store.get_all_new_events_stream( + ) = await self.store.get_all_new_event_ids_stream( last_token, self._last_poked_id, limit=100 ) + event_ids = event_to_received_ts.keys() + event_entries = await self.store.get_unredacted_events_from_cache_or_db( + event_ids + ) + logger.debug( "Handling %i -> %i: %i events to send (current id %i)", last_token, next_token, - len(events), + len(event_entries), self._last_poked_id, ) - if not events and next_token >= self._last_poked_id: + if not event_entries and next_token >= self._last_poked_id: logger.debug("All events processed") break @@ -430,7 +434,23 @@ class FederationSender(AbstractFederationSender): # If there are no prev event IDs then the state is empty # and so no remote servers in the room destinations = set() - else: + + if destinations is None: + # During partial join we use the set of servers that we got + # when beginning the join. It's still possible that we send + # events to servers that left the room in the meantime, but + # we consider that an acceptable risk since it is only our own + # events that we leak and not other server's ones. + partial_state_destinations = ( + await self.store.get_partial_state_servers_at_join( + event.room_id + ) + ) + + if len(partial_state_destinations) > 0: + destinations = partial_state_destinations + + if destinations is None: # We check the external cache for the destinations, which is # stored per state group. @@ -508,8 +528,14 @@ class FederationSender(AbstractFederationSender): await handle_event(event) events_by_room: Dict[str, List[EventBase]] = {} - for event in events: - events_by_room.setdefault(event.room_id, []).append(event) + + for event_id in event_ids: + # `event_entries` is unsorted, so we have to iterate over `event_ids` + # to ensure the events are in the right order + event_cache = event_entries.get(event_id) + if event_cache: + event = event_cache.event + events_by_room.setdefault(event.room_id, []).append(event) await make_deferred_yieldable( defer.gatherResults( @@ -524,9 +550,9 @@ class FederationSender(AbstractFederationSender): logger.debug("Successfully handled up to %i", next_token) await self.store.update_federation_out_pos("events", next_token) - if events: + if event_entries: now = self.clock.time_msec() - ts = event_to_received_ts[events[-1].event_id] + ts = max(t for t in event_to_received_ts.values() if t) assert ts is not None synapse.metrics.event_processing_lag.labels( @@ -536,7 +562,7 @@ class FederationSender(AbstractFederationSender): "federation_sender" ).set(ts) - events_processed_counter.inc(len(events)) + events_processed_counter.inc(len(event_entries)) event_processing_loop_room_count.labels("federation_sender").inc( len(events_by_room) @@ -621,7 +647,7 @@ class FederationSender(AbstractFederationSender): room_id = receipt.room_id # Work out which remote servers should be poked and poke them. - domains_set = await self._storage_controllers.state.get_current_hosts_in_room( + domains_set = await self._storage_controllers.state.get_current_hosts_in_room_or_partial_state_approximation( room_id ) domains = [ diff --git a/synapse/federation/sender/per_destination_queue.py b/synapse/federation/sender/per_destination_queue.py index 084c45a95c..ffc9d95ee7 100644 --- a/synapse/federation/sender/per_destination_queue.py +++ b/synapse/federation/sender/per_destination_queue.py @@ -35,7 +35,7 @@ from synapse.logging import issue9533_logger from synapse.logging.opentracing import SynapseTags, set_tag from synapse.metrics import sent_transactions_counter from synapse.metrics.background_process_metrics import run_as_background_process -from synapse.types import ReadReceipt +from synapse.types import JsonDict, ReadReceipt from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter from synapse.visibility import filter_events_for_server @@ -136,8 +136,11 @@ class PerDestinationQueue: # destination self._pending_presence: Dict[str, UserPresenceState] = {} - # room_id -> receipt_type -> user_id -> receipt_dict - self._pending_rrs: Dict[str, Dict[str, Dict[str, dict]]] = {} + # List of room_id -> receipt_type -> user_id -> receipt_dict, + # + # Each receipt can only have a single receipt per + # (room ID, receipt type, user ID, thread ID) tuple. + self._pending_receipt_edus: List[Dict[str, Dict[str, Dict[str, dict]]]] = [] self._rrs_pending_flush = False # stream_id of last successfully sent to-device message. @@ -202,17 +205,53 @@ class PerDestinationQueue: Args: receipt: receipt to be queued """ - self._pending_rrs.setdefault(receipt.room_id, {}).setdefault( - receipt.receipt_type, {} - )[receipt.user_id] = {"event_ids": receipt.event_ids, "data": receipt.data} + serialized_receipt: JsonDict = { + "event_ids": receipt.event_ids, + "data": receipt.data, + } + if receipt.thread_id is not None: + serialized_receipt["data"]["thread_id"] = receipt.thread_id + + # Find which EDU to add this receipt to. There's three situations depending + # on the (room ID, receipt type, user, thread ID) tuple: + # + # 1. If it fully matches, clobber the information. + # 2. If it is missing, add the information. + # 3. If the subset tuple of (room ID, receipt type, user) matches, check + # the next EDU (or add a new EDU). + for edu in self._pending_receipt_edus: + receipt_content = edu.setdefault(receipt.room_id, {}).setdefault( + receipt.receipt_type, {} + ) + # If this room ID, receipt type, user ID is not in this EDU, OR if + # the full tuple matches, use the current EDU. + if ( + receipt.user_id not in receipt_content + or receipt_content[receipt.user_id].get("thread_id") + == receipt.thread_id + ): + receipt_content[receipt.user_id] = serialized_receipt + break + + # If no matching EDU was found, create a new one. + else: + self._pending_receipt_edus.append( + { + receipt.room_id: { + receipt.receipt_type: {receipt.user_id: serialized_receipt} + } + } + ) def flush_read_receipts_for_room(self, room_id: str) -> None: - # if we don't have any read-receipts for this room, it may be that we've already - # sent them out, so we don't need to flush. - if room_id not in self._pending_rrs: - return - self._rrs_pending_flush = True - self.attempt_new_transaction() + # If there are any pending receipts for this room then force-flush them + # in a new transaction. + for edu in self._pending_receipt_edus: + if room_id in edu: + self._rrs_pending_flush = True + self.attempt_new_transaction() + # No use in checking remaining EDUs if the room was found. + break def send_keyed_edu(self, edu: Edu, key: Hashable) -> None: self._pending_edus_keyed[(edu.edu_type, key)] = edu @@ -351,7 +390,7 @@ class PerDestinationQueue: self._pending_edus = [] self._pending_edus_keyed = {} self._pending_presence = {} - self._pending_rrs = {} + self._pending_receipt_edus = [] self._start_catching_up() except FederationDeniedError as e: @@ -505,6 +544,7 @@ class PerDestinationQueue: new_pdus = await filter_events_for_server( self._storage_controllers, self._destination, + self._server_name, new_pdus, redact=False, ) @@ -542,22 +582,27 @@ class PerDestinationQueue: self._destination, last_successful_stream_ordering ) - def _get_rr_edus(self, force_flush: bool) -> Iterable[Edu]: - if not self._pending_rrs: + def _get_receipt_edus(self, force_flush: bool, limit: int) -> Iterable[Edu]: + if not self._pending_receipt_edus: return if not force_flush and not self._rrs_pending_flush: # not yet time for this lot return - edu = Edu( - origin=self._server_name, - destination=self._destination, - edu_type=EduTypes.RECEIPT, - content=self._pending_rrs, - ) - self._pending_rrs = {} - self._rrs_pending_flush = False - yield edu + # Send at most limit EDUs for receipts. + for content in self._pending_receipt_edus[:limit]: + yield Edu( + origin=self._server_name, + destination=self._destination, + edu_type=EduTypes.RECEIPT, + content=content, + ) + self._pending_receipt_edus = self._pending_receipt_edus[limit:] + + # If there are still pending read-receipts, don't reset the pending flush + # flag. + if not self._pending_receipt_edus: + self._rrs_pending_flush = False def _pop_pending_edus(self, limit: int) -> List[Edu]: pending_edus = self._pending_edus @@ -596,7 +641,7 @@ class PerDestinationQueue: if not message_id: continue - set_tag(SynapseTags.TO_DEVICE_MESSAGE_ID, message_id) + set_tag(SynapseTags.TO_DEVICE_EDU_ID, message_id) edus = [ Edu( @@ -644,27 +689,61 @@ class _TransactionQueueManager: async def __aenter__(self) -> Tuple[List[EventBase], List[Edu]]: # First we calculate the EDUs we want to send, if any. - # We start by fetching device related EDUs, i.e device updates and to - # device messages. We have to keep 2 free slots for presence and rr_edus. - device_edu_limit = MAX_EDUS_PER_TRANSACTION - 2 + # There's a maximum number of EDUs that can be sent with a transaction, + # generally device updates and to-device messages get priority, but we + # want to ensure that there's room for some other EDUs as well. + # + # This is done by: + # + # * Add a presence EDU, if one exists. + # * Add up-to a small limit of read receipt EDUs. + # * Add to-device EDUs, but leave some space for device list updates. + # * Add device list updates EDUs. + # * If there's any remaining room, add other EDUs. + pending_edus = [] + + # Add presence EDU. + if self.queue._pending_presence: + pending_edus.append( + Edu( + origin=self.queue._server_name, + destination=self.queue._destination, + edu_type=EduTypes.PRESENCE, + content={ + "push": [ + format_user_presence_state( + presence, self.queue._clock.time_msec() + ) + for presence in self.queue._pending_presence.values() + ] + }, + ) + ) + self.queue._pending_presence = {} - # We prioritize to-device messages so that existing encryption channels + # Add read receipt EDUs. + pending_edus.extend(self.queue._get_receipt_edus(force_flush=False, limit=5)) + edu_limit = MAX_EDUS_PER_TRANSACTION - len(pending_edus) + + # Next, prioritize to-device messages so that existing encryption channels # work. We also keep a few slots spare (by reducing the limit) so that # we can still trickle out some device list updates. ( to_device_edus, device_stream_id, - ) = await self.queue._get_to_device_message_edus(device_edu_limit - 10) + ) = await self.queue._get_to_device_message_edus(edu_limit - 10) if to_device_edus: self._device_stream_id = device_stream_id else: self.queue._last_device_stream_id = device_stream_id - device_edu_limit -= len(to_device_edus) + pending_edus.extend(to_device_edus) + edu_limit -= len(to_device_edus) + # Add device list update EDUs. device_update_edus, dev_list_id = await self.queue._get_device_update_edus( - device_edu_limit + edu_limit ) if device_update_edus: @@ -672,40 +751,17 @@ class _TransactionQueueManager: else: self.queue._last_device_list_stream_id = dev_list_id - pending_edus = device_update_edus + to_device_edus - - # Now add the read receipt EDU. - pending_edus.extend(self.queue._get_rr_edus(force_flush=False)) - - # And presence EDU. - if self.queue._pending_presence: - pending_edus.append( - Edu( - origin=self.queue._server_name, - destination=self.queue._destination, - edu_type=EduTypes.PRESENCE, - content={ - "push": [ - format_user_presence_state( - presence, self.queue._clock.time_msec() - ) - for presence in self.queue._pending_presence.values() - ] - }, - ) - ) - self.queue._pending_presence = {} + pending_edus.extend(device_update_edus) + edu_limit -= len(device_update_edus) # Finally add any other types of EDUs if there is room. - pending_edus.extend( - self.queue._pop_pending_edus(MAX_EDUS_PER_TRANSACTION - len(pending_edus)) - ) - while ( - len(pending_edus) < MAX_EDUS_PER_TRANSACTION - and self.queue._pending_edus_keyed - ): + other_edus = self.queue._pop_pending_edus(edu_limit) + pending_edus.extend(other_edus) + edu_limit -= len(other_edus) + while edu_limit > 0 and self.queue._pending_edus_keyed: _, val = self.queue._pending_edus_keyed.popitem() pending_edus.append(val) + edu_limit -= 1 # Now we look for any PDUs to send, by getting up to 50 PDUs from the # queue @@ -716,8 +772,10 @@ class _TransactionQueueManager: # if we've decided to send a transaction anyway, and we have room, we # may as well send any pending RRs - if len(pending_edus) < MAX_EDUS_PER_TRANSACTION: - pending_edus.extend(self.queue._get_rr_edus(force_flush=True)) + if edu_limit: + pending_edus.extend( + self.queue._get_receipt_edus(force_flush=True, limit=edu_limit) + ) if self._pdus: self._last_stream_ordering = self._pdus[ diff --git a/synapse/federation/transport/client.py b/synapse/federation/transport/client.py index cd39d4d111..77f1f39cac 100644 --- a/synapse/federation/transport/client.py +++ b/synapse/federation/transport/client.py @@ -185,9 +185,8 @@ class TransportLayerClient: Raises: Various exceptions when the request fails """ - path = _create_path( - FEDERATION_UNSTABLE_PREFIX, - "/org.matrix.msc3030/timestamp_to_event/%s", + path = _create_v1_path( + "/timestamp_to_event/%s", room_id, ) @@ -280,12 +279,11 @@ class TransportLayerClient: Note that this does not append any events to any graphs. Args: - destination (str): address of remote homeserver - room_id (str): room to join/leave - user_id (str): user to be joined/left - membership (str): one of join/leave - params (dict[str, str|Iterable[str]]): Query parameters to include in the - request. + destination: address of remote homeserver + room_id: room to join/leave + user_id: user to be joined/left + membership: one of join/leave + params: Query parameters to include in the request. Returns: Succeeds when we get a 2xx HTTP response. The result diff --git a/synapse/federation/transport/server/__init__.py b/synapse/federation/transport/server/__init__.py index 50623cd385..2725f53cf6 100644 --- a/synapse/federation/transport/server/__init__.py +++ b/synapse/federation/transport/server/__init__.py @@ -25,7 +25,6 @@ from synapse.federation.transport.server._base import ( from synapse.federation.transport.server.federation import ( FEDERATION_SERVLET_CLASSES, FederationAccountStatusServlet, - FederationTimestampLookupServlet, ) from synapse.http.server import HttpServer, JsonResource from synapse.http.servlet import ( @@ -291,13 +290,6 @@ def register_servlets( ) for servletclass in SERVLET_GROUPS[servlet_group]: - # Only allow the `/timestamp_to_event` servlet if msc3030 is enabled - if ( - servletclass == FederationTimestampLookupServlet - and not hs.config.experimental.msc3030_enabled - ): - continue - # Only allow the `/account_status` servlet if msc3720 is enabled if ( servletclass == FederationAccountStatusServlet diff --git a/synapse/federation/transport/server/_base.py b/synapse/federation/transport/server/_base.py index 1db8009d6c..cdaf0d5de7 100644 --- a/synapse/federation/transport/server/_base.py +++ b/synapse/federation/transport/server/_base.py @@ -224,10 +224,10 @@ class BaseFederationServlet: With arguments: - origin (unicode|None): The authenticated server_name of the calling server, + origin (str|None): The authenticated server_name of the calling server, unless REQUIRE_AUTH is set to False and authentication failed. - content (unicode|None): decoded json body of the request. None if the + content (str|None): decoded json body of the request. None if the request was a GET. query (dict[bytes, list[bytes]]): Query params from the request. url-decoded diff --git a/synapse/federation/transport/server/federation.py b/synapse/federation/transport/server/federation.py index 6bb4659c4c..53e77b4bb6 100644 --- a/synapse/federation/transport/server/federation.py +++ b/synapse/federation/transport/server/federation.py @@ -218,14 +218,13 @@ class FederationTimestampLookupServlet(BaseFederationServerServlet): `dir` can be `f` or `b` to indicate forwards and backwards in time from the given timestamp. - GET /_matrix/federation/unstable/org.matrix.msc3030/timestamp_to_event/<roomID>?ts=<timestamp>&dir=<direction> + GET /_matrix/federation/v1/timestamp_to_event/<roomID>?ts=<timestamp>&dir=<direction> { "event_id": ... } """ PATH = "/timestamp_to_event/(?P<room_id>[^/]*)/?" - PREFIX = FEDERATION_UNSTABLE_PREFIX + "/org.matrix.msc3030" async def on_GET( self, @@ -489,7 +488,7 @@ class FederationV2InviteServlet(BaseFederationServerServlet): room_version = content["room_version"] event = content["event"] - invite_room_state = content["invite_room_state"] + invite_room_state = content.get("invite_room_state", []) # Synapse expects invite_room_state to be in unsigned, as it is in v1 # API @@ -499,6 +498,11 @@ class FederationV2InviteServlet(BaseFederationServerServlet): result = await self.handler.on_invite_request( origin, event, room_version_id=room_version ) + + # We only store invite_room_state for internal use, so remove it before + # returning the event to the remote homeserver. + result["event"].get("unsigned", {}).pop("invite_room_state", None) + return 200, result diff --git a/synapse/handlers/account_data.py b/synapse/handlers/account_data.py index 0478448b47..fc21d58001 100644 --- a/synapse/handlers/account_data.py +++ b/synapse/handlers/account_data.py @@ -225,7 +225,7 @@ class AccountDataEventSource(EventSource[int, JsonDict]): self, user: UserID, from_key: int, - limit: Optional[int], + limit: int, room_ids: Collection[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/handlers/admin.py b/synapse/handlers/admin.py index f2989cc4a2..5bf8e86387 100644 --- a/synapse/handlers/admin.py +++ b/synapse/handlers/admin.py @@ -100,6 +100,7 @@ class AdminHandler: user_info_dict["avatar_url"] = profile.avatar_url user_info_dict["threepids"] = threepids user_info_dict["external_ids"] = external_ids + user_info_dict["erased"] = await self.store.is_user_erased(user.to_string()) return user_info_dict diff --git a/synapse/handlers/appservice.py b/synapse/handlers/appservice.py index 203b62e015..5d1d21cdc8 100644 --- a/synapse/handlers/appservice.py +++ b/synapse/handlers/appservice.py @@ -109,10 +109,13 @@ class ApplicationServicesHandler: last_token = await self.store.get_appservice_last_pos() ( upper_bound, - events, event_to_received_ts, - ) = await self.store.get_all_new_events_stream( - last_token, self.current_max, limit=100, get_prev_content=True + ) = await self.store.get_all_new_event_ids_stream( + last_token, self.current_max, limit=100 + ) + + events = await self.store.get_events_as_list( + event_to_received_ts.keys(), get_prev_content=True ) events_by_room: Dict[str, List[EventBase]] = {} @@ -575,9 +578,6 @@ class ApplicationServicesHandler: device_id, ), messages in recipient_device_to_messages.items(): for message_json in messages: - # Remove 'message_id' from the to-device message, as it's an internal ID - message_json.pop("message_id", None) - message_payload.append( { "to_user_id": user_id, @@ -612,8 +612,8 @@ class ApplicationServicesHandler: ) # Fetch the users who have modified their device list since then. - users_with_changed_device_lists = ( - await self.store.get_users_whose_devices_changed(from_key, to_key=new_key) + users_with_changed_device_lists = await self.store.get_all_devices_changed( + from_key, to_key=new_key ) # Filter out any users the application service is not interested in diff --git a/synapse/handlers/auth.py b/synapse/handlers/auth.py index f5f0e0e7a7..8b9ef25d29 100644 --- a/synapse/handlers/auth.py +++ b/synapse/handlers/auth.py @@ -38,6 +38,7 @@ from typing import ( import attr import bcrypt import unpaddedbase64 +from prometheus_client import Counter from twisted.internet.defer import CancelledError from twisted.web.server import Request @@ -48,6 +49,7 @@ from synapse.api.errors import ( Codes, InteractiveAuthIncompleteError, LoginError, + NotFoundError, StoreError, SynapseError, UserDeactivatedError, @@ -63,10 +65,14 @@ from synapse.http.server import finish_request, respond_with_html from synapse.http.site import SynapseRequest from synapse.logging.context import defer_to_thread from synapse.metrics.background_process_metrics import run_as_background_process +from synapse.storage.databases.main.registration import ( + LoginTokenExpired, + LoginTokenLookupResult, + LoginTokenReused, +) from synapse.types import JsonDict, Requester, UserID from synapse.util import stringutils as stringutils from synapse.util.async_helpers import delay_cancellation, maybe_awaitable -from synapse.util.macaroons import LoginTokenAttributes from synapse.util.msisdn import phone_number_to_msisdn from synapse.util.stringutils import base62_encode from synapse.util.threepids import canonicalise_email @@ -80,6 +86,12 @@ logger = logging.getLogger(__name__) INVALID_USERNAME_OR_PASSWORD = "Invalid username or password" +invalid_login_token_counter = Counter( + "synapse_user_login_invalid_login_tokens", + "Counts the number of rejected m.login.token on /login", + ["reason"], +) + def convert_client_dict_legacy_fields_to_identifier( submission: JsonDict, @@ -883,6 +895,25 @@ class AuthHandler: return True + async def create_login_token_for_user_id( + self, + user_id: str, + duration_ms: int = (2 * 60 * 1000), + auth_provider_id: Optional[str] = None, + auth_provider_session_id: Optional[str] = None, + ) -> str: + login_token = self.generate_login_token() + now = self._clock.time_msec() + expiry_ts = now + duration_ms + await self.store.add_login_token_to_user( + user_id=user_id, + token=login_token, + expiry_ts=expiry_ts, + auth_provider_id=auth_provider_id, + auth_provider_session_id=auth_provider_session_id, + ) + return login_token + async def create_refresh_token_for_user_id( self, user_id: str, @@ -1401,6 +1432,18 @@ class AuthHandler: return None return user_id + def generate_login_token(self) -> str: + """Generates an opaque string, for use as an short-term login token""" + + # we use the following format for access tokens: + # syl_<random string>_<base62 crc check> + + random_string = stringutils.random_string(20) + base = f"syl_{random_string}" + + crc = base62_encode(crc32(base.encode("ascii")), minwidth=6) + return f"{base}_{crc}" + def generate_access_token(self, for_user: UserID) -> str: """Generates an opaque string, for use as an access token""" @@ -1427,16 +1470,17 @@ class AuthHandler: crc = base62_encode(crc32(base.encode("ascii")), minwidth=6) return f"{base}_{crc}" - async def validate_short_term_login_token( - self, login_token: str - ) -> LoginTokenAttributes: + async def consume_login_token(self, login_token: str) -> LoginTokenLookupResult: try: - res = self.macaroon_gen.verify_short_term_login_token(login_token) - except Exception: - raise AuthError(403, "Invalid login token", errcode=Codes.FORBIDDEN) + return await self.store.consume_login_token(login_token) + except LoginTokenExpired: + invalid_login_token_counter.labels("expired").inc() + except LoginTokenReused: + invalid_login_token_counter.labels("reused").inc() + except NotFoundError: + invalid_login_token_counter.labels("not found").inc() - await self.auth_blocking.check_auth_blocking(res.user_id) - return res + raise AuthError(403, "Invalid login token", errcode=Codes.FORBIDDEN) async def delete_access_token(self, access_token: str) -> None: """Invalidate a single access token @@ -1711,7 +1755,7 @@ class AuthHandler: ) # Create a login token - login_token = self.macaroon_gen.generate_short_term_login_token( + login_token = await self.create_login_token_for_user_id( registered_user_id, auth_provider_id=auth_provider_id, auth_provider_session_id=auth_provider_session_id, diff --git a/synapse/handlers/deactivate_account.py b/synapse/handlers/deactivate_account.py index 816e1a6d79..d74d135c0c 100644 --- a/synapse/handlers/deactivate_account.py +++ b/synapse/handlers/deactivate_account.py @@ -16,6 +16,7 @@ import logging from typing import TYPE_CHECKING, Optional from synapse.api.errors import SynapseError +from synapse.handlers.device import DeviceHandler from synapse.metrics.background_process_metrics import run_as_background_process from synapse.types import Codes, Requester, UserID, create_requester @@ -76,6 +77,9 @@ class DeactivateAccountHandler: True if identity server supports removing threepids, otherwise False. """ + # This can only be called on the main process. + assert isinstance(self._device_handler, DeviceHandler) + # Check if this user can be deactivated if not await self._third_party_rules.check_can_deactivate_user( user_id, by_admin diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index f9cc5bddbc..d4750a32e6 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -65,6 +65,8 @@ DELETE_STALE_DEVICES_INTERVAL_MS = 24 * 60 * 60 * 1000 class DeviceWorkerHandler: + device_list_updater: "DeviceListWorkerUpdater" + def __init__(self, hs: "HomeServer"): self.clock = hs.get_clock() self.hs = hs @@ -76,6 +78,8 @@ class DeviceWorkerHandler: self.server_name = hs.hostname self._msc3852_enabled = hs.config.experimental.msc3852_enabled + self.device_list_updater = DeviceListWorkerUpdater(hs) + @trace async def get_devices_by_user(self, user_id: str) -> List[JsonDict]: """ @@ -99,6 +103,19 @@ class DeviceWorkerHandler: log_kv(device_map) return devices + 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) + @trace async def get_device(self, user_id: str, device_id: str) -> JsonDict: """Retrieve the given device @@ -127,7 +144,7 @@ class DeviceWorkerHandler: @cancellable async def get_device_changes_in_shared_rooms( self, user_id: str, room_ids: Collection[str], from_token: StreamToken - ) -> Collection[str]: + ) -> Set[str]: """Get the set of users whose devices have changed who share a room with the given user. """ @@ -320,6 +337,8 @@ class DeviceWorkerHandler: class DeviceHandler(DeviceWorkerHandler): + device_list_updater: "DeviceListUpdater" + def __init__(self, hs: "HomeServer"): super().__init__(hs) @@ -606,19 +625,6 @@ class DeviceHandler(DeviceWorkerHandler): await self.delete_devices(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: @@ -682,13 +688,33 @@ class DeviceHandler(DeviceWorkerHandler): hosts_already_sent_to: Set[str] = set() try: + stream_id, room_id = await self.store.get_device_change_last_converted_pos() + while True: self._handle_new_device_update_new_data = False - rows = await self.store.get_uncoverted_outbound_room_pokes() + max_stream_id = self.store.get_device_stream_token() + rows = await self.store.get_uncoverted_outbound_room_pokes( + stream_id, room_id + ) if not rows: # If the DB returned nothing then there is nothing left to # do, *unless* a new device list update happened during the # DB query. + + # Advance `(stream_id, room_id)`. + # `max_stream_id` comes from *before* the query for unconverted + # rows, which means that any unconverted rows must have a larger + # stream ID. + if max_stream_id > stream_id: + stream_id, room_id = max_stream_id, "" + await self.store.set_device_change_last_converted_pos( + stream_id, room_id + ) + else: + assert max_stream_id == stream_id + # Avoid moving `room_id` backwards. + pass + if self._handle_new_device_update_new_data: continue else: @@ -718,7 +744,6 @@ class DeviceHandler(DeviceWorkerHandler): user_id=user_id, device_id=device_id, room_id=room_id, - stream_id=stream_id, hosts=hosts, context=opentracing_context, ) @@ -752,6 +777,12 @@ class DeviceHandler(DeviceWorkerHandler): hosts_already_sent_to.update(hosts) current_stream_id = stream_id + # Advance `(stream_id, room_id)`. + _, _, room_id, stream_id, _ = rows[-1] + await self.store.set_device_change_last_converted_pos( + stream_id, room_id + ) + finally: self._handle_new_device_update_is_processing = False @@ -834,7 +865,6 @@ class DeviceHandler(DeviceWorkerHandler): user_id=user_id, device_id=device_id, room_id=room_id, - stream_id=None, hosts=potentially_changed_hosts, context=None, ) @@ -858,7 +888,36 @@ def _update_device_from_client_ips( ) -class DeviceListUpdater: +class DeviceListWorkerUpdater: + "Handles incoming device list updates from federation and contacts the main process over replication" + + def __init__(self, hs: "HomeServer"): + from synapse.replication.http.devices import ( + ReplicationUserDevicesResyncRestServlet, + ) + + self._user_device_resync_client = ( + ReplicationUserDevicesResyncRestServlet.make_client(hs) + ) + + async def user_device_resync( + self, user_id: str, mark_failed_as_stale: bool = True + ) -> Optional[JsonDict]: + """Fetches all devices for a user and updates the device cache with them. + + Args: + user_id: The user's id whose device_list will be updated. + mark_failed_as_stale: Whether to mark the user's device list as stale + if the attempt to resync failed. + Returns: + A dict with device info as under the "devices" in the result of this + request: + https://matrix.org/docs/spec/server_server/r0.1.2#get-matrix-federation-v1-user-devices-userid + """ + return await self._user_device_resync_client(user_id=user_id) + + +class DeviceListUpdater(DeviceListWorkerUpdater): "Handles incoming device list updates from federation and updates the DB" def __init__(self, hs: "HomeServer", device_handler: DeviceHandler): @@ -937,7 +996,10 @@ class DeviceListUpdater: # Check if we are partially joining any rooms. If so we need to store # all device list updates so that we can handle them correctly once we # know who is in the room. - partial_rooms = await self.store.get_partial_state_rooms_and_servers() + # TODO(faster_joins): this fetches and processes a bunch of data that we don't + # use. Could be replaced by a tighter query e.g. + # SELECT EXISTS(SELECT 1 FROM partial_state_rooms) + partial_rooms = await self.store.get_partial_state_room_resync_info() if partial_rooms: await self.store.add_remote_device_list_to_pending( user_id, diff --git a/synapse/handlers/devicemessage.py b/synapse/handlers/devicemessage.py index 444c08bc2e..75e89850f5 100644 --- a/synapse/handlers/devicemessage.py +++ b/synapse/handlers/devicemessage.py @@ -15,7 +15,7 @@ import logging from typing import TYPE_CHECKING, Any, Dict -from synapse.api.constants import EduTypes, ToDeviceEventTypes +from synapse.api.constants import EduTypes, EventContentFields, ToDeviceEventTypes from synapse.api.errors import SynapseError from synapse.api.ratelimiting import Ratelimiter from synapse.logging.context import run_in_background @@ -216,14 +216,24 @@ class DeviceMessageHandler: """ sender_user_id = requester.user.to_string() - message_id = random_string(16) - set_tag(SynapseTags.TO_DEVICE_MESSAGE_ID, message_id) - - log_kv({"number_of_to_device_messages": len(messages)}) - set_tag("sender", sender_user_id) + set_tag(SynapseTags.TO_DEVICE_TYPE, message_type) + set_tag(SynapseTags.TO_DEVICE_SENDER, sender_user_id) local_messages = {} remote_messages: Dict[str, Dict[str, Dict[str, JsonDict]]] = {} for user_id, by_device in messages.items(): + # add an opentracing log entry for each message + for device_id, message_content in by_device.items(): + log_kv( + { + "event": "send_to_device_message", + "user_id": user_id, + "device_id": device_id, + EventContentFields.TO_DEVICE_MSGID: message_content.get( + EventContentFields.TO_DEVICE_MSGID + ), + } + ) + # Ratelimit local cross-user key requests by the sending device. if ( message_type == ToDeviceEventTypes.RoomKeyRequest @@ -233,6 +243,7 @@ class DeviceMessageHandler: requester, (sender_user_id, requester.device_id) ) if not allowed: + log_kv({"message": f"dropping key requests to {user_id}"}) logger.info( "Dropping room_key_request from %s to %s due to rate limit", sender_user_id, @@ -247,18 +258,11 @@ class DeviceMessageHandler: "content": message_content, "type": message_type, "sender": sender_user_id, - "message_id": message_id, } for device_id, message_content in by_device.items() } if messages_by_device: local_messages[user_id] = messages_by_device - log_kv( - { - "user_id": user_id, - "device_id": list(messages_by_device), - } - ) else: destination = get_domain_from_id(user_id) remote_messages.setdefault(destination, {})[user_id] = by_device @@ -267,7 +271,11 @@ class DeviceMessageHandler: remote_edu_contents = {} for destination, messages in remote_messages.items(): - log_kv({"destination": destination}) + # The EDU contains a "message_id" property which is used for + # idempotence. Make up a random one. + message_id = random_string(16) + log_kv({"destination": destination, "message_id": message_id}) + remote_edu_contents[destination] = { "messages": messages, "sender": sender_user_id, diff --git a/synapse/handlers/directory.py b/synapse/handlers/directory.py index 7127d5aefc..2ea52257cb 100644 --- a/synapse/handlers/directory.py +++ b/synapse/handlers/directory.py @@ -16,6 +16,8 @@ import logging import string from typing import TYPE_CHECKING, Iterable, List, Optional +from typing_extensions import Literal + from synapse.api.constants import MAX_ALIAS_LENGTH, EventTypes from synapse.api.errors import ( AuthError, @@ -83,7 +85,7 @@ class DirectoryHandler: # TODO(erikj): Add transactions. # TODO(erikj): Check if there is a current association. if not servers: - servers = await self._storage_controllers.state.get_current_hosts_in_room( + servers = await self._storage_controllers.state.get_current_hosts_in_room_or_partial_state_approximation( room_id ) @@ -288,7 +290,7 @@ class DirectoryHandler: Codes.NOT_FOUND, ) - extra_servers = await self._storage_controllers.state.get_current_hosts_in_room( + extra_servers = await self._storage_controllers.state.get_current_hosts_in_room_or_partial_state_approximation( room_id ) servers_set = set(extra_servers) | set(servers) @@ -429,7 +431,10 @@ class DirectoryHandler: return await self.auth.check_can_change_room_list(room_id, requester) async def edit_published_room_list( - self, requester: Requester, room_id: str, visibility: str + self, + requester: Requester, + room_id: str, + visibility: Literal["public", "private"], ) -> None: """Edit the entry of the room in the published room list. @@ -451,9 +456,6 @@ class DirectoryHandler: if requester.is_guest: raise AuthError(403, "Guests cannot edit the published room list") - if visibility not in ["public", "private"]: - raise SynapseError(400, "Invalid visibility setting") - if visibility == "public" and not self.enable_room_list_search: # The room list has been disabled. raise AuthError( @@ -505,7 +507,11 @@ class DirectoryHandler: await self.store.set_room_is_public(room_id, making_public) async def edit_published_appservice_room_list( - self, appservice_id: str, network_id: str, room_id: str, visibility: str + self, + appservice_id: str, + network_id: str, + room_id: str, + visibility: Literal["public", "private"], ) -> None: """Add or remove a room from the appservice/network specific public room list. @@ -516,9 +522,6 @@ class DirectoryHandler: room_id visibility: either "public" or "private" """ - if visibility not in ["public", "private"]: - raise SynapseError(400, "Invalid visibility setting") - await self.store.set_room_is_public_appservice( room_id, appservice_id, network_id, visibility == "public" ) diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index 09a2492afc..5fe102e2f2 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -27,9 +27,9 @@ from twisted.internet import defer from synapse.api.constants import EduTypes from synapse.api.errors import CodeMessageException, Codes, NotFoundError, SynapseError +from synapse.handlers.device import DeviceHandler from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.logging.opentracing import log_kv, set_tag, tag_args, trace -from synapse.replication.http.devices import ReplicationUserDevicesResyncRestServlet from synapse.types import ( JsonDict, UserID, @@ -49,33 +49,30 @@ logger = logging.getLogger(__name__) class E2eKeysHandler: def __init__(self, hs: "HomeServer"): + self.config = hs.config self.store = hs.get_datastores().main self.federation = hs.get_federation_client() self.device_handler = hs.get_device_handler() self.is_mine = hs.is_mine self.clock = hs.get_clock() - self._edu_updater = SigningKeyEduUpdater(hs, self) - federation_registry = hs.get_federation_registry() - self._is_master = hs.config.worker.worker_app is None - if not self._is_master: - self._user_device_resync_client = ( - ReplicationUserDevicesResyncRestServlet.make_client(hs) - ) - else: + is_master = hs.config.worker.worker_app is None + if is_master: + edu_updater = SigningKeyEduUpdater(hs) + # Only register this edu handler on master as it requires writing # device updates to the db federation_registry.register_edu_handler( EduTypes.SIGNING_KEY_UPDATE, - self._edu_updater.incoming_signing_key_update, + edu_updater.incoming_signing_key_update, ) # also handle the unstable version # FIXME: remove this when enough servers have upgraded federation_registry.register_edu_handler( EduTypes.UNSTABLE_SIGNING_KEY_UPDATE, - self._edu_updater.incoming_signing_key_update, + edu_updater.incoming_signing_key_update, ) # doesn't really work as part of the generic query API, because the @@ -318,14 +315,13 @@ class E2eKeysHandler: # 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( + 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 - ) + ) + if resync_results is None: + raise ValueError("Device resync failed") # Add the device keys to the results. user_devices = resync_results["devices"] @@ -431,13 +427,17 @@ class E2eKeysHandler: @trace @cancellable async def query_local_devices( - self, query: Mapping[str, Optional[List[str]]] + self, + query: Mapping[str, Optional[List[str]]], + include_displaynames: bool = True, ) -> Dict[str, Dict[str, dict]]: """Get E2E device keys for local users Args: query: map from user_id to a list of devices to query (None for all devices) + include_displaynames: Whether to include device displaynames in the returned + device details. Returns: A map from user_id -> device_id -> device details @@ -469,7 +469,9 @@ class E2eKeysHandler: # make sure that each queried user appears in the result dict result_dict[user_id] = {} - results = await self.store.get_e2e_device_keys_for_cs_api(local_query) + results = await self.store.get_e2e_device_keys_for_cs_api( + local_query, include_displaynames + ) # Build the result structure for user_id, device_keys in results.items(): @@ -482,11 +484,33 @@ class E2eKeysHandler: async def on_federation_query_client_keys( self, query_body: Dict[str, Dict[str, Optional[List[str]]]] ) -> JsonDict: - """Handle a device key query from a federated server""" + """Handle a device key query from a federated server: + + Handles the path: GET /_matrix/federation/v1/users/keys/query + + Args: + query_body: The body of the query request. Should contain a key + "device_keys" that map to a dictionary of user ID's -> list of + device IDs. If the list of device IDs is empty, all devices of + that user will be queried. + + Returns: + A json dictionary containing the following: + - device_keys: A dictionary containing the requested device information. + - master_keys: An optional dictionary of user ID -> master cross-signing + key info. + - self_signing_key: An optional dictionary of user ID -> self-signing + key info. + """ device_keys_query: Dict[str, Optional[List[str]]] = query_body.get( "device_keys", {} ) - res = await self.query_local_devices(device_keys_query) + res = await self.query_local_devices( + device_keys_query, + include_displaynames=( + self.config.federation.allow_device_name_lookup_over_federation + ), + ) ret = {"device_keys": res} # add in the cross-signing keys @@ -576,6 +600,8 @@ class E2eKeysHandler: async def upload_keys_for_user( self, user_id: str, device_id: str, keys: JsonDict ) -> JsonDict: + # This can only be called from the main process. + assert isinstance(self.device_handler, DeviceHandler) time_now = self.clock.time_msec() @@ -703,6 +729,8 @@ class E2eKeysHandler: user_id: the user uploading the keys keys: the signing keys """ + # This can only be called from the main process. + assert isinstance(self.device_handler, DeviceHandler) # if a master key is uploaded, then check it. Otherwise, load the # stored master key, to check signatures on other keys @@ -794,6 +822,9 @@ class E2eKeysHandler: Raises: SynapseError: if the signatures dict is not valid. """ + # This can only be called from the main process. + assert isinstance(self.device_handler, DeviceHandler) + failures = {} # signatures to be stored. Each item will be a SignatureListItem @@ -841,7 +872,7 @@ class E2eKeysHandler: - signatures of the user's master key by the user's devices. Args: - user_id (string): the user uploading the keys + user_id: the user uploading the keys signatures (dict[string, dict]): map of devices to signed keys Returns: @@ -1171,6 +1202,9 @@ class E2eKeysHandler: A tuple of the retrieved key content, the key's ID and the matching VerifyKey. If the key cannot be retrieved, all values in the tuple will instead be None. """ + # This can only be called from the main process. + assert isinstance(self.device_handler, DeviceHandler) + try: remote_result = await self.federation.query_user_devices( user.domain, user.to_string() @@ -1367,11 +1401,14 @@ class SignatureListItem: class SigningKeyEduUpdater: """Handles incoming signing key updates from federation and updates the DB""" - def __init__(self, hs: "HomeServer", e2e_keys_handler: E2eKeysHandler): + def __init__(self, hs: "HomeServer"): self.store = hs.get_datastores().main self.federation = hs.get_federation_client() self.clock = hs.get_clock() - self.e2e_keys_handler = e2e_keys_handler + + device_handler = hs.get_device_handler() + assert isinstance(device_handler, DeviceHandler) + self._device_handler = device_handler self._remote_edu_linearizer = Linearizer(name="remote_signing_key") @@ -1416,9 +1453,6 @@ class SigningKeyEduUpdater: user_id: the user whose updates we are processing """ - device_handler = self.e2e_keys_handler.device_handler - device_list_updater = device_handler.device_list_updater - async with self._remote_edu_linearizer.queue(user_id): pending_updates = self._pending_updates.pop(user_id, []) if not pending_updates: @@ -1430,13 +1464,11 @@ class SigningKeyEduUpdater: logger.info("pending updates: %r", pending_updates) for master_key, self_signing_key in pending_updates: - new_device_ids = ( - await device_list_updater.process_cross_signing_key_update( - user_id, - master_key, - self_signing_key, - ) + new_device_ids = await self._device_handler.device_list_updater.process_cross_signing_key_update( + user_id, + master_key, + self_signing_key, ) device_ids = device_ids + new_device_ids - await device_handler.notify_device_update(user_id, device_ids) + await self._device_handler.notify_device_update(user_id, device_ids) diff --git a/synapse/handlers/e2e_room_keys.py b/synapse/handlers/e2e_room_keys.py index 28dc08c22a..83f53ceb88 100644 --- a/synapse/handlers/e2e_room_keys.py +++ b/synapse/handlers/e2e_room_keys.py @@ -377,8 +377,9 @@ class E2eRoomKeysHandler: """Deletes a given version of the user's e2e_room_keys backup Args: - user_id(str): the user whose current backup version we're deleting - version(str): the version id of the backup being deleted + user_id: the user whose current backup version we're deleting + version: Optional. the version ID of the backup version we're deleting + If missing, we delete the current backup version info. Raises: NotFoundError: if this backup version doesn't exist """ diff --git a/synapse/handlers/event_auth.py b/synapse/handlers/event_auth.py index 8249ca1ed2..f91dbbecb7 100644 --- a/synapse/handlers/event_auth.py +++ b/synapse/handlers/event_auth.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. import logging -from typing import TYPE_CHECKING, Collection, List, Optional, Union +from typing import TYPE_CHECKING, Collection, List, Mapping, Optional, Union from synapse import event_auth from synapse.api.constants import ( @@ -29,7 +29,6 @@ from synapse.event_auth import ( ) from synapse.events import EventBase from synapse.events.builder import EventBuilder -from synapse.events.snapshot import EventContext from synapse.types import StateMap, get_domain_from_id if TYPE_CHECKING: @@ -46,17 +45,27 @@ class EventAuthHandler: def __init__(self, hs: "HomeServer"): self._clock = hs.get_clock() self._store = hs.get_datastores().main + self._state_storage_controller = hs.get_storage_controllers().state self._server_name = hs.hostname async def check_auth_rules_from_context( self, event: EventBase, - context: EventContext, + batched_auth_events: Optional[Mapping[str, EventBase]] = None, ) -> None: - """Check an event passes the auth rules at its own auth events""" - await check_state_independent_auth_rules(self._store, event) + """Check an event passes the auth rules at its own auth events + Args: + event: event to be authed + batched_auth_events: if the event being authed is part of a batch, any events + from the same batch that may be necessary to auth the current event + """ + await check_state_independent_auth_rules( + self._store, event, batched_auth_events + ) auth_event_ids = event.auth_event_ids() auth_events_by_id = await self._store.get_events(auth_event_ids) + if batched_auth_events: + auth_events_by_id.update(batched_auth_events) check_state_dependent_auth_rules(event, auth_events_by_id.values()) def compute_auth_events( @@ -171,17 +180,22 @@ class EventAuthHandler: this function may return an incorrect result as we are not able to fully track server membership in a room without full state. """ - if not allow_partial_state_rooms and await self._store.is_partial_state_room( - room_id - ): - raise AuthError( - 403, - "Unable to authorise you right now; room is partial-stated here.", - errcode=Codes.UNABLE_DUE_TO_PARTIAL_STATE, - ) - - if not await self.is_host_in_room(room_id, host): - raise AuthError(403, "Host not in room.") + if await self._store.is_partial_state_room(room_id): + if allow_partial_state_rooms: + current_hosts = await self._state_storage_controller.get_current_hosts_in_room_or_partial_state_approximation( + room_id + ) + if host not in current_hosts: + raise AuthError(403, "Host not in room (partial-state approx).") + else: + raise AuthError( + 403, + "Unable to authorise you right now; room is partial-stated here.", + errcode=Codes.UNABLE_DUE_TO_PARTIAL_STATE, + ) + else: + if not await self.is_host_in_room(room_id, host): + raise AuthError(403, "Host not in room.") async def check_restricted_join_rules( self, diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 986ffed3d5..b2784d7333 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -45,6 +45,7 @@ from synapse.api.errors import ( Codes, FederationDeniedError, FederationError, + FederationPullAttemptBackoffError, HttpResponseException, LimitExceededError, NotFoundError, @@ -69,8 +70,8 @@ from synapse.replication.http.federation import ( ) from synapse.storage.databases.main.events import PartialStateConflictError from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import JsonDict, get_domain_from_id +from synapse.types.state import StateFilter from synapse.util.async_helpers import Linearizer from synapse.util.retryutils import NotRetryingDestination from synapse.visibility import filter_events_for_server @@ -151,6 +152,7 @@ class FederationHandler: self._federation_event_handler = hs.get_federation_event_handler() self._device_handler = hs.get_device_handler() self._bulk_push_rule_evaluator = hs.get_bulk_push_rule_evaluator() + self._notifier = hs.get_notifier() self._clean_room_for_join_client = ReplicationCleanRoomRestServlet.make_client( hs @@ -378,6 +380,7 @@ class FederationHandler: filtered_extremities = await filter_events_for_server( self._storage_controllers, self.server_name, + self.server_name, events_to_check, redact=False, check_history_visibility_only=True, @@ -441,6 +444,15 @@ class FederationHandler: # appropriate stuff. # TODO: We can probably do something more intelligent here. return True + except NotRetryingDestination as e: + logger.info("_maybe_backfill_inner: %s", e) + continue + except FederationDeniedError: + logger.info( + "_maybe_backfill_inner: Not attempting to backfill from %s because the homeserver is not on our federation whitelist", + dom, + ) + continue except (SynapseError, InvalidResponseError) as e: logger.info("Failed to backfill from %s because %s", dom, e) continue @@ -476,15 +488,9 @@ class FederationHandler: logger.info("Failed to backfill from %s because %s", dom, e) continue - except NotRetryingDestination as e: - logger.info(str(e)) - continue except RequestSendFailed as e: logger.info("Failed to get backfill from %s because %s", dom, e) continue - except FederationDeniedError as e: - logger.info(e) - continue except Exception as e: logger.exception("Failed to backfill from %s because %s", dom, e) continue @@ -631,6 +637,7 @@ class FederationHandler: room_id=room_id, servers=ret.servers_in_room, device_lists_stream_id=self.store.get_device_stream_token(), + joined_via=origin, ) try: @@ -781,15 +788,27 @@ class FederationHandler: # Send the signed event back to the room, and potentially receive some # further information about the room in the form of partial state events - stripped_room_state = await self.federation_client.send_knock( - target_hosts, event - ) + knock_response = await self.federation_client.send_knock(target_hosts, event) # Store any stripped room state events in the "unsigned" key of the event. # This is a bit of a hack and is cribbing off of invites. Basically we # store the room state here and retrieve it again when this event appears # in the invitee's sync stream. It is stripped out for all other local users. - event.unsigned["knock_room_state"] = stripped_room_state["knock_state_events"] + stripped_room_state = ( + knock_response.get("knock_room_state") + # Since v1.37, Synapse incorrectly used "knock_state_events" for this field. + # Thus, we also check for a 'knock_state_events' to support old instances. + # See https://github.com/matrix-org/synapse/issues/14088. + or knock_response.get("knock_state_events") + ) + + if stripped_room_state is None: + raise KeyError( + "Missing 'knock_room_state' (or legacy 'knock_state_events') field in " + "send_knock response" + ) + + event.unsigned["knock_room_state"] = stripped_room_state context = EventContext.for_outlier(self._storage_controllers) stream_id = await self._federation_event_handler.persist_events_and_notify( @@ -928,7 +947,7 @@ class FederationHandler: # The remote hasn't signed it yet, obviously. We'll do the full checks # when we get the event back in `on_send_join_request` - await self._event_auth_handler.check_auth_rules_from_context(event, context) + await self._event_auth_handler.check_auth_rules_from_context(event) return event async def on_invite_request( @@ -1003,7 +1022,9 @@ class FederationHandler: context = EventContext.for_outlier(self._storage_controllers) - await self._bulk_push_rule_evaluator.action_for_event_by_user(event, context) + await self._bulk_push_rule_evaluator.action_for_events_by_user( + [(event, context)] + ) try: await self._federation_event_handler.persist_events_and_notify( event.room_id, [(event, context)] @@ -1109,7 +1130,7 @@ class FederationHandler: try: # The remote hasn't signed it yet, obviously. We'll do the full checks # when we get the event back in `on_send_leave_request` - await self._event_auth_handler.check_auth_rules_from_context(event, context) + await self._event_auth_handler.check_auth_rules_from_context(event) except AuthError as e: logger.warning("Failed to create new leave %r because %s", event, e) raise e @@ -1168,7 +1189,7 @@ class FederationHandler: try: # The remote hasn't signed it yet, obviously. We'll do the full checks # when we get the event back in `on_send_knock_request` - await self._event_auth_handler.check_auth_rules_from_context(event, context) + await self._event_auth_handler.check_auth_rules_from_context(event) except AuthError as e: logger.warning("Failed to create new knock %r because %s", event, e) raise e @@ -1212,7 +1233,9 @@ class FederationHandler: async def on_backfill_request( self, origin: str, room_id: str, pdu_list: List[str], limit: int ) -> List[EventBase]: - await self._event_auth_handler.assert_host_in_room(room_id, origin) + # We allow partially joined rooms since in this case we are filtering out + # non-local events in `filter_events_for_server`. + await self._event_auth_handler.assert_host_in_room(room_id, origin, True) # Synapse asks for 100 events per backfill request. Do not allow more. limit = min(limit, 100) @@ -1233,7 +1256,7 @@ class FederationHandler: ) events = await filter_events_for_server( - self._storage_controllers, origin, events + self._storage_controllers, origin, self.server_name, events ) return events @@ -1264,7 +1287,7 @@ class FederationHandler: await self._event_auth_handler.assert_host_in_room(event.room_id, origin) events = await filter_events_for_server( - self._storage_controllers, origin, [event] + self._storage_controllers, origin, self.server_name, [event] ) event = events[0] return event @@ -1277,7 +1300,9 @@ class FederationHandler: latest_events: List[str], limit: int, ) -> List[EventBase]: - await self._event_auth_handler.assert_host_in_room(room_id, origin) + # We allow partially joined rooms since in this case we are filtering out + # non-local events in `filter_events_for_server`. + await self._event_auth_handler.assert_host_in_room(room_id, origin, True) # Only allow up to 20 events to be retrieved per request. limit = min(limit, 20) @@ -1290,7 +1315,7 @@ class FederationHandler: ) missing_events = await filter_events_for_server( - self._storage_controllers, origin, missing_events + self._storage_controllers, origin, self.server_name, missing_events ) return missing_events @@ -1334,9 +1359,7 @@ class FederationHandler: try: validate_event_for_room_version(event) - await self._event_auth_handler.check_auth_rules_from_context( - event, context - ) + await self._event_auth_handler.check_auth_rules_from_context(event) except AuthError as e: logger.warning("Denying new third party invite %r because %s", event, e) raise e @@ -1386,7 +1409,7 @@ class FederationHandler: try: validate_event_for_room_version(event) - await self._event_auth_handler.check_auth_rules_from_context(event, context) + await self._event_auth_handler.check_auth_rules_from_context(event) except AuthError as e: logger.warning("Denying third party invite %r because %s", event, e) raise e @@ -1579,8 +1602,8 @@ class FederationHandler: Fetch the complexity of a remote room over federation. Args: - remote_room_hosts (list[str]): The remote servers to ask. - room_id (str): The room ID to ask about. + remote_room_hosts: The remote servers to ask. + room_id: The room ID to ask about. Returns: Dict contains the complexity @@ -1602,13 +1625,13 @@ class FederationHandler: """Resumes resyncing of all partial-state rooms after a restart.""" assert not self.config.worker.worker_app - partial_state_rooms = await self.store.get_partial_state_rooms_and_servers() - for room_id, servers_in_room in partial_state_rooms.items(): + partial_state_rooms = await self.store.get_partial_state_room_resync_info() + for room_id, resync_info in partial_state_rooms.items(): run_as_background_process( desc="sync_partial_state_room", func=self._sync_partial_state_room, - initial_destination=None, - other_destinations=servers_in_room, + initial_destination=resync_info.joined_via, + other_destinations=resync_info.servers_in_room, room_id=room_id, ) @@ -1637,28 +1660,12 @@ class FederationHandler: # really leave, that might mean we have difficulty getting the room state over # federation. # https://github.com/matrix-org/synapse/issues/12802 - # - # TODO(faster_joins): we need some way of prioritising which homeservers in - # `other_destinations` to try first, otherwise we'll spend ages trying dead - # homeservers for large rooms. - # https://github.com/matrix-org/synapse/issues/12999 - - if initial_destination is None and len(other_destinations) == 0: - raise ValueError( - f"Cannot resync state of {room_id}: no destinations provided" - ) # Make an infinite iterator of destinations to try. Once we find a working # destination, we'll stick with it until it flakes. - destinations: Collection[str] - if initial_destination is not None: - # Move `initial_destination` to the front of the list. - destinations = list(other_destinations) - if initial_destination in destinations: - destinations.remove(initial_destination) - destinations = [initial_destination] + destinations - else: - destinations = other_destinations + destinations = _prioritise_destinations_for_partial_state_resync( + initial_destination, other_destinations, room_id + ) destination_iter = itertools.cycle(destinations) # `destination` is the current remote homeserver we're pulling from. @@ -1686,6 +1693,9 @@ class FederationHandler: self._storage_controllers.state.notify_room_un_partial_stated( room_id ) + # Poke the notifier so that other workers see the write to + # the un-partial-stated rooms stream. + self._notifier.notify_replication() # TODO(faster_joins) update room stats and user directory? # https://github.com/matrix-org/synapse/issues/12814 @@ -1708,7 +1718,22 @@ class FederationHandler: destination, event ) break + except FederationPullAttemptBackoffError as exc: + # Log a warning about why we failed to process the event (the error message + # for `FederationPullAttemptBackoffError` is pretty good) + logger.warning("_sync_partial_state_room: %s", exc) + # We do not record a failed pull attempt when we backoff fetching a missing + # `prev_event` because not being able to fetch the `prev_events` just means + # we won't be able to de-outlier the pulled event. But we can still use an + # `outlier` in the state/auth chain for another event. So we shouldn't stop + # a downstream event from trying to pull it. + # + # This avoids a cascade of backoff for all events in the DAG downstream from + # one event backoff upstream. except FederationError as e: + # TODO: We should `record_event_failed_pull_attempt` here, + # see https://github.com/matrix-org/synapse/issues/13700 + if attempt == len(destinations) - 1: # We have tried every remote server for this event. Give up. # TODO(faster_joins) giving up isn't the right thing to do @@ -1741,3 +1766,29 @@ class FederationHandler: room_id, destination, ) + + +def _prioritise_destinations_for_partial_state_resync( + initial_destination: Optional[str], + other_destinations: Collection[str], + room_id: str, +) -> Collection[str]: + """Work out the order in which we should ask servers to resync events. + + If an `initial_destination` is given, it takes top priority. Otherwise + all servers are treated equally. + + :raises ValueError: if no destination is provided at all. + """ + if initial_destination is None and len(other_destinations) == 0: + raise ValueError(f"Cannot resync state of {room_id}: no destinations provided") + + if initial_destination is None: + return other_destinations + + # Move `initial_destination` to the front of the list. + destinations = list(other_destinations) + if initial_destination in destinations: + destinations.remove(initial_destination) + destinations = [initial_destination] + destinations + return destinations diff --git a/synapse/handlers/federation_event.py b/synapse/handlers/federation_event.py index da319943cc..66aca2f864 100644 --- a/synapse/handlers/federation_event.py +++ b/synapse/handlers/federation_event.py @@ -43,7 +43,9 @@ from synapse.api.constants import ( from synapse.api.errors import ( AuthError, Codes, + EventSizeError, FederationError, + FederationPullAttemptBackoffError, HttpResponseException, RequestSendFailed, SynapseError, @@ -57,7 +59,7 @@ from synapse.event_auth import ( ) from synapse.events import EventBase from synapse.events.snapshot import EventContext -from synapse.federation.federation_client import InvalidResponseError +from synapse.federation.federation_client import InvalidResponseError, PulledPduInfo from synapse.logging.context import nested_logging_context from synapse.logging.opentracing import ( SynapseTags, @@ -74,7 +76,6 @@ from synapse.replication.http.federation import ( from synapse.state import StateResolutionStore from synapse.storage.databases.main.events import PartialStateConflictError from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import ( PersistedEventPosition, RoomStreamToken, @@ -82,6 +83,7 @@ from synapse.types import ( UserID, get_domain_from_id, ) +from synapse.types.state import StateFilter from synapse.util.async_helpers import Linearizer, concurrently_execute from synapse.util.iterutils import batch_iter from synapse.util.retryutils import NotRetryingDestination @@ -414,7 +416,9 @@ class FederationEventHandler: # First, precalculate the joined hosts so that the federation sender doesn't # need to. - await self._event_creation_handler.cache_joined_hosts_for_event(event, context) + await self._event_creation_handler.cache_joined_hosts_for_events( + [(event, context)] + ) await self._check_for_soft_fail(event, context=context, origin=origin) await self._run_push_actions_and_persist_event(event, context) @@ -565,6 +569,9 @@ class FederationEventHandler: event: partial-state event to be de-partial-stated Raises: + FederationPullAttemptBackoffError if we are are deliberately not attempting + to pull the given event over federation because we've already done so + recently and are backing off. FederationError if we fail to request state from the remote server. """ logger.info("Updating state for %s", event.event_id) @@ -792,9 +799,42 @@ class FederationEventHandler: ], ) + # Check if we already any of these have these events. + # Note: we currently make a lookup in the database directly here rather than + # checking the event cache, due to: + # https://github.com/matrix-org/synapse/issues/13476 + existing_events_map = await self._store._get_events_from_db( + [event.event_id for event in events] + ) + + new_events = [] + for event in events: + event_id = event.event_id + + # If we've already seen this event ID... + if event_id in existing_events_map: + existing_event = existing_events_map[event_id] + + # ...and the event itself was not previously stored as an outlier... + if not existing_event.event.internal_metadata.is_outlier(): + # ...then there's no need to persist it. We have it already. + logger.info( + "_process_pulled_event: Ignoring received event %s which we " + "have already seen", + event.event_id, + ) + continue + + # While we have seen this event before, it was stored as an outlier. + # We'll now persist it as a non-outlier. + logger.info("De-outliering event %s", event_id) + + # Continue on with the events that are new to us. + new_events.append(event) + # We want to sort these by depth so we process them and # tell clients about them in order. - sorted_events = sorted(events, key=lambda x: x.depth) + sorted_events = sorted(new_events, key=lambda x: x.depth) for ev in sorted_events: with nested_logging_context(ev.event_id): await self._process_pulled_event(origin, ev, backfilled=backfilled) @@ -846,18 +886,6 @@ class FederationEventHandler: event_id = event.event_id - existing = await self._store.get_event( - event_id, allow_none=True, allow_rejected=True - ) - if existing: - if not existing.internal_metadata.is_outlier(): - logger.info( - "_process_pulled_event: Ignoring received event %s which we have already seen", - event_id, - ) - return - logger.info("De-outliering event %s", event_id) - try: self._sanity_check_event(event) except SynapseError as err: @@ -899,6 +927,18 @@ class FederationEventHandler: context, backfilled=backfilled, ) + except FederationPullAttemptBackoffError as exc: + # Log a warning about why we failed to process the event (the error message + # for `FederationPullAttemptBackoffError` is pretty good) + logger.warning("_process_pulled_event: %s", exc) + # We do not record a failed pull attempt when we backoff fetching a missing + # `prev_event` because not being able to fetch the `prev_events` just means + # we won't be able to de-outlier the pulled event. But we can still use an + # `outlier` in the state/auth chain for another event. So we shouldn't stop + # a downstream event from trying to pull it. + # + # This avoids a cascade of backoff for all events in the DAG downstream from + # one event backoff upstream. except FederationError as e: await self._store.record_event_failed_pull_attempt( event.room_id, event_id, str(e) @@ -945,6 +985,9 @@ class FederationEventHandler: The event context. Raises: + FederationPullAttemptBackoffError if we are are deliberately not attempting + to pull the given event over federation because we've already done so + recently and are backing off. FederationError if we fail to get the state from the remote server after any missing `prev_event`s. """ @@ -955,6 +998,18 @@ class FederationEventHandler: seen = await self._store.have_events_in_timeline(prevs) missing_prevs = prevs - seen + # If we've already recently attempted to pull this missing event, don't + # try it again so soon. Since we have to fetch all of the prev_events, we can + # bail early here if we find any to ignore. + prevs_to_ignore = await self._store.get_event_ids_to_not_pull_from_backoff( + room_id, missing_prevs + ) + if len(prevs_to_ignore) > 0: + raise FederationPullAttemptBackoffError( + event_ids=prevs_to_ignore, + message=f"While computing context for event={event_id}, not attempting to pull missing prev_event={prevs_to_ignore[0]} because we already tried to pull recently (backing off).", + ) + if not missing_prevs: return await self._state_handler.compute_event_context(event) @@ -1011,10 +1066,9 @@ class FederationEventHandler: state_res_store=StateResolutionStore(self._store), ) - except Exception: + except Exception as e: logger.warning( - "Error attempting to resolve state at missing prev_events", - exc_info=True, + "Error attempting to resolve state at missing prev_events: %s", e ) raise FederationError( "ERROR", @@ -1463,8 +1517,8 @@ class FederationEventHandler: ) async def backfill_event_id( - self, destination: str, room_id: str, event_id: str - ) -> EventBase: + self, destinations: List[str], room_id: str, event_id: str + ) -> PulledPduInfo: """Backfill a single event and persist it as a non-outlier which means we also pull in all of the state and auth events necessary for it. @@ -1476,24 +1530,21 @@ class FederationEventHandler: Raises: FederationError if we are unable to find the event from the destination """ - logger.info( - "backfill_event_id: event_id=%s from destination=%s", event_id, destination - ) + logger.info("backfill_event_id: event_id=%s", event_id) room_version = await self._store.get_room_version(room_id) - event_from_response = await self._federation_client.get_pdu( - [destination], + pulled_pdu_info = await self._federation_client.get_pdu( + destinations, event_id, room_version, ) - if not event_from_response: + if not pulled_pdu_info: raise FederationError( "ERROR", 404, - "Unable to find event_id=%s from destination=%s to backfill." - % (event_id, destination), + f"Unable to find event_id={event_id} from remote servers to backfill.", affected=event_id, ) @@ -1501,13 +1552,13 @@ class FederationEventHandler: # and auth events to de-outlier it. This also sets up the necessary # `state_groups` for the event. await self._process_pulled_events( - destination, - [event_from_response], + pulled_pdu_info.pull_origin, + [pulled_pdu_info.pdu], # Prevent notifications going to clients backfilled=True, ) - return event_from_response + return pulled_pdu_info @trace @tag_args @@ -1530,19 +1581,19 @@ class FederationEventHandler: async def get_event(event_id: str) -> None: with nested_logging_context(event_id): try: - event = await self._federation_client.get_pdu( + pulled_pdu_info = await self._federation_client.get_pdu( [destination], event_id, room_version, ) - if event is None: + if pulled_pdu_info is None: logger.warning( "Server %s didn't return event %s", destination, event_id, ) return - events.append(event) + events.append(pulled_pdu_info.pdu) except Exception as e: logger.warning( @@ -1686,6 +1737,15 @@ class FederationEventHandler: except AuthError as e: logger.warning("Rejecting %r because %s", event, e) context.rejected = RejectedReason.AUTH_ERROR + except EventSizeError as e: + if e.unpersistable: + # This event is completely unpersistable. + raise e + # Otherwise, we are somewhat lenient and just persist the event + # as rejected, for moderate compatibility with older Synapse + # versions. + logger.warning("While validating received event %r: %s", event, e) + context.rejected = RejectedReason.OVERSIZED_EVENT events_and_contexts_to_persist.append((event, context)) @@ -1731,6 +1791,16 @@ class FederationEventHandler: # TODO: use a different rejected reason here? context.rejected = RejectedReason.AUTH_ERROR return + except EventSizeError as e: + if e.unpersistable: + # This event is completely unpersistable. + raise e + # Otherwise, we are somewhat lenient and just persist the event + # as rejected, for moderate compatibility with older Synapse + # versions. + logger.warning("While validating received event %r: %s", event, e) + context.rejected = RejectedReason.OVERSIZED_EVENT + return # next, check that we have all of the event's auth events. # @@ -2117,8 +2187,8 @@ class FederationEventHandler: min_depth, ) else: - await self._bulk_push_rule_evaluator.action_for_event_by_user( - event, context + await self._bulk_push_rule_evaluator.action_for_events_by_user( + [(event, context)] ) try: diff --git a/synapse/handlers/identity.py b/synapse/handlers/identity.py index 93d09e9939..848e46eb9b 100644 --- a/synapse/handlers/identity.py +++ b/synapse/handlers/identity.py @@ -711,7 +711,7 @@ class IdentityHandler: inviter_display_name: The current display name of the inviter. inviter_avatar_url: The URL of the inviter's avatar. - id_access_token (str): The access token to authenticate to the identity + id_access_token: The access token to authenticate to the identity server with Returns: diff --git a/synapse/handlers/initial_sync.py b/synapse/handlers/initial_sync.py index 860c82c110..9c335e6863 100644 --- a/synapse/handlers/initial_sync.py +++ b/synapse/handlers/initial_sync.py @@ -57,13 +57,7 @@ class InitialSyncHandler: self.validator = EventValidator() self.snapshot_cache: ResponseCache[ Tuple[ - str, - Optional[StreamToken], - Optional[StreamToken], - str, - Optional[int], - bool, - bool, + str, Optional[StreamToken], Optional[StreamToken], str, int, bool, bool ] ] = ResponseCache(hs.get_clock(), "initial_sync_cache") self._event_serializer = hs.get_event_client_serializer() @@ -154,11 +148,6 @@ class InitialSyncHandler: public_room_ids = await self.store.get_public_room_ids() - if pagin_config.limit is not None: - limit = pagin_config.limit - else: - limit = 10 - serializer_options = SerializeEventConfig(as_client_event=as_client_event) async def handle_room(event: RoomsForUser) -> None: @@ -210,7 +199,7 @@ class InitialSyncHandler: run_in_background( self.store.get_recent_events_for_room, event.room_id, - limit=limit, + limit=pagin_config.limit, end_token=room_end_token, ), deferred_room_state, @@ -360,15 +349,11 @@ class InitialSyncHandler: member_event_id ) - limit = pagin_config.limit if pagin_config else None - if limit is None: - limit = 10 - leave_position = await self.store.get_position_for_event(member_event_id) stream_token = leave_position.to_room_stream_token() messages, token = await self.store.get_recent_events_for_room( - room_id, limit=limit, end_token=stream_token + room_id, limit=pagin_config.limit, end_token=stream_token ) messages = await filter_events_for_client( @@ -420,10 +405,6 @@ class InitialSyncHandler: now_token = self.hs.get_event_sources().get_current_token() - limit = pagin_config.limit if pagin_config else None - if limit is None: - limit = 10 - room_members = [ m for m in current_state.values() @@ -467,7 +448,7 @@ class InitialSyncHandler: run_in_background( self.store.get_recent_events_for_room, room_id, - limit=limit, + limit=pagin_config.limit, end_token=now_token.room_key, ), ), diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index da1acea275..845f683358 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -50,6 +50,7 @@ from synapse.event_auth import validate_event_for_room_version from synapse.events import EventBase, relation_from_event from synapse.events.builder import EventBuilder from synapse.events.snapshot import EventContext +from synapse.events.utils import maybe_upsert_event_field from synapse.events.validator import EventValidator from synapse.handlers.directory import DirectoryHandler from synapse.logging import opentracing @@ -59,7 +60,6 @@ from synapse.replication.http.send_event import ReplicationSendEventRestServlet from synapse.replication.http.send_events import ReplicationSendEventsRestServlet from synapse.storage.databases.main.events import PartialStateConflictError from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import ( MutableStateMap, PersistedEventPosition, @@ -70,6 +70,7 @@ from synapse.types import ( UserID, create_requester, ) +from synapse.types.state import StateFilter from synapse.util import json_decoder, json_encoder, log_failure, unwrapFirstError from synapse.util.async_helpers import Linearizer, gather_results from synapse.util.caches.expiringcache import ExpiringCache @@ -877,6 +878,36 @@ class EventCreationHandler: return prev_event return None + async def get_event_from_transaction( + self, + requester: Requester, + txn_id: str, + room_id: str, + ) -> Optional[EventBase]: + """For the given transaction ID and room ID, check if there is a matching event. + If so, fetch it and return it. + + Args: + requester: The requester making the request in the context of which we want + to fetch the event. + txn_id: The transaction ID. + room_id: The room ID. + + Returns: + An event if one could be found, None otherwise. + """ + if requester.access_token_id: + existing_event_id = await self.store.get_event_id_from_transaction_id( + room_id, + requester.user.to_string(), + requester.access_token_id, + txn_id, + ) + if existing_event_id: + return await self.store.get_event(existing_event_id) + + return None + async def create_and_send_nonmember_event( self, requester: Requester, @@ -956,18 +987,17 @@ class EventCreationHandler: # extremities to pile up, which in turn leads to state resolution # taking longer. async with self.limiter.queue(event_dict["room_id"]): - if txn_id and requester.access_token_id: - existing_event_id = await self.store.get_event_id_from_transaction_id( - event_dict["room_id"], - requester.user.to_string(), - requester.access_token_id, - txn_id, + if txn_id: + event = await self.get_event_from_transaction( + requester, txn_id, event_dict["room_id"] ) - if existing_event_id: - event = await self.store.get_event(existing_event_id) + if event: # we know it was persisted, so must have a stream ordering assert event.internal_metadata.stream_ordering - return event, event.internal_metadata.stream_ordering + return ( + event, + event.internal_metadata.stream_ordering, + ) event, context = await self.create_event( requester, @@ -1106,11 +1136,13 @@ class EventCreationHandler: ) state_events = await self.store.get_events_as_list(state_event_ids) # Create a StateMap[str] - state_map = {(e.type, e.state_key): e.event_id for e in state_events} + current_state_ids = { + (e.type, e.state_key): e.event_id for e in state_events + } # Actually strip down and only use the necessary auth events auth_event_ids = self._event_auth_handler.compute_auth_events( event=temp_event, - current_state_ids=state_map, + current_state_ids=current_state_ids, for_verification=False, ) @@ -1360,8 +1392,16 @@ class EventCreationHandler: else: try: validate_event_for_room_version(event) + # If we are persisting a batch of events the event(s) needed to auth the + # current event may be part of the batch and will not be in the DB yet + event_id_to_event = {e.event_id: e for e, _ in events_and_context} + batched_auth_events = {} + for event_id in event.auth_event_ids(): + auth_event = event_id_to_event.get(event_id) + if auth_event: + batched_auth_events[event_id] = auth_event await self._event_auth_handler.check_auth_rules_from_context( - event, context + event, batched_auth_events ) except AuthError as err: logger.warning("Denying new event %r because %s", event, err) @@ -1390,7 +1430,7 @@ class EventCreationHandler: extra_users=extra_users, ), run_in_background( - self.cache_joined_hosts_for_event, event, context + self.cache_joined_hosts_for_events, events_and_context ).addErrback( log_failure, "cache_joined_hosts_for_event failed" ), @@ -1425,17 +1465,9 @@ class EventCreationHandler: a room that has been un-partial stated. """ - for event, context in events_and_context: - # Skip push notification actions for historical messages - # because we don't want to notify people about old history back in time. - # The historical messages also do not have the proper `context.current_state_ids` - # and `state_groups` because they have `prev_events` that aren't persisted yet - # (historical messages persisted in reverse-chronological order). - if not event.internal_metadata.is_historical(): - with opentracing.start_active_span("calculate_push_actions"): - await self._bulk_push_rule_evaluator.action_for_event_by_user( - event, context - ) + await self._bulk_push_rule_evaluator.action_for_events_by_user( + events_and_context + ) try: # If we're a worker we need to hit out to the master. @@ -1491,62 +1523,65 @@ class EventCreationHandler: await self.store.remove_push_actions_from_staging(event.event_id) raise - async def cache_joined_hosts_for_event( - self, event: EventBase, context: EventContext + async def cache_joined_hosts_for_events( + self, events_and_context: List[Tuple[EventBase, EventContext]] ) -> None: - """Precalculate the joined hosts at the event, when using Redis, so that + """Precalculate the joined hosts at each of the given events, when using Redis, so that external federation senders don't have to recalculate it themselves. """ - if not self._external_cache.is_enabled(): - return - - # If external cache is enabled we should always have this. - assert self._external_cache_joined_hosts_updates is not None + for event, _ in events_and_context: + if not self._external_cache.is_enabled(): + return - # We actually store two mappings, event ID -> prev state group, - # state group -> joined hosts, which is much more space efficient - # than event ID -> joined hosts. - # - # Note: We have to cache event ID -> prev state group, as we don't - # store that in the DB. - # - # Note: We set the state group -> joined hosts cache if it hasn't been - # set for a while, so that the expiry time is reset. + # If external cache is enabled we should always have this. + assert self._external_cache_joined_hosts_updates is not None - state_entry = await self.state.resolve_state_groups_for_events( - event.room_id, event_ids=event.prev_event_ids() - ) + # We actually store two mappings, event ID -> prev state group, + # state group -> joined hosts, which is much more space efficient + # than event ID -> joined hosts. + # + # Note: We have to cache event ID -> prev state group, as we don't + # store that in the DB. + # + # Note: We set the state group -> joined hosts cache if it hasn't been + # set for a while, so that the expiry time is reset. - if state_entry.state_group: - await self._external_cache.set( - "event_to_prev_state_group", - event.event_id, - state_entry.state_group, - expiry_ms=60 * 60 * 1000, + state_entry = await self.state.resolve_state_groups_for_events( + event.room_id, event_ids=event.prev_event_ids() ) - if state_entry.state_group in self._external_cache_joined_hosts_updates: - return + if state_entry.state_group: + await self._external_cache.set( + "event_to_prev_state_group", + event.event_id, + state_entry.state_group, + expiry_ms=60 * 60 * 1000, + ) - state = await state_entry.get_state( - self._storage_controllers.state, StateFilter.all() - ) - with opentracing.start_active_span("get_joined_hosts"): - joined_hosts = await self.store.get_joined_hosts( - event.room_id, state, state_entry + if state_entry.state_group in self._external_cache_joined_hosts_updates: + return + + state = await state_entry.get_state( + self._storage_controllers.state, StateFilter.all() ) + with opentracing.start_active_span("get_joined_hosts"): + joined_hosts = await self.store.get_joined_hosts( + event.room_id, state, state_entry + ) - # Note that the expiry times must be larger than the expiry time in - # _external_cache_joined_hosts_updates. - await self._external_cache.set( - "get_joined_hosts", - str(state_entry.state_group), - list(joined_hosts), - expiry_ms=60 * 60 * 1000, - ) + # Note that the expiry times must be larger than the expiry time in + # _external_cache_joined_hosts_updates. + await self._external_cache.set( + "get_joined_hosts", + str(state_entry.state_group), + list(joined_hosts), + expiry_ms=60 * 60 * 1000, + ) - self._external_cache_joined_hosts_updates[state_entry.state_group] = None + self._external_cache_joined_hosts_updates[ + state_entry.state_group + ] = None async def _validate_canonical_alias( self, @@ -1705,12 +1740,15 @@ class EventCreationHandler: if event.type == EventTypes.Member: if event.content["membership"] == Membership.INVITE: - event.unsigned[ - "invite_room_state" - ] = await self.store.get_stripped_room_state_from_event_context( - context, - self.room_prejoin_state_types, - membership_user_id=event.sender, + maybe_upsert_event_field( + event, + event.unsigned, + "invite_room_state", + await self.store.get_stripped_room_state_from_event_context( + context, + self.room_prejoin_state_types, + membership_user_id=event.sender, + ), ) invitee = UserID.from_string(event.state_key) @@ -1728,11 +1766,14 @@ class EventCreationHandler: event.signatures.update(returned_invite.signatures) if event.content["membership"] == Membership.KNOCK: - event.unsigned[ - "knock_room_state" - ] = await self.store.get_stripped_room_state_from_event_context( - context, - self.room_prejoin_state_types, + maybe_upsert_event_field( + event, + event.unsigned, + "knock_room_state", + await self.store.get_stripped_room_state_from_event_context( + context, + self.room_prejoin_state_types, + ), ) if event.type == EventTypes.Redaction: diff --git a/synapse/handlers/oidc.py b/synapse/handlers/oidc.py index d7a8226900..03de6a4ba6 100644 --- a/synapse/handlers/oidc.py +++ b/synapse/handlers/oidc.py @@ -12,14 +12,28 @@ # 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 binascii import inspect +import json import logging -from typing import TYPE_CHECKING, Any, Dict, Generic, List, Optional, TypeVar, Union +from typing import ( + TYPE_CHECKING, + Any, + Dict, + Generic, + List, + Optional, + Type, + TypeVar, + Union, +) from urllib.parse import urlencode, urlparse import attr +import unpaddedbase64 from authlib.common.security import generate_token -from authlib.jose import JsonWebToken, jwt +from authlib.jose import JsonWebToken, JWTClaims +from authlib.jose.errors import InvalidClaimError, JoseError, MissingClaimError from authlib.oauth2.auth import ClientAuth from authlib.oauth2.rfc6749.parameters import prepare_grant_uri from authlib.oidc.core import CodeIDToken, UserInfo @@ -35,9 +49,12 @@ from typing_extensions import TypedDict from twisted.web.client import readBody from twisted.web.http_headers import Headers +from synapse.api.errors import SynapseError from synapse.config import ConfigError from synapse.config.oidc import OidcProviderClientSecretJwtKey, OidcProviderConfig from synapse.handlers.sso import MappingException, UserAttributes +from synapse.http.server import finish_request +from synapse.http.servlet import parse_string from synapse.http.site import SynapseRequest from synapse.logging.context import make_deferred_yieldable from synapse.types import JsonDict, UserID, map_username_to_mxid_localpart @@ -88,6 +105,8 @@ class Token(TypedDict): #: there is no real point of doing this in our case. JWK = Dict[str, str] +C = TypeVar("C") + #: A JWK Set, as per RFC7517 sec 5. class JWKS(TypedDict): @@ -247,6 +266,80 @@ class OidcHandler: await oidc_provider.handle_oidc_callback(request, session_data, code) + async def handle_backchannel_logout(self, request: SynapseRequest) -> None: + """Handle an incoming request to /_synapse/client/oidc/backchannel_logout + + This extracts the logout_token from the request and tries to figure out + which OpenID Provider it is comming from. This works by matching the iss claim + with the issuer and the aud claim with the client_id. + + Since at this point we don't know who signed the JWT, we can't just + decode it using authlib since it will always verifies the signature. We + have to decode it manually without validating the signature. The actual JWT + verification is done in the `OidcProvider.handler_backchannel_logout` method, + once we figured out which provider sent the request. + + Args: + request: the incoming request from the browser. + """ + logout_token = parse_string(request, "logout_token") + if logout_token is None: + raise SynapseError(400, "Missing logout_token in request") + + # A JWT looks like this: + # header.payload.signature + # where all parts are encoded with urlsafe base64. + # The aud and iss claims we care about are in the payload part, which + # is a JSON object. + try: + # By destructuring the list after splitting, we ensure that we have + # exactly 3 segments + _, payload, _ = logout_token.split(".") + except ValueError: + raise SynapseError(400, "Invalid logout_token in request") + + try: + payload_bytes = unpaddedbase64.decode_base64(payload) + claims = json_decoder.decode(payload_bytes.decode("utf-8")) + except (json.JSONDecodeError, binascii.Error, UnicodeError): + raise SynapseError(400, "Invalid logout_token payload in request") + + try: + # Let's extract the iss and aud claims + iss = claims["iss"] + aud = claims["aud"] + # The aud claim can be either a string or a list of string. Here we + # normalize it as a list of strings. + if isinstance(aud, str): + aud = [aud] + + # Check that we have the right types for the aud and the iss claims + if not isinstance(iss, str) or not isinstance(aud, list): + raise TypeError() + for a in aud: + if not isinstance(a, str): + raise TypeError() + + # At this point we properly checked both claims types + issuer: str = iss + audience: List[str] = aud + except (TypeError, KeyError): + raise SynapseError(400, "Invalid issuer/audience in logout_token") + + # Now that we know the audience and the issuer, we can figure out from + # what provider it is coming from + oidc_provider: Optional[OidcProvider] = None + for provider in self._providers.values(): + if provider.issuer == issuer and provider.client_id in audience: + oidc_provider = provider + break + + if oidc_provider is None: + raise SynapseError(400, "Could not find the OP that issued this event") + + # Ask the provider to handle the logout request. + await oidc_provider.handle_backchannel_logout(request, logout_token) + class OidcError(Exception): """Used to catch errors when calling the token_endpoint""" @@ -275,6 +368,7 @@ class OidcProvider: provider: OidcProviderConfig, ): self._store = hs.get_datastores().main + self._clock = hs.get_clock() self._macaroon_generaton = macaroon_generator @@ -341,6 +435,7 @@ class OidcProvider: self.idp_brand = provider.idp_brand self._sso_handler = hs.get_sso_handler() + self._device_handler = hs.get_device_handler() self._sso_handler.register_identity_provider(self) @@ -399,6 +494,41 @@ class OidcProvider: # If we're not using userinfo, we need a valid jwks to validate the ID token m.validate_jwks_uri() + if self._config.backchannel_logout_enabled: + if not m.get("backchannel_logout_supported", False): + logger.warning( + "OIDC Back-Channel Logout is enabled for issuer %r" + "but it does not advertise support for it", + self.issuer, + ) + + elif not m.get("backchannel_logout_session_supported", False): + logger.warning( + "OIDC Back-Channel Logout is enabled and supported " + "by issuer %r but it might not send a session ID with " + "logout tokens, which is required for the logouts to work", + self.issuer, + ) + + if not self._config.backchannel_logout_ignore_sub: + # If OIDC backchannel logouts are enabled, the provider mapping provider + # should use the `sub` claim. We verify that by mapping a dumb user and + # see if we get back the sub claim + user = UserInfo({"sub": "thisisasubject"}) + try: + subject = self._user_mapping_provider.get_remote_user_id(user) + if subject != user["sub"]: + raise ValueError("Unexpected subject") + except Exception: + logger.warning( + f"OIDC Back-Channel Logout is enabled for issuer {self.issuer!r} " + "but it looks like the configured `user_mapping_provider` " + "does not use the `sub` claim as subject. If it is the case, " + "and you want Synapse to ignore the `sub` claim in OIDC " + "Back-Channel Logouts, set `backchannel_logout_ignore_sub` " + "to `true` in the issuer config." + ) + @property def _uses_userinfo(self) -> bool: """Returns True if the ``userinfo_endpoint`` should be used. @@ -414,6 +544,16 @@ class OidcProvider: or self._user_profile_method == "userinfo_endpoint" ) + @property + def issuer(self) -> str: + """The issuer identifying this provider.""" + return self._config.issuer + + @property + def client_id(self) -> str: + """The client_id used when interacting with this provider.""" + return self._config.client_id + async def load_metadata(self, force: bool = False) -> OpenIDProviderMetadata: """Return the provider metadata. @@ -647,7 +787,7 @@ class OidcProvider: Must include an ``access_token`` field. Returns: - UserInfo: an object representing the user. + an object representing the user. """ logger.debug("Using the OAuth2 access_token to request userinfo") metadata = await self.load_metadata() @@ -661,61 +801,99 @@ class OidcProvider: return UserInfo(resp) - async def _parse_id_token(self, token: Token, nonce: str) -> CodeIDToken: - """Return an instance of UserInfo from token's ``id_token``. + async def _verify_jwt( + self, + alg_values: List[str], + token: str, + claims_cls: Type[C], + claims_options: Optional[dict] = None, + claims_params: Optional[dict] = None, + ) -> C: + """Decode and validate a JWT, re-fetching the JWKS as needed. Args: - token: the token given by the ``token_endpoint``. - Must include an ``id_token`` field. - nonce: the nonce value originally sent in the initial authorization - request. This value should match the one inside the token. + alg_values: list of `alg` values allowed when verifying the JWT. + token: the JWT. + claims_cls: the JWTClaims class to use to validate the claims. + claims_options: dict of options passed to the `claims_cls` constructor. + claims_params: dict of params passed to the `claims_cls` constructor. Returns: - The decoded claims in the ID token. + The decoded claims in the JWT. """ - metadata = await self.load_metadata() - claims_params = { - "nonce": nonce, - "client_id": self._client_auth.client_id, - } - if "access_token" in token: - # If we got an `access_token`, there should be an `at_hash` claim - # in the `id_token` that we can check against. - claims_params["access_token"] = token["access_token"] - - alg_values = metadata.get("id_token_signing_alg_values_supported", ["RS256"]) jwt = JsonWebToken(alg_values) - claim_options = {"iss": {"values": [metadata["issuer"]]}} - - id_token = token["id_token"] - logger.debug("Attempting to decode JWT id_token %r", id_token) + logger.debug("Attempting to decode JWT (%s) %r", claims_cls.__name__, token) # Try to decode the keys in cache first, then retry by forcing the keys # to be reloaded jwk_set = await self.load_jwks() try: claims = jwt.decode( - id_token, + token, key=jwk_set, - claims_cls=CodeIDToken, - claims_options=claim_options, + claims_cls=claims_cls, + claims_options=claims_options, claims_params=claims_params, ) except ValueError: logger.info("Reloading JWKS after decode error") jwk_set = await self.load_jwks(force=True) # try reloading the jwks claims = jwt.decode( - id_token, + token, key=jwk_set, - claims_cls=CodeIDToken, - claims_options=claim_options, + claims_cls=claims_cls, + claims_options=claims_options, claims_params=claims_params, ) - logger.debug("Decoded id_token JWT %r; validating", claims) + logger.debug("Decoded JWT (%s) %r; validating", claims_cls.__name__, claims) - claims.validate(leeway=120) # allows 2 min of clock skew + claims.validate( + now=self._clock.time(), leeway=120 + ) # allows 2 min of clock skew + return claims + + async def _parse_id_token(self, token: Token, nonce: str) -> CodeIDToken: + """Return an instance of UserInfo from token's ``id_token``. + + Args: + token: the token given by the ``token_endpoint``. + Must include an ``id_token`` field. + nonce: the nonce value originally sent in the initial authorization + request. This value should match the one inside the token. + + Returns: + The decoded claims in the ID token. + """ + id_token = token.get("id_token") + + # That has been theoritically been checked by the caller, so even though + # assertion are not enabled in production, it is mainly here to appease mypy + assert id_token is not None + + metadata = await self.load_metadata() + + claims_params = { + "nonce": nonce, + "client_id": self._client_auth.client_id, + } + if "access_token" in token: + # If we got an `access_token`, there should be an `at_hash` claim + # in the `id_token` that we can check against. + claims_params["access_token"] = token["access_token"] + + claims_options = {"iss": {"values": [metadata["issuer"]]}} + + alg_values = metadata.get("id_token_signing_alg_values_supported", ["RS256"]) + + claims = await self._verify_jwt( + alg_values=alg_values, + token=id_token, + claims_cls=CodeIDToken, + claims_options=claims_options, + claims_params=claims_params, + ) return claims @@ -1036,6 +1214,146 @@ class OidcProvider: # to be strings. return str(remote_user_id) + async def handle_backchannel_logout( + self, request: SynapseRequest, logout_token: str + ) -> None: + """Handle an incoming request to /_synapse/client/oidc/backchannel_logout + + The OIDC Provider posts a logout token to this endpoint when a user + session ends. That token is a JWT signed with the same keys as + ID tokens. The OpenID Connect Back-Channel Logout draft explains how to + validate the JWT and figure out what session to end. + + Args: + request: The request to respond to + logout_token: The logout token (a JWT) extracted from the request body + """ + # Back-Channel Logout can be disabled in the config, hence this check. + # This is not that important for now since Synapse is registered + # manually to the OP, so not specifying the backchannel-logout URI is + # as effective than disabling it here. It might make more sense if we + # support dynamic registration in Synapse at some point. + if not self._config.backchannel_logout_enabled: + logger.warning( + f"Received an OIDC Back-Channel Logout request from issuer {self.issuer!r} but it is disabled in config" + ) + + # TODO: this responds with a 400 status code, which is what the OIDC + # Back-Channel Logout spec expects, but spec also suggests answering with + # a JSON object, with the `error` and `error_description` fields set, which + # we are not doing here. + # See https://openid.net/specs/openid-connect-backchannel-1_0.html#BCResponse + raise SynapseError( + 400, "OpenID Connect Back-Channel Logout is disabled for this provider" + ) + + metadata = await self.load_metadata() + + # As per OIDC Back-Channel Logout 1.0 sec. 2.4: + # A Logout Token MUST be signed and MAY also be encrypted. The same + # keys are used to sign and encrypt Logout Tokens as are used for ID + # Tokens. If the Logout Token is encrypted, it SHOULD replicate the + # iss (issuer) claim in the JWT Header Parameters, as specified in + # Section 5.3 of [JWT]. + alg_values = metadata.get("id_token_signing_alg_values_supported", ["RS256"]) + + # As per sec. 2.6: + # 3. Validate the iss, aud, and iat Claims in the same way they are + # validated in ID Tokens. + # Which means the audience should contain Synapse's client_id and the + # issuer should be the IdP issuer + claims_options = { + "iss": {"values": [metadata["issuer"]]}, + "aud": {"values": [self.client_id]}, + } + + try: + claims = await self._verify_jwt( + alg_values=alg_values, + token=logout_token, + claims_cls=LogoutToken, + claims_options=claims_options, + ) + except JoseError: + logger.exception("Invalid logout_token") + raise SynapseError(400, "Invalid logout_token") + + # As per sec. 2.6: + # 4. Verify that the Logout Token contains a sub Claim, a sid Claim, + # or both. + # 5. Verify that the Logout Token contains an events Claim whose + # value is JSON object containing the member name + # http://schemas.openid.net/event/backchannel-logout. + # 6. Verify that the Logout Token does not contain a nonce Claim. + # This is all verified by the LogoutToken claims class, so at this + # point the `sid` claim exists and is a string. + sid: str = claims.get("sid") + + # If the `sub` claim was included in the logout token, we check that it matches + # that it matches the right user. We can have cases where the `sub` claim is not + # the ID saved in database, so we let admins disable this check in config. + sub: Optional[str] = claims.get("sub") + expected_user_id: Optional[str] = None + if sub is not None and not self._config.backchannel_logout_ignore_sub: + expected_user_id = await self._store.get_user_by_external_id( + self.idp_id, sub + ) + + # Invalidate any running user-mapping sessions, in-flight login tokens and + # active devices + await self._sso_handler.revoke_sessions_for_provider_session_id( + auth_provider_id=self.idp_id, + auth_provider_session_id=sid, + expected_user_id=expected_user_id, + ) + + request.setResponseCode(200) + request.setHeader(b"Cache-Control", b"no-cache, no-store") + request.setHeader(b"Pragma", b"no-cache") + finish_request(request) + + +class LogoutToken(JWTClaims): + """ + Holds and verify claims of a logout token, as per + https://openid.net/specs/openid-connect-backchannel-1_0.html#LogoutToken + """ + + REGISTERED_CLAIMS = ["iss", "sub", "aud", "iat", "jti", "events", "sid"] + + def validate(self, now: Optional[int] = None, leeway: int = 0) -> None: + """Validate everything in claims payload.""" + super().validate(now, leeway) + self.validate_sid() + self.validate_events() + self.validate_nonce() + + def validate_sid(self) -> None: + """Ensure the sid claim is present""" + sid = self.get("sid") + if not sid: + raise MissingClaimError("sid") + + if not isinstance(sid, str): + raise InvalidClaimError("sid") + + def validate_nonce(self) -> None: + """Ensure the nonce claim is absent""" + if "nonce" in self: + raise InvalidClaimError("nonce") + + def validate_events(self) -> None: + """Ensure the events claim is present and with the right value""" + events = self.get("events") + if not events: + raise MissingClaimError("events") + + if not isinstance(events, dict): + raise InvalidClaimError("events") + + if "http://schemas.openid.net/event/backchannel-logout" not in events: + raise InvalidClaimError("events") + # number of seconds a newly-generated client secret should be valid for CLIENT_SECRET_VALIDITY_SECONDS = 3600 @@ -1105,6 +1423,7 @@ class JwtClientSecret: logger.info( "Generating new JWT for %s: %s %s", self._oauth_issuer, header, payload ) + jwt = JsonWebToken(header["alg"]) self._cached_secret = jwt.encode(header, payload, self._key.key) self._cached_secret_replacement_time = ( expires_at - CLIENT_SECRET_MIN_VALIDITY_SECONDS @@ -1116,12 +1435,10 @@ class UserAttributeDict(TypedDict): localpart: Optional[str] confirm_localpart: bool display_name: Optional[str] + picture: Optional[str] # may be omitted by older `OidcMappingProviders` emails: List[str] -C = TypeVar("C") - - class OidcMappingProvider(Generic[C]): """A mapping provider maps a UserInfo object to user attributes. @@ -1204,6 +1521,7 @@ env.filters.update( @attr.s(slots=True, frozen=True, auto_attribs=True) class JinjaOidcMappingConfig: subject_claim: str + picture_claim: str localpart_template: Optional[Template] display_name_template: Optional[Template] email_template: Optional[Template] @@ -1223,6 +1541,7 @@ class JinjaOidcMappingProvider(OidcMappingProvider[JinjaOidcMappingConfig]): @staticmethod def parse_config(config: dict) -> JinjaOidcMappingConfig: subject_claim = config.get("subject_claim", "sub") + picture_claim = config.get("picture_claim", "picture") def parse_template_config(option_name: str) -> Optional[Template]: if option_name not in config: @@ -1256,6 +1575,7 @@ class JinjaOidcMappingProvider(OidcMappingProvider[JinjaOidcMappingConfig]): return JinjaOidcMappingConfig( subject_claim=subject_claim, + picture_claim=picture_claim, localpart_template=localpart_template, display_name_template=display_name_template, email_template=email_template, @@ -1295,10 +1615,13 @@ class JinjaOidcMappingProvider(OidcMappingProvider[JinjaOidcMappingConfig]): if email: emails.append(email) + picture = userinfo.get("picture") + return UserAttributeDict( localpart=localpart, display_name=display_name, emails=emails, + picture=picture, confirm_localpart=self._config.confirm_localpart, ) diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index 1f83bab836..8c8ff18a1a 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -27,9 +27,9 @@ from synapse.handlers.room import ShutdownRoomResponse from synapse.logging.opentracing import trace from synapse.metrics.background_process_metrics import run_as_background_process from synapse.rest.admin._base import assert_user_is_admin -from synapse.storage.state import StateFilter from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, Requester, StreamKeyType +from synapse.types.state import StateFilter from synapse.util.async_helpers import ReadWriteLock from synapse.util.stringutils import random_string from synapse.visibility import filter_events_for_client @@ -448,6 +448,12 @@ class PaginationHandler: if pagin_config.from_token: from_token = pagin_config.from_token + elif pagin_config.direction == "f": + from_token = ( + await self.hs.get_event_sources().get_start_token_for_pagination( + room_id + ) + ) else: from_token = ( await self.hs.get_event_sources().get_current_token_for_pagination( @@ -458,11 +464,6 @@ class PaginationHandler: # `/messages` should still works with live tokens when manually provided. assert from_token.room_key.topological is not None - if pagin_config.limit is None: - # This shouldn't happen as we've set a default limit before this - # gets called. - raise Exception("limit not set") - room_token = from_token.room_key async with self.pagination_lock.read(room_id): diff --git a/synapse/handlers/presence.py b/synapse/handlers/presence.py index 4e575ffbaa..2af90b25a3 100644 --- a/synapse/handlers/presence.py +++ b/synapse/handlers/presence.py @@ -201,7 +201,7 @@ class BasePresenceHandler(abc.ABC): """Get the current presence state for multiple users. Returns: - dict: `user_id` -> `UserPresenceState` + A mapping of `user_id` -> `UserPresenceState` """ states = {} missing = [] @@ -256,7 +256,7 @@ class BasePresenceHandler(abc.ABC): with the app. """ - async def update_external_syncs_row( + async def update_external_syncs_row( # noqa: B027 (no-op by design) self, process_id: str, user_id: str, is_syncing: bool, sync_time_msec: int ) -> None: """Update the syncing users for an external process as a delta. @@ -272,7 +272,9 @@ class BasePresenceHandler(abc.ABC): sync_time_msec: Time in ms when the user was last syncing """ - async def update_external_syncs_clear(self, process_id: str) -> None: + async def update_external_syncs_clear( # noqa: B027 (no-op by design) + self, process_id: str + ) -> None: """Marks all users that had been marked as syncing by a given process as offline. @@ -476,7 +478,7 @@ class WorkerPresenceHandler(BasePresenceHandler): return _NullContextManager() prev_state = await self.current_state_for_user(user_id) - if prev_state != PresenceState.BUSY: + if prev_state.state != PresenceState.BUSY: # We set state here but pass ignore_status_msg = True as we don't want to # cause the status message to be cleared. # Note that this causes last_active_ts to be incremented which is not @@ -1596,7 +1598,9 @@ class PresenceEventSource(EventSource[int, UserPresenceState]): self, user: UserID, from_key: Optional[int], - limit: Optional[int] = None, + # Having a default limit doesn't match the EventSource API, but some + # callers do not provide it. It is unused in this class. + limit: int = 0, room_ids: Optional[Collection[str]] = None, is_guest: bool = False, explicit_room_id: Optional[str] = None, @@ -1688,10 +1692,12 @@ class PresenceEventSource(EventSource[int, UserPresenceState]): if from_key is not None: # First get all users that have had a presence update - updated_users = stream_change_cache.get_all_entities_changed(from_key) + result = stream_change_cache.get_all_entities_changed(from_key) # Cross-reference users we're interested in with those that have had updates. - if updated_users is not None: + if result.hit: + updated_users = result.entities + # If we have the full list of changes for presence we can # simply check which ones share a room with the user. get_updates_counter.labels("stream").inc() @@ -1760,14 +1766,14 @@ class PresenceEventSource(EventSource[int, UserPresenceState]): Returns: A list of presence states for the given user to receive. """ + updated_users = None if from_key: # Only return updates since the last sync - updated_users = self.store.presence_stream_cache.get_all_entities_changed( - from_key - ) - if not updated_users: - updated_users = [] + result = self.store.presence_stream_cache.get_all_entities_changed(from_key) + if result.hit: + updated_users = result.entities + if updated_users is not None: # Get the actual presence update for each change users_to_state = await self.get_presence_handler().current_state_for_users( updated_users diff --git a/synapse/handlers/profile.py b/synapse/handlers/profile.py index d8ff5289b5..4bf9a047a3 100644 --- a/synapse/handlers/profile.py +++ b/synapse/handlers/profile.py @@ -307,7 +307,11 @@ class ProfileHandler: if not self.max_avatar_size and not self.allowed_avatar_mimetypes: return True - server_name, _, media_id = parse_and_validate_mxc_uri(mxc) + host, port, media_id = parse_and_validate_mxc_uri(mxc) + if port is not None: + server_name = host + ":" + str(port) + else: + server_name = host if server_name == self.server_name: media_info = await self.store.get_local_media(media_id) diff --git a/synapse/handlers/receipts.py b/synapse/handlers/receipts.py index 4a7ec9e426..6a4fed1156 100644 --- a/synapse/handlers/receipts.py +++ b/synapse/handlers/receipts.py @@ -92,7 +92,6 @@ class ReceiptsHandler: continue # Check if these receipts apply to a thread. - thread_id = None data = user_values.get("data", {}) thread_id = data.get("thread_id") # If the thread ID is invalid, consider it missing. @@ -257,7 +256,7 @@ class ReceiptEventSource(EventSource[int, JsonDict]): self, user: UserID, from_key: int, - limit: Optional[int], + limit: int, room_ids: Iterable[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/handlers/register.py b/synapse/handlers/register.py index ca1c7a1866..c611efb760 100644 --- a/synapse/handlers/register.py +++ b/synapse/handlers/register.py @@ -38,6 +38,7 @@ from synapse.api.errors import ( ) from synapse.appservice import ApplicationService from synapse.config.server import is_threepid_reserved +from synapse.handlers.device import DeviceHandler from synapse.http.servlet import assert_params_in_dict from synapse.replication.http.login import RegisterDeviceReplicationServlet from synapse.replication.http.register import ( @@ -45,8 +46,8 @@ from synapse.replication.http.register import ( ReplicationRegisterServlet, ) from synapse.spam_checker_api import RegistrationBehaviour -from synapse.storage.state import StateFilter from synapse.types import RoomAlias, UserID, create_requester +from synapse.types.state import StateFilter if TYPE_CHECKING: from synapse.server import HomeServer @@ -841,6 +842,9 @@ class RegistrationHandler: refresh_token = None refresh_token_id = None + # This can only run on the main process. + assert isinstance(self.device_handler, DeviceHandler) + registered_device_id = await self.device_handler.check_device_registered( user_id, device_id, diff --git a/synapse/handlers/relations.py b/synapse/handlers/relations.py index 63bc6a7aa5..e96f9999a8 100644 --- a/synapse/handlers/relations.py +++ b/synapse/handlers/relations.py @@ -11,17 +11,21 @@ # 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 enum import logging -from typing import TYPE_CHECKING, Dict, FrozenSet, Iterable, List, Optional, Tuple +from typing import TYPE_CHECKING, Collection, Dict, FrozenSet, Iterable, List, Optional import attr -from synapse.api.constants import RelationTypes +from synapse.api.constants import EventTypes, RelationTypes from synapse.api.errors import SynapseError from synapse.events import EventBase, relation_from_event +from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.logging.opentracing import trace -from synapse.storage.databases.main.relations import _RelatedEvent -from synapse.types import JsonDict, Requester, StreamToken, UserID +from synapse.storage.databases.main.relations import ThreadsNextBatch, _RelatedEvent +from synapse.streams.config import PaginationConfig +from synapse.types import JsonDict, Requester, UserID +from synapse.util.async_helpers import gather_results from synapse.visibility import filter_events_for_client if TYPE_CHECKING: @@ -31,6 +35,13 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) +class ThreadsListInclude(str, enum.Enum): + """Valid values for the 'include' flag of /threads.""" + + all = "all" + participated = "participated" + + @attr.s(slots=True, frozen=True, auto_attribs=True) class _ThreadAggregation: # The latest event in the thread. @@ -66,19 +77,17 @@ class RelationsHandler: self._clock = hs.get_clock() self._event_handler = hs.get_event_handler() self._event_serializer = hs.get_event_client_serializer() + self._event_creation_handler = hs.get_event_creation_handler() async def get_relations( self, requester: Requester, event_id: str, room_id: str, + pagin_config: PaginationConfig, + include_original_event: bool, relation_type: Optional[str] = None, event_type: Optional[str] = None, - limit: int = 5, - direction: str = "b", - from_token: Optional[StreamToken] = None, - to_token: Optional[StreamToken] = None, - include_original_event: bool = False, ) -> JsonDict: """Get related events of a event, ordered by topological ordering. @@ -88,14 +97,10 @@ class RelationsHandler: requester: The user requesting the relations. event_id: Fetch events that relate to this event ID. room_id: The room the event belongs to. + pagin_config: The pagination config rules to apply, if any. + include_original_event: Whether to include the parent event. relation_type: Only fetch events with this relation type, if given. event_type: Only fetch events with this event type, if given. - limit: Only fetch the most recent `limit` events. - direction: Whether to fetch the most recent first (`"b"`) or the - oldest first (`"f"`). - from_token: Fetch rows from the given token, or from the start if None. - to_token: Fetch rows up to the given token, or up to the end if None. - include_original_event: Whether to include the parent event. Returns: The pagination chunk. @@ -123,10 +128,10 @@ class RelationsHandler: room_id=room_id, relation_type=relation_type, event_type=event_type, - limit=limit, - direction=direction, - from_token=from_token, - to_token=to_token, + limit=pagin_config.limit, + direction=pagin_config.direction, + from_token=pagin_config.from_token, + to_token=pagin_config.to_token, ) events = await self._main_store.get_events_as_list( @@ -162,90 +167,167 @@ class RelationsHandler: if next_token: return_value["next_batch"] = await next_token.to_string(self._main_store) - if from_token: - return_value["prev_batch"] = await from_token.to_string(self._main_store) + if pagin_config.from_token: + return_value["prev_batch"] = await pagin_config.from_token.to_string( + self._main_store + ) return return_value - async def get_relations_for_event( + async def redact_events_related_to( self, + requester: Requester, event_id: str, - event: EventBase, - room_id: str, - relation_type: str, - ignored_users: FrozenSet[str] = frozenset(), - ) -> Tuple[List[_RelatedEvent], Optional[StreamToken]]: - """Get a list of events which relate to an event, ordered by topological ordering. + initial_redaction_event: EventBase, + relation_types: List[str], + ) -> None: + """Redacts all events related to the given event ID with one of the given + relation types. - Args: - event_id: Fetch events that relate to this event ID. - event: The matching EventBase to event_id. - room_id: The room the event belongs to. - relation_type: The type of relation. - ignored_users: The users ignored by the requesting user. + This method is expected to be called when redacting the event referred to by + the given event ID. - Returns: - List of event IDs that match relations requested. The rows are of - the form `{"event_id": "..."}`. - """ + If an event cannot be redacted (e.g. because of insufficient permissions), log + the error and try to redact the next one. - # Call the underlying storage method, which is cached. - related_events, next_token = await self._main_store.get_relations_for_event( - event_id, event, room_id, relation_type, direction="f" + Args: + requester: The requester to redact events on behalf of. + event_id: The event IDs to look and redact relations of. + initial_redaction_event: The redaction for the event referred to by + event_id. + relation_types: The types of relations to look for. + + Raises: + ShadowBanError if the requester is shadow-banned + """ + related_event_ids = ( + await self._main_store.get_all_relations_for_event_with_types( + event_id, relation_types + ) ) - # Filter out ignored users and convert to the expected format. - related_events = [ - event for event in related_events if event.sender not in ignored_users - ] - - return related_events, next_token + for related_event_id in related_event_ids: + try: + await self._event_creation_handler.create_and_send_nonmember_event( + requester, + { + "type": EventTypes.Redaction, + "content": initial_redaction_event.content, + "room_id": initial_redaction_event.room_id, + "sender": requester.user.to_string(), + "redacts": related_event_id, + }, + ratelimit=False, + ) + except SynapseError as e: + logger.warning( + "Failed to redact event %s (related to event %s): %s", + related_event_id, + event_id, + e.msg, + ) - async def get_annotations_for_event( - self, - event_id: str, - room_id: str, - limit: int = 5, - ignored_users: FrozenSet[str] = frozenset(), - ) -> List[JsonDict]: - """Get a list of annotations on the event, grouped by event type and + async def get_annotations_for_events( + self, event_ids: Collection[str], ignored_users: FrozenSet[str] = frozenset() + ) -> Dict[str, List[JsonDict]]: + """Get a list of annotations to the given events, grouped by event type and aggregation key, sorted by count. - This is used e.g. to get the what and how many reactions have happend + This is used e.g. to get the what and how many reactions have happened on an event. Args: - event_id: Fetch events that relate to this event ID. - room_id: The room the event belongs to. - limit: Only fetch the `limit` groups. + event_ids: Fetch events that relate to these event IDs. ignored_users: The users ignored by the requesting user. Returns: - List of groups of annotations that match. Each row is a dict with - `type`, `key` and `count` fields. + A map of event IDs to a list of groups of annotations that match. + Each entry is a dict with `type`, `key` and `count` fields. """ # Get the base results for all users. - full_results = await self._main_store.get_aggregation_groups_for_event( - event_id, room_id, limit + full_results = await self._main_store.get_aggregation_groups_for_events( + event_ids ) + # Avoid additional logic if there are no ignored users. + if not ignored_users: + return { + event_id: results + for event_id, results in full_results.items() + if results + } + # Then subtract off the results for any ignored users. ignored_results = await self._main_store.get_aggregation_groups_for_users( - event_id, room_id, limit, ignored_users + [event_id for event_id, results in full_results.items() if results], + ignored_users, ) - filtered_results = [] - for result in full_results: - key = (result["type"], result["key"]) - if key in ignored_results: - result = result.copy() - result["count"] -= ignored_results[key] - if result["count"] <= 0: - continue - filtered_results.append(result) + filtered_results = {} + for event_id, results in full_results.items(): + # If no annotations, skip. + if not results: + continue + + # If there are not ignored results for this event, copy verbatim. + if event_id not in ignored_results: + filtered_results[event_id] = results + continue + + # Otherwise, subtract out the ignored results. + event_ignored_results = ignored_results[event_id] + for result in results: + key = (result["type"], result["key"]) + if key in event_ignored_results: + # Ensure to not modify the cache. + result = result.copy() + result["count"] -= event_ignored_results[key] + if result["count"] <= 0: + continue + filtered_results.setdefault(event_id, []).append(result) return filtered_results + async def get_references_for_events( + self, event_ids: Collection[str], ignored_users: FrozenSet[str] = frozenset() + ) -> Dict[str, List[_RelatedEvent]]: + """Get a list of references to the given events. + + Args: + event_ids: Fetch events that relate to this event ID. + ignored_users: The users ignored by the requesting user. + + Returns: + A map of event IDs to a list related events. + """ + + related_events = await self._main_store.get_references_for_events(event_ids) + + # Avoid additional logic if there are no ignored users. + if not ignored_users: + return { + event_id: results + for event_id, results in related_events.items() + if results + } + + # Filter out ignored users. + results = {} + for event_id, events in related_events.items(): + # If no references, skip. + if not events: + continue + + # Filter ignored users out. + events = [event for event in events if event.sender not in ignored_users] + # If there are no events left, skip this event. + if not events: + continue + + results[event_id] = events + + return results + async def _get_threads_for_events( self, events_by_id: Dict[str, EventBase], @@ -308,59 +390,66 @@ class RelationsHandler: results = {} for event_id, summary in summaries.items(): - if summary: - thread_count, latest_thread_event = summary - - # Subtract off the count of any ignored users. - for ignored_user in ignored_users: - thread_count -= ignored_results.get((event_id, ignored_user), 0) - - # This is gnarly, but if the latest event is from an ignored user, - # attempt to find one that isn't from an ignored user. - if latest_thread_event.sender in ignored_users: - room_id = latest_thread_event.room_id - - # If the root event is not found, something went wrong, do - # not include a summary of the thread. - event = await self._event_handler.get_event(user, room_id, event_id) - if event is None: - continue + # If no thread, skip. + if not summary: + continue - potential_events, _ = await self.get_relations_for_event( - event_id, - event, - room_id, - RelationTypes.THREAD, - ignored_users, - ) + thread_count, latest_thread_event = summary - # If all found events are from ignored users, do not include - # a summary of the thread. - if not potential_events: - continue + # Subtract off the count of any ignored users. + for ignored_user in ignored_users: + thread_count -= ignored_results.get((event_id, ignored_user), 0) - # The *last* event returned is the one that is cared about. - event = await self._event_handler.get_event( - user, room_id, potential_events[-1].event_id - ) - # It is unexpected that the event will not exist. - if event is None: - logger.warning( - "Unable to fetch latest event in a thread with event ID: %s", - potential_events[-1].event_id, - ) - continue - latest_thread_event = event - - results[event_id] = _ThreadAggregation( - latest_event=latest_thread_event, - count=thread_count, - # If there's a thread summary it must also exist in the - # participated dictionary. - current_user_participated=events_by_id[event_id].sender == user_id - or participated[event_id], + # This is gnarly, but if the latest event is from an ignored user, + # attempt to find one that isn't from an ignored user. + if latest_thread_event.sender in ignored_users: + room_id = latest_thread_event.room_id + + # If the root event is not found, something went wrong, do + # not include a summary of the thread. + event = await self._event_handler.get_event(user, room_id, event_id) + if event is None: + continue + + # Attempt to find another event to use as the latest event. + potential_events, _ = await self._main_store.get_relations_for_event( + event_id, event, room_id, RelationTypes.THREAD, direction="f" ) + # Filter out ignored users. + potential_events = [ + event + for event in potential_events + if event.sender not in ignored_users + ] + + # If all found events are from ignored users, do not include + # a summary of the thread. + if not potential_events: + continue + + # The *last* event returned is the one that is cared about. + event = await self._event_handler.get_event( + user, room_id, potential_events[-1].event_id + ) + # It is unexpected that the event will not exist. + if event is None: + logger.warning( + "Unable to fetch latest event in a thread with event ID: %s", + potential_events[-1].event_id, + ) + continue + latest_thread_event = event + + results[event_id] = _ThreadAggregation( + latest_event=latest_thread_event, + count=thread_count, + # If there's a thread summary it must also exist in the + # participated dictionary. + current_user_participated=events_by_id[event_id].sender == user_id + or participated[event_id], + ) + return results @trace @@ -438,48 +527,131 @@ class RelationsHandler: # (as that is what makes it part of the thread). relations_by_id[latest_thread_event.event_id] = RelationTypes.THREAD - # Fetch other relations per event. - for event in events_by_id.values(): - # Fetch any annotations (ie, reactions) to bundle with this event. - annotations = await self.get_annotations_for_event( - event.event_id, event.room_id, ignored_users=ignored_users + async def _fetch_annotations() -> None: + """Fetch any annotations (ie, reactions) to bundle with this event.""" + annotations_by_event_id = await self.get_annotations_for_events( + events_by_id.keys(), ignored_users=ignored_users ) - if annotations: - results.setdefault( - event.event_id, BundledAggregations() - ).annotations = {"chunk": annotations} - - # Fetch any references to bundle with this event. - references, next_token = await self.get_relations_for_event( - event.event_id, - event, - event.room_id, - RelationTypes.REFERENCE, - ignored_users=ignored_users, + for event_id, annotations in annotations_by_event_id.items(): + if annotations: + results.setdefault(event_id, BundledAggregations()).annotations = { + "chunk": annotations + } + + async def _fetch_references() -> None: + """Fetch any references to bundle with this event.""" + references_by_event_id = await self.get_references_for_events( + events_by_id.keys(), ignored_users=ignored_users + ) + for event_id, references in references_by_event_id.items(): + if references: + results.setdefault(event_id, BundledAggregations()).references = { + "chunk": [{"event_id": ev.event_id} for ev in references] + } + + async def _fetch_edits() -> None: + """ + Fetch any edits (but not for redacted events). + + Note that there is no use in limiting edits by ignored users since the + parent event should be ignored in the first place if the user is ignored. + """ + edits = await self._main_store.get_applicable_edits( + [ + event_id + for event_id, event in events_by_id.items() + if not event.internal_metadata.is_redacted() + ] + ) + for event_id, edit in edits.items(): + results.setdefault(event_id, BundledAggregations()).replace = edit + + # Parallelize the calls for annotations, references, and edits since they + # are unrelated. + await make_deferred_yieldable( + gather_results( + ( + run_in_background(_fetch_annotations), + run_in_background(_fetch_references), + run_in_background(_fetch_edits), + ) ) - if references: - aggregations = results.setdefault(event.event_id, BundledAggregations()) - aggregations.references = { - "chunk": [{"event_id": ev.event_id} for ev in references] - } - - if next_token: - aggregations.references["next_batch"] = await next_token.to_string( - self._main_store - ) - - # Fetch any edits (but not for redacted events). - # - # Note that there is no use in limiting edits by ignored users since the - # parent event should be ignored in the first place if the user is ignored. - edits = await self._main_store.get_applicable_edits( - [ - event_id - for event_id, event in events_by_id.items() - if not event.internal_metadata.is_redacted() - ] ) - for event_id, edit in edits.items(): - results.setdefault(event_id, BundledAggregations()).replace = edit return results + + async def get_threads( + self, + requester: Requester, + room_id: str, + include: ThreadsListInclude, + limit: int = 5, + from_token: Optional[ThreadsNextBatch] = None, + ) -> JsonDict: + """Get related events of a event, ordered by topological ordering. + + Args: + requester: The user requesting the relations. + room_id: The room the event belongs to. + include: One of "all" or "participated" to indicate which threads should + be returned. + limit: Only fetch the most recent `limit` events. + from_token: Fetch rows from the given token, or from the start if None. + + Returns: + The pagination chunk. + """ + + user_id = requester.user.to_string() + + # TODO Properly handle a user leaving a room. + (_, member_event_id) = await self._auth.check_user_in_room_or_world_readable( + room_id, requester, allow_departed_users=True + ) + + # Note that ignored users are not passed into get_threads + # below. Ignored users are handled in filter_events_for_client (and by + # not passing them in here we should get a better cache hit rate). + thread_roots, next_batch = await self._main_store.get_threads( + room_id=room_id, limit=limit, from_token=from_token + ) + + events = await self._main_store.get_events_as_list(thread_roots) + + if include == ThreadsListInclude.participated: + # Pre-seed thread participation with whether the requester sent the event. + participated = {event.event_id: event.sender == user_id for event in events} + # For events the requester did not send, check the database for whether + # the requester sent a threaded reply. + participated.update( + await self._main_store.get_threads_participated( + [eid for eid, p in participated.items() if not p], + user_id, + ) + ) + + # Limit the returned threads to those the user has participated in. + events = [event for event in events if participated[event.event_id]] + + events = await filter_events_for_client( + self._storage_controllers, + user_id, + events, + is_peeking=(member_event_id is None), + ) + + aggregations = await self.get_bundled_aggregations( + events, requester.user.to_string() + ) + + now = self._clock.time_msec() + serialized_events = self._event_serializer.serialize_events( + events, now, bundle_aggregations=aggregations + ) + + return_value: JsonDict = {"chunk": serialized_events} + + if next_batch: + return_value["next_batch"] = str(next_batch) + + return return_value diff --git a/synapse/handlers/room.py b/synapse/handlers/room.py index 57ab05ad25..f81241c2b3 100644 --- a/synapse/handlers/room.py +++ b/synapse/handlers/room.py @@ -49,7 +49,6 @@ from synapse.api.constants import ( from synapse.api.errors import ( AuthError, Codes, - HttpResponseException, LimitExceededError, NotFoundError, StoreError, @@ -60,11 +59,9 @@ from synapse.api.room_versions import KNOWN_ROOM_VERSIONS, RoomVersion from synapse.event_auth import validate_event_for_room_version from synapse.events import EventBase from synapse.events.utils import copy_and_fixup_power_levels_contents -from synapse.federation.federation_client import InvalidResponseError from synapse.handlers.relations import BundledAggregations from synapse.module_api import NOT_SPAM from synapse.rest.admin._base import assert_user_is_admin -from synapse.storage.state import StateFilter from synapse.streams import EventSource from synapse.types import ( JsonDict, @@ -79,6 +76,7 @@ from synapse.types import ( UserID, create_requester, ) +from synapse.types.state import StateFilter from synapse.util import stringutils from synapse.util.caches.response_cache import ResponseCache from synapse.util.stringutils import parse_and_validate_server_name @@ -229,9 +227,7 @@ class RoomCreationHandler: }, ) validate_event_for_room_version(tombstone_event) - await self._event_auth_handler.check_auth_rules_from_context( - tombstone_event, tombstone_context - ) + await self._event_auth_handler.check_auth_rules_from_context(tombstone_event) # Upgrade the room # @@ -561,7 +557,6 @@ class RoomCreationHandler: invite_list=[], initial_state=initial_state, creation_content=creation_content, - ratelimit=False, ) # Transfer membership events @@ -755,6 +750,10 @@ class RoomCreationHandler: ) if ratelimit: + # Rate limit once in advance, but don't rate limit the individual + # events in the room — room creation isn't atomic and it's very + # janky if half the events in the initial state don't make it because + # of rate limiting. await self.request_ratelimiter.ratelimit(requester) room_version_id = config.get( @@ -915,7 +914,6 @@ class RoomCreationHandler: room_alias=room_alias, power_level_content_override=power_level_content_override, creator_join_profile=creator_join_profile, - ratelimit=ratelimit, ) if "name" in config: @@ -1039,7 +1037,6 @@ class RoomCreationHandler: room_alias: Optional[RoomAlias] = None, power_level_content_override: Optional[JsonDict] = None, creator_join_profile: Optional[JsonDict] = None, - ratelimit: bool = True, ) -> Tuple[int, str, int]: """Sends the initial events into a new room. Sends the room creation, membership, and power level events into the room sequentially, then creates and batches up the @@ -1048,6 +1045,8 @@ class RoomCreationHandler: `power_level_content_override` doesn't apply when initial state has power level state event content. + Rate limiting should already have been applied by this point. + Returns: A tuple containing the stream ID, event ID and depth of the last event sent to the room. @@ -1057,9 +1056,6 @@ class RoomCreationHandler: event_keys = {"room_id": room_id, "sender": creator_id, "state_key": ""} depth = 1 - # the last event sent/persisted to the db - last_sent_event_id: Optional[str] = None - # the most recently created event prev_event: List[str] = [] # a map of event types, state keys -> event_ids. We collect these mappings this as events are @@ -1084,6 +1080,19 @@ class RoomCreationHandler: for_batch: bool, **kwargs: Any, ) -> Tuple[EventBase, synapse.events.snapshot.EventContext]: + """ + Creates an event and associated event context. + Args: + etype: the type of event to be created + content: content of the event + for_batch: whether the event is being created for batch persisting. If + bool for_batch is true, this will create an event using the prev_event_ids, + and will create an event context for the event using the parameters state_map + and current_state_group, thus these parameters must be provided in this + case if for_batch is True. The subsequently created event and context + are suitable for being batched up and bulk persisted to the database + with other similarly created events. + """ nonlocal depth nonlocal prev_event @@ -1104,26 +1113,6 @@ class RoomCreationHandler: return new_event, new_context - async def send( - event: EventBase, - context: synapse.events.snapshot.EventContext, - creator: Requester, - ) -> int: - nonlocal last_sent_event_id - - ev = await self.event_creation_handler.handle_new_client_event( - requester=creator, - events_and_context=[(event, context)], - ratelimit=False, - ignore_shadow_ban=True, - ) - - last_sent_event_id = ev.event_id - - # we know it was persisted, so must have a stream ordering - assert ev.internal_metadata.stream_ordering - return ev.internal_metadata.stream_ordering - try: config = self._presets_dict[preset_config] except KeyError: @@ -1137,16 +1126,20 @@ class RoomCreationHandler: ) logger.debug("Sending %s in new room", EventTypes.Member) - await send(creation_event, creation_context, creator) + ev = await self.event_creation_handler.handle_new_client_event( + requester=creator, + events_and_context=[(creation_event, creation_context)], + ratelimit=False, + ignore_shadow_ban=True, + ) + last_sent_event_id = ev.event_id - # Room create event must exist at this point - assert last_sent_event_id is not None member_event_id, _ = await self.room_member_handler.update_membership( creator, creator.user, room_id, "join", - ratelimit=ratelimit, + ratelimit=False, content=creator_join_profile, new_room=True, prev_event_ids=[last_sent_event_id], @@ -1159,15 +1152,24 @@ class RoomCreationHandler: depth += 1 state_map[(EventTypes.Member, creator.user.to_string())] = member_event_id + # we need the state group of the membership event as it is the current state group + event_to_state = ( + await self._storage_controllers.state.get_state_group_for_events( + [member_event_id] + ) + ) + current_state_group = event_to_state[member_event_id] + + events_to_send = [] # We treat the power levels override specially as this needs to be one # of the first events that get sent into a room. pl_content = initial_state.pop((EventTypes.PowerLevels, ""), None) if pl_content is not None: power_event, power_context = await create_event( - EventTypes.PowerLevels, pl_content, False + EventTypes.PowerLevels, pl_content, True ) current_state_group = power_context._state_group - await send(power_event, power_context, creator) + events_to_send.append((power_event, power_context)) else: power_level_content: JsonDict = { "users": {creator_id: 100}, @@ -1213,12 +1215,11 @@ class RoomCreationHandler: pl_event, pl_context = await create_event( EventTypes.PowerLevels, power_level_content, - False, + True, ) current_state_group = pl_context._state_group - await send(pl_event, pl_context, creator) + events_to_send.append((pl_event, pl_context)) - events_to_send = [] if room_alias and (EventTypes.CanonicalAlias, "") not in initial_state: room_alias_event, room_alias_context = await create_event( EventTypes.CanonicalAlias, {"alias": room_alias.to_string()}, True @@ -1271,7 +1272,10 @@ class RoomCreationHandler: events_to_send.append((encryption_event, encryption_context)) last_event = await self.event_creation_handler.handle_new_client_event( - creator, events_to_send, ignore_shadow_ban=True + creator, + events_to_send, + ignore_shadow_ban=True, + ratelimit=False, ) assert last_event.internal_metadata.stream_ordering is not None return last_event.internal_metadata.stream_ordering, last_event.event_id, depth @@ -1447,7 +1451,7 @@ class RoomContextHandler: events_before=events_before, event=event, events_after=events_after, - state=await filter_evts(state_events), + state=state_events, aggregations=aggregations, start=await token.copy_and_replace( StreamKeyType.ROOM, results.start @@ -1493,7 +1497,12 @@ class TimestampLookupHandler: Raises: SynapseError if unable to find any event locally in the given direction """ - + logger.debug( + "get_event_for_timestamp(room_id=%s, timestamp=%s, direction=%s) Finding closest event...", + room_id, + timestamp, + direction, + ) local_event_id = await self.store.get_event_id_for_timestamp( room_id, timestamp, direction ) @@ -1545,85 +1554,54 @@ class TimestampLookupHandler: ) ) - # Loop through each homeserver candidate until we get a succesful response - for domain in likely_domains: - # We don't want to ask our own server for information we don't have - if domain == self.server_name: - continue + remote_response = await self.federation_client.timestamp_to_event( + destinations=likely_domains, + room_id=room_id, + timestamp=timestamp, + direction=direction, + ) + if remote_response is not None: + logger.debug( + "get_event_for_timestamp: remote_response=%s", + remote_response, + ) - try: - remote_response = await self.federation_client.timestamp_to_event( - domain, room_id, timestamp, direction - ) - logger.debug( - "get_event_for_timestamp: response from domain(%s)=%s", - domain, - remote_response, - ) + remote_event_id = remote_response.event_id + remote_origin_server_ts = remote_response.origin_server_ts - remote_event_id = remote_response.event_id - remote_origin_server_ts = remote_response.origin_server_ts - - # Backfill this event so we can get a pagination token for - # it with `/context` and paginate `/messages` from this - # point. - # - # TODO: The requested timestamp may lie in a part of the - # event graph that the remote server *also* didn't have, - # in which case they will have returned another event - # which may be nowhere near the requested timestamp. In - # the future, we may need to reconcile that gap and ask - # other homeservers, and/or extend `/timestamp_to_event` - # to return events on *both* sides of the timestamp to - # help reconcile the gap faster. - remote_event = ( - await self.federation_event_handler.backfill_event_id( - domain, room_id, remote_event_id - ) - ) + # Backfill this event so we can get a pagination token for + # it with `/context` and paginate `/messages` from this + # point. + pulled_pdu_info = await self.federation_event_handler.backfill_event_id( + likely_domains, room_id, remote_event_id + ) + remote_event = pulled_pdu_info.pdu - # XXX: When we see that the remote server is not trustworthy, - # maybe we should not ask them first in the future. - if remote_origin_server_ts != remote_event.origin_server_ts: - logger.info( - "get_event_for_timestamp: Remote server (%s) claimed that remote_event_id=%s occured at remote_origin_server_ts=%s but that isn't true (actually occured at %s). Their claims are dubious and we should consider not trusting them.", - domain, - remote_event_id, - remote_origin_server_ts, - remote_event.origin_server_ts, - ) - - # Only return the remote event if it's closer than the local event - if not local_event or ( - abs(remote_event.origin_server_ts - timestamp) - < abs(local_event.origin_server_ts - timestamp) - ): - logger.info( - "get_event_for_timestamp: returning remote_event_id=%s (%s) since it's closer to timestamp=%s than local_event=%s (%s)", - remote_event_id, - remote_event.origin_server_ts, - timestamp, - local_event.event_id if local_event else None, - local_event.origin_server_ts if local_event else None, - ) - return remote_event_id, remote_origin_server_ts - except (HttpResponseException, InvalidResponseError) as ex: - # Let's not put a high priority on some other homeserver - # failing to respond or giving a random response - logger.debug( - "get_event_for_timestamp: Failed to fetch /timestamp_to_event from %s because of exception(%s) %s args=%s", - domain, - type(ex).__name__, - ex, - ex.args, + # XXX: When we see that the remote server is not trustworthy, + # maybe we should not ask them first in the future. + if remote_origin_server_ts != remote_event.origin_server_ts: + logger.info( + "get_event_for_timestamp: Remote server (%s) claimed that remote_event_id=%s occured at remote_origin_server_ts=%s but that isn't true (actually occured at %s). Their claims are dubious and we should consider not trusting them.", + pulled_pdu_info.pull_origin, + remote_event_id, + remote_origin_server_ts, + remote_event.origin_server_ts, ) - except Exception: - # But we do want to see some exceptions in our code - logger.warning( - "get_event_for_timestamp: Failed to fetch /timestamp_to_event from %s because of exception", - domain, - exc_info=True, + + # Only return the remote event if it's closer than the local event + if not local_event or ( + abs(remote_event.origin_server_ts - timestamp) + < abs(local_event.origin_server_ts - timestamp) + ): + logger.info( + "get_event_for_timestamp: returning remote_event_id=%s (%s) since it's closer to timestamp=%s than local_event=%s (%s)", + remote_event_id, + remote_event.origin_server_ts, + timestamp, + local_event.event_id if local_event else None, + local_event.origin_server_ts if local_event else None, ) + return remote_event_id, remote_origin_server_ts # To appease mypy, we have to add both of these conditions to check for # `None`. We only expect `local_event` to be `None` when @@ -1646,7 +1624,7 @@ class RoomEventSource(EventSource[RoomStreamToken, EventBase]): self, user: UserID, from_key: RoomStreamToken, - limit: Optional[int], + limit: int, room_ids: Collection[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index 6ad2b38b8f..0c39e852a1 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -34,7 +34,6 @@ from synapse.events.snapshot import EventContext from synapse.handlers.profile import MAX_AVATAR_URL_LEN, MAX_DISPLAYNAME_LEN from synapse.logging import opentracing from synapse.module_api import NOT_SPAM -from synapse.storage.state import StateFilter from synapse.types import ( JsonDict, Requester, @@ -45,6 +44,7 @@ from synapse.types import ( create_requester, get_domain_from_id, ) +from synapse.types.state import StateFilter from synapse.util.async_helpers import Linearizer from synapse.util.distributor import user_left_room diff --git a/synapse/handlers/saml.py b/synapse/handlers/saml.py index 9602f0d0bb..874860d461 100644 --- a/synapse/handlers/saml.py +++ b/synapse/handlers/saml.py @@ -441,7 +441,7 @@ class DefaultSamlMappingProvider: client_redirect_url: where the client wants to redirect to Returns: - dict: A dict containing new user attributes. Possible keys: + A dict containing new user attributes. Possible keys: * mxid_localpart (str): Required. The localpart of the user's mxid * displayname (str): The displayname of the user * emails (list[str]): Any emails for the user @@ -483,7 +483,7 @@ class DefaultSamlMappingProvider: Args: config: A dictionary containing configuration options for this provider Returns: - SamlConfig: A custom config object for this module + A custom config object for this module """ # Parse config options and use defaults where necessary mxid_source_attribute = config.get("mxid_source_attribute", "uid") diff --git a/synapse/handlers/search.py b/synapse/handlers/search.py index bcab98c6d5..33115ce488 100644 --- a/synapse/handlers/search.py +++ b/synapse/handlers/search.py @@ -23,8 +23,8 @@ from synapse.api.constants import EventTypes, Membership from synapse.api.errors import NotFoundError, SynapseError from synapse.api.filtering import Filter from synapse.events import EventBase -from synapse.storage.state import StateFilter from synapse.types import JsonDict, StreamKeyType, UserID +from synapse.types.state import StateFilter from synapse.visibility import filter_events_for_client if TYPE_CHECKING: diff --git a/synapse/handlers/set_password.py b/synapse/handlers/set_password.py index 73861bbd40..bd9d0bb34b 100644 --- a/synapse/handlers/set_password.py +++ b/synapse/handlers/set_password.py @@ -15,6 +15,7 @@ import logging from typing import TYPE_CHECKING, Optional from synapse.api.errors import Codes, StoreError, SynapseError +from synapse.handlers.device import DeviceHandler from synapse.types import Requester if TYPE_CHECKING: @@ -29,7 +30,10 @@ class SetPasswordHandler: def __init__(self, hs: "HomeServer"): self.store = hs.get_datastores().main self._auth_handler = hs.get_auth_handler() - self._device_handler = hs.get_device_handler() + # This can only be instantiated on the main process. + device_handler = hs.get_device_handler() + assert isinstance(device_handler, DeviceHandler) + self._device_handler = device_handler async def set_password( self, diff --git a/synapse/handlers/sso.py b/synapse/handlers/sso.py index e035677b8a..44e70fc4b8 100644 --- a/synapse/handlers/sso.py +++ b/synapse/handlers/sso.py @@ -12,6 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. import abc +import hashlib +import io import logging from typing import ( TYPE_CHECKING, @@ -37,6 +39,7 @@ from twisted.web.server import Request from synapse.api.constants import LoginType from synapse.api.errors import Codes, NotFoundError, RedirectException, SynapseError from synapse.config.sso import SsoAttributeRequirement +from synapse.handlers.device import DeviceHandler from synapse.handlers.register import init_counters_for_auth_provider from synapse.handlers.ui_auth import UIAuthSessionDataConstants from synapse.http import get_request_user_agent @@ -137,6 +140,7 @@ class UserAttributes: localpart: Optional[str] confirm_localpart: bool = False display_name: Optional[str] = None + picture: Optional[str] = None emails: Collection[str] = attr.Factory(list) @@ -191,9 +195,14 @@ class SsoHandler: self._server_name = hs.hostname self._registration_handler = hs.get_registration_handler() self._auth_handler = hs.get_auth_handler() + self._device_handler = hs.get_device_handler() self._error_template = hs.config.sso.sso_error_template self._bad_user_template = hs.config.sso.sso_auth_bad_user_template self._profile_handler = hs.get_profile_handler() + self._media_repo = ( + hs.get_media_repository() if hs.config.media.can_load_media_repo else None + ) + self._http_client = hs.get_proxied_blacklisted_http_client() # The following template is shown after a successful user interactive # authentication session. It tells the user they can close the window. @@ -493,6 +502,8 @@ class SsoHandler: await self._profile_handler.set_displayname( user_id_obj, requester, attributes.display_name, True ) + if attributes.picture: + await self.set_avatar(user_id, attributes.picture) await self._auth_handler.complete_sso_login( user_id, @@ -701,8 +712,110 @@ class SsoHandler: await self._store.record_user_external_id( auth_provider_id, remote_user_id, registered_user_id ) + + # Set avatar, if available + if attributes.picture: + await self.set_avatar(registered_user_id, attributes.picture) + return registered_user_id + async def set_avatar(self, user_id: str, picture_https_url: str) -> bool: + """Set avatar of the user. + + This downloads the image file from the URL provided, stores that in + the media repository and then sets the avatar on the user's profile. + + It can detect if the same image is being saved again and bails early by storing + the hash of the file in the `upload_name` of the avatar image. + + Currently, it only supports server configurations which run the media repository + within the same process. + + It silently fails and logs a warning by raising an exception and catching it + internally if: + * it is unable to fetch the image itself (non 200 status code) or + * the image supplied is bigger than max allowed size or + * the image type is not one of the allowed image types. + + Args: + user_id: matrix user ID in the form @localpart:domain as a string. + + picture_https_url: HTTPS url for the picture image file. + + Returns: `True` if the user's avatar has been successfully set to the image at + `picture_https_url`. + """ + if self._media_repo is None: + logger.info( + "failed to set user avatar because out-of-process media repositories " + "are not supported yet " + ) + return False + + try: + uid = UserID.from_string(user_id) + + def is_allowed_mime_type(content_type: str) -> bool: + if ( + self._profile_handler.allowed_avatar_mimetypes + and content_type + not in self._profile_handler.allowed_avatar_mimetypes + ): + return False + return True + + # download picture, enforcing size limit & mime type check + picture = io.BytesIO() + + content_length, headers, uri, code = await self._http_client.get_file( + url=picture_https_url, + output_stream=picture, + max_size=self._profile_handler.max_avatar_size, + is_allowed_content_type=is_allowed_mime_type, + ) + + if code != 200: + raise Exception( + "GET request to download sso avatar image returned {}".format(code) + ) + + # upload name includes hash of the image file's content so that we can + # easily check if it requires an update or not, the next time user logs in + upload_name = "sso_avatar_" + hashlib.sha256(picture.read()).hexdigest() + + # bail if user already has the same avatar + profile = await self._profile_handler.get_profile(user_id) + if profile["avatar_url"] is not None: + server_name = profile["avatar_url"].split("/")[-2] + media_id = profile["avatar_url"].split("/")[-1] + if server_name == self._server_name: + media = await self._media_repo.store.get_local_media(media_id) + if media is not None and upload_name == media["upload_name"]: + logger.info("skipping saving the user avatar") + return True + + # store it in media repository + avatar_mxc_url = await self._media_repo.create_content( + media_type=headers[b"Content-Type"][0].decode("utf-8"), + upload_name=upload_name, + content=picture, + content_length=content_length, + auth_user=uid, + ) + + # save it as user avatar + await self._profile_handler.set_avatar_url( + uid, + create_requester(uid), + str(avatar_mxc_url), + ) + + logger.info("successfully saved the user avatar") + return True + except Exception: + logger.warning("failed to save the user avatar") + return False + async def complete_sso_ui_auth_request( self, auth_provider_id: str, @@ -874,7 +987,7 @@ class SsoHandler: ) async def handle_terms_accepted( - self, request: Request, session_id: str, terms_version: str + self, request: SynapseRequest, session_id: str, terms_version: str ) -> None: """Handle a request to the new-user 'consent' endpoint @@ -1026,6 +1139,84 @@ class SsoHandler: return True + async def revoke_sessions_for_provider_session_id( + self, + auth_provider_id: str, + auth_provider_session_id: str, + expected_user_id: Optional[str] = None, + ) -> None: + """Revoke any devices and in-flight logins tied to a provider session. + + Can only be called from the main process. + + Args: + auth_provider_id: A unique identifier for this SSO provider, e.g. + "oidc" or "saml". + auth_provider_session_id: The session ID from the provider to logout + expected_user_id: The user we're expecting to logout. If set, it will ignore + sessions belonging to other users and log an error. + """ + + # It is expected that this is the main process. + assert isinstance( + self._device_handler, DeviceHandler + ), "revoking SSO sessions can only be called on the main process" + + # Invalidate any running user-mapping sessions + to_delete = [] + for session_id, session in self._username_mapping_sessions.items(): + if ( + session.auth_provider_id == auth_provider_id + and session.auth_provider_session_id == auth_provider_session_id + ): + to_delete.append(session_id) + + for session_id in to_delete: + logger.info("Revoking mapping session %s", session_id) + del self._username_mapping_sessions[session_id] + + # Invalidate any in-flight login tokens + await self._store.invalidate_login_tokens_by_session_id( + auth_provider_id=auth_provider_id, + auth_provider_session_id=auth_provider_session_id, + ) + + # Fetch any device(s) in the store associated with the session ID. + devices = await self._store.get_devices_by_auth_provider_session_id( + auth_provider_id=auth_provider_id, + auth_provider_session_id=auth_provider_session_id, + ) + + # We have no guarantee that all the devices of that session are for the same + # `user_id`. Hence, we have to iterate over the list of devices and log them out + # one by one. + for device in devices: + user_id = device["user_id"] + device_id = device["device_id"] + + # If the user_id associated with that device/session is not the one we got + # out of the `sub` claim, skip that device and show log an error. + if expected_user_id is not None and user_id != expected_user_id: + logger.error( + "Received a logout notification from SSO provider " + f"{auth_provider_id!r} for the user {expected_user_id!r}, but with " + f"a session ID ({auth_provider_session_id!r}) which belongs to " + f"{user_id!r}. This may happen when the SSO provider user mapper " + "uses something else than the standard attribute as mapping ID. " + "For OIDC providers, set `backchannel_logout_ignore_sub` to `true` " + "in the provider config if that is the case." + ) + continue + + logger.info( + "Logging out %r (device %r) via SSO (%r) logout notification (session %r).", + user_id, + device_id, + auth_provider_id, + auth_provider_session_id, + ) + await self._device_handler.delete_devices(user_id, [device_id]) + def get_username_mapping_session_cookie_from_request(request: IRequest) -> str: """Extract the session ID from the cookie diff --git a/synapse/handlers/sync.py b/synapse/handlers/sync.py index 1db5d68021..7d6a653747 100644 --- a/synapse/handlers/sync.py +++ b/synapse/handlers/sync.py @@ -31,18 +31,24 @@ from typing import ( import attr from prometheus_client import Counter -from synapse.api.constants import EventTypes, Membership +from synapse.api.constants import EventContentFields, EventTypes, Membership from synapse.api.filtering import FilterCollection from synapse.api.presence import UserPresenceState from synapse.api.room_versions import KNOWN_ROOM_VERSIONS from synapse.events import EventBase from synapse.handlers.relations import BundledAggregations from synapse.logging.context import current_context -from synapse.logging.opentracing import SynapseTags, log_kv, set_tag, start_active_span +from synapse.logging.opentracing import ( + SynapseTags, + log_kv, + set_tag, + start_active_span, + trace, +) from synapse.push.clientformat import format_push_rules_for_user from synapse.storage.databases.main.event_push_actions import RoomNotifCounts +from synapse.storage.databases.main.roommember import extract_heroes_from_room_summary from synapse.storage.roommember import MemberSummary -from synapse.storage.state import StateFilter from synapse.types import ( DeviceListUpdates, JsonDict, @@ -54,6 +60,7 @@ from synapse.types import ( StreamToken, UserID, ) +from synapse.types.state import StateFilter from synapse.util.async_helpers import concurrently_execute from synapse.util.caches.expiringcache import ExpiringCache from synapse.util.caches.lrucache import LruCache @@ -805,18 +812,6 @@ class SyncHandler: if canonical_alias and canonical_alias.content.get("alias"): return summary - me = sync_config.user.to_string() - - joined_user_ids = [ - r[0] for r in details.get(Membership.JOIN, empty_ms).members if r[0] != me - ] - invited_user_ids = [ - r[0] for r in details.get(Membership.INVITE, empty_ms).members if r[0] != me - ] - gone_user_ids = [ - r[0] for r in details.get(Membership.LEAVE, empty_ms).members if r[0] != me - ] + [r[0] for r in details.get(Membership.BAN, empty_ms).members if r[0] != me] - # FIXME: only build up a member_ids list for our heroes member_ids = {} for membership in ( @@ -828,11 +823,8 @@ class SyncHandler: for user_id, event_id in details.get(membership, empty_ms).members: member_ids[user_id] = event_id - # FIXME: order by stream ordering rather than as returned by SQL - if joined_user_ids or invited_user_ids: - summary["m.heroes"] = sorted(joined_user_ids + invited_user_ids)[0:5] - else: - summary["m.heroes"] = sorted(gone_user_ids)[0:5] + me = sync_config.user.to_string() + summary["m.heroes"] = extract_heroes_from_room_summary(details, me) if not sync_config.filter_collection.lazy_load_members(): return summary @@ -1440,14 +1432,14 @@ class SyncHandler: logger.debug("Fetching OTK data") device_id = sync_config.device_id - one_time_key_counts: JsonDict = {} + one_time_keys_count: JsonDict = {} unused_fallback_key_types: List[str] = [] if device_id: # TODO: We should have a way to let clients differentiate between the states of: # * no change in OTK count since the provided since token # * the server has zero OTKs left for this device # Spec issue: https://github.com/matrix-org/matrix-doc/issues/3298 - one_time_key_counts = await self.store.count_e2e_one_time_keys( + one_time_keys_count = await self.store.count_e2e_one_time_keys( user_id, device_id ) unused_fallback_key_types = ( @@ -1477,7 +1469,7 @@ class SyncHandler: archived=sync_result_builder.archived, to_device=sync_result_builder.to_device, device_lists=device_lists, - device_one_time_keys_count=one_time_key_counts, + device_one_time_keys_count=one_time_keys_count, device_unused_fallback_key_types=unused_fallback_key_types, next_batch=sync_result_builder.now_token, ) @@ -1542,10 +1534,12 @@ class SyncHandler: # # If we don't have that info cached then we get all the users that # share a room with our user and check if those users have changed. - changed_users = self.store.get_cached_device_list_changes( + cache_result = self.store.get_cached_device_list_changes( since_token.device_list_key ) - if changed_users is not None: + if cache_result.hit: + changed_users = cache_result.entities + result = await self.store.get_rooms_for_users(changed_users) for changed_user_id, entries in result.items(): @@ -1598,6 +1592,7 @@ class SyncHandler: else: return DeviceListUpdates() + @trace async def _generate_sync_entry_for_to_device( self, sync_result_builder: "SyncResultBuilder" ) -> None: @@ -1617,11 +1612,16 @@ class SyncHandler: ) for message in messages: - # We pop here as we shouldn't be sending the message ID down - # `/sync` - message_id = message.pop("message_id", None) - if message_id: - set_tag(SynapseTags.TO_DEVICE_MESSAGE_ID, message_id) + log_kv( + { + "event": "to_device_message", + "sender": message["sender"], + "type": message["type"], + EventContentFields.TO_DEVICE_MSGID: message["content"].get( + EventContentFields.TO_DEVICE_MSGID + ), + } + ) logger.debug( "Returning %d to-device messages between %d and %d (current token: %d)", diff --git a/synapse/handlers/typing.py b/synapse/handlers/typing.py index f953691669..3f656ea4f5 100644 --- a/synapse/handlers/typing.py +++ b/synapse/handlers/typing.py @@ -420,11 +420,11 @@ class TypingWriterHandler(FollowerTypingHandler): if last_id == current_id: return [], current_id, False - changed_rooms: Optional[ - Iterable[str] - ] = self._typing_stream_change_cache.get_all_entities_changed(last_id) + result = self._typing_stream_change_cache.get_all_entities_changed(last_id) - if changed_rooms is None: + if result.hit: + changed_rooms: Iterable[str] = result.entities + else: changed_rooms = self._room_serials rows = [] @@ -513,7 +513,7 @@ class TypingNotificationEventSource(EventSource[int, JsonDict]): self, user: UserID, from_key: int, - limit: Optional[int], + limit: int, room_ids: Iterable[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/http/additional_resource.py b/synapse/http/additional_resource.py index 6a9f6635d2..8729630581 100644 --- a/synapse/http/additional_resource.py +++ b/synapse/http/additional_resource.py @@ -45,8 +45,7 @@ class AdditionalResource(DirectServeJsonResource): Args: hs: homeserver - handler ((twisted.web.server.Request) -> twisted.internet.defer.Deferred): - function to be called to handle the request. + handler: function to be called to handle the request. """ super().__init__() self._handler = handler diff --git a/synapse/http/client.py b/synapse/http/client.py index 084d0a5b84..4eb740c040 100644 --- a/synapse/http/client.py +++ b/synapse/http/client.py @@ -25,7 +25,6 @@ from typing import ( List, Mapping, Optional, - Sequence, Tuple, Union, ) @@ -90,14 +89,29 @@ incoming_responses_counter = Counter( "synapse_http_client_responses", "", ["method", "code"] ) -# the type of the headers list, to be passed to the t.w.h.Headers. -# Actually we can mix str and bytes keys, but Mapping treats 'key' as invariant so -# we simplify. +# the type of the headers map, to be passed to the t.w.h.Headers. +# +# The actual type accepted by Twisted is +# Mapping[Union[str, bytes], Sequence[Union[str, bytes]] , +# allowing us to mix and match str and bytes freely. However: any str is also a +# Sequence[str]; passing a header string value which is a +# standalone str is interpreted as a sequence of 1-codepoint strings. This is a disastrous footgun. +# We use a narrower value type (RawHeaderValue) to avoid this footgun. +# +# We also simplify the keys to be either all str or all bytes. This helps because +# Dict[K, V] is invariant in K (and indeed V). RawHeaders = Union[Mapping[str, "RawHeaderValue"], Mapping[bytes, "RawHeaderValue"]] # the value actually has to be a List, but List is invariant so we can't specify that # the entries can either be Lists or bytes. -RawHeaderValue = Sequence[Union[str, bytes]] +RawHeaderValue = Union[ + List[str], + List[bytes], + List[Union[str, bytes]], + Tuple[str, ...], + Tuple[bytes, ...], + Tuple[Union[str, bytes], ...], +] def check_against_blacklist( diff --git a/synapse/http/federation/matrix_federation_agent.py b/synapse/http/federation/matrix_federation_agent.py index 2f0177f1e2..0359231e7d 100644 --- a/synapse/http/federation/matrix_federation_agent.py +++ b/synapse/http/federation/matrix_federation_agent.py @@ -155,11 +155,10 @@ class MatrixFederationAgent: a file for a file upload). Or None if the request is to have no body. Returns: - Deferred[twisted.web.iweb.IResponse]: - fires when the header of the response has been received (regardless of the - response status code). Fails if there is any problem which prevents that - response from being received (including problems that prevent the request - from being sent). + A deferred which fires when the header of the response has been received + (regardless of the response status code). Fails if there is any problem + which prevents that response from being received (including problems that + prevent the request from being sent). """ # We use urlparse as that will set `port` to None if there is no # explicit port. diff --git a/synapse/http/matrixfederationclient.py b/synapse/http/matrixfederationclient.py index 3c35b1d2c7..b92f1d3d1a 100644 --- a/synapse/http/matrixfederationclient.py +++ b/synapse/http/matrixfederationclient.py @@ -951,8 +951,7 @@ class MatrixFederationHttpClient: args: query params Returns: - dict|list: Succeeds when we get a 2xx HTTP response. The - result will be the decoded JSON body. + Succeeds when we get a 2xx HTTP response. The result will be the decoded JSON body. Raises: HttpResponseException: If we get an HTTP response code >= 300 diff --git a/synapse/http/proxyagent.py b/synapse/http/proxyagent.py index 1f8227896f..18899bc6d1 100644 --- a/synapse/http/proxyagent.py +++ b/synapse/http/proxyagent.py @@ -34,7 +34,7 @@ from twisted.web.client import ( ) from twisted.web.error import SchemeNotSupported from twisted.web.http_headers import Headers -from twisted.web.iweb import IAgent, IBodyProducer, IPolicyForHTTPS +from twisted.web.iweb import IAgent, IBodyProducer, IPolicyForHTTPS, IResponse from synapse.http import redact_uri from synapse.http.connectproxyclient import HTTPConnectProxyEndpoint, ProxyCredentials @@ -134,7 +134,7 @@ class ProxyAgent(_AgentBase): uri: bytes, headers: Optional[Headers] = None, bodyProducer: Optional[IBodyProducer] = None, - ) -> defer.Deferred: + ) -> "defer.Deferred[IResponse]": """ Issue a request to the server indicated by the given uri. @@ -157,17 +157,17 @@ class ProxyAgent(_AgentBase): a file upload). Or, None if the request is to have no body. Returns: - Deferred[IResponse]: completes when the header of the response has - been received (regardless of the response status code). + A deferred which completes when the header of the response has + been received (regardless of the response status code). - Can fail with: - SchemeNotSupported: if the uri is not http or https + Can fail with: + SchemeNotSupported: if the uri is not http or https - twisted.internet.error.TimeoutError if the server we are connecting - to (proxy or destination) does not accept a connection before - connectTimeout. + twisted.internet.error.TimeoutError if the server we are connecting + to (proxy or destination) does not accept a connection before + connectTimeout. - ... other things too. + ... other things too. """ uri = uri.strip() if not _VALID_URI.match(uri): diff --git a/synapse/http/server.py b/synapse/http/server.py index bcbfac2c9f..2563858f3c 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -19,6 +19,7 @@ import logging import types import urllib from http import HTTPStatus +from http.client import FOUND from inspect import isawaitable from typing import ( TYPE_CHECKING, @@ -266,7 +267,7 @@ class HttpServer(Protocol): request. The first argument will be the request object and subsequent arguments will be any matched groups from the regex. This should return either tuple of (code, response), or None. - servlet_classname (str): The name of the handler to be used in prometheus + servlet_classname: The name of the handler to be used in prometheus and opentracing logs. """ @@ -339,7 +340,7 @@ class _AsyncResource(resource.Resource, metaclass=abc.ABCMeta): return callback_return - _unrecognised_request_handler(request) + return _unrecognised_request_handler(request) @abc.abstractmethod def _send_response( @@ -576,7 +577,24 @@ def _unrecognised_request_handler(request: Request) -> NoReturn: Args: request: Unused, but passed in to match the signature of ServletCallback. """ - raise UnrecognizedRequestError() + raise UnrecognizedRequestError(code=404) + + +class UnrecognizedRequestResource(resource.Resource): + """ + Similar to twisted.web.resource.NoResource, but returns a JSON 404 with an + errcode of M_UNRECOGNIZED. + """ + + def render(self, request: SynapseRequest) -> int: + f = failure.Failure(UnrecognizedRequestError(code=404)) + return_json_error(f, request, None) + # A response has already been sent but Twisted requires either NOT_DONE_YET + # or the response bytes as a return value. + return NOT_DONE_YET + + def getChild(self, name: str, request: Request) -> resource.Resource: + return self class RootRedirect(resource.Resource): @@ -598,7 +616,7 @@ class RootRedirect(resource.Resource): class OptionsResource(resource.Resource): """Responds to OPTION requests for itself and all children.""" - def render_OPTIONS(self, request: Request) -> bytes: + def render_OPTIONS(self, request: SynapseRequest) -> bytes: request.setResponseCode(204) request.setHeader(b"Content-Length", b"0") @@ -763,7 +781,7 @@ def respond_with_json( def respond_with_json_bytes( - request: Request, + request: SynapseRequest, code: int, json_bytes: bytes, send_cors: bool = False, @@ -859,7 +877,7 @@ def _write_bytes_to_request(request: Request, bytes_to_write: bytes) -> None: _ByteProducer(request, bytes_generator) -def set_cors_headers(request: Request) -> None: +def set_cors_headers(request: SynapseRequest) -> None: """Set the CORS headers so that javascript running in a web browsers can use this API @@ -870,10 +888,20 @@ def set_cors_headers(request: Request) -> None: request.setHeader( b"Access-Control-Allow-Methods", b"GET, HEAD, POST, PUT, DELETE, OPTIONS" ) - request.setHeader( - b"Access-Control-Allow-Headers", - b"X-Requested-With, Content-Type, Authorization, Date", - ) + if request.experimental_cors_msc3886: + request.setHeader( + b"Access-Control-Allow-Headers", + b"X-Requested-With, Content-Type, Authorization, Date, If-Match, If-None-Match", + ) + request.setHeader( + b"Access-Control-Expose-Headers", + b"ETag, Location, X-Max-Bytes", + ) + else: + request.setHeader( + b"Access-Control-Allow-Headers", + b"X-Requested-With, Content-Type, Authorization, Date", + ) def set_corp_headers(request: Request) -> None: @@ -942,10 +970,25 @@ def set_clickjacking_protection_headers(request: Request) -> None: request.setHeader(b"Content-Security-Policy", b"frame-ancestors 'none';") -def respond_with_redirect(request: Request, url: bytes) -> None: - """Write a 302 response to the request, if it is still alive.""" +def respond_with_redirect( + request: SynapseRequest, url: bytes, statusCode: int = FOUND, cors: bool = False +) -> None: + """ + Write a 302 (or other specified status code) response to the request, if it is still alive. + + Args: + request: The http request to respond to. + url: The URL to redirect to. + statusCode: The HTTP status code to use for the redirect (defaults to 302). + cors: Whether to set CORS headers on the response. + """ logger.debug("Redirect to %s", url.decode("utf-8")) - request.redirect(url) + + if cors: + set_cors_headers(request) + + request.setResponseCode(statusCode) + request.setHeader(b"location", url) finish_request(request) diff --git a/synapse/http/servlet.py b/synapse/http/servlet.py index 80acbdcf3c..dead02cd5c 100644 --- a/synapse/http/servlet.py +++ b/synapse/http/servlet.py @@ -35,6 +35,7 @@ from typing_extensions import Literal from twisted.web.server import Request from synapse.api.errors import Codes, SynapseError +from synapse.http import redact_uri from synapse.http.server import HttpServer from synapse.types import JsonDict, RoomAlias, RoomID from synapse.util import json_decoder @@ -664,7 +665,13 @@ def parse_json_value_from_request( try: content = json_decoder.decode(content_bytes.decode("utf-8")) except Exception as e: - logger.warning("Unable to parse JSON: %s (%s)", e, content_bytes) + logger.warning( + "Unable to parse JSON from %s %s response: %s (%s)", + request.method.decode("ascii", errors="replace"), + redact_uri(request.uri.decode("ascii", errors="replace")), + e, + content_bytes, + ) raise SynapseError( HTTPStatus.BAD_REQUEST, "Content not JSON.", errcode=Codes.NOT_JSON ) diff --git a/synapse/http/site.py b/synapse/http/site.py index 55a6afce35..6a1dbf7f33 100644 --- a/synapse/http/site.py +++ b/synapse/http/site.py @@ -82,6 +82,7 @@ class SynapseRequest(Request): self.reactor = site.reactor self._channel = channel # this is used by the tests self.start_time = 0.0 + self.experimental_cors_msc3886 = site.experimental_cors_msc3886 # The requester, if authenticated. For federation requests this is the # server name, for client requests this is the Requester object. @@ -399,7 +400,7 @@ class SynapseRequest(Request): be sure to call finished_processing. Args: - servlet_name (str): the name of the servlet which will be + servlet_name: the name of the servlet which will be processing this request. This is used in the metrics. It is possible to update this afterwards by updating @@ -622,6 +623,8 @@ class SynapseSite(Site): request_id_header = config.http_options.request_id_header + self.experimental_cors_msc3886 = config.http_options.experimental_cors_msc3886 + def request_factory(channel: HTTPChannel, queued: bool) -> Request: return request_class( channel, diff --git a/synapse/logging/context.py b/synapse/logging/context.py index 6a08ffed64..f62bea968f 100644 --- a/synapse/logging/context.py +++ b/synapse/logging/context.py @@ -117,8 +117,7 @@ class ContextResourceUsage: """Create a new ContextResourceUsage Args: - copy_from (ContextResourceUsage|None): if not None, an object to - copy stats from + copy_from: if not None, an object to copy stats from """ if copy_from is None: self.reset() @@ -162,7 +161,7 @@ class ContextResourceUsage: """Add another ContextResourceUsage's stats to this one's. Args: - other (ContextResourceUsage): the other resource usage object + other: the other resource usage object """ self.ru_utime += other.ru_utime self.ru_stime += other.ru_stime @@ -342,7 +341,7 @@ class LoggingContext: called directly. Returns: - LoggingContext: the current logging context + The current logging context """ warnings.warn( "synapse.logging.context.LoggingContext.current_context() is deprecated " @@ -362,7 +361,8 @@ class LoggingContext: called directly. Args: - context(LoggingContext): The context to activate. + context: The context to activate. + Returns: The context that was previously active """ @@ -474,8 +474,7 @@ class LoggingContext: """Get resources used by this logcontext so far. Returns: - ContextResourceUsage: a *copy* of the object tracking resource - usage so far + A *copy* of the object tracking resource usage so far """ # we always return a copy, for consistency res = self._resource_usage.copy() @@ -663,7 +662,8 @@ def current_context() -> LoggingContextOrSentinel: def set_current_context(context: LoggingContextOrSentinel) -> LoggingContextOrSentinel: """Set the current logging context in thread local storage Args: - context(LoggingContext): The context to activate. + context: The context to activate. + Returns: The context that was previously active """ @@ -700,7 +700,7 @@ def nested_logging_context(suffix: str) -> LoggingContext: suffix: suffix to add to the parent context's 'name'. Returns: - LoggingContext: new logging context. + A new logging context. """ curr_context = current_context() if not curr_context: @@ -898,20 +898,19 @@ def defer_to_thread( on it. Args: - reactor (twisted.internet.base.ReactorBase): The reactor in whose main thread - the Deferred will be invoked, and whose threadpool we should use for the - function. + reactor: The reactor in whose main thread the Deferred will be invoked, + and whose threadpool we should use for the function. Normally this will be hs.get_reactor(). - f (callable): The function to call. + f: The function to call. args: positional arguments to pass to f. kwargs: keyword arguments to pass to f. Returns: - Deferred: A Deferred which fires a callback with the result of `f`, or an + A Deferred which fires a callback with the result of `f`, or an errback if `f` throws an exception. """ return defer_to_threadpool(reactor, reactor.getThreadPool(), f, *args, **kwargs) @@ -939,20 +938,20 @@ def defer_to_threadpool( on it. Args: - reactor (twisted.internet.base.ReactorBase): The reactor in whose main thread - the Deferred will be invoked. Normally this will be hs.get_reactor(). + reactor: The reactor in whose main thread the Deferred will be invoked. + Normally this will be hs.get_reactor(). - threadpool (twisted.python.threadpool.ThreadPool): The threadpool to use for - running `f`. Normally this will be hs.get_reactor().getThreadPool(). + threadpool: The threadpool to use for running `f`. Normally this will be + hs.get_reactor().getThreadPool(). - f (callable): The function to call. + f: The function to call. args: positional arguments to pass to f. kwargs: keyword arguments to pass to f. Returns: - Deferred: A Deferred which fires a callback with the result of `f`, or an + A Deferred which fires a callback with the result of `f`, or an errback if `f` throws an exception. """ curr_context = current_context() diff --git a/synapse/logging/opentracing.py b/synapse/logging/opentracing.py index 8ce5a2a338..a705af8356 100644 --- a/synapse/logging/opentracing.py +++ b/synapse/logging/opentracing.py @@ -292,8 +292,15 @@ logger = logging.getLogger(__name__) class SynapseTags: - # The message ID of any to_device message processed - TO_DEVICE_MESSAGE_ID = "to_device.message_id" + # The message ID of any to_device EDU processed + TO_DEVICE_EDU_ID = "to_device.edu_id" + + # Details about to-device messages + TO_DEVICE_TYPE = "to_device.type" + TO_DEVICE_SENDER = "to_device.sender" + TO_DEVICE_RECIPIENT = "to_device.recipient" + TO_DEVICE_RECIPIENT_DEVICE = "to_device.recipient_device" + TO_DEVICE_MSGID = "to_device.msgid" # client-generated ID # Whether the sync response has new data to be returned to the client. SYNC_RESULT = "sync.new_data" @@ -721,7 +728,7 @@ def inject_header_dict( destination: address of entity receiving the span context. Must be given unless check_destination is False. The context will only be injected if the destination matches the opentracing whitelist - check_destination (bool): If false, destination will be ignored and the context + check_destination: If false, destination will be ignored and the context will always be injected. Note: @@ -780,7 +787,7 @@ def get_active_span_text_map(destination: Optional[str] = None) -> Dict[str, str destination: the name of the remote server. Returns: - dict: the active span's context if opentracing is enabled, otherwise empty. + the active span's context if opentracing is enabled, otherwise empty. """ if destination and not whitelisted_homeserver(destination): diff --git a/synapse/metrics/__init__.py b/synapse/metrics/__init__.py index c3d3daf877..b01372565d 100644 --- a/synapse/metrics/__init__.py +++ b/synapse/metrics/__init__.py @@ -47,11 +47,7 @@ from twisted.python.threadpool import ThreadPool # This module is imported for its side effects; flake8 needn't warn that it's unused. import synapse.metrics._reactor_metrics # noqa: F401 from synapse.metrics._gc import MIN_TIME_BETWEEN_GCS, install_gc_manager -from synapse.metrics._legacy_exposition import ( - MetricsResource, - generate_latest, - start_http_server, -) +from synapse.metrics._twisted_exposition import MetricsResource, generate_latest from synapse.metrics._types import Collector from synapse.util import SYNAPSE_VERSION @@ -474,7 +470,6 @@ __all__ = [ "Collector", "MetricsResource", "generate_latest", - "start_http_server", "LaterGauge", "InFlightGauge", "GaugeBucketCollector", diff --git a/synapse/metrics/_legacy_exposition.py b/synapse/metrics/_legacy_exposition.py deleted file mode 100644 index 1459f9d224..0000000000 --- a/synapse/metrics/_legacy_exposition.py +++ /dev/null @@ -1,288 +0,0 @@ -# Copyright 2015-2019 Prometheus Python Client Developers -# Copyright 2019 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. - -""" -This code is based off `prometheus_client/exposition.py` from version 0.7.1. - -Due to the renaming of metrics in prometheus_client 0.4.0, this customised -vendoring of the code will emit both the old versions that Synapse dashboards -expect, and the newer "best practice" version of the up-to-date official client. -""" -import logging -import math -import threading -from http.server import BaseHTTPRequestHandler, HTTPServer -from socketserver import ThreadingMixIn -from typing import Any, Dict, List, Type, Union -from urllib.parse import parse_qs, urlparse - -from prometheus_client import REGISTRY, CollectorRegistry -from prometheus_client.core import Sample - -from twisted.web.resource import Resource -from twisted.web.server import Request - -logger = logging.getLogger(__name__) -CONTENT_TYPE_LATEST = "text/plain; version=0.0.4; charset=utf-8" - - -def floatToGoString(d: Union[int, float]) -> str: - d = float(d) - if d == math.inf: - return "+Inf" - elif d == -math.inf: - return "-Inf" - elif math.isnan(d): - return "NaN" - else: - s = repr(d) - dot = s.find(".") - # Go switches to exponents sooner than Python. - # We only need to care about positive values for le/quantile. - if d > 0 and dot > 6: - mantissa = f"{s[0]}.{s[1:dot]}{s[dot + 1 :]}".rstrip("0.") - return f"{mantissa}e+0{dot - 1}" - return s - - -def sample_line(line: Sample, name: str) -> str: - if line.labels: - labelstr = "{{{0}}}".format( - ",".join( - [ - '{}="{}"'.format( - k, - v.replace("\\", r"\\").replace("\n", r"\n").replace('"', r"\""), - ) - for k, v in sorted(line.labels.items()) - ] - ) - ) - else: - labelstr = "" - timestamp = "" - if line.timestamp is not None: - # Convert to milliseconds. - timestamp = f" {int(float(line.timestamp) * 1000):d}" - return "{}{} {}{}\n".format(name, labelstr, floatToGoString(line.value), timestamp) - - -# Mapping from new metric names to legacy metric names. -# We translate these back to their old names when exposing them through our -# legacy vendored exporter. -# Only this legacy exposition module applies these name changes. -LEGACY_METRIC_NAMES = { - "synapse_util_caches_cache_hits": "synapse_util_caches_cache:hits", - "synapse_util_caches_cache_size": "synapse_util_caches_cache:size", - "synapse_util_caches_cache_evicted_size": "synapse_util_caches_cache:evicted_size", - "synapse_util_caches_cache": "synapse_util_caches_cache:total", - "synapse_util_caches_response_cache_size": "synapse_util_caches_response_cache:size", - "synapse_util_caches_response_cache_hits": "synapse_util_caches_response_cache:hits", - "synapse_util_caches_response_cache_evicted_size": "synapse_util_caches_response_cache:evicted_size", - "synapse_util_caches_response_cache": "synapse_util_caches_response_cache:total", - "synapse_federation_client_sent_pdu_destinations": "synapse_federation_client_sent_pdu_destinations:total", - "synapse_federation_client_sent_pdu_destinations_count": "synapse_federation_client_sent_pdu_destinations:count", - "synapse_admin_mau_current": "synapse_admin_mau:current", - "synapse_admin_mau_max": "synapse_admin_mau:max", - "synapse_admin_mau_registered_reserved_users": "synapse_admin_mau:registered_reserved_users", -} - - -def generate_latest(registry: CollectorRegistry, emit_help: bool = False) -> bytes: - """ - Generate metrics in legacy format. Modern metrics are generated directly - by prometheus-client. - """ - - output = [] - - for metric in registry.collect(): - if not metric.samples: - # No samples, don't bother. - continue - - # Translate to legacy metric name if it has one. - mname = LEGACY_METRIC_NAMES.get(metric.name, metric.name) - mnewname = metric.name - mtype = metric.type - - # OpenMetrics -> Prometheus - if mtype == "counter": - mnewname = mnewname + "_total" - elif mtype == "info": - mtype = "gauge" - mnewname = mnewname + "_info" - elif mtype == "stateset": - mtype = "gauge" - elif mtype == "gaugehistogram": - mtype = "histogram" - elif mtype == "unknown": - mtype = "untyped" - - # Output in the old format for compatibility. - if emit_help: - output.append( - "# HELP {} {}\n".format( - mname, - metric.documentation.replace("\\", r"\\").replace("\n", r"\n"), - ) - ) - output.append(f"# TYPE {mname} {mtype}\n") - - om_samples: Dict[str, List[str]] = {} - for s in metric.samples: - for suffix in ["_created", "_gsum", "_gcount"]: - if s.name == mname + suffix: - # OpenMetrics specific sample, put in a gauge at the end. - # (these come from gaugehistograms which don't get renamed, - # so no need to faff with mnewname) - om_samples.setdefault(suffix, []).append(sample_line(s, s.name)) - break - else: - newname = s.name.replace(mnewname, mname) - if ":" in newname and newname.endswith("_total"): - newname = newname[: -len("_total")] - output.append(sample_line(s, newname)) - - for suffix, lines in sorted(om_samples.items()): - if emit_help: - output.append( - "# HELP {}{} {}\n".format( - mname, - suffix, - metric.documentation.replace("\\", r"\\").replace("\n", r"\n"), - ) - ) - output.append(f"# TYPE {mname}{suffix} gauge\n") - output.extend(lines) - - # Get rid of the weird colon things while we're at it - if mtype == "counter": - mnewname = mnewname.replace(":total", "") - mnewname = mnewname.replace(":", "_") - - if mname == mnewname: - continue - - # Also output in the new format, if it's different. - if emit_help: - output.append( - "# HELP {} {}\n".format( - mnewname, - metric.documentation.replace("\\", r"\\").replace("\n", r"\n"), - ) - ) - output.append(f"# TYPE {mnewname} {mtype}\n") - - for s in metric.samples: - # Get rid of the OpenMetrics specific samples (we should already have - # dealt with them above anyway.) - for suffix in ["_created", "_gsum", "_gcount"]: - if s.name == mname + suffix: - break - else: - sample_name = LEGACY_METRIC_NAMES.get(s.name, s.name) - output.append( - sample_line(s, sample_name.replace(":total", "").replace(":", "_")) - ) - - return "".join(output).encode("utf-8") - - -class MetricsHandler(BaseHTTPRequestHandler): - """HTTP handler that gives metrics from ``REGISTRY``.""" - - registry = REGISTRY - - def do_GET(self) -> None: - registry = self.registry - params = parse_qs(urlparse(self.path).query) - - if "help" in params: - emit_help = True - else: - emit_help = False - - try: - output = generate_latest(registry, emit_help=emit_help) - except Exception: - self.send_error(500, "error generating metric output") - raise - try: - self.send_response(200) - self.send_header("Content-Type", CONTENT_TYPE_LATEST) - self.send_header("Content-Length", str(len(output))) - self.end_headers() - self.wfile.write(output) - except BrokenPipeError as e: - logger.warning( - "BrokenPipeError when serving metrics (%s). Did Prometheus restart?", e - ) - - def log_message(self, format: str, *args: Any) -> None: - """Log nothing.""" - - @classmethod - def factory(cls, registry: CollectorRegistry) -> Type: - """Returns a dynamic MetricsHandler class tied - to the passed registry. - """ - # This implementation relies on MetricsHandler.registry - # (defined above and defaulted to REGISTRY). - - # As we have unicode_literals, we need to create a str() - # object for type(). - cls_name = str(cls.__name__) - MyMetricsHandler = type(cls_name, (cls, object), {"registry": registry}) - return MyMetricsHandler - - -class _ThreadingSimpleServer(ThreadingMixIn, HTTPServer): - """Thread per request HTTP server.""" - - # Make worker threads "fire and forget". Beginning with Python 3.7 this - # prevents a memory leak because ``ThreadingMixIn`` starts to gather all - # non-daemon threads in a list in order to join on them at server close. - # Enabling daemon threads virtually makes ``_ThreadingSimpleServer`` the - # same as Python 3.7's ``ThreadingHTTPServer``. - daemon_threads = True - - -def start_http_server( - port: int, addr: str = "", registry: CollectorRegistry = REGISTRY -) -> None: - """Starts an HTTP server for prometheus metrics as a daemon thread""" - CustomMetricsHandler = MetricsHandler.factory(registry) - httpd = _ThreadingSimpleServer((addr, port), CustomMetricsHandler) - t = threading.Thread(target=httpd.serve_forever) - t.daemon = True - t.start() - - -class MetricsResource(Resource): - """ - Twisted ``Resource`` that serves prometheus metrics. - """ - - isLeaf = True - - def __init__(self, registry: CollectorRegistry = REGISTRY): - self.registry = registry - - def render_GET(self, request: Request) -> bytes: - request.setHeader(b"Content-Type", CONTENT_TYPE_LATEST.encode("ascii")) - response = generate_latest(self.registry) - request.setHeader(b"Content-Length", str(len(response))) - return response diff --git a/synapse/metrics/_twisted_exposition.py b/synapse/metrics/_twisted_exposition.py new file mode 100644 index 0000000000..0abcd14953 --- /dev/null +++ b/synapse/metrics/_twisted_exposition.py @@ -0,0 +1,38 @@ +# Copyright 2015-2019 Prometheus Python Client Developers +# Copyright 2019 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. + +from prometheus_client import REGISTRY, CollectorRegistry, generate_latest + +from twisted.web.resource import Resource +from twisted.web.server import Request + +CONTENT_TYPE_LATEST = "text/plain; version=0.0.4; charset=utf-8" + + +class MetricsResource(Resource): + """ + Twisted ``Resource`` that serves prometheus metrics. + """ + + isLeaf = True + + def __init__(self, registry: CollectorRegistry = REGISTRY): + self.registry = registry + + def render_GET(self, request: Request) -> bytes: + request.setHeader(b"Content-Type", CONTENT_TYPE_LATEST.encode("ascii")) + response = generate_latest(self.registry) + request.setHeader(b"Content-Length", str(len(response))) + return response diff --git a/synapse/metrics/background_process_metrics.py b/synapse/metrics/background_process_metrics.py index 7a1516d3a8..9ea4e23b31 100644 --- a/synapse/metrics/background_process_metrics.py +++ b/synapse/metrics/background_process_metrics.py @@ -174,8 +174,10 @@ class _BackgroundProcess: diff = new_stats - self._reported_stats self._reported_stats = new_stats - _background_process_ru_utime.labels(self.desc).inc(diff.ru_utime) - _background_process_ru_stime.labels(self.desc).inc(diff.ru_stime) + # For unknown reasons, the difference in times can be negative. See comment in + # synapse.http.request_metrics.RequestMetrics.update_metrics. + _background_process_ru_utime.labels(self.desc).inc(max(diff.ru_utime, 0)) + _background_process_ru_stime.labels(self.desc).inc(max(diff.ru_stime, 0)) _background_process_db_txn_count.labels(self.desc).inc(diff.db_txn_count) _background_process_db_txn_duration.labels(self.desc).inc( diff.db_txn_duration_sec diff --git a/synapse/metrics/common_usage_metrics.py b/synapse/metrics/common_usage_metrics.py index 0a22ea3d92..6e05b043d3 100644 --- a/synapse/metrics/common_usage_metrics.py +++ b/synapse/metrics/common_usage_metrics.py @@ -54,7 +54,9 @@ class CommonUsageMetricsManager: async def setup(self) -> None: """Keep the gauges for common usage metrics up to date.""" - await self._update_gauges() + run_as_background_process( + desc="common_usage_metrics_update_gauges", func=self._update_gauges + ) self._clock.looping_call( run_as_background_process, 5 * 60 * 1000, diff --git a/synapse/module_api/__init__.py b/synapse/module_api/__init__.py index 6a6ae208d1..0092a03c59 100644 --- a/synapse/module_api/__init__.py +++ b/synapse/module_api/__init__.py @@ -86,6 +86,7 @@ from synapse.handlers.auth import ( ON_LOGGED_OUT_CALLBACK, AuthHandler, ) +from synapse.handlers.device import DeviceHandler from synapse.handlers.push_rules import RuleSpec, check_actions from synapse.http.client import SimpleHttpClient from synapse.http.server import ( @@ -110,7 +111,6 @@ from synapse.storage.background_updates import ( ) from synapse.storage.database import DatabasePool, LoggingTransaction from synapse.storage.databases.main.roommember import ProfileInfo -from synapse.storage.state import StateFilter from synapse.types import ( DomainSpecificString, JsonDict, @@ -123,6 +123,7 @@ from synapse.types import ( UserProfile, create_requester, ) +from synapse.types.state import StateFilter from synapse.util import Clock from synapse.util.async_helpers import maybe_awaitable from synapse.util.caches.descriptors import CachedFunction, cached @@ -207,6 +208,7 @@ class ModuleApi: self._registration_handler = hs.get_registration_handler() self._send_email_handler = hs.get_send_email_handler() self._push_rules_handler = hs.get_push_rules_handler() + self._device_handler = hs.get_device_handler() self.custom_template_dir = hs.config.server.custom_template_directory try: @@ -771,50 +773,11 @@ class ModuleApi: auth_provider_session_id: The session ID got during login from the SSO IdP, if any. """ - # The deprecated `generate_short_term_login_token` method defaulted to an empty - # string for the `auth_provider_id` because of how the underlying macaroon was - # generated. This will change to a proper NULL-able field when the tokens get - # moved to the database. - return self._hs.get_macaroon_generator().generate_short_term_login_token( + return await self._hs.get_auth_handler().create_login_token_for_user_id( user_id, - auth_provider_id or "", - auth_provider_session_id, duration_in_ms, - ) - - def generate_short_term_login_token( - self, - user_id: str, - duration_in_ms: int = (2 * 60 * 1000), - auth_provider_id: str = "", - auth_provider_session_id: Optional[str] = None, - ) -> str: - """Generate a login token suitable for m.login.token authentication - - Added in Synapse v1.9.0. - - This was deprecated in Synapse v1.69.0 in favor of create_login_token, and will - be removed in Synapse 1.71.0. - - Args: - user_id: gives the ID of the user that the token is for - - duration_in_ms: the time that the token will be valid for - - auth_provider_id: the ID of the SSO IdP that the user used to authenticate - to get this token, if any. This is encoded in the token so that - /login can report stats on number of successful logins by IdP. - """ - logger.warn( - "A module configured on this server uses ModuleApi.generate_short_term_login_token(), " - "which is deprecated in favor of ModuleApi.create_login_token(), and will be removed in " - "Synapse 1.71.0", - ) - return self._hs.get_macaroon_generator().generate_short_term_login_token( - user_id, auth_provider_id, auth_provider_session_id, - duration_in_ms, ) @defer.inlineCallbacks @@ -823,10 +786,12 @@ class ModuleApi: ) -> Generator["defer.Deferred[Any]", Any, None]: """Invalidate an access token for a user + Can only be called from the main process. + Added in Synapse v0.25.0. Args: - access_token(str): access token + access_token: access token Returns: twisted.internet.defer.Deferred - resolves once the access token @@ -835,6 +800,10 @@ class ModuleApi: Raises: synapse.api.errors.AuthError: the access token is invalid """ + assert isinstance( + self._device_handler, DeviceHandler + ), "invalidate_access_token can only be called on the main process" + # see if the access token corresponds to a device user_info = yield defer.ensureDeferred( self._auth.get_user_by_access_token(access_token) @@ -844,7 +813,7 @@ class ModuleApi: if device_id: # delete the device, which will also delete its access tokens yield defer.ensureDeferred( - self._hs.get_device_handler().delete_devices(user_id, [device_id]) + self._device_handler.delete_devices(user_id, [device_id]) ) else: # no associated device. Just delete the access token. @@ -871,7 +840,7 @@ class ModuleApi: **kwargs: named args to be passed to func Returns: - Deferred[object]: result of func + Result of func """ # type-ignore: See https://github.com/python/mypy/issues/8862 return defer.ensureDeferred( @@ -963,8 +932,7 @@ class ModuleApi: 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. + The filtered state events in the room. """ state_ids = yield defer.ensureDeferred( self._storage_controllers.state.get_current_state_ids( diff --git a/synapse/push/bulk_push_rule_evaluator.py b/synapse/push/bulk_push_rule_evaluator.py index eced182fd5..f27ba64d53 100644 --- a/synapse/push/bulk_push_rule_evaluator.py +++ b/synapse/push/bulk_push_rule_evaluator.py @@ -13,18 +13,15 @@ # See the License for the specific language governing permissions and # limitations under the License. -import itertools import logging from typing import ( TYPE_CHECKING, Any, Collection, Dict, - Iterable, List, Mapping, Optional, - Set, Tuple, Union, ) @@ -32,13 +29,14 @@ from typing import ( from prometheus_client import Counter from synapse.api.constants import MAIN_TIMELINE, EventTypes, Membership, RelationTypes +from synapse.api.room_versions import PushRuleRoomFlag, RoomVersion from synapse.event_auth import auth_types_for_event, get_user_power_level from synapse.events import EventBase, relation_from_event from synapse.events.snapshot import EventContext from synapse.state import POWER_KEY from synapse.storage.databases.main.roommember import EventIdMembership -from synapse.storage.state import StateFilter -from synapse.synapse_rust.push import FilteredPushRules, PushRule, PushRuleEvaluator +from synapse.synapse_rust.push import FilteredPushRules, PushRuleEvaluator +from synapse.types.state import StateFilter from synapse.util.caches import register_cache from synapse.util.metrics import measure_func from synapse.visibility import filter_event_for_clients_with_state @@ -48,7 +46,6 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) - push_rules_invalidation_counter = Counter( "synapse_push_bulk_push_rule_evaluator_push_rules_invalidation_counter", "" ) @@ -109,6 +106,9 @@ class BulkPushRuleEvaluator: self.store = hs.get_datastores().main self.clock = hs.get_clock() self._event_auth_handler = hs.get_event_auth_handler() + self.should_calculate_push_rules = self.hs.config.push.enable_push + + self._related_event_match_enabled = self.hs.config.experimental.msc3664_enabled self.room_push_rule_cache_metrics = register_cache( "cache", @@ -117,9 +117,6 @@ class BulkPushRuleEvaluator: resizable=False, ) - # Whether to support MSC3772 is supported. - self._relations_match_enabled = self.hs.config.experimental.msc3772_enabled - async def _get_rules_for_event( self, event: EventBase, @@ -171,8 +168,21 @@ class BulkPushRuleEvaluator: return rules_by_user async def _get_power_levels_and_sender_level( - self, event: EventBase, context: EventContext + self, + event: EventBase, + context: EventContext, + event_id_to_event: Mapping[str, EventBase], ) -> Tuple[dict, Optional[int]]: + """ + Given an event and an event context, get the power level event relevant to the event + and the power level of the sender of the event. + Args: + event: event to check + context: context of event to check + event_id_to_event: a mapping of event_id to event for a set of events being + batch persisted. This is needed as the sought-after power level event may + be in this batch rather than the DB + """ # There are no power levels and sender levels possible to get from outlier if event.internal_metadata.is_outlier(): return {}, None @@ -183,15 +193,26 @@ class BulkPushRuleEvaluator: ) pl_event_id = prev_state_ids.get(POWER_KEY) + # fastpath: if there's a power level event, that's all we need, and + # not having a power level event is an extreme edge case if pl_event_id: - # fastpath: if there's a power level event, that's all we need, and - # not having a power level event is an extreme edge case - auth_events = {POWER_KEY: await self.store.get_event(pl_event_id)} + # Get the power level event from the batch, or fall back to the database. + pl_event = event_id_to_event.get(pl_event_id) + if pl_event: + auth_events = {POWER_KEY: pl_event} + else: + auth_events = {POWER_KEY: await self.store.get_event(pl_event_id)} else: auth_events_ids = self._event_auth_handler.compute_auth_events( event, prev_state_ids, for_verification=False ) auth_events_dict = await self.store.get_events(auth_events_ids) + # Some needed auth events might be in the batch, combine them with those + # fetched from the database. + for auth_event_id in auth_events_ids: + auth_event = event_id_to_event.get(auth_event_id) + if auth_event: + auth_events_dict[auth_event_id] = auth_event auth_events = {(e.type, e.state_key): e for e in auth_events_dict.values()} sender_level = get_user_power_level(event.sender, auth_events) @@ -200,61 +221,82 @@ class BulkPushRuleEvaluator: return pl_event.content if pl_event else {}, sender_level - async def _get_mutual_relations( - self, parent_id: str, rules: Iterable[Tuple[PushRule, bool]] - ) -> Dict[str, Set[Tuple[str, str]]]: - """ - Fetch event metadata for events which related to the same event as the given event. - - If the given event has no relation information, returns an empty dictionary. - - Args: - parent_id: The event ID which is targeted by relations. - rules: The push rules which will be processed for this event. + async def _related_events(self, event: EventBase) -> Dict[str, Dict[str, str]]: + """Fetches the related events for 'event'. Sets the im.vector.is_falling_back key if the event is from a fallback relation Returns: - A dictionary of relation type to: - A set of tuples of: - The sender - The event type + Mapping of relation type to flattened events. """ + related_events: Dict[str, Dict[str, str]] = {} + if self._related_event_match_enabled: + related_event_id = event.content.get("m.relates_to", {}).get("event_id") + relation_type = event.content.get("m.relates_to", {}).get("rel_type") + if related_event_id is not None and relation_type is not None: + related_event = await self.store.get_event( + related_event_id, allow_none=True + ) + if related_event is not None: + related_events[relation_type] = _flatten_dict(related_event) + + reply_event_id = ( + event.content.get("m.relates_to", {}) + .get("m.in_reply_to", {}) + .get("event_id") + ) - # If the experimental feature is not enabled, skip fetching relations. - if not self._relations_match_enabled: - return {} + # convert replies to pseudo relations + if reply_event_id is not None: + related_event = await self.store.get_event( + reply_event_id, allow_none=True + ) - # Pre-filter to figure out which relation types are interesting. - rel_types = set() - for rule, enabled in rules: - if not enabled: - continue + if related_event is not None: + related_events["m.in_reply_to"] = _flatten_dict(related_event) - for condition in rule.conditions: - if condition["kind"] != "org.matrix.msc3772.relation_match": - continue + # indicate that this is from a fallback relation. + if relation_type == "m.thread" and event.content.get( + "m.relates_to", {} + ).get("is_falling_back", False): + related_events["m.in_reply_to"][ + "im.vector.is_falling_back" + ] = "" - # rel_type is required. - rel_type = condition.get("rel_type") - if rel_type: - rel_types.add(rel_type) + return related_events - # If no valid rules were found, no mutual relations. - if not rel_types: - return {} - - # If any valid rules were found, fetch the mutual relations. - return await self.store.get_mutual_event_relations(parent_id, rel_types) + async def action_for_events_by_user( + self, events_and_context: List[Tuple[EventBase, EventContext]] + ) -> None: + """Given a list of events and their associated contexts, evaluate the push rules + for each event, check if the message should increment the unread count, and + insert the results into the event_push_actions_staging table. + """ + if not self.should_calculate_push_rules: + return + # For batched events the power level events may not have been persisted yet, + # so we pass in the batched events. Thus if the event cannot be found in the + # database we can check in the batch. + event_id_to_event = {e.event_id: e for e, _ in events_and_context} + for event, context in events_and_context: + await self._action_for_event_by_user(event, context, event_id_to_event) @measure_func("action_for_event_by_user") - async def action_for_event_by_user( - self, event: EventBase, context: EventContext + async def _action_for_event_by_user( + self, + event: EventBase, + context: EventContext, + event_id_to_event: Mapping[str, EventBase], ) -> None: - """Given an event and context, evaluate the push rules, check if the message - should increment the unread count, and insert the results into the - event_push_actions_staging table. - """ - if not event.internal_metadata.is_notifiable(): - # Push rules for events that aren't notifiable can't be processed by this + + if ( + not event.internal_metadata.is_notifiable() + or event.internal_metadata.is_historical() + ): + # Push rules for events that aren't notifiable can't be processed by this and + # we want to skip push notification actions for historical messages + # because we don't want to notify people about old history back in time. + # The historical messages also do not have the proper `context.current_state_ids` + # and `state_groups` because they have `prev_events` that aren't persisted yet + # (historical messages persisted in reverse-chronological order). return # Disable counting as unread unless the experimental configuration is @@ -274,25 +316,24 @@ class BulkPushRuleEvaluator: ( power_levels, sender_power_level, - ) = await self._get_power_levels_and_sender_level(event, context) + ) = await self._get_power_levels_and_sender_level( + event, context, event_id_to_event + ) + # Find the event's thread ID. relation = relation_from_event(event) - # If the event does not have a relation, then cannot have any mutual - # relations or thread ID. - relations = {} + # If the event does not have a relation, then it cannot have a thread ID. thread_id = MAIN_TIMELINE if relation: - relations = await self._get_mutual_relations( - relation.parent_id, - itertools.chain(*(r.rules() for r in rules_by_user.values())), - ) # Recursively attempt to find the thread this event relates to. if relation.rel_type == RelationTypes.THREAD: thread_id = relation.parent_id else: # Since the event has not yet been persisted we check whether # the parent is part of a thread. - thread_id = await self.store.get_thread_id(relation.parent_id) or "main" + thread_id = await self.store.get_thread_id(relation.parent_id) + + related_events = await self._related_events(event) # It's possible that old room versions have non-integer power levels (floats or # strings). Workaround this by explicitly converting to int. @@ -302,12 +343,14 @@ class BulkPushRuleEvaluator: notification_levels[user_id] = int(level) evaluator = PushRuleEvaluator( - _flatten_dict(event), + _flatten_dict(event, room_version=event.room_version), room_member_count, sender_power_level, notification_levels, - relations, - self._relations_match_enabled, + related_events, + self._related_event_match_enabled, + event.room_version.msc3931_push_features, + self.hs.config.experimental.msc1767_enabled, # MSC3931 flag ) users = rules_by_user.keys() @@ -383,6 +426,7 @@ StateGroup = Union[object, int] def _flatten_dict( d: Union[EventBase, Mapping[str, Any]], + room_version: Optional[RoomVersion] = None, prefix: Optional[List[str]] = None, result: Optional[Dict[str, str]] = None, ) -> Dict[str, str]: @@ -394,6 +438,31 @@ def _flatten_dict( if isinstance(value, str): result[".".join(prefix + [key])] = value.lower() elif isinstance(value, Mapping): + # do not set `room_version` due to recursion considerations below _flatten_dict(value, prefix=(prefix + [key]), result=result) + # `room_version` should only ever be set when looking at the top level of an event + if ( + room_version is not None + and PushRuleRoomFlag.EXTENSIBLE_EVENTS in room_version.msc3931_push_features + and isinstance(d, EventBase) + ): + # Room supports extensible events: replace `content.body` with the plain text + # representation from `m.markup`, as per MSC1767. + markup = d.get("content").get("m.markup") + if room_version.identifier.startswith("org.matrix.msc1767."): + markup = d.get("content").get("org.matrix.msc1767.markup") + if markup is not None and isinstance(markup, list): + text = "" + for rep in markup: + if not isinstance(rep, dict): + # invalid markup - skip all processing + break + if rep.get("mimetype", "text/plain") == "text/plain": + rep_text = rep.get("body") + if rep_text is not None and isinstance(rep_text, str): + text = rep_text.lower() + break + result["content.body"] = text + return result diff --git a/synapse/push/clientformat.py b/synapse/push/clientformat.py index 7095ae83f9..622a1e35c5 100644 --- a/synapse/push/clientformat.py +++ b/synapse/push/clientformat.py @@ -44,6 +44,12 @@ def format_push_rules_for_user( rulearray.append(template_rule) + pattern_type = template_rule.pop("pattern_type", None) + if pattern_type == "user_id": + template_rule["pattern"] = user.to_string() + elif pattern_type == "user_localpart": + template_rule["pattern"] = user.localpart + template_rule["enabled"] = enabled if "conditions" not in template_rule: @@ -93,10 +99,14 @@ def _rule_to_template(rule: PushRule) -> Optional[Dict[str, Any]]: if len(rule.conditions) != 1: return None thecond = rule.conditions[0] - if "pattern" not in thecond: - return None + templaterule = {"actions": rule.actions} - templaterule["pattern"] = thecond["pattern"] + if "pattern" in thecond: + templaterule["pattern"] = thecond["pattern"] + elif "pattern_type" in thecond: + templaterule["pattern_type"] = thecond["pattern_type"] + else: + return None else: # This should not be reached unless this function is not kept in sync # with PRIORITY_CLASS_INVERSE_MAP. diff --git a/synapse/push/mailer.py b/synapse/push/mailer.py index c2575ba3d9..93b255ced5 100644 --- a/synapse/push/mailer.py +++ b/synapse/push/mailer.py @@ -37,8 +37,8 @@ from synapse.push.push_types import ( TemplateVars, ) from synapse.storage.databases.main.event_push_actions import EmailPushAction -from synapse.storage.state import StateFilter from synapse.types import StateMap, UserID +from synapse.types.state import StateFilter from synapse.util.async_helpers import concurrently_execute from synapse.visibility import filter_events_for_client diff --git a/synapse/push/push_tools.py b/synapse/push/push_tools.py index edeba27a45..7ee07e4bee 100644 --- a/synapse/push/push_tools.py +++ b/synapse/push/push_tools.py @@ -17,7 +17,6 @@ from synapse.events import EventBase from synapse.push.presentable_names import calculate_room_name, name_from_member_event from synapse.storage.controllers import StorageControllers from synapse.storage.databases.main import DataStore -from synapse.util.async_helpers import concurrently_execute async def get_badge_count(store: DataStore, user_id: str, group_by_room: bool) -> int: @@ -26,23 +25,12 @@ async def get_badge_count(store: DataStore, user_id: str, group_by_room: bool) - badge = len(invites) - room_notifs = [] - - async def get_room_unread_count(room_id: str) -> None: - room_notifs.append( - await store.get_unread_event_push_actions_by_room_for_user( - room_id, - user_id, - ) - ) - - await concurrently_execute(get_room_unread_count, joins, 10) - - for notifs in room_notifs: - # Combine the counts from all the threads. - notify_count = notifs.main_timeline.notify_count + sum( - n.notify_count for n in notifs.threads.values() - ) + room_to_count = await store.get_unread_counts_by_room_for_user(user_id) + for room_id, notify_count in room_to_count.items(): + # room_to_count may include rooms which the user has left, + # ignore those. + if room_id not in joins: + continue if notify_count == 0: continue @@ -51,8 +39,10 @@ async def get_badge_count(store: DataStore, user_id: str, group_by_room: bool) - # return one badge count per conversation badge += 1 else: - # increment the badge count by the number of unread messages in the room + # Increase badge by number of notifications in room + # NOTE: this includes threaded and unthreaded notifications. badge += notify_count + return badge diff --git a/synapse/replication/http/_base.py b/synapse/replication/http/_base.py index acb0bd18f7..3f4d3fc51a 100644 --- a/synapse/replication/http/_base.py +++ b/synapse/replication/http/_base.py @@ -153,7 +153,7 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta): argument list. Returns: - dict: If POST/PUT request then dictionary must be JSON serialisable, + If POST/PUT request then dictionary must be JSON serialisable, otherwise must be appropriate for adding as query args. """ return {} @@ -184,8 +184,10 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta): client = hs.get_simple_http_client() local_instance_name = hs.get_instance_name() + # The value of these option should match the replication listener settings master_host = hs.config.worker.worker_replication_host master_port = hs.config.worker.worker_replication_http_port + master_tls = hs.config.worker.worker_replication_http_tls instance_map = hs.config.worker.instance_map @@ -205,9 +207,11 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta): if instance_name == "master": host = master_host port = master_port + tls = master_tls elif instance_name in instance_map: host = instance_map[instance_name].host port = instance_map[instance_name].port + tls = instance_map[instance_name].tls else: raise Exception( "Instance %r not in 'instance_map' config" % (instance_name,) @@ -238,7 +242,11 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta): "Unknown METHOD on %s replication endpoint" % (cls.NAME,) ) - uri = "http://%s:%s/_synapse/replication/%s/%s" % ( + # Here the protocol is hard coded to be http by default or https in case the replication + # port is set to have tls true. + scheme = "https" if tls else "http" + uri = "%s://%s:%s/_synapse/replication/%s/%s" % ( + scheme, host, port, cls.NAME, diff --git a/synapse/replication/http/devices.py b/synapse/replication/http/devices.py index 3d63645726..7c4941c3d3 100644 --- a/synapse/replication/http/devices.py +++ b/synapse/replication/http/devices.py @@ -13,11 +13,12 @@ # limitations under the License. import logging -from typing import TYPE_CHECKING, Tuple +from typing import TYPE_CHECKING, Optional, Tuple from twisted.web.server import Request from synapse.http.server import HttpServer +from synapse.http.servlet import parse_json_object_from_request from synapse.replication.http._base import ReplicationEndpoint from synapse.types import JsonDict @@ -62,7 +63,12 @@ class ReplicationUserDevicesResyncRestServlet(ReplicationEndpoint): def __init__(self, hs: "HomeServer"): super().__init__(hs) - self.device_list_updater = hs.get_device_handler().device_list_updater + from synapse.handlers.device import DeviceHandler + + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_list_updater = handler.device_list_updater + self.store = hs.get_datastores().main self.clock = hs.get_clock() @@ -72,11 +78,77 @@ class ReplicationUserDevicesResyncRestServlet(ReplicationEndpoint): async def _handle_request( # type: ignore[override] self, request: Request, user_id: str - ) -> Tuple[int, JsonDict]: + ) -> Tuple[int, Optional[JsonDict]]: user_devices = await self.device_list_updater.user_device_resync(user_id) return 200, user_devices +class ReplicationUploadKeysForUserRestServlet(ReplicationEndpoint): + """Ask master to upload keys for the user and send them out over federation to + update other servers. + + For now, only the master is permitted to handle key upload requests; + any worker can handle key query requests (since they're read-only). + + Calls to e2e_keys_handler.upload_keys_for_user(user_id, device_id, keys) on + the main process to accomplish this. + + Defined in https://spec.matrix.org/v1.4/client-server-api/#post_matrixclientv3keysupload + Request format(borrowed and expanded from KeyUploadServlet): + + POST /_synapse/replication/upload_keys_for_user + + { + "user_id": "<user_id>", + "device_id": "<device_id>", + "keys": { + ....this part can be found in KeyUploadServlet in rest/client/keys.py.... + } + } + + Response is equivalent to ` /_matrix/client/v3/keys/upload` found in KeyUploadServlet + + """ + + NAME = "upload_keys_for_user" + PATH_ARGS = () + CACHE = False + + def __init__(self, hs: "HomeServer"): + super().__init__(hs) + + self.e2e_keys_handler = hs.get_e2e_keys_handler() + self.store = hs.get_datastores().main + self.clock = hs.get_clock() + + @staticmethod + async def _serialize_payload( # type: ignore[override] + user_id: str, device_id: str, keys: JsonDict + ) -> JsonDict: + + return { + "user_id": user_id, + "device_id": device_id, + "keys": keys, + } + + async def _handle_request( # type: ignore[override] + self, request: Request + ) -> Tuple[int, JsonDict]: + content = parse_json_object_from_request(request) + + user_id = content["user_id"] + device_id = content["device_id"] + keys = content["keys"] + + results = await self.e2e_keys_handler.upload_keys_for_user( + user_id, device_id, keys + ) + + return 200, results + + def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: ReplicationUserDevicesResyncRestServlet(hs).register(http_server) + ReplicationUploadKeysForUserRestServlet(hs).register(http_server) diff --git a/synapse/replication/http/register.py b/synapse/replication/http/register.py index 61abb529c8..976c283360 100644 --- a/synapse/replication/http/register.py +++ b/synapse/replication/http/register.py @@ -39,6 +39,16 @@ class ReplicationRegisterServlet(ReplicationEndpoint): self.store = hs.get_datastores().main self.registration_handler = hs.get_registration_handler() + # Default value if the worker that sent the replication request did not include + # an 'approved' property. + if ( + hs.config.experimental.msc3866.enabled + and hs.config.experimental.msc3866.require_approval_for_new_accounts + ): + self._approval_default = False + else: + self._approval_default = True + @staticmethod async def _serialize_payload( # type: ignore[override] user_id: str, @@ -92,6 +102,12 @@ class ReplicationRegisterServlet(ReplicationEndpoint): await self.registration_handler.check_registration_ratelimit(content["address"]) + # Always default admin users to approved (since it means they were created by + # an admin). + approved_default = self._approval_default + if content["admin"]: + approved_default = True + await self.registration_handler.register_with_store( user_id=user_id, password_hash=content["password_hash"], @@ -103,7 +119,7 @@ class ReplicationRegisterServlet(ReplicationEndpoint): user_type=content["user_type"], address=content["address"], shadow_banned=content["shadow_banned"], - approved=content["approved"], + approved=content.get("approved", approved_default), ) return 200, {} diff --git a/synapse/replication/slave/__init__.py b/synapse/replication/slave/__init__.py deleted file mode 100644 index f43a360a80..0000000000 --- a/synapse/replication/slave/__init__.py +++ /dev/null @@ -1,13 +0,0 @@ -# Copyright 2016 OpenMarket Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. diff --git a/synapse/replication/slave/storage/__init__.py b/synapse/replication/slave/storage/__init__.py deleted file mode 100644 index f43a360a80..0000000000 --- a/synapse/replication/slave/storage/__init__.py +++ /dev/null @@ -1,13 +0,0 @@ -# Copyright 2016 OpenMarket Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. diff --git a/synapse/replication/slave/storage/_slaved_id_tracker.py b/synapse/replication/slave/storage/_slaved_id_tracker.py deleted file mode 100644 index 8f3f953ed4..0000000000 --- a/synapse/replication/slave/storage/_slaved_id_tracker.py +++ /dev/null @@ -1,50 +0,0 @@ -# Copyright 2016 OpenMarket Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -from typing import List, Optional, Tuple - -from synapse.storage.database import LoggingDatabaseConnection -from synapse.storage.util.id_generators import AbstractStreamIdTracker, _load_current_id - - -class SlavedIdTracker(AbstractStreamIdTracker): - """Tracks the "current" stream ID of a stream with a single writer. - - See `AbstractStreamIdTracker` for more details. - - Note that this class does not work correctly when there are multiple - writers. - """ - - def __init__( - self, - db_conn: LoggingDatabaseConnection, - table: str, - column: str, - extra_tables: Optional[List[Tuple[str, str]]] = None, - step: int = 1, - ): - self.step = step - self._current = _load_current_id(db_conn, table, column, step) - if extra_tables: - for table, column in extra_tables: - self.advance(None, _load_current_id(db_conn, table, column)) - - def advance(self, instance_name: Optional[str], new_id: int) -> None: - self._current = (max if self.step > 0 else min)(self._current, new_id) - - def get_current_token(self) -> int: - return self._current - - def get_current_token_for_writer(self, instance_name: str) -> int: - return self.get_current_token() diff --git a/synapse/replication/slave/storage/devices.py b/synapse/replication/slave/storage/devices.py deleted file mode 100644 index 6fcade510a..0000000000 --- a/synapse/replication/slave/storage/devices.py +++ /dev/null @@ -1,79 +0,0 @@ -# Copyright 2016 OpenMarket Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from typing import TYPE_CHECKING, Any, Iterable - -from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker -from synapse.replication.tcp.streams._base import DeviceListsStream, UserSignatureStream -from synapse.storage.database import DatabasePool, LoggingDatabaseConnection -from synapse.storage.databases.main.devices import DeviceWorkerStore - -if TYPE_CHECKING: - from synapse.server import HomeServer - - -class SlavedDeviceStore(DeviceWorkerStore): - def __init__( - self, - database: DatabasePool, - db_conn: LoggingDatabaseConnection, - hs: "HomeServer", - ): - self.hs = hs - - self._device_list_id_gen = SlavedIdTracker( - db_conn, - "device_lists_stream", - "stream_id", - extra_tables=[ - ("user_signature_stream", "stream_id"), - ("device_lists_outbound_pokes", "stream_id"), - ("device_lists_changes_in_room", "stream_id"), - ], - ) - - super().__init__(database, db_conn, hs) - - def get_device_stream_token(self) -> int: - return self._device_list_id_gen.get_current_token() - - def process_replication_rows( - self, stream_name: str, instance_name: str, token: int, rows: Iterable[Any] - ) -> None: - if stream_name == DeviceListsStream.NAME: - self._device_list_id_gen.advance(instance_name, token) - self._invalidate_caches_for_devices(token, rows) - elif stream_name == UserSignatureStream.NAME: - self._device_list_id_gen.advance(instance_name, token) - for row in rows: - self._user_signature_stream_cache.entity_has_changed(row.user_id, token) - return super().process_replication_rows(stream_name, instance_name, token, rows) - - def _invalidate_caches_for_devices( - self, token: int, rows: Iterable[DeviceListsStream.DeviceListsStreamRow] - ) -> None: - for row in rows: - # The entities are either user IDs (starting with '@') whose devices - # have changed, or remote servers that we need to tell about - # changes. - if row.entity.startswith("@"): - self._device_list_stream_cache.entity_has_changed(row.entity, token) - self.get_cached_devices_for_user.invalidate((row.entity,)) - self._get_cached_user_device.invalidate((row.entity,)) - self.get_device_list_last_stream_id_for_remote.invalidate((row.entity,)) - - else: - self._device_list_federation_stream_cache.entity_has_changed( - row.entity, token - ) diff --git a/synapse/replication/slave/storage/events.py b/synapse/replication/slave/storage/events.py deleted file mode 100644 index fe47778cb1..0000000000 --- a/synapse/replication/slave/storage/events.py +++ /dev/null @@ -1,79 +0,0 @@ -# Copyright 2016 OpenMarket Ltd -# Copyright 2018 New Vector Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# 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 - -from synapse.storage.database import DatabasePool, LoggingDatabaseConnection -from synapse.storage.databases.main.event_federation import EventFederationWorkerStore -from synapse.storage.databases.main.event_push_actions import ( - EventPushActionsWorkerStore, -) -from synapse.storage.databases.main.events_worker import EventsWorkerStore -from synapse.storage.databases.main.relations import RelationsWorkerStore -from synapse.storage.databases.main.roommember import RoomMemberWorkerStore -from synapse.storage.databases.main.signatures import SignatureWorkerStore -from synapse.storage.databases.main.state import StateGroupWorkerStore -from synapse.storage.databases.main.stream import StreamWorkerStore -from synapse.storage.databases.main.user_erasure_store import UserErasureWorkerStore -from synapse.util.caches.stream_change_cache import StreamChangeCache - -if TYPE_CHECKING: - from synapse.server import HomeServer - -logger = logging.getLogger(__name__) - - -# So, um, we want to borrow a load of functions intended for reading from -# a DataStore, but we don't want to take functions that either write to the -# DataStore or are cached and don't have cache invalidation logic. -# -# Rather than write duplicate versions of those functions, or lift them to -# a common base class, we going to grab the underlying __func__ object from -# the method descriptor on the DataStore and chuck them into our class. - - -class SlavedEventStore( - EventFederationWorkerStore, - RoomMemberWorkerStore, - EventPushActionsWorkerStore, - StreamWorkerStore, - StateGroupWorkerStore, - SignatureWorkerStore, - EventsWorkerStore, - UserErasureWorkerStore, - RelationsWorkerStore, -): - def __init__( - self, - database: DatabasePool, - db_conn: LoggingDatabaseConnection, - hs: "HomeServer", - ): - super().__init__(database, db_conn, hs) - - events_max = self._stream_id_gen.get_current_token() - curr_state_delta_prefill, min_curr_state_delta_id = self.db_pool.get_cache_dict( - db_conn, - "current_state_delta_stream", - entity_column="room_id", - stream_column="stream_id", - max_value=events_max, # As we share the stream id with events token - limit=1000, - ) - self._curr_state_delta_stream_cache = StreamChangeCache( - "_curr_state_delta_stream_cache", - min_curr_state_delta_id, - prefilled_cache=curr_state_delta_prefill, - ) diff --git a/synapse/replication/slave/storage/push_rule.py b/synapse/replication/slave/storage/push_rule.py deleted file mode 100644 index 5e65eaf1e0..0000000000 --- a/synapse/replication/slave/storage/push_rule.py +++ /dev/null @@ -1,35 +0,0 @@ -# Copyright 2015, 2016 OpenMarket Ltd -# Copyright 2018 New Vector Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -from typing import Any, Iterable - -from synapse.replication.tcp.streams import PushRulesStream -from synapse.storage.databases.main.push_rule import PushRulesWorkerStore - -from .events import SlavedEventStore - - -class SlavedPushRuleStore(SlavedEventStore, PushRulesWorkerStore): - def get_max_push_rules_stream_id(self) -> int: - return self._push_rules_stream_id_gen.get_current_token() - - def process_replication_rows( - self, stream_name: str, instance_name: str, token: int, rows: Iterable[Any] - ) -> None: - if stream_name == PushRulesStream.NAME: - self._push_rules_stream_id_gen.advance(instance_name, token) - for row in rows: - self.get_push_rules_for_user.invalidate((row.user_id,)) - self.push_rules_stream_cache.entity_has_changed(row.user_id, token) - return super().process_replication_rows(stream_name, instance_name, token, rows) diff --git a/synapse/replication/slave/storage/pushers.py b/synapse/replication/slave/storage/pushers.py deleted file mode 100644 index 44ed20e424..0000000000 --- a/synapse/replication/slave/storage/pushers.py +++ /dev/null @@ -1,47 +0,0 @@ -# Copyright 2016 OpenMarket Ltd -# Copyright 2018 New Vector Ltd -# -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# -# http://www.apache.org/licenses/LICENSE-2.0 -# -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. -from typing import TYPE_CHECKING, Any, Iterable - -from synapse.replication.tcp.streams import PushersStream -from synapse.storage.database import DatabasePool, LoggingDatabaseConnection -from synapse.storage.databases.main.pusher import PusherWorkerStore - -from ._slaved_id_tracker import SlavedIdTracker - -if TYPE_CHECKING: - from synapse.server import HomeServer - - -class SlavedPusherStore(PusherWorkerStore): - def __init__( - self, - database: DatabasePool, - db_conn: LoggingDatabaseConnection, - hs: "HomeServer", - ): - super().__init__(database, db_conn, hs) - self._pushers_id_gen = SlavedIdTracker( # type: ignore - db_conn, "pushers", "id", extra_tables=[("deleted_pushers", "stream_id")] - ) - - def get_pushers_stream_token(self) -> int: - return self._pushers_id_gen.get_current_token() - - def process_replication_rows( - self, stream_name: str, instance_name: str, token: int, rows: Iterable[Any] - ) -> None: - if stream_name == PushersStream.NAME: - self._pushers_id_gen.advance(instance_name, token) - return super().process_replication_rows(stream_name, instance_name, token, rows) diff --git a/synapse/replication/tcp/client.py b/synapse/replication/tcp/client.py index 18252a2958..b4dad47b45 100644 --- a/synapse/replication/tcp/client.py +++ b/synapse/replication/tcp/client.py @@ -36,12 +36,14 @@ from synapse.replication.tcp.streams import ( TagAccountDataStream, ToDeviceStream, TypingStream, + UnPartialStatedRoomStream, ) from synapse.replication.tcp.streams.events import ( EventsStream, EventsStreamEventRow, EventsStreamRow, ) +from synapse.replication.tcp.streams.partial_state import UnPartialStatedRoomStreamRow from synapse.types import PersistedEventPosition, ReadReceipt, StreamKeyType, UserID from synapse.util.async_helpers import Linearizer, timeout_deferred from synapse.util.metrics import Measure @@ -117,6 +119,7 @@ class ReplicationDataHandler: self._streams = hs.get_replication_streams() self._instance_name = hs.get_instance_name() self._typing_handler = hs.get_typing_handler() + self._state_storage_controller = hs.get_storage_controllers().state self._notify_pushers = hs.config.worker.start_pushers self._pusher_pool = hs.get_pusherpool() @@ -236,6 +239,14 @@ class ReplicationDataHandler: self.notifier.notify_user_joined_room( row.data.event_id, row.data.room_id ) + elif stream_name == UnPartialStatedRoomStream.NAME: + for row in rows: + assert isinstance(row, UnPartialStatedRoomStreamRow) + + # Wake up any tasks waiting for the room to be un-partial-stated. + self._state_storage_controller.notify_room_un_partial_stated( + row.room_id + ) await self._presence_handler.process_replication_rows( stream_name, instance_name, token, rows diff --git a/synapse/replication/tcp/protocol.py b/synapse/replication/tcp/protocol.py index 7763ffb2d0..56a5c21910 100644 --- a/synapse/replication/tcp/protocol.py +++ b/synapse/replication/tcp/protocol.py @@ -245,7 +245,7 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver): self._parse_and_dispatch_line(line) def _parse_and_dispatch_line(self, line: bytes) -> None: - if line.strip() == "": + if line.strip() == b"": # Ignore blank lines return diff --git a/synapse/replication/tcp/streams/__init__.py b/synapse/replication/tcp/streams/__init__.py index b1cd55bf6f..8575666d9c 100644 --- a/synapse/replication/tcp/streams/__init__.py +++ b/synapse/replication/tcp/streams/__init__.py @@ -42,6 +42,7 @@ from synapse.replication.tcp.streams._base import ( ) from synapse.replication.tcp.streams.events import EventsStream from synapse.replication.tcp.streams.federation import FederationStream +from synapse.replication.tcp.streams.partial_state import UnPartialStatedRoomStream STREAMS_MAP = { stream.NAME: stream @@ -61,6 +62,7 @@ STREAMS_MAP = { TagAccountDataStream, AccountDataStream, UserSignatureStream, + UnPartialStatedRoomStream, ) } @@ -80,4 +82,5 @@ __all__ = [ "TagAccountDataStream", "AccountDataStream", "UserSignatureStream", + "UnPartialStatedRoomStream", ] diff --git a/synapse/replication/tcp/streams/partial_state.py b/synapse/replication/tcp/streams/partial_state.py new file mode 100644 index 0000000000..18f087ffa2 --- /dev/null +++ b/synapse/replication/tcp/streams/partial_state.py @@ -0,0 +1,48 @@ +# Copyright 2022 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. +from typing import TYPE_CHECKING + +import attr + +from synapse.replication.tcp.streams import Stream +from synapse.replication.tcp.streams._base import current_token_without_instance + +if TYPE_CHECKING: + from synapse.server import HomeServer + + +@attr.s(slots=True, frozen=True, auto_attribs=True) +class UnPartialStatedRoomStreamRow: + # ID of the room that has been un-partial-stated. + room_id: str + + +class UnPartialStatedRoomStream(Stream): + """ + Stream to notify about rooms becoming un-partial-stated; + that is, when the background sync finishes such that we now have full state for + the room. + """ + + NAME = "un_partial_stated_room" + ROW_TYPE = UnPartialStatedRoomStreamRow + + def __init__(self, hs: "HomeServer"): + store = hs.get_datastores().main + super().__init__( + hs.get_instance_name(), + # TODO(faster_joins, multiple writers): we need to account for instance names + current_token_without_instance(store.get_un_partial_stated_rooms_token), + store.get_un_partial_stated_rooms_from_stream, + ) diff --git a/synapse/res/templates/_base.html b/synapse/res/templates/_base.html new file mode 100644 index 0000000000..4b5cc7bcb6 --- /dev/null +++ b/synapse/res/templates/_base.html @@ -0,0 +1,29 @@ +<!DOCTYPE html> +<html lang="en"> +<head> + <meta charset="UTF-8"> + <meta http-equiv="X-UA-Compatible" content="IE=edge"> + <meta name="viewport" content="width=device-width, initial-scale=1.0"> + <title>{% block title %}{% endblock %}</title> + <style type="text/css"> + {%- include 'style.css' without context %} + </style> + {% block header %}{% endblock %} +</head> +<body> +<header class="mx_Header"> + {% if app_name == "Riot" %} + <img src="https://riot.im/img/external/riot-logo-email.png" width="83" height="83" alt="[Riot]"/> + {% elif app_name == "Vector" %} + <img src="https://matrix.org/img/vector-logo-email.png" width="64" height="83" alt="[Vector]"/> + {% elif app_name == "Element" %} + <img src="https://static.element.io/images/email-logo.png" width="83" height="83" alt="[Element]"/> + {% else %} + <img src="https://matrix.org/img/matrix-120x51.png" width="120" height="51" alt="[matrix]"/> + {% endif %} +</header> + +{% block body %}{% endblock %} + +</body> +</html> diff --git a/synapse/res/templates/account_previously_renewed.html b/synapse/res/templates/account_previously_renewed.html index bd4f7cea97..91582a8af0 100644 --- a/synapse/res/templates/account_previously_renewed.html +++ b/synapse/res/templates/account_previously_renewed.html @@ -1,12 +1,6 @@ -<!DOCTYPE html> -<html lang="en"> -<head> - <meta charset="UTF-8"> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <title>Your account is valid until {{ expiration_ts|format_ts("%d-%m-%Y") }}.</title> -</head> -<body> - Your account is valid until {{ expiration_ts|format_ts("%d-%m-%Y") }}. -</body> -</html> \ No newline at end of file +{% extends "_base.html" %} +{% block title %}Your account is valid until {{ expiration_ts|format_ts("%d-%m-%Y") }}.{% endblock %} + +{% block body %} +<p>Your account is valid until {{ expiration_ts|format_ts("%d-%m-%Y") }}.</p> +{% endblock %} diff --git a/synapse/res/templates/account_renewed.html b/synapse/res/templates/account_renewed.html index 57b319f375..18a57833f1 100644 --- a/synapse/res/templates/account_renewed.html +++ b/synapse/res/templates/account_renewed.html @@ -1,12 +1,6 @@ -<!DOCTYPE html> -<html lang="en"> -<head> - <meta charset="UTF-8"> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <title>Your account has been successfully renewed and is valid until {{ expiration_ts|format_ts("%d-%m-%Y") }}.</title> -</head> -<body> - Your account has been successfully renewed and is valid until {{ expiration_ts|format_ts("%d-%m-%Y") }}. -</body> -</html> \ No newline at end of file +{% extends "_base.html" %} +{% block title %}Your account has been successfully renewed and is valid until {{ expiration_ts|format_ts("%d-%m-%Y") }}.{% endblock %} + +{% block body %} +<p>Your account has been successfully renewed and is valid until {{ expiration_ts|format_ts("%d-%m-%Y") }}.</p> +{% endblock %} diff --git a/synapse/res/templates/add_threepid.html b/synapse/res/templates/add_threepid.html index 71f2215b7a..33c883936a 100644 --- a/synapse/res/templates/add_threepid.html +++ b/synapse/res/templates/add_threepid.html @@ -1,14 +1,8 @@ -<!DOCTYPE html> -<html lang="en"> -<head> - <meta charset="UTF-8"> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <title>Request to add an email address to your Matrix account</title> -</head> -<body> - <p>A request to add an email address to your Matrix account has been received. If this was you, please click the link below to confirm adding this email:</p> - <a href="{{ link }}">{{ link }}</a> - <p>If this was not you, you can safely ignore this email. Thank you.</p> -</body> -</html> +{% extends "_base.html" %} +{% block title %}Request to add an email address to your Matrix account{% endblock %} + +{% block body %} +<p>A request to add an email address to your Matrix account has been received. If this was you, please click the link below to confirm adding this email:</p> +<a href="{{ link }}">{{ link }}</a> +<p>If this was not you, you can safely ignore this email. Thank you.</p> +{% endblock %} diff --git a/synapse/res/templates/add_threepid_failure.html b/synapse/res/templates/add_threepid_failure.html index bd627ee9ce..f6d7e33825 100644 --- a/synapse/res/templates/add_threepid_failure.html +++ b/synapse/res/templates/add_threepid_failure.html @@ -1,13 +1,7 @@ -<!DOCTYPE html> -<html lang="en"> -<head> - <meta charset="UTF-8"> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <title>Request failed</title> -</head> -<body> - <p>The request failed for the following reason: {{ failure_reason }}.</p> - <p>No changes have been made to your account.</p> -</body> -</html> +{% extends "_base.html" %} +{% block title %}Request failed{% endblock %} + +{% block body %} +<p>The request failed for the following reason: {{ failure_reason }}.</p> +<p>No changes have been made to your account.</p> +{% endblock %} diff --git a/synapse/res/templates/add_threepid_success.html b/synapse/res/templates/add_threepid_success.html index 49170c138e..6d45111796 100644 --- a/synapse/res/templates/add_threepid_success.html +++ b/synapse/res/templates/add_threepid_success.html @@ -1,12 +1,6 @@ -<!DOCTYPE html> -<html lang="en"> -<head> - <meta charset="UTF-8"> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <title>Your email has now been validated</title> -</head> -<body> - <p>Your email has now been validated, please return to your client. You may now close this window.</p> -</body> -</html> \ No newline at end of file +{% extends "_base.html" %} +{% block title %}Your email has now been validated{% endblock %} + +{% block body %} +<p>Your email has now been validated, please return to your client. You may now close this window.</p> +{% endblock %} diff --git a/synapse/res/templates/auth_success.html b/synapse/res/templates/auth_success.html index 2d6ac44a0e..9178332f59 100644 --- a/synapse/res/templates/auth_success.html +++ b/synapse/res/templates/auth_success.html @@ -1,21 +1,21 @@ -<html> -<head> -<title>Success!</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> +{% extends "_base.html" %} +{% block title %}Success!{% endblock %} + +{% block header %} <link rel="stylesheet" href="/_matrix/static/client/register/style.css"> <script> if (window.onAuthDone) { window.onAuthDone(); } else if (window.opener && window.opener.postMessage) { - window.opener.postMessage("authDone", "*"); + window.opener.postMessage("authDone", "*"); } </script> -</head> -<body> - <div> - <p>Thank you</p> - <p>You may now close this window and return to the application</p> - </div> -</body> -</html> +{% endblock %} + +{% block body %} +<div> + <p>Thank you</p> + <p>You may now close this window and return to the application</p> +</div> + +{% endblock %} diff --git a/synapse/res/templates/invalid_token.html b/synapse/res/templates/invalid_token.html index 2c7c384fe3..b19e3023a1 100644 --- a/synapse/res/templates/invalid_token.html +++ b/synapse/res/templates/invalid_token.html @@ -1,12 +1,6 @@ -<!DOCTYPE html> -<html lang="en"> -<head> - <meta charset="UTF-8"> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <title>Invalid renewal token.</title> -</head> -<body> - Invalid renewal token. -</body> -</html> +{% extends "_base.html" %} +{% block title %}Invalid renewal token.{% endblock %} + +{% block body %} +<p>Invalid renewal token.</p> +{% endblock %} diff --git a/synapse/res/templates/notice_expiry.html b/synapse/res/templates/notice_expiry.html index 865f9f7ada..f62038e111 100644 --- a/synapse/res/templates/notice_expiry.html +++ b/synapse/res/templates/notice_expiry.html @@ -1,47 +1,46 @@ -<!doctype html> -<html lang="en"> - <head> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <style type="text/css"> - {% include 'mail.css' without context %} - {% include "mail-%s.css" % app_name ignore missing without context %} - {% include 'mail-expiry.css' without context %} - </style> - </head> - <body> - <table id="page"> - <tr> - <td> </td> - <td id="inner"> - <table class="header"> - <tr> - <td> - <div class="salutation">Hi {{ display_name }},</div> - </td> - <td class="logo"> - {% if app_name == "Riot" %} - <img src="http://riot.im/img/external/riot-logo-email.png" width="83" height="83" alt="[Riot]"/> - {% elif app_name == "Vector" %} - <img src="http://matrix.org/img/vector-logo-email.png" width="64" height="83" alt="[Vector]"/> - {% elif app_name == "Element" %} - <img src="https://static.element.io/images/email-logo.png" width="83" height="83" alt="[Element]"/> - {% else %} - <img src="http://matrix.org/img/matrix-120x51.png" width="120" height="51" alt="[matrix]"/> - {% endif %} - </td> - </tr> - <tr> - <td colspan="2"> - <div class="noticetext">Your account will expire on {{ expiration_ts|format_ts("%d-%m-%Y") }}. This means that you will lose access to your account after this date.</div> - <div class="noticetext">To extend the validity of your account, please click on the link below (or copy and paste it into a new browser tab):</div> - <div class="noticetext"><a href="{{ url }}">{{ url }}</a></div> - </td> - </tr> - </table> - </td> - <td> </td> - </tr> - </table> - </body> -</html> +{% extends "_base.html" %} +{% block title %}Notice of expiry{% endblock %} + +{% block header %} +<style type="text/css"> + {% include 'mail.css' without context %} + {% include "mail-%s.css" % app_name ignore missing without context %} + {% include 'mail-expiry.css' without context %} +</style> +{% endblock %} + +{% block body %} +<table id="page"> + <tr> + <td> </td> + <td id="inner"> + <table class="header"> + <tr> + <td> + <div class="salutation">Hi {{ display_name }},</div> + </td> + <td class="logo"> + {% if app_name == "Riot" %} + <img src="https://riot.im/img/external/riot-logo-email.png" width="83" height="83" alt="[Riot]"/> + {% elif app_name == "Vector" %} + <img src="https://matrix.org/img/vector-logo-email.png" width="64" height="83" alt="[Vector]"/> + {% elif app_name == "Element" %} + <img src="https://static.element.io/images/email-logo.png" width="83" height="83" alt="[Element]"/> + {% else %} + <img src="https://matrix.org/img/matrix-120x51.png" width="120" height="51" alt="[matrix]"/> + {% endif %} + </td> + </tr> + <tr> + <td colspan="2"> + <div class="noticetext">Your account will expire on {{ expiration_ts|format_ts("%d-%m-%Y") }}. This means that you will lose access to your account after this date.</div> + <div class="noticetext">To extend the validity of your account, please click on the link below (or copy and paste it into a new browser tab):</div> + <div class="noticetext"><a href="{{ url }}">{{ url }}</a></div> + </td> + </tr> + </table> + </td> + <td> </td> + </tr> +</table> +{% endblock %} diff --git a/synapse/res/templates/notif_mail.html b/synapse/res/templates/notif_mail.html index 9dba0c0253..7da0fff5e9 100644 --- a/synapse/res/templates/notif_mail.html +++ b/synapse/res/templates/notif_mail.html @@ -1,59 +1,59 @@ -<!doctype html> -<html lang="en"> - <head> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <style type="text/css"> - {%- include 'mail.css' without context %} - {%- include "mail-%s.css" % app_name ignore missing without context %} - </style> - </head> - <body> - <table id="page"> - <tr> - <td> </td> - <td id="inner"> - <table class="header"> - <tr> - <td> - <div class="salutation">Hi {{ user_display_name }},</div> - <div class="summarytext">{{ summary_text }}</div> - </td> - <td class="logo"> - {%- if app_name == "Riot" %} - <img src="http://riot.im/img/external/riot-logo-email.png" width="83" height="83" alt="[Riot]"/> - {%- elif app_name == "Vector" %} - <img src="http://matrix.org/img/vector-logo-email.png" width="64" height="83" alt="[Vector]"/> - {%- elif app_name == "Element" %} - <img src="https://static.element.io/images/email-logo.png" width="83" height="83" alt="[Element]"/> - {%- else %} - <img src="http://matrix.org/img/matrix-120x51.png" width="120" height="51" alt="[matrix]"/> - {%- endif %} - </td> - </tr> - </table> - {%- for room in rooms %} - {%- include 'room.html' with context %} - {%- endfor %} - <div class="footer"> - <a href="{{ unsubscribe_link }}">Unsubscribe</a> - <br/> - <br/> - <div class="debug"> - Sending email at {{ reason.now|format_ts("%c") }} due to activity in room {{ reason.room_name }} because - an event was received at {{ reason.received_at|format_ts("%c") }} - which is more than {{ "%.1f"|format(reason.delay_before_mail_ms / (60*1000)) }} ({{ reason.delay_before_mail_ms }}) mins ago, - {%- if reason.last_sent_ts %} - and the last time we sent a mail for this room was {{ reason.last_sent_ts|format_ts("%c") }}, - which is more than {{ "%.1f"|format(reason.throttle_ms / (60*1000)) }} (current throttle_ms) mins ago. - {%- else %} - and we don't have a last time we sent a mail for this room. - {%- endif %} - </div> - </div> - </td> - <td> </td> - </tr> - </table> - </body> -</html> +{% extends "_base.html" %} + +{% block title %}New activity in room{% endblock %} + +{% block header %} +<style type="text/css"> + {%- include 'mail.css' without context %} + {%- include "mail-%s.css" % app_name ignore missing without context %} +</style> +{% endblock %} + +{% block body %} +<table id="page"> + <tr> + <td> </td> + <td id="inner"> + <table class="header"> + <tr> + <td> + <div class="salutation">Hi {{ user_display_name }},</div> + <div class="summarytext">{{ summary_text }}</div> + </td> + <td class="logo"> + {%- if app_name == "Riot" %} + <img src="https://riot.im/img/external/riot-logo-email.png" width="83" height="83" alt="[Riot]"/> + {%- elif app_name == "Vector" %} + <img src="https://matrix.org/img/vector-logo-email.png" width="64" height="83" alt="[Vector]"/> + {%- elif app_name == "Element" %} + <img src="https://static.element.io/images/email-logo.png" width="83" height="83" alt="[Element]"/> + {%- else %} + <img src="https://matrix.org/img/matrix-120x51.png" width="120" height="51" alt="[matrix]"/> + {%- endif %} + </td> + </tr> + </table> + {%- for room in rooms %} + {%- include 'room.html' with context %} + {%- endfor %} + <div class="footer"> + <a href="{{ unsubscribe_link }}">Unsubscribe</a> + <br/> + <br/> + <div class="debug"> + Sending email at {{ reason.now|format_ts("%c") }} due to activity in room {{ reason.room_name }} because + an event was received at {{ reason.received_at|format_ts("%c") }} + which is more than {{ "%.1f"|format(reason.delay_before_mail_ms / (60*1000)) }} ({{ reason.delay_before_mail_ms }}) mins ago, + {%- if reason.last_sent_ts %} + and the last time we sent a mail for this room was {{ reason.last_sent_ts|format_ts("%c") }}, + which is more than {{ "%.1f"|format(reason.throttle_ms / (60*1000)) }} (current throttle_ms) mins ago. + {%- else %} + and we don't have a last time we sent a mail for this room. + {%- endif %} + </div> + </div> + </td> + <td> </td> + </tr> +</table> +{% endblock %} diff --git a/synapse/res/templates/password_reset.html b/synapse/res/templates/password_reset.html index a8bdce357b..1f267946c8 100644 --- a/synapse/res/templates/password_reset.html +++ b/synapse/res/templates/password_reset.html @@ -1,14 +1,10 @@ -<html lang="en"> - <head> - <title>Password reset</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - </head> -<body> - <p>A password reset request has been received for your Matrix account. If this was you, please click the link below to confirm resetting your password:</p> +{% extends "_base.html" %} +{% block title %}Password reset{% endblock %} - <a href="{{ link }}">{{ link }}</a> +{% block body %} +<p>A password reset request has been received for your Matrix account. If this was you, please click the link below to confirm resetting your password:</p> - <p>If this was not you, <strong>do not</strong> click the link above and instead contact your server administrator. Thank you.</p> -</body> -</html> +<a href="{{ link }}">{{ link }}</a> + +<p>If this was not you, <strong>do not</strong> click the link above and instead contact your server administrator. Thank you.</p> +{% endblock %} diff --git a/synapse/res/templates/password_reset_confirmation.html b/synapse/res/templates/password_reset_confirmation.html index 2e3fd2ec1e..fabb9a6ed5 100644 --- a/synapse/res/templates/password_reset_confirmation.html +++ b/synapse/res/templates/password_reset_confirmation.html @@ -1,10 +1,7 @@ -<html lang="en"> -<head> - <title>Password reset confirmation</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> -</head> -<body> +{% extends "_base.html" %} +{% block title %}Password reset confirmation{% endblock %} + +{% block body %} <!--Use a hidden form to resubmit the information necessary to reset the password--> <form method="post"> <input type="hidden" name="sid" value="{{ sid }}"> @@ -15,6 +12,4 @@ If you did not mean to do this, please close this page and your password will not be changed.</p> <p><button type="submit">Confirm changing my password</button></p> </form> -</body> -</html> - +{% endblock %} diff --git a/synapse/res/templates/password_reset_failure.html b/synapse/res/templates/password_reset_failure.html index 2d59c463f0..9990e860f9 100644 --- a/synapse/res/templates/password_reset_failure.html +++ b/synapse/res/templates/password_reset_failure.html @@ -1,12 +1,7 @@ -<html lang="en"> -<head> - <title>Password reset failure</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> -</head> -<body> -<p>The request failed for the following reason: {{ failure_reason }}.</p> +{% extends "_base.html" %} +{% block title %}Password reset failure{% endblock %} +{% block body %} +<p>The request failed for the following reason: {{ failure_reason }}.</p> <p>Your password has not been reset.</p> -</body> -</html> +{% endblock %} diff --git a/synapse/res/templates/password_reset_success.html b/synapse/res/templates/password_reset_success.html index 5165bd1fa2..edada513ab 100644 --- a/synapse/res/templates/password_reset_success.html +++ b/synapse/res/templates/password_reset_success.html @@ -1,9 +1,6 @@ -<html lang="en"> -<head> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> -</head> -<body> +{% extends "_base.html" %} +{% block title %}Password reset success{% endblock %} + +{% block body %} <p>Your email has now been validated, please return to your client to reset your password. You may now close this window.</p> -</body> -</html> +{% endblock %} diff --git a/synapse/res/templates/recaptcha.html b/synapse/res/templates/recaptcha.html index 615d3239c6..8204928cdf 100644 --- a/synapse/res/templates/recaptcha.html +++ b/synapse/res/templates/recaptcha.html @@ -1,10 +1,8 @@ -<html> -<head> -<title>Authentication</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> -<script src="https://www.recaptcha.net/recaptcha/api.js" - async defer></script> +{% extends "_base.html" %} +{% block title %}Authentication{% endblock %} + +{% block header %} +<script src="https://www.recaptcha.net/recaptcha/api.js" async defer></script> <script src="//code.jquery.com/jquery-1.11.2.min.js"></script> <link rel="stylesheet" href="/_matrix/static/client/register/style.css"> <script> @@ -12,8 +10,9 @@ function captchaDone() { $('#registrationForm').submit(); } </script> -</head> -<body> +{% endblock %} + +{% block body %} <form id="registrationForm" method="post" action="{{ myurl }}"> <div> {% if error is defined %} @@ -37,5 +36,4 @@ function captchaDone() { </div> </div> </form> -</body> -</html> +{% endblock %} \ No newline at end of file diff --git a/synapse/res/templates/registration.html b/synapse/res/templates/registration.html index 20e831ff4a..cdb815665e 100644 --- a/synapse/res/templates/registration.html +++ b/synapse/res/templates/registration.html @@ -1,16 +1,12 @@ -<html lang="en"> -<head> - <title>Registration</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> -</head> -<body> - <p>You have asked us to register this email with a new Matrix account. If this was you, please click the link below to confirm your email address:</p> +{% extends "_base.html" %} +{% block title %}Registration{% endblock %} - <a href="{{ link }}">Verify Your Email Address</a> +{% block body %} +<p>You have asked us to register this email with a new Matrix account. If this was you, please click the link below to confirm your email address:</p> - <p>If this was not you, you can safely disregard this email.</p> +<a href="{{ link }}">Verify Your Email Address</a> - <p>Thank you.</p> -</body> -</html> +<p>If this was not you, you can safely disregard this email.</p> + +<p>Thank you.</p> +{% endblock %} diff --git a/synapse/res/templates/registration_failure.html b/synapse/res/templates/registration_failure.html index a6ed22bc90..ae2a9cae2c 100644 --- a/synapse/res/templates/registration_failure.html +++ b/synapse/res/templates/registration_failure.html @@ -1,9 +1,6 @@ -<html lang="en"> -<head> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> -</head> -<body> +{% extends "_base.html" %} +{% block title %}Registration failure{% endblock %} + +{% block body %} <p>Validation failed for the following reason: {{ failure_reason }}.</p> -</body> -</html> +{% endblock %} diff --git a/synapse/res/templates/registration_success.html b/synapse/res/templates/registration_success.html index d51d5549d8..6d45111796 100644 --- a/synapse/res/templates/registration_success.html +++ b/synapse/res/templates/registration_success.html @@ -1,10 +1,6 @@ -<html lang="en"> -<head> - <title>Your email has now been validated</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> -</head> -<body> +{% extends "_base.html" %} +{% block title %}Your email has now been validated{% endblock %} + +{% block body %} <p>Your email has now been validated, please return to your client. You may now close this window.</p> -</body> -</html> +{% endblock %} diff --git a/synapse/res/templates/registration_token.html b/synapse/res/templates/registration_token.html index 59a98f564c..ee4e5295e7 100644 --- a/synapse/res/templates/registration_token.html +++ b/synapse/res/templates/registration_token.html @@ -1,11 +1,11 @@ -<html lang="en"> -<head> -<title>Authentication</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> +{% extends "_base.html" %} +{% block title %}Authentication{% endblock %} + +{% block header %} <link rel="stylesheet" href="/_matrix/static/client/register/style.css"> -</head> -<body> +{% endblock %} + +{% block body %} <form id="registrationForm" method="post" action="{{ myurl }}"> <div> {% if error is defined %} @@ -19,5 +19,4 @@ <input type="submit" value="Authenticate" /> </div> </form> -</body> -</html> +{% endblock %} diff --git a/synapse/res/templates/sso_account_deactivated.html b/synapse/res/templates/sso_account_deactivated.html index 075f801cec..b85d96cc74 100644 --- a/synapse/res/templates/sso_account_deactivated.html +++ b/synapse/res/templates/sso_account_deactivated.html @@ -1,25 +1,25 @@ -<!DOCTYPE html> -<html lang="en"> - <head> - <meta charset="UTF-8"> - <title>SSO account deactivated</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> <style type="text/css"> - {% include "sso.css" without context %} - </style> - </head> - <body class="error_page"> - <header> - <h1>Your account has been deactivated</h1> - <p> - <strong>No account found</strong> - </p> - <p> - Your account might have been deactivated by the server administrator. - You can either try to create a new account or contact the server’s - administrator. - </p> - </header> - {% include "sso_footer.html" without context %} - </body> -</html> +{% extends "_base.html" %} +{% block title %}SSO account deactivated{% endblock %} + +{% block header %} +<style type="text/css"> + {% include "sso.css" without context %} +</style> +{% endblock %} + +{% block body %} +<div class="error_page"> + <header> + <h1>Your account has been deactivated</h1> + <p> + <strong>No account found</strong> + </p> + <p> + Your account might have been deactivated by the server administrator. + You can either try to create a new account or contact the server’s + administrator. + </p> + </header> +</div> +{% include "sso_footer.html" without context %} +{% endblock %} diff --git a/synapse/res/templates/sso_auth_account_details.html b/synapse/res/templates/sso_auth_account_details.html index 2d1db386e1..11636d7f5d 100644 --- a/synapse/res/templates/sso_auth_account_details.html +++ b/synapse/res/templates/sso_auth_account_details.html @@ -1,189 +1,186 @@ -<!DOCTYPE html> -<html lang="en"> - <head> - <title>Create your account</title> - <meta charset="utf-8"> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <script type="text/javascript"> - let wasKeyboard = false; - document.addEventListener("mousedown", function() { wasKeyboard = false; }); - document.addEventListener("keydown", function() { wasKeyboard = true; }); - document.addEventListener("focusin", function() { - if (wasKeyboard) { - document.body.classList.add("keyboard-focus"); - } else { - document.body.classList.remove("keyboard-focus"); - } - }); - </script> - <style type="text/css"> - {% include "sso.css" without context %} - - body.keyboard-focus :focus, body.keyboard-focus .username_input:focus-within { - outline: 3px solid #17191C; - outline-offset: 4px; - } - - .username_input { - display: flex; - border: 2px solid #418DED; - border-radius: 8px; - padding: 12px; - position: relative; - margin: 16px 0; - align-items: center; - font-size: 12px; - } - - .username_input.invalid { - border-color: #FE2928; - } - - .username_input.invalid input, .username_input.invalid label { - color: #FE2928; - } - - .username_input div, .username_input input { - line-height: 18px; - font-size: 14px; - } - - .username_input label { - position: absolute; - top: -5px; - left: 14px; - font-size: 10px; - line-height: 10px; - background: white; - padding: 0 2px; - } - - .username_input input { - flex: 1; - display: block; - min-width: 0; - border: none; - } - - /* only clear the outline if we know it will be shown on the parent div using :focus-within */ - @supports selector(:focus-within) { - .username_input input { - outline: none !important; - } - } - - .username_input div { - color: #8D99A5; - } - - .idp-pick-details { - border: 1px solid #E9ECF1; - border-radius: 8px; - margin: 24px 0; - } - - .idp-pick-details h2 { - margin: 0; - padding: 8px 12px; - } - - .idp-pick-details .idp-detail { - border-top: 1px solid #E9ECF1; - padding: 12px; - display: block; - } - .idp-pick-details .check-row { - display: flex; - align-items: center; - } - - .idp-pick-details .check-row .name { - flex: 1; - } - - .idp-pick-details .use, .idp-pick-details .idp-value { - color: #737D8C; - } - - .idp-pick-details .idp-value { - margin: 0; - margin-top: 8px; - } - - .idp-pick-details .avatar { - width: 53px; - height: 53px; - border-radius: 100%; - display: block; - margin-top: 8px; - } - - output { - padding: 0 14px; - display: block; - } - - output.error { - color: #FE2928; - } - </style> - </head> - <body> - <header> - <h1>Create your account</h1> - <p>This is required. Continue to create your account on {{ server_name }}. You can't change this later.</p> - </header> - <main> - <form method="post" class="form__input" id="form"> - <div class="username_input" id="username_input"> - <label for="field-username">Username (required)</label> - <div class="prefix">@</div> - <input type="text" name="username" id="field-username" value="{{ user_attributes.localpart }}" autofocus autocorrect="off" autocapitalize="none"> - <div class="postfix">:{{ server_name }}</div> +{% extends "_base.html" %} +{% block title %}Create your account{% endblock %} + +{% block header %} +<script type="text/javascript"> + let wasKeyboard = false; + document.addEventListener("mousedown", function() { wasKeyboard = false; }); + document.addEventListener("keydown", function() { wasKeyboard = true; }); + document.addEventListener("focusin", function() { + if (wasKeyboard) { + document.body.classList.add("keyboard-focus"); + } else { + document.body.classList.remove("keyboard-focus"); + } + }); +</script> +<style type="text/css"> + {% include "sso.css" without context %} + + body.keyboard-focus :focus, body.keyboard-focus .username_input:focus-within { + outline: 3px solid #17191C; + outline-offset: 4px; + } + + .username_input { + display: flex; + border: 2px solid #418DED; + border-radius: 8px; + padding: 12px; + position: relative; + margin: 16px 0; + align-items: center; + font-size: 12px; + } + + .username_input.invalid { + border-color: #FE2928; + } + + .username_input.invalid input, .username_input.invalid label { + color: #FE2928; + } + + .username_input div, .username_input input { + line-height: 18px; + font-size: 14px; + } + + .username_input label { + position: absolute; + top: -5px; + left: 14px; + font-size: 10px; + line-height: 10px; + background: white; + padding: 0 2px; + } + + .username_input input { + flex: 1; + display: block; + min-width: 0; + border: none; + } + + /* only clear the outline if we know it will be shown on the parent div using :focus-within */ + @supports selector(:focus-within) { + .username_input input { + outline: none !important; + } + } + + .username_input div { + color: #8D99A5; + } + + .idp-pick-details { + border: 1px solid #E9ECF1; + border-radius: 8px; + margin: 24px 0; + } + + .idp-pick-details h2 { + margin: 0; + padding: 8px 12px; + } + + .idp-pick-details .idp-detail { + border-top: 1px solid #E9ECF1; + padding: 12px; + display: block; + } + .idp-pick-details .check-row { + display: flex; + align-items: center; + } + + .idp-pick-details .check-row .name { + flex: 1; + } + + .idp-pick-details .use, .idp-pick-details .idp-value { + color: #737D8C; + } + + .idp-pick-details .idp-value { + margin: 0; + margin-top: 8px; + } + + .idp-pick-details .avatar { + width: 53px; + height: 53px; + border-radius: 100%; + display: block; + margin-top: 8px; + } + + output { + padding: 0 14px; + display: block; + } + + output.error { + color: #FE2928; + } +</style> +{% endblock %} + +{% block body %} +<header> + <h1>Create your account</h1> + <p>This is required. Continue to create your account on {{ server_name }}. You can't change this later.</p> +</header> +<main> + <form method="post" class="form__input" id="form"> + <div class="username_input" id="username_input"> + <label for="field-username">Username (required)</label> + <div class="prefix">@</div> + <input type="text" name="username" id="field-username" value="{{ user_attributes.localpart }}" autofocus autocorrect="off" autocapitalize="none"> + <div class="postfix">:{{ server_name }}</div> + </div> + <output for="username_input" id="field-username-output"></output> + <input type="submit" value="Continue" class="primary-button"> + {% if user_attributes.avatar_url or user_attributes.display_name or user_attributes.emails %} + <section class="idp-pick-details"> + <h2>{% if idp.idp_icon %}<img src="{{ idp.idp_icon | mxc_to_http(24, 24) }}"/>{% endif %}Optional data from {{ idp.idp_name }}</h2> + {% if user_attributes.avatar_url %} + <label class="idp-detail idp-avatar" for="idp-avatar"> + <div class="check-row"> + <span class="name">Avatar</span> + <span class="use">Use</span> + <input type="checkbox" name="use_avatar" id="idp-avatar" value="true" checked> </div> - <output for="username_input" id="field-username-output"></output> - <input type="submit" value="Continue" class="primary-button"> - {% if user_attributes.avatar_url or user_attributes.display_name or user_attributes.emails %} - <section class="idp-pick-details"> - <h2>{% if idp.idp_icon %}<img src="{{ idp.idp_icon | mxc_to_http(24, 24) }}"/>{% endif %}Optional data from {{ idp.idp_name }}</h2> - {% if user_attributes.avatar_url %} - <label class="idp-detail idp-avatar" for="idp-avatar"> - <div class="check-row"> - <span class="name">Avatar</span> - <span class="use">Use</span> - <input type="checkbox" name="use_avatar" id="idp-avatar" value="true" checked> - </div> - <img src="{{ user_attributes.avatar_url }}" class="avatar" /> - </label> - {% endif %} - {% if user_attributes.display_name %} - <label class="idp-detail" for="idp-displayname"> - <div class="check-row"> - <span class="name">Display name</span> - <span class="use">Use</span> - <input type="checkbox" name="use_display_name" id="idp-displayname" value="true" checked> - </div> - <p class="idp-value">{{ user_attributes.display_name }}</p> - </label> - {% endif %} - {% for email in user_attributes.emails %} - <label class="idp-detail" for="idp-email{{ loop.index }}"> - <div class="check-row"> - <span class="name">E-mail</span> - <span class="use">Use</span> - <input type="checkbox" name="use_email" id="idp-email{{ loop.index }}" value="{{ email }}" checked> - </div> - <p class="idp-value">{{ email }}</p> - </label> - {% endfor %} - </section> - {% endif %} - </form> - </main> - {% include "sso_footer.html" without context %} - <script type="text/javascript"> - {% include "sso_auth_account_details.js" without context %} - </script> - </body> -</html> + <img src="{{ user_attributes.avatar_url }}" class="avatar" /> + </label> + {% endif %} + {% if user_attributes.display_name %} + <label class="idp-detail" for="idp-displayname"> + <div class="check-row"> + <span class="name">Display name</span> + <span class="use">Use</span> + <input type="checkbox" name="use_display_name" id="idp-displayname" value="true" checked> + </div> + <p class="idp-value">{{ user_attributes.display_name }}</p> + </label> + {% endif %} + {% for email in user_attributes.emails %} + <label class="idp-detail" for="idp-email{{ loop.index }}"> + <div class="check-row"> + <span class="name">E-mail</span> + <span class="use">Use</span> + <input type="checkbox" name="use_email" id="idp-email{{ loop.index }}" value="{{ email }}" checked> + </div> + <p class="idp-value">{{ email }}</p> + </label> + {% endfor %} + </section> + {% endif %} + </form> +</main> +{% include "sso_footer.html" without context %} +<script type="text/javascript"> + {% include "sso_auth_account_details.js" without context %} +</script> +{% endblock %} diff --git a/synapse/res/templates/sso_auth_bad_user.html b/synapse/res/templates/sso_auth_bad_user.html index 94403fc3ce..819d79a461 100644 --- a/synapse/res/templates/sso_auth_bad_user.html +++ b/synapse/res/templates/sso_auth_bad_user.html @@ -1,27 +1,26 @@ -<!DOCTYPE html> -<html lang="en"> - <head> - <meta charset="UTF-8"> - <title>Authentication failed</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <style type="text/css"> - {% include "sso.css" without context %} - </style> - </head> - <body class="error_page"> - <header> - <h1>That doesn't look right</h1> - <p> - <strong>We were unable to validate your {{ server_name }} account</strong> - via single sign‑on (SSO), because the SSO Identity - Provider returned different details than when you logged in. - </p> - <p> - Try the operation again, and ensure that you use the same details on - the Identity Provider as when you log into your account. - </p> - </header> - {% include "sso_footer.html" without context %} - </body> -</html> +{% extends "_base.html" %} +{% block title %}Authentication failed{% endblock %} + +{% block header %} +<style type="text/css"> + {% include "sso.css" without context %} +</style> +{% endblock %} + +{% block body %} +<div class="error_page"> + <header> + <h1>That doesn't look right</h1> + <p> + <strong>We were unable to validate your {{ server_name }} account</strong> + via single sign‑on (SSO), because the SSO Identity + Provider returned different details than when you logged in. + </p> + <p> + Try the operation again, and ensure that you use the same details on + the Identity Provider as when you log into your account. + </p> + </header> +</div> +{% include "sso_footer.html" without context %} +{% endblock %} diff --git a/synapse/res/templates/sso_auth_confirm.html b/synapse/res/templates/sso_auth_confirm.html index aa1c974a6b..3927d6eda3 100644 --- a/synapse/res/templates/sso_auth_confirm.html +++ b/synapse/res/templates/sso_auth_confirm.html @@ -1,30 +1,27 @@ -<!DOCTYPE html> -<html lang="en"> - <head> - <meta charset="UTF-8"> - <title>Confirm it's you</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <style type="text/css"> - {% include "sso.css" without context %} - </style> - </head> - <body> - <header> - <h1>Confirm it's you to continue</h1> - <p> - A client is trying to {{ description }}. To confirm this action - re-authorize your account with single sign-on. - </p> - <p><strong> - If you did not expect this, your account may be compromised. - </strong></p> - </header> - <main> - <a href="{{ redirect_url }}" class="primary-button"> - Continue with {{ idp.idp_name }} - </a> - </main> - {% include "sso_footer.html" without context %} - </body> -</html> +{% extends "_base.html" %} +{% block title %}Confirm it's you{% endblock %} + +{% block header %} +<style type="text/css"> + {% include "sso.css" without context %} +</style> +{% endblock %} + +{% block body %} +<header> + <h1>Confirm it's you to continue</h1> + <p> + A client is trying to {{ description }}. To confirm this action + re-authorize your account with single sign-on. + </p> + <p><strong> + If you did not expect this, your account may be compromised. + </strong></p> +</header> +<main> + <a href="{{ redirect_url }}" class="primary-button"> + Continue with {{ idp.idp_name }} + </a> +</main> +{% include "sso_footer.html" without context %} +{% endblock %} diff --git a/synapse/res/templates/sso_auth_success.html b/synapse/res/templates/sso_auth_success.html index 4898af6011..afeffb7191 100644 --- a/synapse/res/templates/sso_auth_success.html +++ b/synapse/res/templates/sso_auth_success.html @@ -1,29 +1,26 @@ -<!DOCTYPE html> -<html lang="en"> - <head> - <meta charset="UTF-8"> - <title>Authentication successful</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <style type="text/css"> - {% include "sso.css" without context %} - </style> - <script> - if (window.onAuthDone) { - window.onAuthDone(); - } else if (window.opener && window.opener.postMessage) { - window.opener.postMessage("authDone", "*"); - } - </script> - </head> - <body> - <header> - <h1>Thank you</h1> - <p> - Now we know it’s you, you can close this window and return to the - application. - </p> - </header> - {% include "sso_footer.html" without context %} - </body> -</html> +{% extends "_base.html" %} +{% block title %}Authentication successful{% endblock %} + +{% block header %} +<style type="text/css"> + {% include "sso.css" without context %} +</style> +<script> + if (window.onAuthDone) { + window.onAuthDone(); + } else if (window.opener && window.opener.postMessage) { + window.opener.postMessage("authDone", "*"); + } +</script> +{% endblock %} + +{% block body %} +<header> + <h1>Thank you</h1> + <p> + Now we know it’s you, you can close this window and return to the + application. + </p> +</header> +{% include "sso_footer.html" without context %} +{% endblock %} diff --git a/synapse/res/templates/sso_error.html b/synapse/res/templates/sso_error.html index 19992ff2ad..6fa36c11c9 100644 --- a/synapse/res/templates/sso_error.html +++ b/synapse/res/templates/sso_error.html @@ -1,19 +1,20 @@ -<!DOCTYPE html> -<html lang="en"> - <head> - <meta charset="UTF-8"> - <title>Authentication failed</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <style type="text/css"> - {% include "sso.css" without context %} +{% extends "_base.html" %} +{% block title %}Authentication failed{% endblock %} - #error_code { - margin-top: 56px; - } - </style> - </head> - <body class="error_page"> +{% block header %} +{% if error == "unauthorised" %} +<style type="text/css"> + {% include "sso.css" without context %} + + #error_code { + margin-top: 56px; + } +</style> +{% endif %} +{% endblock %} + +{% block body %} +<div class="error_page"> {# If an error of unauthorised is returned it means we have actively rejected their login #} {% if error == "unauthorised" %} <header> @@ -66,5 +67,5 @@ } </script> {% endif %} -</body> -</html> +</div> +{% endblock %} diff --git a/synapse/res/templates/sso_login_idp_picker.html b/synapse/res/templates/sso_login_idp_picker.html index 56fabfa3d2..58b0b3121c 100644 --- a/synapse/res/templates/sso_login_idp_picker.html +++ b/synapse/res/templates/sso_login_idp_picker.html @@ -1,63 +1,60 @@ -<!DOCTYPE html> -<html lang="en"> - <head> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <meta charset="UTF-8"> - <title>Choose identity provider</title> - <style type="text/css"> - {% include "sso.css" without context %} +{% extends "_base.html" %} +{% block title %}Choose identity provider{% endblock %} - .providers { - list-style: none; - padding: 0; - } +{% block header %} +<style type="text/css"> + {% include "sso.css" without context %} - .providers li { - margin: 12px; - } + .providers { + list-style: none; + padding: 0; + } - .providers a { - display: block; - border-radius: 4px; - border: 1px solid #17191C; - padding: 8px; - text-align: center; - text-decoration: none; - color: #17191C; - display: flex; - align-items: center; - font-weight: bold; - } + .providers li { + margin: 12px; + } - .providers a img { - width: 24px; - height: 24px; - } - .providers a span { - flex: 1; - } - </style> - </head> - <body> - <header> - <h1>Log in to {{ server_name }} </h1> - <p>Choose an identity provider to log in</p> - </header> - <main> - <ul class="providers"> - {% for p in providers %} - <li> - <a href="pick_idp?idp={{ p.idp_id }}&redirectUrl={{ redirect_url | urlencode }}"> - {% if p.idp_icon %} - <img src="{{ p.idp_icon | mxc_to_http(32, 32) }}"/> - {% endif %} - <span>{{ p.idp_name }}</span> - </a> - </li> - {% endfor %} - </ul> - </main> - {% include "sso_footer.html" without context %} - </body> -</html> + .providers a { + display: block; + border-radius: 4px; + border: 1px solid #17191C; + padding: 8px; + text-align: center; + text-decoration: none; + color: #17191C; + display: flex; + align-items: center; + font-weight: bold; + } + + .providers a img { + width: 24px; + height: 24px; + } + .providers a span { + flex: 1; + } +</style> +{% endblock %} + +{% block body %} +<header> + <h1>Log in to {{ server_name }} </h1> + <p>Choose an identity provider to log in</p> +</header> +<main> + <ul class="providers"> + {% for p in providers %} + <li> + <a href="pick_idp?idp={{ p.idp_id }}&redirectUrl={{ redirect_url | urlencode }}"> + {% if p.idp_icon %} + <img src="{{ p.idp_icon | mxc_to_http(32, 32) }}"/> + {% endif %} + <span>{{ p.idp_name }}</span> + </a> + </li> + {% endfor %} + </ul> +</main> +{% include "sso_footer.html" without context %} +{% endblock %} diff --git a/synapse/res/templates/sso_new_user_consent.html b/synapse/res/templates/sso_new_user_consent.html index 523f64c4fc..fda29928d1 100644 --- a/synapse/res/templates/sso_new_user_consent.html +++ b/synapse/res/templates/sso_new_user_consent.html @@ -1,33 +1,30 @@ -<!DOCTYPE html> -<html lang="en"> -<head> - <meta charset="UTF-8"> - <title>Agree to terms and conditions</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <style type="text/css"> - {% include "sso.css" without context %} +{% extends "_base.html" %} +{% block title %}Agree to terms and conditions{% endblock %} - #consent_form { - margin-top: 56px; - } - </style> -</head> - <body> - <header> - <h1>Your account is nearly ready</h1> - <p>Agree to the terms to create your account.</p> - </header> - <main> - {% include "sso_partial_profile.html" %} - <form method="post" action="{{my_url}}" id="consent_form"> - <p> - <input id="accepted_version" type="checkbox" name="accepted_version" value="{{ consent_version }}" required> - <label for="accepted_version">I have read and agree to the <a href="{{ terms_url }}" target="_blank" rel="noopener">terms and conditions</a>.</label> - </p> - <input type="submit" class="primary-button" value="Continue"/> - </form> - </main> - {% include "sso_footer.html" without context %} - </body> -</html> +{% block header %} +<style type="text/css"> + {% include "sso.css" without context %} + + #consent_form { + margin-top: 56px; + } +</style> +{% endblock %} + +{% block body %} +<header> + <h1>Your account is nearly ready</h1> + <p>Agree to the terms to create your account.</p> +</header> +<main> + {% include "sso_partial_profile.html" %} + <form method="post" action="{{my_url}}" id="consent_form"> + <p> + <input id="accepted_version" type="checkbox" name="accepted_version" value="{{ consent_version }}" required> + <label for="accepted_version">I have read and agree to the <a href="{{ terms_url }}" target="_blank" rel="noopener">terms and conditions</a>.</label> + </p> + <input type="submit" class="primary-button" value="Continue"/> + </form> +</main> +{% include "sso_footer.html" without context %} +{% endblock %} diff --git a/synapse/res/templates/sso_redirect_confirm.html b/synapse/res/templates/sso_redirect_confirm.html index 1049a9bd92..cc2e7b3a5b 100644 --- a/synapse/res/templates/sso_redirect_confirm.html +++ b/synapse/res/templates/sso_redirect_confirm.html @@ -1,41 +1,39 @@ -<!DOCTYPE html> -<html lang="en"> -<head> - <meta charset="UTF-8"> - <title>Continue to your account</title> - <meta http-equiv="X-UA-Compatible" content="IE=edge"> - <meta name="viewport" content="width=device-width, initial-scale=1.0"> - <style type="text/css"> - {% include "sso.css" without context %} +{% extends "_base.html" %} +{% block title %}Continue to your account{% endblock %} - .confirm-trust { - margin: 34px 0; - color: #8D99A5; - } - .confirm-trust strong { - color: #17191C; - } +{% block header %} +<style type="text/css"> + {% include "sso.css" without context %} - .confirm-trust::before { - content: ""; - background-image: url('data:image/svg+xml;base64,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'); - background-repeat: no-repeat; - width: 24px; - height: 24px; - display: block; - float: left; - } - </style> -</head> - <body> - <header> - <h1>Continue to your account</h1> - </header> - <main> - {% include "sso_partial_profile.html" %} - <p class="confirm-trust">Continuing will grant <strong>{{ display_url }}</strong> access to your account.</p> - <a href="{{ redirect_url }}" class="primary-button">Continue</a> - </main> - {% include "sso_footer.html" without context %} - </body> -</html> + .confirm-trust { + margin: 34px 0; + color: #8D99A5; + } + .confirm-trust strong { + color: #17191C; + } + + .confirm-trust::before { + content: ""; + background-image: url('data:image/svg+xml;base64,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'); + background-repeat: no-repeat; + width: 24px; + height: 24px; + display: block; + float: left; + } +</style> +{% endblock %} + +{% block body %} +<header> + <h1>Continue to your account</h1> +</header> +<main> + {% include "sso_partial_profile.html" %} + <p class="confirm-trust">Continuing will grant <strong>{{ display_url }}</strong> access to your account.</p> + <a href="{{ redirect_url }}" class="primary-button">Continue</a> +</main> +{% include "sso_footer.html" without context %} + +{% endblock %} diff --git a/synapse/res/templates/style.css b/synapse/res/templates/style.css new file mode 100644 index 0000000000..097b235ae5 --- /dev/null +++ b/synapse/res/templates/style.css @@ -0,0 +1,29 @@ +html { + height: 100%; +} + +body { + background: #f9fafb; + max-width: 680px; + margin: auto; + font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, Helvetica, Arial, sans-serif, "Apple Color Emoji", "Segoe UI Emoji", "Segoe UI Symbol"; +} + +.mx_Header { + border-bottom: 3px solid #ddd; + margin-bottom: 1rem; + padding-top: 1rem; + padding-bottom: 1rem; + text-align: center; +} + +@media screen and (max-width: 1120px) { + body { + font-size: 20px; + } + + h1 { font-size: 1rem; } + h2 { font-size: .9rem; } + h3 { font-size: .85rem; } + h4 { font-size: .8rem; } +} diff --git a/synapse/res/templates/terms.html b/synapse/res/templates/terms.html index 2081d990ab..ffabebdd8b 100644 --- a/synapse/res/templates/terms.html +++ b/synapse/res/templates/terms.html @@ -1,11 +1,11 @@ -<html> -<head> -<title>Authentication</title> -<meta http-equiv="X-UA-Compatible" content="IE=edge"> -<meta name="viewport" content="width=device-width, initial-scale=1.0"> +{% extends "_base.html" %} +{% block title %}Authentication{% endblock %} + +{% block header %} <link rel="stylesheet" href="/_matrix/static/client/register/style.css"> -</head> -<body> +{% endblock %} + +{% block body %} <form id="registrationForm" method="post" action="{{ myurl }}"> <div> {% if error is defined %} @@ -19,5 +19,4 @@ <input type="submit" value="Agree" /> </div> </form> -</body> -</html> +{% endblock %} diff --git a/synapse/rest/__init__.py b/synapse/rest/__init__.py index 9a2ab99ede..14c4e6ebbb 100644 --- a/synapse/rest/__init__.py +++ b/synapse/rest/__init__.py @@ -29,7 +29,7 @@ from synapse.rest.client import ( initial_sync, keys, knock, - login as v1_login, + login, login_token_request, logout, mutual_rooms, @@ -44,6 +44,7 @@ from synapse.rest.client import ( receipts, register, relations, + rendezvous, report_event, room, room_batch, @@ -81,6 +82,10 @@ class ClientRestResource(JsonResource): @staticmethod def register_servlets(client_resource: HttpServer, hs: "HomeServer") -> None: + # Some servlets are only registered on the main process (and not worker + # processes). + is_main_process = hs.config.worker.worker_app is None + versions.register_servlets(hs, client_resource) # Deprecated in r0 @@ -91,44 +96,58 @@ class ClientRestResource(JsonResource): events.register_servlets(hs, client_resource) room.register_servlets(hs, client_resource) - v1_login.register_servlets(hs, client_resource) + login.register_servlets(hs, client_resource) profile.register_servlets(hs, client_resource) presence.register_servlets(hs, client_resource) - directory.register_servlets(hs, client_resource) + if is_main_process: + directory.register_servlets(hs, client_resource) voip.register_servlets(hs, client_resource) - pusher.register_servlets(hs, client_resource) + if is_main_process: + pusher.register_servlets(hs, client_resource) push_rule.register_servlets(hs, client_resource) - logout.register_servlets(hs, client_resource) + if is_main_process: + logout.register_servlets(hs, client_resource) sync.register_servlets(hs, client_resource) - filter.register_servlets(hs, client_resource) + if is_main_process: + filter.register_servlets(hs, client_resource) account.register_servlets(hs, client_resource) register.register_servlets(hs, client_resource) - auth.register_servlets(hs, client_resource) + if is_main_process: + auth.register_servlets(hs, client_resource) receipts.register_servlets(hs, client_resource) read_marker.register_servlets(hs, client_resource) room_keys.register_servlets(hs, client_resource) keys.register_servlets(hs, client_resource) - tokenrefresh.register_servlets(hs, client_resource) + if is_main_process: + tokenrefresh.register_servlets(hs, client_resource) tags.register_servlets(hs, client_resource) account_data.register_servlets(hs, client_resource) - report_event.register_servlets(hs, client_resource) - openid.register_servlets(hs, client_resource) - notifications.register_servlets(hs, client_resource) + if is_main_process: + report_event.register_servlets(hs, client_resource) + openid.register_servlets(hs, client_resource) + notifications.register_servlets(hs, client_resource) devices.register_servlets(hs, client_resource) - thirdparty.register_servlets(hs, client_resource) + if is_main_process: + thirdparty.register_servlets(hs, client_resource) sendtodevice.register_servlets(hs, client_resource) user_directory.register_servlets(hs, client_resource) - room_upgrade_rest_servlet.register_servlets(hs, client_resource) + if is_main_process: + room_upgrade_rest_servlet.register_servlets(hs, client_resource) room_batch.register_servlets(hs, client_resource) - capabilities.register_servlets(hs, client_resource) - account_validity.register_servlets(hs, client_resource) + if is_main_process: + capabilities.register_servlets(hs, client_resource) + account_validity.register_servlets(hs, client_resource) relations.register_servlets(hs, client_resource) - password_policy.register_servlets(hs, client_resource) - knock.register_servlets(hs, client_resource) + if is_main_process: + password_policy.register_servlets(hs, client_resource) + knock.register_servlets(hs, client_resource) # moving to /_synapse/admin - admin.register_servlets_for_client_rest_resource(hs, client_resource) + if is_main_process: + admin.register_servlets_for_client_rest_resource(hs, client_resource) # unstable - mutual_rooms.register_servlets(hs, client_resource) - login_token_request.register_servlets(hs, client_resource) + if is_main_process: + mutual_rooms.register_servlets(hs, client_resource) + login_token_request.register_servlets(hs, client_resource) + rendezvous.register_servlets(hs, client_resource) diff --git a/synapse/rest/admin/__init__.py b/synapse/rest/admin/__init__.py index 885669f9c7..fb73886df0 100644 --- a/synapse/rest/admin/__init__.py +++ b/synapse/rest/admin/__init__.py @@ -81,6 +81,7 @@ from synapse.rest.admin.users import ( ShadowBanRestServlet, UserAdminServlet, UserByExternalId, + UserByThreePid, UserMembershipRestServlet, UserRegisterServlet, UserRestServletV2, @@ -237,6 +238,10 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: """ Register all the admin servlets. """ + # Admin servlets aren't registered on workers. + if hs.config.worker.worker_app is not None: + return + register_servlets_for_client_rest_resource(hs, http_server) BlockRoomRestServlet(hs).register(http_server) ListRoomRestServlet(hs).register(http_server) @@ -253,9 +258,6 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: UserTokenRestServlet(hs).register(http_server) UserRestServletV2(hs).register(http_server) UsersRestServletV2(hs).register(http_server) - DeviceRestServlet(hs).register(http_server) - DevicesRestServlet(hs).register(http_server) - DeleteDevicesRestServlet(hs).register(http_server) UserMediaStatisticsRestServlet(hs).register(http_server) EventReportDetailRestServlet(hs).register(http_server) EventReportsRestServlet(hs).register(http_server) @@ -277,13 +279,15 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: RoomMessagesRestServlet(hs).register(http_server) RoomTimestampToEventRestServlet(hs).register(http_server) UserByExternalId(hs).register(http_server) + UserByThreePid(hs).register(http_server) - # 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) - BackgroundUpdateStartJobRestServlet(hs).register(http_server) + DeviceRestServlet(hs).register(http_server) + DevicesRestServlet(hs).register(http_server) + DeleteDevicesRestServlet(hs).register(http_server) + SendServerNoticeServlet(hs).register(http_server) + BackgroundUpdateEnabledRestServlet(hs).register(http_server) + BackgroundUpdateRestServlet(hs).register(http_server) + BackgroundUpdateStartJobRestServlet(hs).register(http_server) def register_servlets_for_client_rest_resource( @@ -292,9 +296,11 @@ def register_servlets_for_client_rest_resource( """Register only the servlets which need to be exposed on /_matrix/client/xxx""" WhoisRestServlet(hs).register(http_server) PurgeHistoryStatusRestServlet(hs).register(http_server) - DeactivateAccountRestServlet(hs).register(http_server) PurgeHistoryRestServlet(hs).register(http_server) - ResetPasswordRestServlet(hs).register(http_server) + # The following resources can only be run on the main process. + if hs.config.worker.worker_app is None: + DeactivateAccountRestServlet(hs).register(http_server) + ResetPasswordRestServlet(hs).register(http_server) SearchUsersRestServlet(hs).register(http_server) UserRegisterServlet(hs).register(http_server) AccountValidityRenewServlet(hs).register(http_server) diff --git a/synapse/rest/admin/devices.py b/synapse/rest/admin/devices.py index d934880102..3b2f2d9abb 100644 --- a/synapse/rest/admin/devices.py +++ b/synapse/rest/admin/devices.py @@ -16,6 +16,7 @@ from http import HTTPStatus from typing import TYPE_CHECKING, Tuple from synapse.api.errors import NotFoundError, SynapseError +from synapse.handlers.device import DeviceHandler from synapse.http.servlet import ( RestServlet, assert_params_in_dict, @@ -43,7 +44,9 @@ class DeviceRestServlet(RestServlet): def __init__(self, hs: "HomeServer"): super().__init__() self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.store = hs.get_datastores().main self.is_mine = hs.is_mine @@ -112,7 +115,9 @@ class DevicesRestServlet(RestServlet): def __init__(self, hs: "HomeServer"): self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.store = hs.get_datastores().main self.is_mine = hs.is_mine @@ -143,7 +148,9 @@ class DeleteDevicesRestServlet(RestServlet): def __init__(self, hs: "HomeServer"): self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.store = hs.get_datastores().main self.is_mine = hs.is_mine diff --git a/synapse/rest/admin/rooms.py b/synapse/rest/admin/rooms.py index 747e6fda83..e957aa28ca 100644 --- a/synapse/rest/admin/rooms.py +++ b/synapse/rest/admin/rooms.py @@ -34,9 +34,9 @@ from synapse.rest.admin._base import ( assert_user_is_admin, ) from synapse.storage.databases.main.room import RoomSortOrder -from synapse.storage.state import StateFilter from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, RoomID, UserID, create_requester +from synapse.types.state import StateFilter from synapse.util import json_decoder if TYPE_CHECKING: diff --git a/synapse/rest/admin/users.py b/synapse/rest/admin/users.py index 15ac2059aa..6e0c44be2a 100644 --- a/synapse/rest/admin/users.py +++ b/synapse/rest/admin/users.py @@ -903,8 +903,9 @@ class PushersRestServlet(RestServlet): @user:server/pushers Returns: - pushers: Dictionary containing pushers information. - total: Number of pushers in dictionary `pushers`. + A dictionary with keys: + pushers: Dictionary containing pushers information. + total: Number of pushers in dictionary `pushers`. """ PATTERNS = admin_patterns("/users/(?P<user_id>[^/]*)/pushers$") @@ -1224,3 +1225,28 @@ class UserByExternalId(RestServlet): raise NotFoundError("User not found") return HTTPStatus.OK, {"user_id": user_id} + + +class UserByThreePid(RestServlet): + """Find a user based on 3PID of a particular medium""" + + PATTERNS = admin_patterns("/threepid/(?P<medium>[^/]*)/users/(?P<address>[^/]*)") + + def __init__(self, hs: "HomeServer"): + self._auth = hs.get_auth() + self._store = hs.get_datastores().main + + async def on_GET( + self, + request: SynapseRequest, + medium: str, + address: str, + ) -> Tuple[int, JsonDict]: + await assert_requester_is_admin(self._auth, request) + + user_id = await self._store.get_user_id_by_threepid(medium, address) + + if user_id is None: + raise NotFoundError("User not found") + + return HTTPStatus.OK, {"user_id": user_id} diff --git a/synapse/rest/client/account.py b/synapse/rest/client/account.py index 44f622bcce..b4b92f0c99 100644 --- a/synapse/rest/client/account.py +++ b/synapse/rest/client/account.py @@ -875,19 +875,21 @@ class AccountStatusRestServlet(RestServlet): def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: - EmailPasswordRequestTokenRestServlet(hs).register(http_server) - PasswordRestServlet(hs).register(http_server) - DeactivateAccountRestServlet(hs).register(http_server) - EmailThreepidRequestTokenRestServlet(hs).register(http_server) - MsisdnThreepidRequestTokenRestServlet(hs).register(http_server) - AddThreepidEmailSubmitTokenServlet(hs).register(http_server) - AddThreepidMsisdnSubmitTokenServlet(hs).register(http_server) + if hs.config.worker.worker_app is None: + EmailPasswordRequestTokenRestServlet(hs).register(http_server) + PasswordRestServlet(hs).register(http_server) + DeactivateAccountRestServlet(hs).register(http_server) + EmailThreepidRequestTokenRestServlet(hs).register(http_server) + MsisdnThreepidRequestTokenRestServlet(hs).register(http_server) + AddThreepidEmailSubmitTokenServlet(hs).register(http_server) + AddThreepidMsisdnSubmitTokenServlet(hs).register(http_server) ThreepidRestServlet(hs).register(http_server) - ThreepidAddRestServlet(hs).register(http_server) - ThreepidBindRestServlet(hs).register(http_server) - ThreepidUnbindRestServlet(hs).register(http_server) - ThreepidDeleteRestServlet(hs).register(http_server) + if hs.config.worker.worker_app is None: + ThreepidAddRestServlet(hs).register(http_server) + ThreepidBindRestServlet(hs).register(http_server) + ThreepidUnbindRestServlet(hs).register(http_server) + ThreepidDeleteRestServlet(hs).register(http_server) WhoamiRestServlet(hs).register(http_server) - if hs.config.experimental.msc3720_enabled: + if hs.config.worker.worker_app is None and hs.config.experimental.msc3720_enabled: AccountStatusRestServlet(hs).register(http_server) diff --git a/synapse/rest/client/capabilities.py b/synapse/rest/client/capabilities.py index 4237071c61..e84dde31b1 100644 --- a/synapse/rest/client/capabilities.py +++ b/synapse/rest/client/capabilities.py @@ -77,6 +77,11 @@ class CapabilitiesRestServlet(RestServlet): "enabled": True, } + if self.config.experimental.msc3664_enabled: + response["capabilities"]["im.nheko.msc3664.related_event_match"] = { + "enabled": self.config.experimental.msc3664_enabled, + } + return HTTPStatus.OK, response diff --git a/synapse/rest/client/devices.py b/synapse/rest/client/devices.py index 90828c95c4..486c6dbbc5 100644 --- a/synapse/rest/client/devices.py +++ b/synapse/rest/client/devices.py @@ -20,6 +20,7 @@ from pydantic import Extra, StrictStr from synapse.api import errors from synapse.api.errors import NotFoundError +from synapse.handlers.device import DeviceHandler from synapse.http.server import HttpServer from synapse.http.servlet import ( RestServlet, @@ -80,7 +81,9 @@ class DeleteDevicesRestServlet(RestServlet): super().__init__() self.hs = hs self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.auth_handler = hs.get_auth_handler() class PostBody(RequestBodyModel): @@ -125,7 +128,9 @@ class DeviceRestServlet(RestServlet): super().__init__() self.hs = hs self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler self.auth_handler = hs.get_auth_handler() self._msc3852_enabled = hs.config.experimental.msc3852_enabled @@ -231,7 +236,7 @@ class DehydratedDeviceServlet(RestServlet): } } - PUT /org.matrix.msc2697/dehydrated_device + PUT /org.matrix.msc2697.v2/dehydrated_device Content-Type: application/json { @@ -256,7 +261,9 @@ class DehydratedDeviceServlet(RestServlet): super().__init__() self.hs = hs self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler async def on_GET(self, request: SynapseRequest) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request) @@ -271,7 +278,6 @@ class DehydratedDeviceServlet(RestServlet): raise errors.NotFoundError("No dehydrated device available") class PutBody(RequestBodyModel): - device_id: StrictStr device_data: DehydratedDeviceDataModel initial_device_display_name: Optional[StrictStr] @@ -281,7 +287,7 @@ class DehydratedDeviceServlet(RestServlet): device_id = await self.device_handler.store_dehydrated_device( requester.user.to_string(), - submission.device_data, + submission.device_data.dict(), submission.initial_device_display_name, ) return 200, {"device_id": device_id} @@ -314,7 +320,9 @@ class ClaimDehydratedDeviceServlet(RestServlet): super().__init__() self.hs = hs self.auth = hs.get_auth() - self.device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self.device_handler = handler class PostBody(RequestBodyModel): device_id: StrictStr @@ -334,8 +342,10 @@ class ClaimDehydratedDeviceServlet(RestServlet): def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: - DeleteDevicesRestServlet(hs).register(http_server) + if hs.config.worker.worker_app is None: + 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) + if hs.config.worker.worker_app is None: + DeviceRestServlet(hs).register(http_server) + DehydratedDeviceServlet(hs).register(http_server) + ClaimDehydratedDeviceServlet(hs).register(http_server) diff --git a/synapse/rest/client/directory.py b/synapse/rest/client/directory.py index bc1b18c92d..f17b4c8d22 100644 --- a/synapse/rest/client/directory.py +++ b/synapse/rest/client/directory.py @@ -13,15 +13,22 @@ # limitations under the License. import logging -from typing import TYPE_CHECKING, Tuple +from typing import TYPE_CHECKING, List, Optional, Tuple + +from pydantic import StrictStr +from typing_extensions import Literal from twisted.web.server import Request from synapse.api.errors import AuthError, Codes, NotFoundError, SynapseError from synapse.http.server import HttpServer -from synapse.http.servlet import RestServlet, parse_json_object_from_request +from synapse.http.servlet import ( + RestServlet, + parse_and_validate_json_object_from_request, +) from synapse.http.site import SynapseRequest from synapse.rest.client._base import client_patterns +from synapse.rest.models import RequestBodyModel from synapse.types import JsonDict, RoomAlias if TYPE_CHECKING: @@ -54,6 +61,12 @@ class ClientDirectoryServer(RestServlet): return 200, res + class PutBody(RequestBodyModel): + # TODO: get Pydantic to validate that this is a valid room id? + room_id: StrictStr + # `servers` is unspecced + servers: Optional[List[StrictStr]] = None + async def on_PUT( self, request: SynapseRequest, room_alias: str ) -> Tuple[int, JsonDict]: @@ -61,31 +74,22 @@ class ClientDirectoryServer(RestServlet): raise SynapseError(400, "Room alias invalid", errcode=Codes.INVALID_PARAM) room_alias_obj = RoomAlias.from_string(room_alias) - content = parse_json_object_from_request(request) - if "room_id" not in content: - raise SynapseError( - 400, 'Missing params: ["room_id"]', errcode=Codes.BAD_JSON - ) + content = parse_and_validate_json_object_from_request(request, self.PutBody) logger.debug("Got content: %s", content) logger.debug("Got room name: %s", room_alias_obj.to_string()) - room_id = content["room_id"] - servers = content["servers"] if "servers" in content else None - - logger.debug("Got room_id: %s", room_id) - logger.debug("Got servers: %s", servers) + logger.debug("Got room_id: %s", content.room_id) + logger.debug("Got servers: %s", content.servers) - # TODO(erikj): Check types. - - room = await self.store.get_room(room_id) + room = await self.store.get_room(content.room_id) if room is None: raise SynapseError(400, "Room does not exist") requester = await self.auth.get_user_by_req(request) await self.directory_handler.create_association( - requester, room_alias_obj, room_id, servers + requester, room_alias_obj, content.room_id, content.servers ) return 200, {} @@ -137,16 +141,18 @@ class ClientDirectoryListServer(RestServlet): return 200, {"visibility": "public" if room["is_public"] else "private"} + class PutBody(RequestBodyModel): + visibility: Literal["public", "private"] = "public" + async def on_PUT( self, request: SynapseRequest, room_id: str ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request) - content = parse_json_object_from_request(request) - visibility = content.get("visibility", "public") + content = parse_and_validate_json_object_from_request(request, self.PutBody) await self.directory_handler.edit_published_room_list( - requester, room_id, visibility + requester, room_id, content.visibility ) return 200, {} @@ -163,12 +169,14 @@ class ClientAppserviceDirectoryListServer(RestServlet): self.directory_handler = hs.get_directory_handler() self.auth = hs.get_auth() + class PutBody(RequestBodyModel): + visibility: Literal["public", "private"] = "public" + async def on_PUT( self, request: SynapseRequest, network_id: str, room_id: str ) -> Tuple[int, JsonDict]: - content = parse_json_object_from_request(request) - visibility = content.get("visibility", "public") - return await self._edit(request, network_id, room_id, visibility) + content = parse_and_validate_json_object_from_request(request, self.PutBody) + return await self._edit(request, network_id, room_id, content.visibility) async def on_DELETE( self, request: SynapseRequest, network_id: str, room_id: str @@ -176,7 +184,11 @@ class ClientAppserviceDirectoryListServer(RestServlet): return await self._edit(request, network_id, room_id, "private") async def _edit( - self, request: SynapseRequest, network_id: str, room_id: str, visibility: str + self, + request: SynapseRequest, + network_id: str, + room_id: str, + visibility: Literal["public", "private"], ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request) if not requester.app_service: diff --git a/synapse/rest/client/events.py b/synapse/rest/client/events.py index 916f5230f1..782e7d14e8 100644 --- a/synapse/rest/client/events.py +++ b/synapse/rest/client/events.py @@ -50,7 +50,9 @@ class EventStreamRestServlet(RestServlet): raise SynapseError(400, "Guest users must specify room_id param") room_id = parse_string(request, "room_id") - pagin_config = await PaginationConfig.from_request(self.store, request) + pagin_config = await PaginationConfig.from_request( + self.store, request, default_limit=10 + ) timeout = EventStreamRestServlet.DEFAULT_LONGPOLL_TIME_MS if b"timeout" in args: try: diff --git a/synapse/rest/client/initial_sync.py b/synapse/rest/client/initial_sync.py index cfadcb8e50..9b1bb8b521 100644 --- a/synapse/rest/client/initial_sync.py +++ b/synapse/rest/client/initial_sync.py @@ -39,7 +39,9 @@ class InitialSyncRestServlet(RestServlet): requester = await self.auth.get_user_by_req(request) args: Dict[bytes, List[bytes]] = request.args # type: ignore as_client_event = b"raw" not in args - pagination_config = await PaginationConfig.from_request(self.store, request) + pagination_config = await PaginationConfig.from_request( + self.store, request, default_limit=10 + ) include_archived = parse_boolean(request, "archived", default=False) content = await self.initial_sync_handler.snapshot_all_rooms( user_id=requester.user.to_string(), diff --git a/synapse/rest/client/keys.py b/synapse/rest/client/keys.py index f653d2a3e1..7873b363c0 100644 --- a/synapse/rest/client/keys.py +++ b/synapse/rest/client/keys.py @@ -27,6 +27,7 @@ from synapse.http.servlet import ( ) from synapse.http.site import SynapseRequest from synapse.logging.opentracing import log_kv, set_tag +from synapse.replication.http.devices import ReplicationUploadKeysForUserRestServlet from synapse.rest.client._base import client_patterns, interactive_auth_handler from synapse.types import JsonDict, StreamToken from synapse.util.cancellation import cancellable @@ -43,24 +44,48 @@ class KeyUploadServlet(RestServlet): Content-Type: application/json { - "device_keys": { - "user_id": "<user_id>", - "device_id": "<device_id>", - "valid_until_ts": <millisecond_timestamp>, - "algorithms": [ - "m.olm.curve25519-aes-sha2", - ] - "keys": { - "<algorithm>:<device_id>": "<key_base64>", + "device_keys": { + "user_id": "<user_id>", + "device_id": "<device_id>", + "valid_until_ts": <millisecond_timestamp>, + "algorithms": [ + "m.olm.curve25519-aes-sha2", + ] + "keys": { + "<algorithm>:<device_id>": "<key_base64>", + }, + "signatures:" { + "<user_id>" { + "<algorithm>:<device_id>": "<signature_base64>" + } + } + }, + "fallback_keys": { + "<algorithm>:<device_id>": "<key_base64>", + "signed_<algorithm>:<device_id>": { + "fallback": true, + "key": "<key_base64>", + "signatures": { + "<user_id>": { + "<algorithm>:<device_id>": "<key_base64>" + } + } + } + } + "one_time_keys": { + "<algorithm>:<key_id>": "<key_base64>" }, - "signatures:" { - "<user_id>" { - "<algorithm>:<device_id>": "<signature_base64>" - } } }, - "one_time_keys": { - "<algorithm>:<key_id>": "<key_base64>" - }, } + + response, e.g.: + + { + "one_time_key_counts": { + "curve25519": 10, + "signed_curve25519": 20 + } + } + """ PATTERNS = client_patterns("/keys/upload(/(?P<device_id>[^/]+))?$") @@ -71,6 +96,13 @@ class KeyUploadServlet(RestServlet): self.e2e_keys_handler = hs.get_e2e_keys_handler() self.device_handler = hs.get_device_handler() + if hs.config.worker.worker_app is None: + # if main process + self.key_uploader = self.e2e_keys_handler.upload_keys_for_user + else: + # then a worker + self.key_uploader = ReplicationUploadKeysForUserRestServlet.make_client(hs) + async def on_POST( self, request: SynapseRequest, device_id: Optional[str] ) -> Tuple[int, JsonDict]: @@ -109,8 +141,8 @@ class KeyUploadServlet(RestServlet): 400, "To upload keys, you must pass device_id when authenticating" ) - result = await self.e2e_keys_handler.upload_keys_for_user( - user_id, device_id, body + result = await self.key_uploader( + user_id=user_id, device_id=device_id, keys=body ) return 200, result @@ -344,5 +376,6 @@ def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: KeyQueryServlet(hs).register(http_server) KeyChangesServlet(hs).register(http_server) OneTimeKeyServlet(hs).register(http_server) - SigningKeyUploadServlet(hs).register(http_server) - SignaturesUploadServlet(hs).register(http_server) + if hs.config.worker.worker_app is None: + SigningKeyUploadServlet(hs).register(http_server) + SignaturesUploadServlet(hs).register(http_server) diff --git a/synapse/rest/client/login.py b/synapse/rest/client/login.py index f554586ac3..8adced41e5 100644 --- a/synapse/rest/client/login.py +++ b/synapse/rest/client/login.py @@ -350,7 +350,7 @@ class LoginRestServlet(RestServlet): auth_provider_session_id: The session ID got during login from the SSO IdP. Returns: - result: Dictionary of account information after successful login. + Dictionary of account information after successful login. """ # Before we actually log them in we check if they've already logged in @@ -436,8 +436,7 @@ class LoginRestServlet(RestServlet): The body of the JSON response. """ token = login_submission["token"] - auth_handler = self.auth_handler - res = await auth_handler.validate_short_term_login_token(token) + res = await self.auth_handler.consume_login_token(token) return await self._complete_login( res.user_id, @@ -537,7 +536,7 @@ def _get_auth_flow_dict_for_idp(idp: SsoIdentityProvider) -> JsonDict: class RefreshTokenServlet(RestServlet): - PATTERNS = (re.compile("^/_matrix/client/v1/refresh$"),) + PATTERNS = client_patterns("/refresh$") def __init__(self, hs: "HomeServer"): self._auth_handler = hs.get_auth_handler() diff --git a/synapse/rest/client/login_token_request.py b/synapse/rest/client/login_token_request.py index 277b20fb63..43ea21d5e6 100644 --- a/synapse/rest/client/login_token_request.py +++ b/synapse/rest/client/login_token_request.py @@ -57,7 +57,6 @@ class LoginTokenRequestServlet(RestServlet): self.store = hs.get_datastores().main self.clock = hs.get_clock() self.server_name = hs.config.server.server_name - self.macaroon_gen = hs.get_macaroon_generator() self.auth_handler = hs.get_auth_handler() self.token_timeout = hs.config.experimental.msc3882_token_timeout self.ui_auth = hs.config.experimental.msc3882_ui_auth @@ -76,10 +75,10 @@ class LoginTokenRequestServlet(RestServlet): can_skip_ui_auth=False, # Don't allow skipping of UI auth ) - login_token = self.macaroon_gen.generate_short_term_login_token( + login_token = await self.auth_handler.create_login_token_for_user_id( user_id=requester.user.to_string(), auth_provider_id="org.matrix.msc3882.login_token_request", - duration_in_ms=self.token_timeout, + duration_ms=self.token_timeout, ) return ( diff --git a/synapse/rest/client/logout.py b/synapse/rest/client/logout.py index 23dfa4518f..6d34625ad5 100644 --- a/synapse/rest/client/logout.py +++ b/synapse/rest/client/logout.py @@ -15,6 +15,7 @@ import logging from typing import TYPE_CHECKING, Tuple +from synapse.handlers.device import DeviceHandler from synapse.http.server import HttpServer from synapse.http.servlet import RestServlet from synapse.http.site import SynapseRequest @@ -34,7 +35,9 @@ class LogoutRestServlet(RestServlet): super().__init__() self.auth = hs.get_auth() self._auth_handler = hs.get_auth_handler() - self._device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self._device_handler = handler async def on_POST(self, request: SynapseRequest) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request, allow_expired=True) @@ -59,7 +62,9 @@ class LogoutAllRestServlet(RestServlet): super().__init__() self.auth = hs.get_auth() self._auth_handler = hs.get_auth_handler() - self._device_handler = hs.get_device_handler() + handler = hs.get_device_handler() + assert isinstance(handler, DeviceHandler) + self._device_handler = handler async def on_POST(self, request: SynapseRequest) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request, allow_expired=True) diff --git a/synapse/rest/client/receipts.py b/synapse/rest/client/receipts.py index 14dec7ac4e..28b7d30ea8 100644 --- a/synapse/rest/client/receipts.py +++ b/synapse/rest/client/receipts.py @@ -15,12 +15,12 @@ import logging from typing import TYPE_CHECKING, Tuple -from synapse.api.constants import ReceiptTypes +from synapse.api.constants import MAIN_TIMELINE, ReceiptTypes from synapse.api.errors import Codes, SynapseError from synapse.http.server import HttpServer from synapse.http.servlet import RestServlet, parse_json_object_from_request from synapse.http.site import SynapseRequest -from synapse.types import JsonDict +from synapse.types import EventID, JsonDict, RoomID from ._base import client_patterns @@ -56,6 +56,9 @@ class ReceiptRestServlet(RestServlet): ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request) + if not RoomID.is_valid(room_id) or not event_id.startswith(EventID.SIGIL): + raise SynapseError(400, "A valid room ID and event ID must be specified") + if receipt_type not in self._known_receipt_types: raise SynapseError( 400, @@ -83,7 +86,7 @@ class ReceiptRestServlet(RestServlet): ) # Ensure the event ID roughly correlates to the thread ID. - if thread_id != await self._main_store.get_thread_id(event_id): + if not await self._is_event_in_thread(event_id, thread_id): raise SynapseError( 400, f"event_id {event_id} is not related to thread {thread_id}", @@ -109,6 +112,46 @@ class ReceiptRestServlet(RestServlet): return 200, {} + async def _is_event_in_thread(self, event_id: str, thread_id: str) -> bool: + """ + The event must be related to the thread ID (in a vague sense) to ensure + clients aren't sending bogus receipts. + + A thread ID is considered valid for a given event E if: + + 1. E has a thread relation which matches the thread ID; + 2. E has another event which has a thread relation to E matching the + thread ID; or + 3. E is recursively related (via any rel_type) to an event which + satisfies 1 or 2. + + Given the following DAG: + + A <---[m.thread]-- B <--[m.annotation]-- C + ^ + |--[m.reference]-- D <--[m.annotation]-- E + + It is valid to send a receipt for thread A on A, B, C, D, or E. + + It is valid to send a receipt for the main timeline on A, D, and E. + + Args: + event_id: The event ID to check. + thread_id: The thread ID the event is potentially part of. + + Returns: + True if the event belongs to the given thread, otherwise False. + """ + + # If the receipt is on the main timeline, it is enough to check whether + # the event is directly related to a thread. + if thread_id == MAIN_TIMELINE: + return MAIN_TIMELINE == await self._main_store.get_thread_id(event_id) + + # Otherwise, check if the event is directly part of a thread, or is the + # root message (or related to the root message) of a thread. + return thread_id == await self._main_store.get_thread_id_for_receipts(event_id) + def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: ReceiptRestServlet(hs).register(http_server) diff --git a/synapse/rest/client/register.py b/synapse/rest/client/register.py index de810ae3ec..3cb1e7e375 100644 --- a/synapse/rest/client/register.py +++ b/synapse/rest/client/register.py @@ -949,9 +949,10 @@ def _calculate_registration_flows( def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: - EmailRegisterRequestTokenRestServlet(hs).register(http_server) - MsisdnRegisterRequestTokenRestServlet(hs).register(http_server) - UsernameAvailabilityRestServlet(hs).register(http_server) - RegistrationSubmitTokenServlet(hs).register(http_server) + if hs.config.worker.worker_app is None: + EmailRegisterRequestTokenRestServlet(hs).register(http_server) + MsisdnRegisterRequestTokenRestServlet(hs).register(http_server) + UsernameAvailabilityRestServlet(hs).register(http_server) + RegistrationSubmitTokenServlet(hs).register(http_server) RegistrationTokenValidityRestServlet(hs).register(http_server) RegisterRestServlet(hs).register(http_server) diff --git a/synapse/rest/client/relations.py b/synapse/rest/client/relations.py index 7a25de5c85..9dd59196d9 100644 --- a/synapse/rest/client/relations.py +++ b/synapse/rest/client/relations.py @@ -13,13 +13,17 @@ # limitations under the License. import logging +import re from typing import TYPE_CHECKING, Optional, Tuple +from synapse.handlers.relations import ThreadsListInclude from synapse.http.server import HttpServer from synapse.http.servlet import RestServlet, parse_integer, parse_string from synapse.http.site import SynapseRequest from synapse.rest.client._base import client_patterns -from synapse.types import JsonDict, StreamToken +from synapse.storage.databases.main.relations import ThreadsNextBatch +from synapse.streams.config import PaginationConfig +from synapse.types import JsonDict if TYPE_CHECKING: from synapse.server import HomeServer @@ -41,9 +45,8 @@ class RelationPaginationServlet(RestServlet): def __init__(self, hs: "HomeServer"): super().__init__() self.auth = hs.get_auth() - self.store = hs.get_datastores().main + self._store = hs.get_datastores().main self._relations_handler = hs.get_relations_handler() - self._msc3715_enabled = hs.config.experimental.msc3715_enabled async def on_GET( self, @@ -55,49 +58,63 @@ class RelationPaginationServlet(RestServlet): ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request, allow_guest=True) - limit = parse_integer(request, "limit", default=5) - # Fetch the direction parameter, if provided. - # - # TODO Use PaginationConfig.from_request when the unstable parameter is - # no longer needed. - direction = parse_string(request, "dir", allowed_values=["f", "b"]) - if direction is None: - if self._msc3715_enabled: - direction = parse_string( - request, - "org.matrix.msc3715.dir", - default="b", - allowed_values=["f", "b"], - ) - else: - direction = "b" - from_token_str = parse_string(request, "from") - to_token_str = parse_string(request, "to") - - # Return the relations - from_token = None - if from_token_str: - from_token = await StreamToken.from_string(self.store, from_token_str) - to_token = None - if to_token_str: - to_token = await StreamToken.from_string(self.store, to_token_str) + pagination_config = await PaginationConfig.from_request( + self._store, request, default_limit=5, default_dir="b" + ) # The unstable version of this API returns an extra field for client # compatibility, see https://github.com/matrix-org/synapse/issues/12930. assert request.path is not None include_original_event = request.path.startswith(b"/_matrix/client/unstable/") + # Return the relations result = await self._relations_handler.get_relations( requester=requester, event_id=parent_id, room_id=room_id, + pagin_config=pagination_config, + include_original_event=include_original_event, relation_type=relation_type, event_type=event_type, + ) + + return 200, result + + +class ThreadsServlet(RestServlet): + PATTERNS = (re.compile("^/_matrix/client/v1/rooms/(?P<room_id>[^/]*)/threads"),) + + def __init__(self, hs: "HomeServer"): + super().__init__() + self.auth = hs.get_auth() + self.store = hs.get_datastores().main + self._relations_handler = hs.get_relations_handler() + + async def on_GET( + self, request: SynapseRequest, room_id: str + ) -> Tuple[int, JsonDict]: + requester = await self.auth.get_user_by_req(request) + + limit = parse_integer(request, "limit", default=5) + from_token_str = parse_string(request, "from") + include = parse_string( + request, + "include", + default=ThreadsListInclude.all.value, + allowed_values=[v.value for v in ThreadsListInclude], + ) + + # Return the relations + from_token = None + if from_token_str: + from_token = ThreadsNextBatch.from_string(from_token_str) + + result = await self._relations_handler.get_threads( + requester=requester, + room_id=room_id, + include=ThreadsListInclude(include), limit=limit, - direction=direction, from_token=from_token, - to_token=to_token, - include_original_event=include_original_event, ) return 200, result @@ -105,3 +122,4 @@ class RelationPaginationServlet(RestServlet): def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: RelationPaginationServlet(hs).register(http_server) + ThreadsServlet(hs).register(http_server) diff --git a/synapse/rest/client/rendezvous.py b/synapse/rest/client/rendezvous.py new file mode 100644 index 0000000000..89176b1ffa --- /dev/null +++ b/synapse/rest/client/rendezvous.py @@ -0,0 +1,74 @@ +# Copyright 2022 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 http.client import TEMPORARY_REDIRECT +from typing import TYPE_CHECKING, Optional + +from synapse.http.server import HttpServer, respond_with_redirect +from synapse.http.servlet import RestServlet +from synapse.http.site import SynapseRequest +from synapse.rest.client._base import client_patterns + +if TYPE_CHECKING: + from synapse.server import HomeServer + +logger = logging.getLogger(__name__) + + +class RendezvousServlet(RestServlet): + """ + This is a placeholder implementation of [MSC3886](https://github.com/matrix-org/matrix-spec-proposals/pull/3886) + simple client rendezvous capability that is used by the "Sign in with QR" functionality. + + This implementation only serves as a 307 redirect to a configured server rather than being a full implementation. + + A module that implements the full functionality is available at: https://pypi.org/project/matrix-http-rendezvous-synapse/. + + Request: + + POST /rendezvous HTTP/1.1 + Content-Type: ... + + ... + + Response: + + HTTP/1.1 307 + Location: <configured endpoint> + """ + + PATTERNS = client_patterns( + "/org.matrix.msc3886/rendezvous$", releases=[], v1=False, unstable=True + ) + + def __init__(self, hs: "HomeServer"): + super().__init__() + redirection_target: Optional[str] = hs.config.experimental.msc3886_endpoint + assert ( + redirection_target is not None + ), "Servlet is only registered if there is a redirection target" + self.endpoint = redirection_target.encode("utf-8") + + async def on_POST(self, request: SynapseRequest) -> None: + respond_with_redirect( + request, self.endpoint, statusCode=TEMPORARY_REDIRECT, cors=True + ) + + # PUT, GET and DELETE are not implemented as they should be fulfilled by the redirect target. + + +def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: + if hs.config.experimental.msc3886_endpoint is not None: + RendezvousServlet(hs).register(http_server) diff --git a/synapse/rest/client/room.py b/synapse/rest/client/room.py index b6dedbed04..790614d721 100644 --- a/synapse/rest/client/room.py +++ b/synapse/rest/client/room.py @@ -52,11 +52,12 @@ from synapse.http.servlet import ( from synapse.http.site import SynapseRequest from synapse.logging.context import make_deferred_yieldable, run_in_background from synapse.logging.opentracing import set_tag +from synapse.metrics.background_process_metrics import run_as_background_process from synapse.rest.client._base import client_patterns from synapse.rest.client.transactions import HttpTransactionCache -from synapse.storage.state import StateFilter from synapse.streams.config import PaginationConfig from synapse.types import JsonDict, StreamToken, ThirdPartyInstanceID, UserID +from synapse.types.state import StateFilter from synapse.util import json_decoder from synapse.util.cancellation import cancellable from synapse.util.stringutils import parse_and_validate_server_name, random_string @@ -395,12 +396,7 @@ class JoinRoomAliasServlet(ResolveRoomIdMixin, TransactionRestServlet): ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request, allow_guest=True) - try: - content = parse_json_object_from_request(request) - except Exception: - # Turns out we used to ignore the body entirely, and some clients - # cheekily send invalid bodies. - content = {} + content = parse_json_object_from_request(request, allow_empty_body=True) # twisted.web.server.Request.args is incorrectly defined as Optional[Any] args: Dict[bytes, List[bytes]] = request.args # type: ignore @@ -729,7 +725,9 @@ class RoomInitialSyncRestServlet(RestServlet): self, request: SynapseRequest, room_id: str ) -> Tuple[int, JsonDict]: requester = await self.auth.get_user_by_req(request, allow_guest=True) - pagination_config = await PaginationConfig.from_request(self.store, request) + pagination_config = await PaginationConfig.from_request( + self.store, request, default_limit=10 + ) content = await self.initial_sync_handler.room_initial_sync( room_id=room_id, requester=requester, pagin_config=pagination_config ) @@ -949,12 +947,7 @@ class RoomMembershipRestServlet(TransactionRestServlet): }: raise AuthError(403, "Guest access not allowed") - try: - content = parse_json_object_from_request(request) - except Exception: - # Turns out we used to ignore the body entirely, and some clients - # cheekily send invalid bodies. - content = {} + content = parse_json_object_from_request(request, allow_empty_body=True) if membership_action == "invite" and all( key in content for key in ("medium", "address") @@ -1027,6 +1020,8 @@ class RoomRedactEventRestServlet(TransactionRestServlet): super().__init__(hs) self.event_creation_handler = hs.get_event_creation_handler() self.auth = hs.get_auth() + self._relation_handler = hs.get_relations_handler() + self._msc3912_enabled = hs.config.experimental.msc3912_enabled def register(self, http_server: HttpServer) -> None: PATTERNS = "/rooms/(?P<room_id>[^/]*)/redact/(?P<event_id>[^/]*)" @@ -1043,20 +1038,46 @@ class RoomRedactEventRestServlet(TransactionRestServlet): content = parse_json_object_from_request(request) try: - ( - event, - _, - ) = await self.event_creation_handler.create_and_send_nonmember_event( - requester, - { - "type": EventTypes.Redaction, - "content": content, - "room_id": room_id, - "sender": requester.user.to_string(), - "redacts": event_id, - }, - txn_id=txn_id, - ) + with_relations = None + if self._msc3912_enabled and "org.matrix.msc3912.with_relations" in content: + with_relations = content["org.matrix.msc3912.with_relations"] + del content["org.matrix.msc3912.with_relations"] + + # Check if there's an existing event for this transaction now (even though + # create_and_send_nonmember_event also does it) because, if there's one, + # then we want to skip the call to redact_events_related_to. + event = None + if txn_id: + event = await self.event_creation_handler.get_event_from_transaction( + requester, txn_id, room_id + ) + + if event is None: + ( + event, + _, + ) = await self.event_creation_handler.create_and_send_nonmember_event( + requester, + { + "type": EventTypes.Redaction, + "content": content, + "room_id": room_id, + "sender": requester.user.to_string(), + "redacts": event_id, + }, + txn_id=txn_id, + ) + + if with_relations: + run_as_background_process( + "redact_related_events", + self._relation_handler.redact_events_related_to, + requester=requester, + event_id=event_id, + initial_redaction_event=event, + relation_types=with_relations, + ) + event_id = event.event_id except ShadowBanError: event_id = "$" + random_string(43) @@ -1253,17 +1274,14 @@ class TimestampLookupRestServlet(RestServlet): `dir` can be `f` or `b` to indicate forwards and backwards in time from the given timestamp. - GET /_matrix/client/unstable/org.matrix.msc3030/rooms/<roomID>/timestamp_to_event?ts=<timestamp>&dir=<direction> + GET /_matrix/client/v1/rooms/<roomID>/timestamp_to_event?ts=<timestamp>&dir=<direction> { "event_id": ... } """ PATTERNS = ( - re.compile( - "^/_matrix/client/unstable/org.matrix.msc3030" - "/rooms/(?P<room_id>[^/]*)/timestamp_to_event$" - ), + re.compile("^/_matrix/client/v1/rooms/(?P<room_id>[^/]*)/timestamp_to_event$"), ) def __init__(self, hs: "HomeServer"): @@ -1367,9 +1385,7 @@ class RoomSummaryRestServlet(ResolveRoomIdMixin, RestServlet): ) -def register_servlets( - hs: "HomeServer", http_server: HttpServer, is_worker: bool = False -) -> None: +def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: RoomStateEventRestServlet(hs).register(http_server) RoomMemberListRestServlet(hs).register(http_server) JoinedRoomMemberListRestServlet(hs).register(http_server) @@ -1390,11 +1406,10 @@ def register_servlets( RoomAliasListServlet(hs).register(http_server) SearchRestServlet(hs).register(http_server) RoomCreateRestServlet(hs).register(http_server) - if hs.config.experimental.msc3030_enabled: - TimestampLookupRestServlet(hs).register(http_server) + TimestampLookupRestServlet(hs).register(http_server) # Some servlets only get registered for the main process. - if not is_worker: + if hs.config.worker.worker_app is None: RoomForgetRestServlet(hs).register(http_server) diff --git a/synapse/rest/client/room_batch.py b/synapse/rest/client/room_batch.py index dd91dabedd..10be4a781b 100644 --- a/synapse/rest/client/room_batch.py +++ b/synapse/rest/client/room_batch.py @@ -108,6 +108,13 @@ class RoomBatchSendEventRestServlet(RestServlet): errcode=Codes.MISSING_PARAM, ) + if await self.store.is_partial_state_room(room_id): + raise SynapseError( + HTTPStatus.BAD_REQUEST, + "Cannot insert history batches until we have fully joined the room", + errcode=Codes.UNABLE_DUE_TO_PARTIAL_STATE, + ) + # Verify the batch_id_from_query corresponds to an actual insertion event # and have the batch connected. if batch_id_from_query: diff --git a/synapse/rest/client/sendtodevice.py b/synapse/rest/client/sendtodevice.py index 46a8b03829..55d52f0b28 100644 --- a/synapse/rest/client/sendtodevice.py +++ b/synapse/rest/client/sendtodevice.py @@ -46,7 +46,6 @@ class SendToDeviceRestServlet(servlet.RestServlet): def on_PUT( self, request: SynapseRequest, message_type: str, txn_id: str ) -> Awaitable[Tuple[int, JsonDict]]: - set_tag("message_type", message_type) set_tag("txn_id", txn_id) return self.txns.fetch_or_execute_request( request, self._put, request, message_type, txn_id diff --git a/synapse/rest/client/sync.py b/synapse/rest/client/sync.py index 8a16459105..f2013faeb2 100644 --- a/synapse/rest/client/sync.py +++ b/synapse/rest/client/sync.py @@ -146,12 +146,12 @@ class SyncRestServlet(RestServlet): elif filter_id.startswith("{"): try: filter_object = json_decoder.decode(filter_id) - set_timeline_upper_limit( - filter_object, self.hs.config.server.filter_timeline_limit - ) except Exception: - raise SynapseError(400, "Invalid filter JSON") + raise SynapseError(400, "Invalid filter JSON", errcode=Codes.NOT_JSON) self.filtering.check_valid_filter(filter_object) + set_timeline_upper_limit( + filter_object, self.hs.config.server.filter_timeline_limit + ) filter_collection = FilterCollection(self.hs, filter_object) else: try: diff --git a/synapse/rest/client/user_directory.py b/synapse/rest/client/user_directory.py index 116c982ce6..4670fad608 100644 --- a/synapse/rest/client/user_directory.py +++ b/synapse/rest/client/user_directory.py @@ -63,8 +63,8 @@ class UserDirectorySearchRestServlet(RestServlet): body = parse_json_object_from_request(request) - limit = body.get("limit", 10) - limit = min(limit, 50) + limit = int(body.get("limit", 10)) + limit = max(min(limit, 50), 0) try: search_term = body["search_term"] diff --git a/synapse/rest/client/versions.py b/synapse/rest/client/versions.py index d1d2e5f7e3..e19c0946c0 100644 --- a/synapse/rest/client/versions.py +++ b/synapse/rest/client/versions.py @@ -76,6 +76,8 @@ class VersionsRestServlet(RestServlet): "v1.1", "v1.2", "v1.3", + "v1.4", + "v1.5", ], # as per MSC1497: "unstable_features": { @@ -100,8 +102,6 @@ class VersionsRestServlet(RestServlet): "org.matrix.msc3827.stable": True, # Adds support for importing historical messages as per MSC2716 "org.matrix.msc2716": self.config.experimental.msc2716_enabled, - # Adds support for jump to date endpoints (/timestamp_to_event) as per MSC3030 - "org.matrix.msc3030": self.config.experimental.msc3030_enabled, # Adds support for thread relations, per MSC3440. "org.matrix.msc3440.stable": True, # TODO: remove when "v1.3" is added above # Support for thread read receipts & notification counts. @@ -113,6 +113,13 @@ class VersionsRestServlet(RestServlet): "org.matrix.msc3882": self.config.experimental.msc3882_enabled, # Adds support for remotely enabling/disabling pushers, as per MSC3881 "org.matrix.msc3881": self.config.experimental.msc3881_enabled, + # Adds support for filtering /messages by event relation. + "org.matrix.msc3874": self.config.experimental.msc3874_enabled, + # Adds support for simple HTTP rendezvous as per MSC3886 + "org.matrix.msc3886": self.config.experimental.msc3886_endpoint + is not None, + # Adds support for relation-based redactions as per MSC3912. + "org.matrix.msc3912": self.config.experimental.msc3912_enabled, }, }, ) diff --git a/synapse/rest/key/v2/__init__.py b/synapse/rest/key/v2/__init__.py index 7f8c1de1ff..26403facb8 100644 --- a/synapse/rest/key/v2/__init__.py +++ b/synapse/rest/key/v2/__init__.py @@ -14,17 +14,20 @@ from typing import TYPE_CHECKING -from twisted.web.resource import Resource - -from .local_key_resource import LocalKey -from .remote_key_resource import RemoteKey +from synapse.http.server import HttpServer, JsonResource +from synapse.rest.key.v2.local_key_resource import LocalKey +from synapse.rest.key.v2.remote_key_resource import RemoteKey if TYPE_CHECKING: from synapse.server import HomeServer -class KeyApiV2Resource(Resource): +class KeyResource(JsonResource): def __init__(self, hs: "HomeServer"): - Resource.__init__(self) - self.putChild(b"server", LocalKey(hs)) - self.putChild(b"query", RemoteKey(hs)) + super().__init__(hs, canonical_json=True) + self.register_servlets(self, hs) + + @staticmethod + def register_servlets(http_server: HttpServer, hs: "HomeServer") -> None: + LocalKey(hs).register(http_server) + RemoteKey(hs).register(http_server) diff --git a/synapse/rest/key/v2/local_key_resource.py b/synapse/rest/key/v2/local_key_resource.py index 0c9f042c84..d03e728d42 100644 --- a/synapse/rest/key/v2/local_key_resource.py +++ b/synapse/rest/key/v2/local_key_resource.py @@ -13,16 +13,15 @@ # limitations under the License. import logging -from typing import TYPE_CHECKING, Optional +import re +from typing import TYPE_CHECKING, Optional, Tuple -from canonicaljson import encode_canonical_json from signedjson.sign import sign_json from unpaddedbase64 import encode_base64 -from twisted.web.resource import Resource from twisted.web.server import Request -from synapse.http.server import respond_with_json_bytes +from synapse.http.servlet import RestServlet from synapse.types import JsonDict if TYPE_CHECKING: @@ -31,7 +30,7 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) -class LocalKey(Resource): +class LocalKey(RestServlet): """HTTP resource containing encoding the TLS X.509 certificate and NACL signature verification keys for this server:: @@ -61,18 +60,17 @@ class LocalKey(Resource): } """ - isLeaf = True + PATTERNS = (re.compile("^/_matrix/key/v2/server(/(?P<key_id>[^/]*))?$"),) def __init__(self, hs: "HomeServer"): self.config = hs.config self.clock = hs.get_clock() self.update_response_body(self.clock.time_msec()) - Resource.__init__(self) def update_response_body(self, time_now_msec: int) -> None: refresh_interval = self.config.key.key_refresh_interval self.valid_until_ts = int(time_now_msec + refresh_interval) - self.response_body = encode_canonical_json(self.response_json_object()) + self.response_body = self.response_json_object() def response_json_object(self) -> JsonDict: verify_keys = {} @@ -99,9 +97,11 @@ class LocalKey(Resource): json_object = sign_json(json_object, self.config.server.server_name, key) return json_object - def render_GET(self, request: Request) -> Optional[int]: + def on_GET( + self, request: Request, key_id: Optional[str] = None + ) -> Tuple[int, JsonDict]: time_now = self.clock.time_msec() # Update the expiry time if less than half the interval remains. if time_now + self.config.key.key_refresh_interval / 2 > self.valid_until_ts: self.update_response_body(time_now) - return respond_with_json_bytes(request, 200, self.response_body) + return 200, self.response_body diff --git a/synapse/rest/key/v2/remote_key_resource.py b/synapse/rest/key/v2/remote_key_resource.py index 7f8ad29566..19820886f5 100644 --- a/synapse/rest/key/v2/remote_key_resource.py +++ b/synapse/rest/key/v2/remote_key_resource.py @@ -13,15 +13,20 @@ # limitations under the License. import logging -from typing import TYPE_CHECKING, Dict, Set +import re +from typing import TYPE_CHECKING, Dict, Optional, Set, Tuple from signedjson.sign import sign_json -from synapse.api.errors import Codes, SynapseError +from twisted.web.server import Request + from synapse.crypto.keyring import ServerKeyFetcher -from synapse.http.server import DirectServeJsonResource, respond_with_json -from synapse.http.servlet import parse_integer, parse_json_object_from_request -from synapse.http.site import SynapseRequest +from synapse.http.server import HttpServer +from synapse.http.servlet import ( + RestServlet, + parse_integer, + parse_json_object_from_request, +) from synapse.types import JsonDict from synapse.util import json_decoder from synapse.util.async_helpers import yieldable_gather_results @@ -32,7 +37,7 @@ if TYPE_CHECKING: logger = logging.getLogger(__name__) -class RemoteKey(DirectServeJsonResource): +class RemoteKey(RestServlet): """HTTP resource for retrieving the TLS certificate and NACL signature verification keys for a collection of servers. Checks that the reported X.509 TLS certificate matches the one used in the HTTPS connection. Checks @@ -88,11 +93,7 @@ class RemoteKey(DirectServeJsonResource): } """ - isLeaf = True - def __init__(self, hs: "HomeServer"): - super().__init__() - self.fetcher = ServerKeyFetcher(hs) self.store = hs.get_datastores().main self.clock = hs.get_clock() @@ -101,36 +102,48 @@ class RemoteKey(DirectServeJsonResource): ) self.config = hs.config - async def _async_render_GET(self, request: SynapseRequest) -> None: - assert request.postpath is not None - if len(request.postpath) == 1: - (server,) = request.postpath - query: dict = {server.decode("ascii"): {}} - elif len(request.postpath) == 2: - server, key_id = request.postpath + def register(self, http_server: HttpServer) -> None: + http_server.register_paths( + "GET", + ( + re.compile( + "^/_matrix/key/v2/query/(?P<server>[^/]*)(/(?P<key_id>[^/]*))?$" + ), + ), + self.on_GET, + self.__class__.__name__, + ) + http_server.register_paths( + "POST", + (re.compile("^/_matrix/key/v2/query$"),), + self.on_POST, + self.__class__.__name__, + ) + + async def on_GET( + self, request: Request, server: str, key_id: Optional[str] = None + ) -> Tuple[int, JsonDict]: + if server and key_id: minimum_valid_until_ts = parse_integer(request, "minimum_valid_until_ts") arguments = {} if minimum_valid_until_ts is not None: arguments["minimum_valid_until_ts"] = minimum_valid_until_ts - query = {server.decode("ascii"): {key_id.decode("ascii"): arguments}} + query = {server: {key_id: arguments}} else: - raise SynapseError(404, "Not found %r" % request.postpath, Codes.NOT_FOUND) + query = {server: {}} - await self.query_keys(request, query, query_remote_on_cache_miss=True) + return 200, await self.query_keys(query, query_remote_on_cache_miss=True) - async def _async_render_POST(self, request: SynapseRequest) -> None: + async def on_POST(self, request: Request) -> Tuple[int, JsonDict]: content = parse_json_object_from_request(request) query = content["server_keys"] - await self.query_keys(request, query, query_remote_on_cache_miss=True) + return 200, await self.query_keys(query, query_remote_on_cache_miss=True) async def query_keys( - self, - request: SynapseRequest, - query: JsonDict, - query_remote_on_cache_miss: bool = False, - ) -> None: + self, query: JsonDict, query_remote_on_cache_miss: bool = False + ) -> JsonDict: logger.info("Handling query for keys %r", query) store_queries = [] @@ -232,7 +245,7 @@ class RemoteKey(DirectServeJsonResource): for server_name, keys in cache_misses.items() ), ) - await self.query_keys(request, query, query_remote_on_cache_miss=False) + return await self.query_keys(query, query_remote_on_cache_miss=False) else: signed_keys = [] for key_json_raw in json_results: @@ -244,6 +257,4 @@ class RemoteKey(DirectServeJsonResource): signed_keys.append(key_json) - response = {"server_keys": signed_keys} - - respond_with_json(request, 200, response, canonical_json=True) + return {"server_keys": signed_keys} diff --git a/synapse/rest/media/v1/media_repository.py b/synapse/rest/media/v1/media_repository.py index 328c0c5477..c70e1837af 100644 --- a/synapse/rest/media/v1/media_repository.py +++ b/synapse/rest/media/v1/media_repository.py @@ -24,7 +24,6 @@ from matrix_common.types.mxc_uri import MXCUri import twisted.internet.error import twisted.web.http from twisted.internet.defer import Deferred -from twisted.web.resource import Resource from synapse.api.errors import ( FederationDeniedError, @@ -35,6 +34,7 @@ from synapse.api.errors import ( ) from synapse.config._base import ConfigError from synapse.config.repository import ThumbnailRequirement +from synapse.http.server import UnrecognizedRequestResource from synapse.http.site import SynapseRequest from synapse.logging.context import defer_to_thread from synapse.metrics.background_process_metrics import run_as_background_process @@ -344,8 +344,8 @@ class MediaRepository: download from remote server. Args: - server_name (str): Remote server_name where the media originated. - media_id (str): The media ID of the content (as defined by the + server_name: Remote server_name where the media originated. + media_id: The media ID of the content (as defined by the remote server). Returns: @@ -1046,7 +1046,7 @@ class MediaRepository: return removed_media, len(removed_media) -class MediaRepositoryResource(Resource): +class MediaRepositoryResource(UnrecognizedRequestResource): """File uploading and downloading. Uploads are POSTed to a resource which returns a token which is used to GET diff --git a/synapse/rest/media/v1/thumbnailer.py b/synapse/rest/media/v1/thumbnailer.py index 9b93b9b4f6..a48a4de92a 100644 --- a/synapse/rest/media/v1/thumbnailer.py +++ b/synapse/rest/media/v1/thumbnailer.py @@ -138,7 +138,7 @@ class Thumbnailer: """Rescales the image to the given dimensions. Returns: - BytesIO: the bytes of the encoded image ready to be written to disk + The bytes of the encoded image ready to be written to disk """ with self._resize(width, height) as scaled: return self._encode_image(scaled, output_type) @@ -155,7 +155,7 @@ class Thumbnailer: max_height: The largest possible height. Returns: - BytesIO: the bytes of the encoded image ready to be written to disk + The bytes of the encoded image ready to be written to disk """ if width * self.height > height * self.width: scaled_width = width diff --git a/synapse/rest/synapse/client/new_user_consent.py b/synapse/rest/synapse/client/new_user_consent.py index 1c1c7b3613..22784157e6 100644 --- a/synapse/rest/synapse/client/new_user_consent.py +++ b/synapse/rest/synapse/client/new_user_consent.py @@ -20,6 +20,7 @@ from synapse.api.errors import SynapseError from synapse.handlers.sso import get_username_mapping_session_cookie_from_request from synapse.http.server import DirectServeHtmlResource, respond_with_html from synapse.http.servlet import parse_string +from synapse.http.site import SynapseRequest from synapse.types import UserID from synapse.util.templates import build_jinja_env @@ -88,7 +89,7 @@ class NewUserConsentResource(DirectServeHtmlResource): html = template.render(template_params) respond_with_html(request, 200, html) - async def _async_render_POST(self, request: Request) -> None: + async def _async_render_POST(self, request: SynapseRequest) -> None: try: session_id = get_username_mapping_session_cookie_from_request(request) except SynapseError as e: diff --git a/synapse/rest/synapse/client/oidc/__init__.py b/synapse/rest/synapse/client/oidc/__init__.py index 81fec39659..e4b28ce3df 100644 --- a/synapse/rest/synapse/client/oidc/__init__.py +++ b/synapse/rest/synapse/client/oidc/__init__.py @@ -17,6 +17,9 @@ from typing import TYPE_CHECKING from twisted.web.resource import Resource +from synapse.rest.synapse.client.oidc.backchannel_logout_resource import ( + OIDCBackchannelLogoutResource, +) from synapse.rest.synapse.client.oidc.callback_resource import OIDCCallbackResource if TYPE_CHECKING: @@ -29,6 +32,7 @@ class OIDCResource(Resource): def __init__(self, hs: "HomeServer"): Resource.__init__(self) self.putChild(b"callback", OIDCCallbackResource(hs)) + self.putChild(b"backchannel_logout", OIDCBackchannelLogoutResource(hs)) __all__ = ["OIDCResource"] diff --git a/synapse/replication/slave/storage/filtering.py b/synapse/rest/synapse/client/oidc/backchannel_logout_resource.py index c52679cd60..e07e76855a 100644 --- a/synapse/replication/slave/storage/filtering.py +++ b/synapse/rest/synapse/client/oidc/backchannel_logout_resource.py @@ -1,4 +1,4 @@ -# Copyright 2015, 2016 OpenMarket Ltd +# Copyright 2022 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. @@ -12,24 +12,24 @@ # See the License for the specific language governing permissions and # limitations under the License. +import logging from typing import TYPE_CHECKING -from synapse.storage._base import SQLBaseStore -from synapse.storage.database import DatabasePool, LoggingDatabaseConnection -from synapse.storage.databases.main.filtering import FilteringStore +from synapse.http.server import DirectServeJsonResource +from synapse.http.site import SynapseRequest if TYPE_CHECKING: from synapse.server import HomeServer +logger = logging.getLogger(__name__) -class SlavedFilteringStore(SQLBaseStore): - def __init__( - self, - database: DatabasePool, - db_conn: LoggingDatabaseConnection, - hs: "HomeServer", - ): - super().__init__(database, db_conn, hs) - # Filters are immutable so this cache doesn't need to be expired - get_user_filter = FilteringStore.__dict__["get_user_filter"] +class OIDCBackchannelLogoutResource(DirectServeJsonResource): + isLeaf = 1 + + def __init__(self, hs: "HomeServer"): + super().__init__() + self._oidc_handler = hs.get_oidc_handler() + + async def _async_render_POST(self, request: SynapseRequest) -> None: + await self._oidc_handler.handle_backchannel_logout(request) diff --git a/synapse/rest/well_known.py b/synapse/rest/well_known.py index 6f7ac54c65..e2174fdfea 100644 --- a/synapse/rest/well_known.py +++ b/synapse/rest/well_known.py @@ -18,6 +18,7 @@ from twisted.web.resource import Resource from twisted.web.server import Request from synapse.http.server import set_cors_headers +from synapse.http.site import SynapseRequest from synapse.types import JsonDict from synapse.util import json_encoder from synapse.util.stringutils import parse_server_name @@ -63,7 +64,7 @@ class ClientWellKnownResource(Resource): Resource.__init__(self) self._well_known_builder = WellKnownBuilder(hs) - def render_GET(self, request: Request) -> bytes: + def render_GET(self, request: SynapseRequest) -> bytes: set_cors_headers(request) r = self._well_known_builder.get_well_known() if not r: diff --git a/synapse/server.py b/synapse/server.py index df3a1cb405..5baae2325e 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -221,8 +221,6 @@ class HomeServer(metaclass=abc.ABCMeta): # instantiated during setup() for future return by get_datastores() DATASTORE_CLASS = abc.abstractproperty() - tls_server_context_factory: Optional[IOpenSSLContextFactory] - def __init__( self, hostname: str, @@ -258,6 +256,9 @@ class HomeServer(metaclass=abc.ABCMeta): self._module_web_resources: Dict[str, Resource] = {} self._module_web_resources_consumed = False + # This attribute is set by the free function `refresh_certificate`. + self.tls_server_context_factory: Optional[IOpenSSLContextFactory] = None + def register_module_web_resource(self, path: str, resource: Resource) -> None: """Allows a module to register a web resource to be served at the given path. @@ -315,7 +316,7 @@ class HomeServer(metaclass=abc.ABCMeta): if self.config.worker.run_background_tasks: self.setup_background_tasks() - def start_listening(self) -> None: + def start_listening(self) -> None: # noqa: B027 (no-op by design) """Start the HTTP, manhole, metrics, etc listeners Does nothing in this base class; overridden in derived classes to start the @@ -509,7 +510,7 @@ class HomeServer(metaclass=abc.ABCMeta): ) @cache_in_self - def get_device_handler(self): + def get_device_handler(self) -> DeviceWorkerHandler: if self.config.worker.worker_app: return DeviceWorkerHandler(self) else: diff --git a/synapse/server_notices/consent_server_notices.py b/synapse/server_notices/consent_server_notices.py index 698ca742ed..94025ba41f 100644 --- a/synapse/server_notices/consent_server_notices.py +++ b/synapse/server_notices/consent_server_notices.py @@ -113,9 +113,8 @@ def copy_with_str_subst(x: Any, substitutions: Any) -> Any: """Deep-copy a structure, carrying out string substitutions on any strings Args: - x (object): structure to be copied - substitutions (object): substitutions to be made - passed into the - string '%' operator + x: structure to be copied + substitutions: substitutions to be made - passed into the string '%' operator Returns: copy of x diff --git a/synapse/server_notices/resource_limits_server_notices.py b/synapse/server_notices/resource_limits_server_notices.py index 3134cd2d3d..a31a2c99a7 100644 --- a/synapse/server_notices/resource_limits_server_notices.py +++ b/synapse/server_notices/resource_limits_server_notices.py @@ -170,11 +170,13 @@ class ResourceLimitsServerNotices: room_id: The room id of the server notices room Returns: - bool: Is the room currently blocked - list: The list of pinned event IDs that are unrelated to limit blocking - This list can be used as a convenience in the case where the block - is to be lifted and the remaining pinned event references need to be - preserved + Tuple of: + Is the room currently blocked + + The list of pinned event IDs that are unrelated to limit blocking + This list can be used as a convenience in the case where the block + is to be lifted and the remaining pinned event references need to be + preserved """ currently_blocked = False pinned_state_event = None diff --git a/synapse/state/__init__.py b/synapse/state/__init__.py index 6f3dd0463e..ee5469d5a8 100644 --- a/synapse/state/__init__.py +++ b/synapse/state/__init__.py @@ -44,8 +44,8 @@ from synapse.logging.context import ContextResourceUsage from synapse.replication.http.state import ReplicationUpdateCurrentStateRestServlet from synapse.state import v1, v2 from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import StateMap +from synapse.types.state import StateFilter from synapse.util.async_helpers import Linearizer from synapse.util.caches.expiringcache import ExpiringCache from synapse.util.metrics import Measure, measure_func @@ -190,6 +190,7 @@ class StateHandler: room_id: str, event_ids: Collection[str], state_filter: Optional[StateFilter] = None, + await_full_state: bool = True, ) -> StateMap[str]: """Fetch the state after each of the given event IDs. Resolve them and return. @@ -200,13 +201,18 @@ class StateHandler: Args: room_id: the room_id containing the given events. event_ids: the events whose state should be fetched and resolved. + await_full_state: if `True`, will block if we do not yet have complete state + at the given `event_id`s, regardless of whether `state_filter` is + satisfied by partial state. Returns: the state dict (a mapping from (event_type, state_key) -> event_id) which holds the resolution of the states after the given event IDs. """ logger.debug("calling resolve_state_groups from compute_state_after_events") - ret = await self.resolve_state_groups_for_events(room_id, event_ids) + ret = await self.resolve_state_groups_for_events( + room_id, event_ids, await_full_state + ) return await ret.get_state(self._state_storage_controller, state_filter) async def get_current_user_ids_in_room( diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py index bf42aeb8d1..69abf6fa87 100644 --- a/synapse/storage/_base.py +++ b/synapse/storage/_base.py @@ -50,7 +50,7 @@ class SQLBaseStore(metaclass=ABCMeta): self.external_cached_functions: Dict[str, CachedFunction] = {} - def process_replication_rows( + def process_replication_rows( # noqa: B027 (no-op by design) self, stream_name: str, instance_name: str, diff --git a/synapse/storage/background_updates.py b/synapse/storage/background_updates.py index 2056ecb2c3..a99aea8926 100644 --- a/synapse/storage/background_updates.py +++ b/synapse/storage/background_updates.py @@ -544,6 +544,48 @@ class BackgroundUpdater: The named index will be dropped upon completion of the new index. """ + async def updater(progress: JsonDict, batch_size: int) -> int: + await self.create_index_in_background( + index_name=index_name, + table=table, + columns=columns, + where_clause=where_clause, + unique=unique, + psql_only=psql_only, + replaces_index=replaces_index, + ) + await self._end_background_update(update_name) + return 1 + + self._background_update_handlers[update_name] = _BackgroundUpdateHandler( + updater, oneshot=True + ) + + async def create_index_in_background( + self, + index_name: str, + table: str, + columns: Iterable[str], + where_clause: Optional[str] = None, + unique: bool = False, + psql_only: bool = False, + replaces_index: Optional[str] = None, + ) -> None: + """Add an index in the background. + + Args: + update_name: update_name to register for + index_name: name of index to add + table: table to add index to + columns: columns/expressions to include in index + where_clause: A WHERE clause to specify a partial unique index. + unique: true to make a UNIQUE index + psql_only: true to only create this index on psql databases (useful + for virtual sqlite tables) + replaces_index: The name of an index that this index replaces. + The named index will be dropped upon completion of the new index. + """ + def create_index_psql(conn: Connection) -> None: conn.rollback() # postgres insists on autocommit for the index @@ -618,16 +660,11 @@ class BackgroundUpdater: else: runner = create_index_sqlite - async def updater(progress: JsonDict, batch_size: int) -> int: - if runner is not None: - logger.info("Adding index %s to %s", index_name, table) - await self.db_pool.runWithConnection(runner) - await self._end_background_update(update_name) - return 1 + if runner is None: + return - self._background_update_handlers[update_name] = _BackgroundUpdateHandler( - updater, oneshot=True - ) + logger.info("Adding index %s to %s", index_name, table) + await self.db_pool.runWithConnection(runner) async def _end_background_update(self, update_name: str) -> None: """Removes a completed background update task from the queue. diff --git a/synapse/storage/controllers/persist_events.py b/synapse/storage/controllers/persist_events.py index 06e71a8053..f1d2c71c91 100644 --- a/synapse/storage/controllers/persist_events.py +++ b/synapse/storage/controllers/persist_events.py @@ -58,13 +58,13 @@ from synapse.storage.controllers.state import StateStorageController from synapse.storage.databases import Databases from synapse.storage.databases.main.events import DeltaState from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.state import StateFilter from synapse.types import ( PersistedEventPosition, RoomStreamToken, StateMap, get_domain_from_id, ) +from synapse.types.state import StateFilter from synapse.util.async_helpers import ObservableDeferred, yieldable_gather_results from synapse.util.metrics import Measure @@ -204,9 +204,8 @@ class _EventPeristenceQueue(Generic[_PersistResult]): process to to so, calling the per_item_callback for each item. Args: - room_id (str): - task (_EventPersistQueueTask): A _PersistEventsTask or - _UpdateCurrentStateTask to process. + room_id: + task: A _PersistEventsTask or _UpdateCurrentStateTask to process. Returns: the result returned by the `_per_item_callback` passed to @@ -716,8 +715,6 @@ class EventsPersistenceStorageController: ) if not is_still_joined: logger.info("Server no longer in room %s", room_id) - latest_event_ids = set() - current_state = {} delta.no_longer_in_room = True state_delta_for_room[room_id] = delta diff --git a/synapse/storage/controllers/state.py b/synapse/storage/controllers/state.py index 2b31ce54bb..26d79c6e62 100644 --- a/synapse/storage/controllers/state.py +++ b/synapse/storage/controllers/state.py @@ -31,12 +31,12 @@ from synapse.api.constants import EventTypes from synapse.events import EventBase from synapse.logging.opentracing import tag_args, trace from synapse.storage.roommember import ProfileInfo -from synapse.storage.state import StateFilter from synapse.storage.util.partial_state_events_tracker import ( PartialCurrentStateTracker, PartialStateEventsTracker, ) from synapse.types import MutableStateMap, StateMap +from synapse.types.state import StateFilter from synapse.util.cancellation import cancellable if TYPE_CHECKING: diff --git a/synapse/storage/database.py b/synapse/storage/database.py index 7bb21f8f81..0b29e67b94 100644 --- a/synapse/storage/database.py +++ b/synapse/storage/database.py @@ -569,15 +569,15 @@ class DatabasePool: retcols=["update_name"], desc="check_background_updates", ) - updates = [x["update_name"] for x in updates] + background_update_names = [x["update_name"] for x in updates] for table, update_name in UNIQUE_INDEX_BACKGROUND_UPDATES.items(): - if update_name not in updates: + if update_name not in background_update_names: logger.debug("Now safe to upsert in %s", table) self._unsafe_to_upsert_tables.discard(table) # If there's any updates still running, reschedule to run. - if updates: + if background_update_names: self._clock.call_later( 15.0, run_as_background_process, @@ -667,7 +667,8 @@ class DatabasePool: ) # also check variables referenced in func's closure if inspect.isfunction(func): - f = cast(types.FunctionType, func) + # Keep the cast for now---it helps PyCharm to understand what `func` is. + f = cast(types.FunctionType, func) # type: ignore[redundant-cast] if f.__closure__: for i, cell in enumerate(f.__closure__): if inspect.isgenerator(cell.cell_contents): @@ -1129,7 +1130,6 @@ class DatabasePool: values: Dict[str, Any], insertion_values: Optional[Dict[str, Any]] = None, desc: str = "simple_upsert", - lock: bool = True, ) -> bool: """Insert a row with values + insertion_values; on conflict, update with values. @@ -1154,21 +1154,12 @@ class DatabasePool: requiring that a unique index exist on the column names used to detect a conflict (i.e. `keyvalues.keys()`). - If there is no such index, we can "emulate" an upsert with a SELECT followed - by either an INSERT or an UPDATE. This is unsafe: we cannot make the same - atomicity guarantees that a native upsert can and are very vulnerable to races - and crashes. Therefore if we wish to upsert without an appropriate unique index, - we must either: - - 1. Acquire a table-level lock before the emulated upsert (`lock=True`), or - 2. VERY CAREFULLY ensure that we are the only thread and worker which will be - writing to this table, in which case we can proceed without a lock - (`lock=False`). - - Generally speaking, you should use `lock=True`. If the table in question has a - unique index[*], this class will use a native upsert (which is atomic and so can - ignore the `lock` argument). Otherwise this class will use an emulated upsert, - in which case we want the safer option unless we been VERY CAREFUL. + If there is no such index yet[*], we can "emulate" an upsert with a SELECT + followed by either an INSERT or an UPDATE. This is unsafe unless *all* upserters + run at the SERIALIZABLE isolation level: we cannot make the same atomicity + guarantees that a native upsert can and are very vulnerable to races and + crashes. Therefore to upsert without an appropriate unique index, we acquire a + table-level lock before the emulated upsert. [*]: Some tables have unique indices added to them in the background. Those tables `T` are keys in the dictionary UNIQUE_INDEX_BACKGROUND_UPDATES, @@ -1189,7 +1180,6 @@ class DatabasePool: values: The nonunique columns and their new values insertion_values: additional key/values to use only when inserting desc: description of the transaction, for logging and metrics - lock: True to lock the table when doing the upsert. Returns: Returns True if a row was inserted or updated (i.e. if `values` is not empty then this always returns True) @@ -1209,7 +1199,6 @@ class DatabasePool: keyvalues, values, insertion_values, - lock=lock, db_autocommit=autocommit, ) except self.engine.module.IntegrityError as e: @@ -1232,7 +1221,6 @@ class DatabasePool: values: Dict[str, Any], insertion_values: Optional[Dict[str, Any]] = None, where_clause: Optional[str] = None, - lock: bool = True, ) -> bool: """ Pick the UPSERT method which works best on the platform. Either the @@ -1245,8 +1233,6 @@ class DatabasePool: values: The nonunique columns and their new values insertion_values: additional key/values to use only when inserting where_clause: An index predicate to apply to the upsert. - lock: True to lock the table when doing the upsert. Unused when performing - a native upsert. Returns: Returns True if a row was inserted or updated (i.e. if `values` is not empty then this always returns True) @@ -1270,7 +1256,6 @@ class DatabasePool: values, insertion_values=insertion_values, where_clause=where_clause, - lock=lock, ) def simple_upsert_txn_emulated( @@ -1291,14 +1276,15 @@ class DatabasePool: insertion_values: additional key/values to use only when inserting where_clause: An index predicate to apply to the upsert. lock: True to lock the table when doing the upsert. + Must not be False unless the table has already been locked. Returns: Returns True if a row was inserted or updated (i.e. if `values` is not empty then this always returns True) """ insertion_values = insertion_values or {} - # We need to lock the table :(, unless we're *really* careful if lock: + # We need to lock the table :( self.engine.lock_table(txn, table) def _getwhere(key: str) -> str: @@ -1406,7 +1392,6 @@ class DatabasePool: value_names: Collection[str], value_values: Collection[Collection[Any]], desc: str, - lock: bool = True, ) -> None: """ Upsert, many times. @@ -1418,8 +1403,6 @@ class DatabasePool: value_names: The value column names value_values: A list of each row's value column values. Ignored if value_names is empty. - lock: True to lock the table when doing the upsert. Unused when performing - a native upsert. """ # We can autocommit if it safe to upsert @@ -1433,7 +1416,6 @@ class DatabasePool: key_values, value_names, value_values, - lock=lock, db_autocommit=autocommit, ) @@ -1445,7 +1427,6 @@ class DatabasePool: key_values: Collection[Iterable[Any]], value_names: Collection[str], value_values: Iterable[Iterable[Any]], - lock: bool = True, ) -> None: """ Upsert, many times. @@ -1457,8 +1438,6 @@ class DatabasePool: value_names: The value column names value_values: A list of each row's value column values. Ignored if value_names is empty. - lock: True to lock the table when doing the upsert. Unused when performing - a native upsert. """ if table not in self._unsafe_to_upsert_tables: return self.simple_upsert_many_txn_native_upsert( @@ -1466,7 +1445,12 @@ class DatabasePool: ) else: return self.simple_upsert_many_txn_emulated( - txn, table, key_names, key_values, value_names, value_values, lock=lock + txn, + table, + key_names, + key_values, + value_names, + value_values, ) def simple_upsert_many_txn_emulated( @@ -1477,7 +1461,6 @@ class DatabasePool: key_values: Collection[Iterable[Any]], value_names: Collection[str], value_values: Iterable[Iterable[Any]], - lock: bool = True, ) -> None: """ Upsert, many times, but without native UPSERT support or batching. @@ -1489,18 +1472,16 @@ class DatabasePool: value_names: The value column names value_values: A list of each row's value column values. Ignored if value_names is empty. - lock: True to lock the table when doing the upsert. """ # No value columns, therefore make a blank list so that the following # zip() works correctly. if not value_names: value_values = [() for x in range(len(key_values))] - if lock: - # Lock the table just once, to prevent it being done once per row. - # Note that, according to Postgres' documentation, once obtained, - # the lock is held for the remainder of the current transaction. - self.engine.lock_table(txn, "user_ips") + # Lock the table just once, to prevent it being done once per row. + # Note that, according to Postgres' documentation, once obtained, + # the lock is held for the remainder of the current transaction. + self.engine.lock_table(txn, "user_ips") for keyv, valv in zip(key_values, value_values): _keys = {x: y for x, y in zip(key_names, keyv)} @@ -1658,7 +1639,7 @@ class DatabasePool: table: string giving the table name keyvalues: dict of column names and values to select the row with retcol: string giving the name of the column to return - allow_none: If true, return None instead of failing if the SELECT + allow_none: If true, return None instead of raising StoreError if the SELECT statement returns no rows desc: description of the transaction, for logging and metrics """ @@ -2075,13 +2056,14 @@ class DatabasePool: retcols: Collection[str], allow_none: bool = False, ) -> Optional[Dict[str, Any]]: - select_sql = "SELECT %s FROM %s WHERE %s" % ( - ", ".join(retcols), - table, - " AND ".join("%s = ?" % (k,) for k in keyvalues), - ) + select_sql = "SELECT %s FROM %s" % (", ".join(retcols), table) + + if keyvalues: + select_sql += " WHERE %s" % (" AND ".join("%s = ?" % k for k in keyvalues),) + txn.execute(select_sql, list(keyvalues.values())) + else: + txn.execute(select_sql) - txn.execute(select_sql, list(keyvalues.values())) row = txn.fetchone() if not row: diff --git a/synapse/storage/databases/main/__init__.py b/synapse/storage/databases/main/__init__.py index a62b4abd4e..0e47592be3 100644 --- a/synapse/storage/databases/main/__init__.py +++ b/synapse/storage/databases/main/__init__.py @@ -26,9 +26,7 @@ from synapse.storage.database import ( from synapse.storage.databases.main.stats import UserSortOrder from synapse.storage.engines import BaseDatabaseEngine from synapse.storage.types import Cursor -from synapse.storage.util.id_generators import StreamIdGenerator from synapse.types import JsonDict, get_domain_from_id -from synapse.util.caches.stream_change_cache import StreamChangeCache from .account_data import AccountDataStore from .appservice import ApplicationServiceStore, ApplicationServiceTransactionStore @@ -138,41 +136,8 @@ class DataStore( self._clock = hs.get_clock() self.database_engine = database.engine - self._device_list_id_gen = StreamIdGenerator( - db_conn, - "device_lists_stream", - "stream_id", - extra_tables=[ - ("user_signature_stream", "stream_id"), - ("device_lists_outbound_pokes", "stream_id"), - ("device_lists_changes_in_room", "stream_id"), - ], - ) - super().__init__(database, db_conn, hs) - events_max = self._stream_id_gen.get_current_token() - curr_state_delta_prefill, min_curr_state_delta_id = self.db_pool.get_cache_dict( - db_conn, - "current_state_delta_stream", - entity_column="room_id", - stream_column="stream_id", - max_value=events_max, # As we share the stream id with events token - limit=1000, - ) - self._curr_state_delta_stream_cache = StreamChangeCache( - "_curr_state_delta_stream_cache", - min_curr_state_delta_id, - prefilled_cache=curr_state_delta_prefill, - ) - - self._stream_order_on_start = self.get_room_max_stream_ordering() - self._min_stream_order_on_start = self.get_room_min_stream_ordering() - - def get_device_stream_token(self) -> int: - # TODO: shouldn't this be moved to `DeviceWorkerStore`? - return self._device_list_id_gen.get_current_token() - async def get_users(self) -> List[JsonDict]: """Function to retrieve a list of users in users table. @@ -201,7 +166,7 @@ class DataStore( name: Optional[str] = None, guests: bool = True, deactivated: bool = False, - order_by: str = UserSortOrder.USER_ID.value, + order_by: str = UserSortOrder.NAME.value, direction: str = "f", approved: bool = True, ) -> Tuple[List[JsonDict], int]: @@ -261,6 +226,7 @@ class DataStore( sql_base = f""" FROM users as u LEFT JOIN profiles AS p ON u.name = '@' || p.user_id || ':' || ? + LEFT JOIN erased_users AS eu ON u.name = eu.user_id {where_clause} """ sql = "SELECT COUNT(*) as total_users " + sql_base @@ -269,7 +235,8 @@ class DataStore( sql = f""" SELECT name, user_type, is_guest, admin, deactivated, shadow_banned, - displayname, avatar_url, creation_ts * 1000 as creation_ts, approved + displayname, avatar_url, creation_ts * 1000 as creation_ts, approved, + eu.user_id is not null as erased {sql_base} ORDER BY {order_by_column} {order}, u.name ASC LIMIT ? OFFSET ? @@ -277,6 +244,13 @@ class DataStore( args += [limit, start] txn.execute(sql, args) users = self.db_pool.cursor_to_dict(txn) + + # some of those boolean values are returned as integers when we're on SQLite + columns_to_boolify = ["erased"] + for user in users: + for column in columns_to_boolify: + user[column] = bool(user[column]) + return users, count return await self.db_pool.runInteraction( diff --git a/synapse/storage/databases/main/account_data.py b/synapse/storage/databases/main/account_data.py index c38b8a9e5a..07908c41d9 100644 --- a/synapse/storage/databases/main/account_data.py +++ b/synapse/storage/databases/main/account_data.py @@ -27,7 +27,6 @@ from typing import ( ) from synapse.api.constants import AccountDataTypes -from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker from synapse.replication.tcp.streams import AccountDataStream, TagAccountDataStream from synapse.storage._base import db_to_json from synapse.storage.database import ( @@ -68,12 +67,11 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) # to write account data. A value of `True` implies that `_account_data_id_gen` # is an `AbstractStreamIdGenerator` and not just a tracker. self._account_data_id_gen: AbstractStreamIdTracker + self._can_write_to_account_data = ( + self._instance_name in hs.config.worker.writers.account_data + ) if isinstance(database.engine, PostgresEngine): - self._can_write_to_account_data = ( - self._instance_name in hs.config.worker.writers.account_data - ) - self._account_data_id_gen = MultiWriterIdGenerator( db_conn=db_conn, db=database, @@ -95,21 +93,13 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) # `StreamIdGenerator`, otherwise we use `SlavedIdTracker` which gets # updated over replication. (Multiple writers are not supported for # SQLite). - if self._instance_name in hs.config.worker.writers.account_data: - self._can_write_to_account_data = True - self._account_data_id_gen = StreamIdGenerator( - db_conn, - "room_account_data", - "stream_id", - extra_tables=[("room_tags_revisions", "stream_id")], - ) - else: - self._account_data_id_gen = SlavedIdTracker( - db_conn, - "room_account_data", - "stream_id", - extra_tables=[("room_tags_revisions", "stream_id")], - ) + self._account_data_id_gen = StreamIdGenerator( + db_conn, + "room_account_data", + "stream_id", + extra_tables=[("room_tags_revisions", "stream_id")], + is_writer=self._instance_name in hs.config.worker.writers.account_data, + ) account_max = self.get_max_account_data_stream_id() self._account_data_stream_cache = StreamChangeCache( @@ -459,9 +449,6 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) content_json = json_encoder.encode(content) async with self._account_data_id_gen.get_next() as next_id: - # no need to lock here as room_account_data has a unique constraint - # on (user_id, room_id, account_data_type) so simple_upsert will - # retry if there is a conflict. await self.db_pool.simple_upsert( desc="add_room_account_data", table="room_account_data", @@ -471,7 +458,6 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) "account_data_type": account_data_type, }, values={"stream_id": next_id, "content": content_json}, - lock=False, ) self._account_data_stream_cache.entity_has_changed(user_id, next_id) @@ -527,15 +513,11 @@ class AccountDataWorkerStore(PushRulesWorkerStore, CacheInvalidationWorkerStore) ) -> None: content_json = json_encoder.encode(content) - # no need to lock here as account_data has a unique constraint on - # (user_id, account_data_type) so simple_upsert will retry if - # there is a conflict. self.db_pool.simple_upsert_txn( txn, table="account_data", keyvalues={"user_id": user_id, "account_data_type": account_data_type}, values={"stream_id": next_id, "content": content_json}, - lock=False, ) # Ignored users get denormalized into a separate table as an optimisation. diff --git a/synapse/storage/databases/main/appservice.py b/synapse/storage/databases/main/appservice.py index 64b70a7b28..c2c8018ee2 100644 --- a/synapse/storage/databases/main/appservice.py +++ b/synapse/storage/databases/main/appservice.py @@ -20,7 +20,7 @@ from synapse.appservice import ( ApplicationService, ApplicationServiceState, AppServiceTransaction, - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.config.appservice import load_appservices @@ -157,10 +157,23 @@ class ApplicationServiceWorkerStore(RoomMemberWorkerStore): app_service: "ApplicationService", cache_context: _CacheContext, ) -> List[str]: - users_in_room = await self.get_users_in_room( + """ + Get all users in a room that the appservice controls. + + Args: + room_id: The room to check in. + app_service: The application service to check interest/control against + + Returns: + List of user IDs that the appservice controls. + """ + # We can use `get_local_users_in_room(...)` here because an application service + # can only be interested in local users of the server it's on (ignore any remote + # users that might match the user namespace regex). + local_users_in_room = await self.get_local_users_in_room( room_id, on_invalidate=cache_context.invalidate ) - return list(filter(app_service.is_interested_in_user, users_in_room)) + return list(filter(app_service.is_interested_in_user, local_users_in_room)) class ApplicationServiceStore(ApplicationServiceWorkerStore): @@ -247,7 +260,7 @@ class ApplicationServiceTransactionWorkerStore( events: List[EventBase], ephemeral: List[JsonDict], to_device_messages: List[JsonDict], - one_time_key_counts: TransactionOneTimeKeyCounts, + one_time_keys_count: TransactionOneTimeKeysCount, unused_fallback_keys: TransactionUnusedFallbackKeys, device_list_summary: DeviceListUpdates, ) -> AppServiceTransaction: @@ -260,7 +273,7 @@ class ApplicationServiceTransactionWorkerStore( events: A list of persistent events to put in the transaction. ephemeral: A list of ephemeral events to put in the transaction. to_device_messages: A list of to-device messages to put in the transaction. - one_time_key_counts: Counts of remaining one-time keys for relevant + one_time_keys_count: Counts of remaining one-time keys for relevant appservice devices in the transaction. unused_fallback_keys: Lists of unused fallback keys for relevant appservice devices in the transaction. @@ -286,7 +299,7 @@ class ApplicationServiceTransactionWorkerStore( events=events, ephemeral=ephemeral, to_device_messages=to_device_messages, - one_time_key_counts=one_time_key_counts, + one_time_keys_count=one_time_keys_count, unused_fallback_keys=unused_fallback_keys, device_list_summary=device_list_summary, ) @@ -366,7 +379,7 @@ class ApplicationServiceTransactionWorkerStore( events=events, ephemeral=[], to_device_messages=[], - one_time_key_counts={}, + one_time_keys_count={}, unused_fallback_keys={}, device_list_summary=DeviceListUpdates(), ) @@ -438,8 +451,6 @@ class ApplicationServiceTransactionWorkerStore( table="application_services_state", keyvalues={"as_id": service.id}, values={f"{stream_type}_stream_id": pos}, - # no need to lock when emulating upsert: as_id is a unique key - lock=False, desc="set_appservice_stream_type_pos", ) diff --git a/synapse/storage/databases/main/cache.py b/synapse/storage/databases/main/cache.py index 3b8ed1f7ee..a58668a380 100644 --- a/synapse/storage/databases/main/cache.py +++ b/synapse/storage/databases/main/cache.py @@ -244,24 +244,29 @@ class CacheInvalidationWorkerStore(SQLBaseStore): # redacted. self._attempt_to_invalidate_cache("get_relations_for_event", (redacts,)) self._attempt_to_invalidate_cache("get_applicable_edit", (redacts,)) + self._attempt_to_invalidate_cache("get_thread_id", (redacts,)) + self._attempt_to_invalidate_cache("get_thread_id_for_receipts", (redacts,)) if etype == EventTypes.Member: self._membership_stream_cache.entity_has_changed(state_key, stream_ordering) self._attempt_to_invalidate_cache( "get_invited_rooms_for_local_user", (state_key,) ) + self._attempt_to_invalidate_cache( + "get_rooms_for_user_with_stream_ordering", (state_key,) + ) + self._attempt_to_invalidate_cache("get_rooms_for_user", (state_key,)) if relates_to: self._attempt_to_invalidate_cache("get_relations_for_event", (relates_to,)) + self._attempt_to_invalidate_cache("get_references_for_event", (relates_to,)) self._attempt_to_invalidate_cache( "get_aggregation_groups_for_event", (relates_to,) ) self._attempt_to_invalidate_cache("get_applicable_edit", (relates_to,)) self._attempt_to_invalidate_cache("get_thread_summary", (relates_to,)) self._attempt_to_invalidate_cache("get_thread_participated", (relates_to,)) - self._attempt_to_invalidate_cache( - "get_mutual_event_relations_for_rel_type", (relates_to,) - ) + self._attempt_to_invalidate_cache("get_threads", (room_id,)) async def invalidate_cache_and_stream( self, cache_name: str, keys: Tuple[Any, ...] diff --git a/synapse/storage/databases/main/deviceinbox.py b/synapse/storage/databases/main/deviceinbox.py index 73c95ffb6f..48a54d9cb8 100644 --- a/synapse/storage/databases/main/deviceinbox.py +++ b/synapse/storage/databases/main/deviceinbox.py @@ -26,8 +26,15 @@ from typing import ( cast, ) +from synapse.api.constants import EventContentFields from synapse.logging import issue9533_logger -from synapse.logging.opentracing import log_kv, set_tag, trace +from synapse.logging.opentracing import ( + SynapseTags, + log_kv, + set_tag, + start_active_span, + trace, +) from synapse.replication.tcp.streams import ToDeviceStream from synapse.storage._base import SQLBaseStore, db_to_json from synapse.storage.database import ( @@ -397,6 +404,17 @@ class DeviceInboxWorkerStore(SQLBaseStore): (recipient_user_id, recipient_device_id), [] ).append(message_dict) + # start a new span for each message, so that we can tag each separately + with start_active_span("get_to_device_message"): + set_tag(SynapseTags.TO_DEVICE_TYPE, message_dict["type"]) + set_tag(SynapseTags.TO_DEVICE_SENDER, message_dict["sender"]) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT, recipient_user_id) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT_DEVICE, recipient_device_id) + set_tag( + SynapseTags.TO_DEVICE_MSGID, + message_dict["content"].get(EventContentFields.TO_DEVICE_MSGID), + ) + if limit is not None and rowcount == limit: # We ended up bumping up against the message limit. There may be more messages # to retrieve. Return what we have, as well as the last stream position that @@ -678,12 +696,35 @@ class DeviceInboxWorkerStore(SQLBaseStore): ], ) - if remote_messages_by_destination: - issue9533_logger.debug( - "Queued outgoing to-device messages with stream_id %i for %s", - stream_id, - list(remote_messages_by_destination.keys()), - ) + for destination, edu in remote_messages_by_destination.items(): + if issue9533_logger.isEnabledFor(logging.DEBUG): + issue9533_logger.debug( + "Queued outgoing to-device messages with " + "stream_id %i, EDU message_id %s, type %s for %s: %s", + stream_id, + edu["message_id"], + edu["type"], + destination, + [ + f"{user_id}/{device_id} (msgid " + f"{msg.get(EventContentFields.TO_DEVICE_MSGID)})" + for (user_id, messages_by_device) in edu["messages"].items() + for (device_id, msg) in messages_by_device.items() + ], + ) + + for (user_id, messages_by_device) in edu["messages"].items(): + for (device_id, msg) in messages_by_device.items(): + with start_active_span("store_outgoing_to_device_message"): + set_tag(SynapseTags.TO_DEVICE_EDU_ID, edu["sender"]) + set_tag(SynapseTags.TO_DEVICE_EDU_ID, edu["message_id"]) + set_tag(SynapseTags.TO_DEVICE_TYPE, edu["type"]) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT, user_id) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT_DEVICE, device_id) + set_tag( + SynapseTags.TO_DEVICE_MSGID, + msg.get(EventContentFields.TO_DEVICE_MSGID), + ) async with self._device_inbox_id_gen.get_next() as stream_id: now_ms = self._clock.time_msec() @@ -801,7 +842,19 @@ class DeviceInboxWorkerStore(SQLBaseStore): # Only insert into the local inbox if the device exists on # this server device_id = row["device_id"] - message_json = json_encoder.encode(messages_by_device[device_id]) + + with start_active_span("serialise_to_device_message"): + msg = messages_by_device[device_id] + set_tag(SynapseTags.TO_DEVICE_TYPE, msg["type"]) + set_tag(SynapseTags.TO_DEVICE_SENDER, msg["sender"]) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT, user_id) + set_tag(SynapseTags.TO_DEVICE_RECIPIENT_DEVICE, device_id) + set_tag( + SynapseTags.TO_DEVICE_MSGID, + msg["content"].get(EventContentFields.TO_DEVICE_MSGID), + ) + message_json = json_encoder.encode(msg) + messages_json_for_user[device_id] = message_json if messages_json_for_user: @@ -821,15 +874,20 @@ class DeviceInboxWorkerStore(SQLBaseStore): ], ) - issue9533_logger.debug( - "Stored to-device messages with stream_id %i for %s", - stream_id, - [ - (user_id, device_id) - for (user_id, messages_by_device) in local_by_user_then_device.items() - for device_id in messages_by_device.keys() - ], - ) + if issue9533_logger.isEnabledFor(logging.DEBUG): + issue9533_logger.debug( + "Stored to-device messages with stream_id %i: %s", + stream_id, + [ + f"{user_id}/{device_id} (msgid " + f"{msg['content'].get(EventContentFields.TO_DEVICE_MSGID)})" + for ( + user_id, + messages_by_device, + ) in messages_by_user_then_device.items() + for (device_id, msg) in messages_by_device.items() + ], + ) class DeviceInboxBackgroundUpdateStore(SQLBaseStore): diff --git a/synapse/storage/databases/main/devices.py b/synapse/storage/databases/main/devices.py index 18358eca46..a5bb4d404e 100644 --- a/synapse/storage/databases/main/devices.py +++ b/synapse/storage/databases/main/devices.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 abc import logging from typing import ( TYPE_CHECKING, @@ -39,6 +38,7 @@ from synapse.logging.opentracing import ( whitelisted_homeserver, ) from synapse.metrics.background_process_metrics import wrap_as_background_process +from synapse.replication.tcp.streams._base import DeviceListsStream, UserSignatureStream from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause from synapse.storage.database import ( DatabasePool, @@ -49,11 +49,19 @@ from synapse.storage.database import ( from synapse.storage.databases.main.end_to_end_keys import EndToEndKeyWorkerStore from synapse.storage.databases.main.roommember import RoomMemberWorkerStore from synapse.storage.types import Cursor +from synapse.storage.util.id_generators import ( + AbstractStreamIdGenerator, + AbstractStreamIdTracker, + StreamIdGenerator, +) from synapse.types import JsonDict, get_verify_key_from_cross_signing_key from synapse.util import json_decoder, json_encoder from synapse.util.caches.descriptors import cached, cachedList from synapse.util.caches.lrucache import LruCache -from synapse.util.caches.stream_change_cache import StreamChangeCache +from synapse.util.caches.stream_change_cache import ( + AllEntitiesChangedResult, + StreamChangeCache, +) from synapse.util.cancellation import cancellable from synapse.util.iterutils import batch_iter from synapse.util.stringutils import shortstr @@ -80,9 +88,23 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): ): super().__init__(database, db_conn, hs) + # In the worker store this is an ID tracker which we overwrite in the non-worker + # class below that is used on the main process. + self._device_list_id_gen: AbstractStreamIdTracker = StreamIdGenerator( + db_conn, + "device_lists_stream", + "stream_id", + extra_tables=[ + ("user_signature_stream", "stream_id"), + ("device_lists_outbound_pokes", "stream_id"), + ("device_lists_changes_in_room", "stream_id"), + ], + is_writer=hs.config.worker.worker_app is None, + ) + # Type-ignore: _device_list_id_gen is mixed in from either DataStore (as a # StreamIdGenerator) or SlavedDataStore (as a SlavedIdTracker). - device_list_max = self._device_list_id_gen.get_current_token() # type: ignore[attr-defined] + device_list_max = self._device_list_id_gen.get_current_token() device_list_prefill, min_device_list_id = self.db_pool.get_cache_dict( db_conn, "device_lists_stream", @@ -136,6 +158,39 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): self._prune_old_outbound_device_pokes, 60 * 60 * 1000 ) + def process_replication_rows( + self, stream_name: str, instance_name: str, token: int, rows: Iterable[Any] + ) -> None: + if stream_name == DeviceListsStream.NAME: + self._device_list_id_gen.advance(instance_name, token) + self._invalidate_caches_for_devices(token, rows) + elif stream_name == UserSignatureStream.NAME: + self._device_list_id_gen.advance(instance_name, token) + for row in rows: + self._user_signature_stream_cache.entity_has_changed(row.user_id, token) + return super().process_replication_rows(stream_name, instance_name, token, rows) + + def _invalidate_caches_for_devices( + self, token: int, rows: Iterable[DeviceListsStream.DeviceListsStreamRow] + ) -> None: + for row in rows: + # The entities are either user IDs (starting with '@') whose devices + # have changed, or remote servers that we need to tell about + # changes. + if row.entity.startswith("@"): + self._device_list_stream_cache.entity_has_changed(row.entity, token) + self.get_cached_devices_for_user.invalidate((row.entity,)) + self._get_cached_user_device.invalidate((row.entity,)) + self.get_device_list_last_stream_id_for_remote.invalidate((row.entity,)) + + else: + self._device_list_federation_stream_cache.entity_has_changed( + row.entity, token + ) + + def get_device_stream_token(self) -> int: + return self._device_list_id_gen.get_current_token() + async def count_devices_by_users(self, user_ids: Optional[List[str]] = None) -> int: """Retrieve number of all devices of given users. Only returns number of devices that are not marked as hidden. @@ -274,6 +329,13 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): destination, int(from_stream_id) ) if not has_changed: + # debugging for https://github.com/matrix-org/synapse/issues/14251 + issue_8631_logger.debug( + "%s: no change between %i and %i", + destination, + from_stream_id, + now_stream_id, + ) return now_stream_id, [] updates = await self.db_pool.runInteraction( @@ -466,7 +528,7 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): limit: Maximum number of device updates to return Returns: - List: List of device update tuples: + List of device update tuples: - user_id - device_id - stream_id @@ -539,9 +601,11 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): "device_id": device_id, "prev_id": [prev_id] if prev_id else [], "stream_id": stream_id, - "org.matrix.opentracing_context": opentracing_context, } + if opentracing_context != "{}": + result["org.matrix.opentracing_context"] = opentracing_context + prev_id = stream_id if device is not None: @@ -549,7 +613,11 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): if keys: result["keys"] = keys - device_display_name = device.display_name + device_display_name = None + if ( + self.hs.config.federation.allow_device_name_lookup_over_federation + ): + device_display_name = device.display_name if device_display_name: result["device_display_name"] = device_display_name else: @@ -664,11 +732,6 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): }, ) - @abc.abstractmethod - def get_device_stream_token(self) -> int: - """Get the current stream id from the _device_list_id_gen""" - ... - @trace @cancellable async def get_user_devices_from_cache( @@ -739,7 +802,7 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): def get_cached_device_list_changes( self, from_key: int, - ) -> Optional[List[str]]: + ) -> AllEntitiesChangedResult: """Get set of users whose devices have changed since `from_key`, or None if that information is not in our cache. """ @@ -747,10 +810,58 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): return self._device_list_stream_cache.get_all_entities_changed(from_key) @cancellable + async def get_all_devices_changed( + self, + from_key: int, + to_key: int, + ) -> Set[str]: + """Get all users whose devices have changed in the given range. + + Args: + from_key: The minimum device lists stream token to query device list + changes for, exclusive. + to_key: The maximum device lists stream token to query device list + changes for, inclusive. + + Returns: + The set of user_ids whose devices have changed since `from_key` + (exclusive) until `to_key` (inclusive). + """ + + result = self._device_list_stream_cache.get_all_entities_changed(from_key) + + if result.hit: + # We know which users might have changed devices. + if not result.entities: + # If no users then we can return early. + return set() + + # Otherwise we need to filter down the list + return await self.get_users_whose_devices_changed( + from_key, result.entities, to_key + ) + + # If the cache didn't tell us anything, we just need to query the full + # range. + sql = """ + SELECT DISTINCT user_id FROM device_lists_stream + WHERE ? < stream_id AND stream_id <= ? + """ + + rows = await self.db_pool.execute( + "get_all_devices_changed", + None, + sql, + from_key, + to_key, + ) + return {u for u, in rows} + + @cancellable async def get_users_whose_devices_changed( self, from_key: int, - user_ids: Optional[Collection[str]] = None, + user_ids: Collection[str], to_key: Optional[int] = None, ) -> Set[str]: """Get set of users whose devices have changed since `from_key` that @@ -770,46 +881,31 @@ class DeviceWorkerStore(RoomMemberWorkerStore, EndToEndKeyWorkerStore): """ # Get set of users who *may* have changed. Users not in the returned # list have definitely not changed. - user_ids_to_check: Optional[Collection[str]] - if user_ids is None: - # Get set of all users that have had device list changes since 'from_key' - user_ids_to_check = self._device_list_stream_cache.get_all_entities_changed( - from_key - ) - else: - # The same as above, but filter results to only those users in 'user_ids' - user_ids_to_check = self._device_list_stream_cache.get_entities_changed( - user_ids, from_key - ) + user_ids_to_check = self._device_list_stream_cache.get_entities_changed( + user_ids, from_key + ) + # If an empty set was returned, there's nothing to do. if not user_ids_to_check: return set() - def _get_users_whose_devices_changed_txn(txn: LoggingTransaction) -> Set[str]: - changes: Set[str] = set() - - stream_id_where_clause = "stream_id > ?" - sql_args = [from_key] - - if to_key: - stream_id_where_clause += " AND stream_id <= ?" - sql_args.append(to_key) + if to_key is None: + to_key = self._device_list_id_gen.get_current_token() - sql = f""" + def _get_users_whose_devices_changed_txn(txn: LoggingTransaction) -> Set[str]: + sql = """ SELECT DISTINCT user_id FROM device_lists_stream - WHERE {stream_id_where_clause} - AND + WHERE ? < stream_id AND stream_id <= ? AND %s """ + changes: Set[str] = set() + # Query device changes with a batch of users at a time - # Assertion for mypy's benefit; see also - # https://mypy.readthedocs.io/en/stable/common_issues.html#narrowing-and-inner-functions - assert user_ids_to_check is not None for chunk in batch_iter(user_ids_to_check, 100): clause, args = make_in_list_sql_clause( txn.database_engine, "user_id", chunk ) - txn.execute(sql + clause, sql_args + args) + txn.execute(sql % (clause,), [from_key, to_key] + args) changes.update(user_id for user_id, in txn) return changes @@ -1381,6 +1477,13 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): self._remove_duplicate_outbound_pokes, ) + self.db_pool.updates.register_background_index_update( + "device_lists_changes_in_room_by_room_index", + index_name="device_lists_changes_in_room_by_room_idx", + table="device_lists_changes_in_room", + columns=["room_id", "stream_id"], + ) + async def _drop_device_list_streams_non_unique_indexes( self, progress: JsonDict, batch_size: int ) -> int: @@ -1468,6 +1571,10 @@ class DeviceBackgroundUpdateStore(SQLBaseStore): class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): + # Because we have write access, this will be a StreamIdGenerator + # (see DeviceWorkerStore.__init__) + _device_list_id_gen: AbstractStreamIdGenerator + def __init__( self, database: DatabasePool, @@ -1673,9 +1780,6 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): table="device_lists_remote_cache", keyvalues={"user_id": user_id, "device_id": device_id}, values={"content": json_encoder.encode(content)}, - # we don't need to lock, because we assume we are the only thread - # updating this user's devices. - lock=False, ) txn.call_after(self._get_cached_user_device.invalidate, (user_id, device_id)) @@ -1689,9 +1793,6 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): table="device_lists_remote_extremeties", keyvalues={"user_id": user_id}, values={"stream_id": stream_id}, - # again, we can assume we are the only thread updating this user's - # extremity. - lock=False, ) async def update_remote_device_list_cache( @@ -1744,9 +1845,6 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): table="device_lists_remote_extremeties", keyvalues={"user_id": user_id}, values={"stream_id": stream_id}, - # we don't need to lock, because we can assume we are the only thread - # updating this user's extremity. - lock=False, ) async def add_device_change_to_streams( @@ -1792,7 +1890,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): context, ) - async with self._device_list_id_gen.get_next_mult( # type: ignore[attr-defined] + async with self._device_list_id_gen.get_next_mult( len(device_ids) ) as stream_ids: await self.db_pool.runInteraction( @@ -1842,7 +1940,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): self, txn: LoggingTransaction, user_id: str, - device_ids: Iterable[str], + device_id: str, hosts: Collection[str], stream_ids: List[int], context: Optional[Dict[str, str]], @@ -1858,6 +1956,21 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): stream_id_iterator = iter(stream_ids) encoded_context = json_encoder.encode(context) + mark_sent = not self.hs.is_mine_id(user_id) + + values = [ + ( + destination, + next(stream_id_iterator), + user_id, + device_id, + mark_sent, + now, + encoded_context if whitelisted_homeserver(destination) else "{}", + ) + for destination in hosts + ] + self.db_pool.simple_insert_many_txn( txn, table="device_lists_outbound_pokes", @@ -1870,23 +1983,21 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): "ts", "opentracing_context", ), - values=[ - ( - destination, - next(stream_id_iterator), - user_id, - device_id, - not self.hs.is_mine_id( - user_id - ), # We only need to send out update for *our* users - now, - encoded_context if whitelisted_homeserver(destination) else "{}", - ) - for destination in hosts - for device_id in device_ids - ], + values=values, ) + # debugging for https://github.com/matrix-org/synapse/issues/14251 + if issue_8631_logger.isEnabledFor(logging.DEBUG): + issue_8631_logger.debug( + "Recorded outbound pokes for %s:%s with device stream ids %s", + user_id, + device_id, + { + stream_id: destination + for (destination, stream_id, _, _, _, _, _) in values + }, + ) + def _add_device_outbound_room_poke_txn( self, txn: LoggingTransaction, @@ -1931,27 +2042,48 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): ) async def get_uncoverted_outbound_room_pokes( - self, limit: int = 10 + self, start_stream_id: int, start_room_id: str, limit: int = 10 ) -> List[Tuple[str, str, str, int, Optional[Dict[str, str]]]]: """Get device list changes by room that have not yet been handled and written to `device_lists_outbound_pokes`. + Args: + start_stream_id: Together with `start_room_id`, indicates the position after + which to return device list changes. + start_room_id: Together with `start_stream_id`, indicates the position after + which to return device list changes. + limit: The maximum number of device list changes to return. + Returns: - A list of user ID, device ID, room ID, stream ID and optional opentracing context. + A list of user ID, device ID, room ID, stream ID and optional opentracing + context, in order of ascending (stream ID, room ID). """ sql = """ SELECT user_id, device_id, room_id, stream_id, opentracing_context FROM device_lists_changes_in_room - WHERE NOT converted_to_destinations - ORDER BY stream_id + WHERE + (stream_id, room_id) > (?, ?) AND + stream_id <= ? AND + NOT converted_to_destinations + ORDER BY stream_id ASC, room_id ASC LIMIT ? """ def get_uncoverted_outbound_room_pokes_txn( txn: LoggingTransaction, ) -> List[Tuple[str, str, str, int, Optional[Dict[str, str]]]]: - txn.execute(sql, (limit,)) + txn.execute( + sql, + ( + start_stream_id, + start_room_id, + # Avoid returning rows if there may be uncommitted device list + # changes with smaller stream IDs. + self._device_list_id_gen.get_current_token(), + limit, + ), + ) return [ ( @@ -1973,52 +2105,28 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): user_id: str, device_id: str, room_id: str, - stream_id: Optional[int], hosts: Collection[str], context: Optional[Dict[str, str]], ) -> None: """Queue the device update to be sent to the given set of hosts, calculated from the room ID. - - Marks the associated row in `device_lists_changes_in_room` as handled, - if `stream_id` is provided. """ + if not hosts: + return def add_device_list_outbound_pokes_txn( txn: LoggingTransaction, stream_ids: List[int] ) -> None: - if hosts: - self._add_device_outbound_poke_to_stream_txn( - txn, - user_id=user_id, - device_ids=[device_id], - hosts=hosts, - stream_ids=stream_ids, - context=context, - ) - - if stream_id: - self.db_pool.simple_update_txn( - txn, - table="device_lists_changes_in_room", - keyvalues={ - "user_id": user_id, - "device_id": device_id, - "stream_id": stream_id, - "room_id": room_id, - }, - updatevalues={"converted_to_destinations": True}, - ) - - if not hosts: - # If there are no hosts then we don't try and generate stream IDs. - return await self.db_pool.runInteraction( - "add_device_list_outbound_pokes", - add_device_list_outbound_pokes_txn, - [], + self._add_device_outbound_poke_to_stream_txn( + txn, + user_id=user_id, + device_id=device_id, + hosts=hosts, + stream_ids=stream_ids, + context=context, ) - async with self._device_list_id_gen.get_next_mult(len(hosts)) as stream_ids: # type: ignore[attr-defined] + async with self._device_list_id_gen.get_next_mult(len(hosts)) as stream_ids: return await self.db_pool.runInteraction( "add_device_list_outbound_pokes", add_device_list_outbound_pokes_txn, @@ -2032,7 +2140,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): updates during partial joins. """ - async with self._device_list_id_gen.get_next() as stream_id: # type: ignore[attr-defined] + async with self._device_list_id_gen.get_next() as stream_id: await self.db_pool.simple_upsert( table="device_lists_remote_pending", keyvalues={ @@ -2079,3 +2187,37 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore): "get_pending_remote_device_list_updates_for_room", get_pending_remote_device_list_updates_for_room_txn, ) + + async def get_device_change_last_converted_pos(self) -> Tuple[int, str]: + """ + Get the position of the last row in `device_list_changes_in_room` that has been + converted to `device_lists_outbound_pokes`. + + Rows with a strictly greater position where `converted_to_destinations` is + `FALSE` have not been converted. + """ + + row = await self.db_pool.simple_select_one( + table="device_lists_changes_converted_stream_position", + keyvalues={}, + retcols=["stream_id", "room_id"], + desc="get_device_change_last_converted_pos", + ) + return row["stream_id"], row["room_id"] + + async def set_device_change_last_converted_pos( + self, + stream_id: int, + room_id: str, + ) -> None: + """ + Set the position of the last row in `device_list_changes_in_room` that has been + converted to `device_lists_outbound_pokes`. + """ + + await self.db_pool.simple_update_one( + table="device_lists_changes_converted_stream_position", + keyvalues={}, + updatevalues={"stream_id": stream_id, "room_id": room_id}, + desc="set_device_change_last_converted_pos", + ) diff --git a/synapse/storage/databases/main/e2e_room_keys.py b/synapse/storage/databases/main/e2e_room_keys.py index af59be6b48..6240f9a75e 100644 --- a/synapse/storage/databases/main/e2e_room_keys.py +++ b/synapse/storage/databases/main/e2e_room_keys.py @@ -391,10 +391,10 @@ class EndToEndRoomKeyStore(SQLBaseStore): Returns: A dict giving the info metadata for this backup version, with fields including: - version(str) - algorithm(str) - auth_data(object): opaque dict supplied by the client - etag(int): tag of the keys in the backup + version (str) + algorithm (str) + auth_data (object): opaque dict supplied by the client + etag (int): tag of the keys in the backup """ def _get_e2e_room_keys_version_info_txn(txn: LoggingTransaction) -> JsonDict: diff --git a/synapse/storage/databases/main/end_to_end_keys.py b/synapse/storage/databases/main/end_to_end_keys.py index 8a10ae800c..4c691642e2 100644 --- a/synapse/storage/databases/main/end_to_end_keys.py +++ b/synapse/storage/databases/main/end_to_end_keys.py @@ -33,7 +33,7 @@ from typing_extensions import Literal from synapse.api.constants import DeviceKeyAlgorithms from synapse.appservice import ( - TransactionOneTimeKeyCounts, + TransactionOneTimeKeysCount, TransactionUnusedFallbackKeys, ) from synapse.logging.opentracing import log_kv, set_tag, trace @@ -139,11 +139,15 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker @trace @cancellable async def get_e2e_device_keys_for_cs_api( - self, query_list: List[Tuple[str, Optional[str]]] + self, + query_list: Collection[Tuple[str, Optional[str]]], + include_displaynames: bool = True, ) -> Dict[str, Dict[str, JsonDict]]: """Fetch a list of device keys, formatted suitably for the C/S API. Args: - query_list(list): List of pairs of user_ids and device_ids. + query_list: List of pairs of user_ids and device_ids. + include_displaynames: Whether to include the displayname of returned devices + (if one exists). Returns: Dict mapping from user-id to dict mapping from device_id to key data. The key data will be a dict in the same format as the @@ -166,9 +170,12 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker continue r["unsigned"] = {} - display_name = device_info.display_name - if display_name is not None: - r["unsigned"]["device_display_name"] = display_name + if include_displaynames: + # Include the device's display name in the "unsigned" dictionary + display_name = device_info.display_name + if display_name is not None: + r["unsigned"]["device_display_name"] = display_name + rv[user_id][device_id] = r return rv @@ -405,10 +412,9 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker """Retrieve a number of one-time keys for a user Args: - user_id(str): id of user to get keys for - device_id(str): id of device to get keys for - key_ids(list[str]): list of key ids (excluding algorithm) to - retrieve + user_id: id of user to get keys for + device_id: id of device to get keys for + key_ids: list of key ids (excluding algorithm) to retrieve Returns: A map from (algorithm, key_id) to json string for key @@ -508,7 +514,7 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker async def count_bulk_e2e_one_time_keys_for_as( self, user_ids: Collection[str] - ) -> TransactionOneTimeKeyCounts: + ) -> TransactionOneTimeKeysCount: """ Counts, in bulk, the one-time keys for all the users specified. Intended to be used by application services for populating OTK counts in @@ -522,7 +528,7 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker def _count_bulk_e2e_one_time_keys_txn( txn: LoggingTransaction, - ) -> TransactionOneTimeKeyCounts: + ) -> TransactionOneTimeKeysCount: user_in_where_clause, user_parameters = make_in_list_sql_clause( self.database_engine, "user_id", user_ids ) @@ -535,7 +541,7 @@ class EndToEndKeyWorkerStore(EndToEndKeyBackgroundStore, CacheInvalidationWorker """ txn.execute(sql, user_parameters) - result: TransactionOneTimeKeyCounts = {} + result: TransactionOneTimeKeysCount = {} for user_id, device_id, algorithm, count in txn: # We deliberately construct empty dictionaries for diff --git a/synapse/storage/databases/main/event_federation.py b/synapse/storage/databases/main/event_federation.py index 6b9a629edd..bbee02ab18 100644 --- a/synapse/storage/databases/main/event_federation.py +++ b/synapse/storage/databases/main/event_federation.py @@ -1501,6 +1501,12 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas event_id: The event that failed to be fetched or processed cause: The error message or reason that we failed to pull the event """ + logger.debug( + "record_event_failed_pull_attempt room_id=%s, event_id=%s, cause=%s", + room_id, + event_id, + cause, + ) await self.db_pool.runInteraction( "record_event_failed_pull_attempt", self._record_event_failed_pull_attempt_upsert_txn, @@ -1530,6 +1536,54 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas txn.execute(sql, (room_id, event_id, 1, self._clock.time_msec(), cause)) + @trace + async def get_event_ids_to_not_pull_from_backoff( + self, + room_id: str, + event_ids: Collection[str], + ) -> List[str]: + """ + Filter down the events to ones that we've failed to pull before recently. Uses + exponential backoff. + + Args: + room_id: The room that the events belong to + event_ids: A list of events to filter down + + Returns: + List of event_ids that should not be attempted to be pulled + """ + event_failed_pull_attempts = await self.db_pool.simple_select_many_batch( + table="event_failed_pull_attempts", + column="event_id", + iterable=event_ids, + keyvalues={}, + retcols=( + "event_id", + "last_attempt_ts", + "num_attempts", + ), + desc="get_event_ids_to_not_pull_from_backoff", + ) + + current_time = self._clock.time_msec() + return [ + event_failed_pull_attempt["event_id"] + for event_failed_pull_attempt in event_failed_pull_attempts + # Exponential back-off (up to the upper bound) so we don't try to + # pull the same event over and over. ex. 2hr, 4hr, 8hr, 16hr, etc. + if current_time + < event_failed_pull_attempt["last_attempt_ts"] + + ( + 2 + ** min( + event_failed_pull_attempt["num_attempts"], + BACKFILL_EVENT_EXPONENTIAL_BACKOFF_MAXIMUM_DOUBLING_STEPS, + ) + ) + * BACKFILL_EVENT_EXPONENTIAL_BACKOFF_STEP_MILLISECONDS + ] + async def get_missing_events( self, room_id: str, @@ -1632,7 +1686,6 @@ class EventFederationWorkerStore(SignatureWorkerStore, EventsWorkerStore, SQLBas }, insertion_values={}, desc="insert_insertion_extremity", - lock=False, ) async def insert_received_event_to_staging( diff --git a/synapse/storage/databases/main/event_push_actions.py b/synapse/storage/databases/main/event_push_actions.py index 332e13d1c9..7ebe34f773 100644 --- a/synapse/storage/databases/main/event_push_actions.py +++ b/synapse/storage/databases/main/event_push_actions.py @@ -74,6 +74,7 @@ receipt. """ import logging +from collections import defaultdict from typing import ( TYPE_CHECKING, Collection, @@ -95,6 +96,7 @@ from synapse.storage.database import ( DatabasePool, LoggingDatabaseConnection, LoggingTransaction, + PostgresEngine, ) from synapse.storage.databases.main.receipts import ReceiptsWorkerStore from synapse.storage.databases.main.stream import StreamWorkerStore @@ -294,6 +296,44 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas self._background_backfill_thread_id, ) + # Indexes which will be used to quickly make the thread_id column non-null. + self.db_pool.updates.register_background_index_update( + "event_push_actions_thread_id_null", + index_name="event_push_actions_thread_id_null", + table="event_push_actions", + columns=["thread_id"], + where_clause="thread_id IS NULL", + ) + self.db_pool.updates.register_background_index_update( + "event_push_summary_thread_id_null", + index_name="event_push_summary_thread_id_null", + table="event_push_summary", + columns=["thread_id"], + where_clause="thread_id IS NULL", + ) + + # Check ASAP (and then later, every 1s) to see if we have finished + # background updates the event_push_actions and event_push_summary tables. + self._clock.call_later(0.0, self._check_event_push_backfill_thread_id) + self._event_push_backfill_thread_id_done = False + + @wrap_as_background_process("check_event_push_backfill_thread_id") + async def _check_event_push_backfill_thread_id(self) -> None: + """ + Has thread_id finished backfilling? + + If not, we need to just-in-time update it so the queries work. + """ + done = await self.db_pool.updates.has_completed_background_update( + "event_push_backfill_thread_id" + ) + + if done: + self._event_push_backfill_thread_id_done = True + else: + # Reschedule to run. + self._clock.call_later(15.0, self._check_event_push_backfill_thread_id) + async def _background_backfill_thread_id( self, progress: JsonDict, batch_size: int ) -> int: @@ -310,11 +350,11 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas event_push_actions_done = progress.get("event_push_actions_done", False) def add_thread_id_txn( - txn: LoggingTransaction, table_name: str, start_stream_ordering: int + txn: LoggingTransaction, start_stream_ordering: int ) -> int: - sql = f""" + sql = """ SELECT stream_ordering - FROM {table_name} + FROM event_push_actions WHERE thread_id IS NULL AND stream_ordering > ? @@ -326,7 +366,7 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas # No more rows to process. rows = txn.fetchall() if not rows: - progress[f"{table_name}_done"] = True + progress["event_push_actions_done"] = True self.db_pool.updates._background_update_progress_txn( txn, "event_push_backfill_thread_id", progress ) @@ -335,16 +375,65 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas # Update the thread ID for any of those rows. max_stream_ordering = rows[-1][0] - sql = f""" - UPDATE {table_name} + sql = """ + UPDATE event_push_actions SET thread_id = 'main' - WHERE stream_ordering <= ? AND thread_id IS NULL + WHERE ? < stream_ordering AND stream_ordering <= ? AND thread_id IS NULL """ - txn.execute(sql, (max_stream_ordering,)) + txn.execute( + sql, + ( + start_stream_ordering, + max_stream_ordering, + ), + ) # Update progress. processed_rows = txn.rowcount - progress[f"max_{table_name}_stream_ordering"] = max_stream_ordering + progress["max_event_push_actions_stream_ordering"] = max_stream_ordering + self.db_pool.updates._background_update_progress_txn( + txn, "event_push_backfill_thread_id", progress + ) + + return processed_rows + + def add_thread_id_summary_txn(txn: LoggingTransaction) -> int: + min_user_id = progress.get("max_summary_user_id", "") + min_room_id = progress.get("max_summary_room_id", "") + + # Slightly overcomplicated query for getting the Nth user ID / room + # ID tuple, or the last if there are less than N remaining. + sql = """ + SELECT user_id, room_id FROM ( + SELECT user_id, room_id FROM event_push_summary + WHERE (user_id, room_id) > (?, ?) + AND thread_id IS NULL + ORDER BY user_id, room_id + LIMIT ? + ) AS e + ORDER BY user_id DESC, room_id DESC + LIMIT 1 + """ + + txn.execute(sql, (min_user_id, min_room_id, batch_size)) + row = txn.fetchone() + if not row: + return 0 + + max_user_id, max_room_id = row + + sql = """ + UPDATE event_push_summary + SET thread_id = 'main' + WHERE + (?, ?) < (user_id, room_id) AND (user_id, room_id) <= (?, ?) + AND thread_id IS NULL + """ + txn.execute(sql, (min_user_id, min_room_id, max_user_id, max_room_id)) + processed_rows = txn.rowcount + + progress["max_summary_user_id"] = max_user_id + progress["max_summary_room_id"] = max_room_id self.db_pool.updates._background_update_progress_txn( txn, "event_push_backfill_thread_id", progress ) @@ -360,15 +449,12 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas result = await self.db_pool.runInteraction( "event_push_backfill_thread_id", add_thread_id_txn, - "event_push_actions", progress.get("max_event_push_actions_stream_ordering", 0), ) else: result = await self.db_pool.runInteraction( "event_push_backfill_thread_id", - add_thread_id_txn, - "event_push_summary", - progress.get("max_event_push_summary_stream_ordering", 0), + add_thread_id_summary_txn, ) # Only done after the event_push_summary table is done. @@ -379,6 +465,153 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas return result + async def get_unread_counts_by_room_for_user(self, user_id: str) -> Dict[str, int]: + """Get the notification count by room for a user. Only considers notifications, + not highlight or unread counts, and threads are currently aggregated under their room. + + This function is intentionally not cached because it is called to calculate the + unread badge for push notifications and thus the result is expected to change. + + Note that this function assumes the user is a member of the room. Because + summary rows are not removed when a user leaves a room, the caller must + filter out those results from the result. + + Returns: + A map of room ID to notification counts for the given user. + """ + return await self.db_pool.runInteraction( + "get_unread_counts_by_room_for_user", + self._get_unread_counts_by_room_for_user_txn, + user_id, + ) + + def _get_unread_counts_by_room_for_user_txn( + self, txn: LoggingTransaction, user_id: str + ) -> Dict[str, int]: + receipt_types_clause, args = make_in_list_sql_clause( + self.database_engine, + "receipt_type", + (ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE), + ) + args.extend([user_id, user_id]) + + receipts_cte = f""" + WITH all_receipts AS ( + SELECT room_id, thread_id, MAX(event_stream_ordering) AS max_receipt_stream_ordering + FROM receipts_linearized + LEFT JOIN events USING (room_id, event_id) + WHERE + {receipt_types_clause} + AND user_id = ? + GROUP BY room_id, thread_id + ) + """ + + receipts_joins = """ + LEFT JOIN ( + SELECT room_id, thread_id, + max_receipt_stream_ordering AS threaded_receipt_stream_ordering + FROM all_receipts + WHERE thread_id IS NOT NULL + ) AS threaded_receipts USING (room_id, thread_id) + LEFT JOIN ( + SELECT room_id, thread_id, + max_receipt_stream_ordering AS unthreaded_receipt_stream_ordering + FROM all_receipts + WHERE thread_id IS NULL + ) AS unthreaded_receipts USING (room_id) + """ + + # First get summary counts by room / thread for the user. We use the max receipt + # stream ordering of both threaded & unthreaded receipts to compare against the + # summary table. + # + # PostgreSQL and SQLite differ in comparing scalar numerics. + if isinstance(self.database_engine, PostgresEngine): + # GREATEST ignores NULLs. + max_clause = """GREATEST( + threaded_receipt_stream_ordering, + unthreaded_receipt_stream_ordering + )""" + else: + # MAX returns NULL if any are NULL, so COALESCE to 0 first. + max_clause = """MAX( + COALESCE(threaded_receipt_stream_ordering, 0), + COALESCE(unthreaded_receipt_stream_ordering, 0) + )""" + + sql = f""" + {receipts_cte} + SELECT eps.room_id, eps.thread_id, notif_count + FROM event_push_summary AS eps + {receipts_joins} + WHERE user_id = ? + AND notif_count != 0 + AND ( + (last_receipt_stream_ordering IS NULL AND stream_ordering > {max_clause}) + OR last_receipt_stream_ordering = {max_clause} + ) + """ + txn.execute(sql, args) + + seen_thread_ids = set() + room_to_count: Dict[str, int] = defaultdict(int) + + for room_id, thread_id, notif_count in txn: + room_to_count[room_id] += notif_count + seen_thread_ids.add(thread_id) + + # Now get any event push actions that haven't been rotated using the same OR + # join and filter by receipt and event push summary rotated up to stream ordering. + sql = f""" + {receipts_cte} + SELECT epa.room_id, epa.thread_id, COUNT(CASE WHEN epa.notif = 1 THEN 1 END) AS notif_count + FROM event_push_actions AS epa + {receipts_joins} + WHERE user_id = ? + AND epa.notif = 1 + AND stream_ordering > (SELECT stream_ordering FROM event_push_summary_stream_ordering) + AND (threaded_receipt_stream_ordering IS NULL OR stream_ordering > threaded_receipt_stream_ordering) + AND (unthreaded_receipt_stream_ordering IS NULL OR stream_ordering > unthreaded_receipt_stream_ordering) + GROUP BY epa.room_id, epa.thread_id + """ + txn.execute(sql, args) + + for room_id, thread_id, notif_count in txn: + # Note: only count push actions we have valid summaries for with up to date receipt. + if thread_id not in seen_thread_ids: + continue + room_to_count[room_id] += notif_count + + thread_id_clause, thread_ids_args = make_in_list_sql_clause( + self.database_engine, "epa.thread_id", seen_thread_ids + ) + + # Finally re-check event_push_actions for any rooms not in the summary, ignoring + # the rotated up-to position. This handles the case where a read receipt has arrived + # but not been rotated meaning the summary table is out of date, so we go back to + # the push actions table. + sql = f""" + {receipts_cte} + SELECT epa.room_id, COUNT(CASE WHEN epa.notif = 1 THEN 1 END) AS notif_count + FROM event_push_actions AS epa + {receipts_joins} + WHERE user_id = ? + AND NOT {thread_id_clause} + AND epa.notif = 1 + AND (threaded_receipt_stream_ordering IS NULL OR stream_ordering > threaded_receipt_stream_ordering) + AND (unthreaded_receipt_stream_ordering IS NULL OR stream_ordering > unthreaded_receipt_stream_ordering) + GROUP BY epa.room_id + """ + + args.extend(thread_ids_args) + txn.execute(sql, args) + + for room_id, notif_count in txn: + room_to_count[room_id] += notif_count + + return room_to_count + @cached(tree=True, max_entries=5000, iterable=True) async def get_unread_event_push_actions_by_room_for_user( self, @@ -480,6 +713,25 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas (ReceiptTypes.READ, ReceiptTypes.READ_PRIVATE), ) + # First ensure that the existing rows have an updated thread_id field. + if not self._event_push_backfill_thread_id_done: + txn.execute( + """ + UPDATE event_push_summary + SET thread_id = ? + WHERE room_id = ? AND user_id = ? AND thread_id is NULL + """, + (MAIN_TIMELINE, room_id, user_id), + ) + txn.execute( + """ + UPDATE event_push_actions + SET thread_id = ? + WHERE room_id = ? AND user_id = ? AND thread_id is NULL + """, + (MAIN_TIMELINE, room_id, user_id), + ) + # First we pull the counts from the summary table. # # We check that `last_receipt_stream_ordering` matches the stream ordering of the @@ -1295,6 +1547,25 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas (room_id, user_id, stream_ordering, *thread_args), ) + # First ensure that the existing rows have an updated thread_id field. + if not self._event_push_backfill_thread_id_done: + txn.execute( + """ + UPDATE event_push_summary + SET thread_id = ? + WHERE room_id = ? AND user_id = ? AND thread_id is NULL + """, + (MAIN_TIMELINE, room_id, user_id), + ) + txn.execute( + """ + UPDATE event_push_actions + SET thread_id = ? + WHERE room_id = ? AND user_id = ? AND thread_id is NULL + """, + (MAIN_TIMELINE, room_id, user_id), + ) + # Fetch the notification counts between the stream ordering of the # latest receipt and what was previously summarised. unread_counts = self._get_notif_unread_count_for_user_room( @@ -1429,6 +1700,19 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas rotate_to_stream_ordering: The new maximum event stream ordering to summarise. """ + # Ensure that any new actions have an updated thread_id. + if not self._event_push_backfill_thread_id_done: + txn.execute( + """ + UPDATE event_push_actions + SET thread_id = ? + WHERE ? < stream_ordering AND stream_ordering <= ? AND thread_id IS NULL + """, + (MAIN_TIMELINE, old_rotate_stream_ordering, rotate_to_stream_ordering), + ) + + # XXX Do we need to update summaries here too? + # Calculate the new counts that should be upserted into event_push_summary sql = """ SELECT user_id, room_id, thread_id, @@ -1491,6 +1775,20 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, StreamWorkerStore, SQLBas logger.info("Rotating notifications, handling %d rows", len(summaries)) + # Ensure that any updated threads have the proper thread_id. + if not self._event_push_backfill_thread_id_done: + txn.execute_batch( + """ + UPDATE event_push_summary + SET thread_id = ? + WHERE room_id = ? AND user_id = ? AND thread_id is NULL + """, + [ + (MAIN_TIMELINE, room_id, user_id) + for user_id, room_id, _ in summaries + ], + ) + self.db_pool.simple_upsert_many_txn( txn, table="event_push_summary", diff --git a/synapse/storage/databases/main/events.py b/synapse/storage/databases/main/events.py index 3e15827986..0f097a2927 100644 --- a/synapse/storage/databases/main/events.py +++ b/synapse/storage/databases/main/events.py @@ -35,7 +35,7 @@ import attr from prometheus_client import Counter import synapse.metrics -from synapse.api.constants import EventContentFields, EventTypes +from synapse.api.constants import EventContentFields, EventTypes, RelationTypes from synapse.api.errors import Codes, SynapseError from synapse.api.room_versions import RoomVersions from synapse.events import EventBase, relation_from_event @@ -355,9 +355,9 @@ class PersistEventsStore: txn: LoggingTransaction, *, events_and_contexts: List[Tuple[EventBase, EventContext]], - inhibit_local_membership_updates: bool = False, - state_delta_for_room: Optional[Dict[str, DeltaState]] = None, - new_forward_extremities: Optional[Dict[str, Set[str]]] = None, + inhibit_local_membership_updates: bool, + state_delta_for_room: Dict[str, DeltaState], + new_forward_extremities: Dict[str, Set[str]], ) -> None: """Insert some number of room events into the necessary database tables. @@ -384,9 +384,6 @@ class PersistEventsStore: PartialStateConflictError: if attempting to persist a partial state event in a room that has been un-partial stated. """ - state_delta_for_room = state_delta_for_room or {} - new_forward_extremities = new_forward_extremities or {} - all_events_and_contexts = events_and_contexts min_stream_order = events_and_contexts[0][0].internal_metadata.stream_ordering @@ -1282,9 +1279,10 @@ class PersistEventsStore: Pick the earliest non-outlier if there is one, else the earliest one. Args: - events_and_contexts (list[(EventBase, EventContext)]): + events_and_contexts: + Returns: - list[(EventBase, EventContext)]: filtered list + filtered list """ new_events_and_contexts: OrderedDict[ str, Tuple[EventBase, EventContext] @@ -1310,9 +1308,8 @@ class PersistEventsStore: """Update min_depth for each room Args: - txn (twisted.enterprise.adbapi.Connection): db connection - events_and_contexts (list[(EventBase, EventContext)]): events - we are persisting + txn: db connection + events_and_contexts: events we are persisting """ depth_updates: Dict[str, int] = {} for event, context in events_and_contexts: @@ -1583,13 +1580,11 @@ class PersistEventsStore: """Update all the miscellaneous tables for new events Args: - txn (twisted.enterprise.adbapi.Connection): db connection - events_and_contexts (list[(EventBase, EventContext)]): events - we are persisting - all_events_and_contexts (list[(EventBase, EventContext)]): all - events that we were going to persist. This includes events - we've already persisted, etc, that wouldn't appear in - events_and_context. + txn: db connection + events_and_contexts: events we are persisting + all_events_and_contexts: all events that we were going to persist. + This includes events we've already persisted, etc, that wouldn't + appear in events_and_context. inhibit_local_membership_updates: Stop the local_current_membership from being updated by these events. This should be set to True for backfilled events because backfilled events in the past do @@ -1616,7 +1611,7 @@ class PersistEventsStore: ) # Remove from relations table. - self._handle_redact_relations(txn, event.redacts) + self._handle_redact_relations(txn, event.room_id, event.redacts) # Update the event_forward_extremities, event_backward_extremities and # event_edges tables. @@ -1866,6 +1861,34 @@ class PersistEventsStore: }, ) + if relation.rel_type == RelationTypes.THREAD: + # Upsert into the threads table, but only overwrite the value if the + # new event is of a later topological order OR if the topological + # ordering is equal, but the stream ordering is later. + sql = """ + INSERT INTO threads (room_id, thread_id, latest_event_id, topological_ordering, stream_ordering) + VALUES (?, ?, ?, ?, ?) + ON CONFLICT (room_id, thread_id) + DO UPDATE SET + latest_event_id = excluded.latest_event_id, + topological_ordering = excluded.topological_ordering, + stream_ordering = excluded.stream_ordering + WHERE + threads.topological_ordering <= excluded.topological_ordering AND + threads.stream_ordering < excluded.stream_ordering + """ + + txn.execute( + sql, + ( + event.room_id, + relation.parent_id, + event.event_id, + event.depth, + event.internal_metadata.stream_ordering, + ), + ) + def _handle_insertion_event( self, txn: LoggingTransaction, event: EventBase ) -> None: @@ -1989,35 +2012,52 @@ class PersistEventsStore: txn.execute(sql, (batch_id,)) def _handle_redact_relations( - self, txn: LoggingTransaction, redacted_event_id: str + self, txn: LoggingTransaction, room_id: str, redacted_event_id: str ) -> None: """Handles receiving a redaction and checking whether the redacted event has any relations which must be removed from the database. Args: txn + room_id: The room ID of the event that was redacted. redacted_event_id: The event that was redacted. """ - # Fetch the current relation of the event being redacted. - redacted_relates_to = self.db_pool.simple_select_one_onecol_txn( + # Fetch the relation of the event being redacted. + row = self.db_pool.simple_select_one_txn( txn, table="event_relations", keyvalues={"event_id": redacted_event_id}, - retcol="relates_to_id", + retcols=("relates_to_id", "relation_type"), allow_none=True, ) + # Nothing to do if no relation is found. + if row is None: + return + + redacted_relates_to = row["relates_to_id"] + rel_type = row["relation_type"] + self.db_pool.simple_delete_txn( + txn, table="event_relations", keyvalues={"event_id": redacted_event_id} + ) + # Any relation information for the related event must be cleared. - if redacted_relates_to is not None: + self.store._invalidate_cache_and_stream( + txn, self.store.get_relations_for_event, (redacted_relates_to,) + ) + if rel_type == RelationTypes.ANNOTATION: self.store._invalidate_cache_and_stream( - txn, self.store.get_relations_for_event, (redacted_relates_to,) + txn, self.store.get_aggregation_groups_for_event, (redacted_relates_to,) ) + if rel_type == RelationTypes.REFERENCE: self.store._invalidate_cache_and_stream( - txn, self.store.get_aggregation_groups_for_event, (redacted_relates_to,) + txn, self.store.get_references_for_event, (redacted_relates_to,) ) + if rel_type == RelationTypes.REPLACE: self.store._invalidate_cache_and_stream( txn, self.store.get_applicable_edit, (redacted_relates_to,) ) + if rel_type == RelationTypes.THREAD: self.store._invalidate_cache_and_stream( txn, self.store.get_thread_summary, (redacted_relates_to,) ) @@ -2025,14 +2065,41 @@ class PersistEventsStore: txn, self.store.get_thread_participated, (redacted_relates_to,) ) self.store._invalidate_cache_and_stream( - txn, - self.store.get_mutual_event_relations_for_rel_type, - (redacted_relates_to,), + txn, self.store.get_threads, (room_id,) ) - self.db_pool.simple_delete_txn( - txn, table="event_relations", keyvalues={"event_id": redacted_event_id} - ) + # Find the new latest event in the thread. + sql = """ + SELECT event_id, topological_ordering, stream_ordering + FROM event_relations + INNER JOIN events USING (event_id) + WHERE relates_to_id = ? AND relation_type = ? + ORDER BY topological_ordering DESC, stream_ordering DESC + LIMIT 1 + """ + txn.execute(sql, (redacted_relates_to, RelationTypes.THREAD)) + + # If a latest event is found, update the threads table, this might + # be the same current latest event (if an earlier event in the thread + # was redacted). + latest_event_row = txn.fetchone() + if latest_event_row: + self.db_pool.simple_upsert_txn( + txn, + table="threads", + keyvalues={"room_id": room_id, "thread_id": redacted_relates_to}, + values={ + "latest_event_id": latest_event_row[0], + "topological_ordering": latest_event_row[1], + "stream_ordering": latest_event_row[2], + }, + ) + + # Otherwise, delete the thread: it no longer exists. + else: + self.db_pool.simple_delete_one_txn( + txn, table="threads", keyvalues={"thread_id": redacted_relates_to} + ) def _store_room_topic_txn(self, txn: LoggingTransaction, event: EventBase) -> None: if isinstance(event.content.get("topic"), str): diff --git a/synapse/storage/databases/main/events_bg_updates.py b/synapse/storage/databases/main/events_bg_updates.py index 6e8aeed7b4..9e31798ab1 100644 --- a/synapse/storage/databases/main/events_bg_updates.py +++ b/synapse/storage/databases/main/events_bg_updates.py @@ -1435,16 +1435,16 @@ class EventsBackgroundUpdatesStore(SQLBaseStore): ), ) - endpoint = None row = txn.fetchone() if row: endpoint = row[0] + else: + # if the query didn't return a row, we must be almost done. We just + # need to go up to the recorded max_stream_ordering. + endpoint = max_stream_ordering_inclusive - where_clause = "stream_ordering > ?" - args = [min_stream_ordering_exclusive] - if endpoint: - where_clause += " AND stream_ordering <= ?" - args.append(endpoint) + where_clause = "stream_ordering > ? AND stream_ordering <= ?" + args = [min_stream_ordering_exclusive, endpoint] # now do the updates. txn.execute( @@ -1458,13 +1458,13 @@ class EventsBackgroundUpdatesStore(SQLBaseStore): ) logger.info( - "populated new `events` columns up to %s/%i: updated %i rows", + "populated new `events` columns up to %i/%i: updated %i rows", endpoint, max_stream_ordering_inclusive, txn.rowcount, ) - if endpoint is None: + if endpoint >= max_stream_ordering_inclusive: # we're done return True diff --git a/synapse/storage/databases/main/events_worker.py b/synapse/storage/databases/main/events_worker.py index 7cdc9fe98f..318fd7dc71 100644 --- a/synapse/storage/databases/main/events_worker.py +++ b/synapse/storage/databases/main/events_worker.py @@ -16,11 +16,11 @@ import logging import threading import weakref from enum import Enum, auto +from itertools import chain from typing import ( TYPE_CHECKING, Any, Collection, - Container, Dict, Iterable, List, @@ -59,7 +59,6 @@ from synapse.metrics.background_process_metrics import ( run_as_background_process, wrap_as_background_process, ) -from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker from synapse.replication.tcp.streams import BackfillStream from synapse.replication.tcp.streams.events import EventsStream from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause @@ -77,10 +76,12 @@ from synapse.storage.util.id_generators import ( ) from synapse.storage.util.sequence import build_sequence_generator from synapse.types import JsonDict, get_domain_from_id +from synapse.types.state import StateFilter from synapse.util import unwrapFirstError from synapse.util.async_helpers import ObservableDeferred, delay_cancellation from synapse.util.caches.descriptors import cached, cachedList from synapse.util.caches.lrucache import AsyncLruCache +from synapse.util.caches.stream_change_cache import StreamChangeCache from synapse.util.cancellation import cancellable from synapse.util.iterutils import batch_iter from synapse.util.metrics import Measure @@ -212,26 +213,35 @@ class EventsWorkerStore(SQLBaseStore): # `StreamIdGenerator`, otherwise we use `SlavedIdTracker` which gets # updated over replication. (Multiple writers are not supported for # SQLite). - if hs.get_instance_name() in hs.config.worker.writers.events: - self._stream_id_gen = StreamIdGenerator( - db_conn, - "events", - "stream_ordering", - ) - self._backfill_id_gen = StreamIdGenerator( - db_conn, - "events", - "stream_ordering", - step=-1, - extra_tables=[("ex_outlier_stream", "event_stream_ordering")], - ) - else: - self._stream_id_gen = SlavedIdTracker( - db_conn, "events", "stream_ordering" - ) - self._backfill_id_gen = SlavedIdTracker( - db_conn, "events", "stream_ordering", step=-1 - ) + self._stream_id_gen = StreamIdGenerator( + db_conn, + "events", + "stream_ordering", + is_writer=hs.get_instance_name() in hs.config.worker.writers.events, + ) + self._backfill_id_gen = StreamIdGenerator( + db_conn, + "events", + "stream_ordering", + step=-1, + extra_tables=[("ex_outlier_stream", "event_stream_ordering")], + is_writer=hs.get_instance_name() in hs.config.worker.writers.events, + ) + + events_max = self._stream_id_gen.get_current_token() + curr_state_delta_prefill, min_curr_state_delta_id = self.db_pool.get_cache_dict( + db_conn, + "current_state_delta_stream", + entity_column="room_id", + stream_column="stream_id", + max_value=events_max, # As we share the stream id with events token + limit=1000, + ) + self._curr_state_delta_stream_cache: StreamChangeCache = StreamChangeCache( + "_curr_state_delta_stream_cache", + min_curr_state_delta_id, + prefilled_cache=curr_state_delta_prefill, + ) if hs.config.worker.run_background_tasks: # We periodically clean out old transaction ID mappings @@ -374,7 +384,7 @@ class EventsWorkerStore(SQLBaseStore): If there is a mismatch, behave as per allow_none. Returns: - The event, or None if the event was not found. + The event, or None if the event was not found and allow_none is `True`. """ if not isinstance(event_id, str): raise TypeError("Invalid event event_id %r" % (event_id,)) @@ -474,7 +484,7 @@ class EventsWorkerStore(SQLBaseStore): return [] # there may be duplicates so we cast the list to a set - event_entry_map = await self._get_events_from_cache_or_db( + event_entry_map = await self.get_unredacted_events_from_cache_or_db( set(event_ids), allow_rejected=allow_rejected ) @@ -509,7 +519,9 @@ class EventsWorkerStore(SQLBaseStore): continue redacted_event_id = entry.event.redacts - event_map = await self._get_events_from_cache_or_db([redacted_event_id]) + event_map = await self.get_unredacted_events_from_cache_or_db( + [redacted_event_id] + ) original_event_entry = event_map.get(redacted_event_id) if not original_event_entry: # we don't have the redacted event (or it was rejected). @@ -588,11 +600,16 @@ class EventsWorkerStore(SQLBaseStore): return events @cancellable - async def _get_events_from_cache_or_db( - self, event_ids: Iterable[str], allow_rejected: bool = False + async def get_unredacted_events_from_cache_or_db( + self, + event_ids: Iterable[str], + allow_rejected: bool = False, ) -> Dict[str, EventCacheEntry]: """Fetch a bunch of events from the cache or the database. + Note that the events pulled by this function will not have any redactions + applied, and no guarantee is made about the ordering of the events returned. + If events are pulled from the database, they will be cached for future lookups. Unknown events are omitted from the response. @@ -863,7 +880,7 @@ class EventsWorkerStore(SQLBaseStore): async def get_stripped_room_state_from_event_context( self, context: EventContext, - state_types_to_include: Container[str], + state_keys_to_include: StateFilter, membership_user_id: Optional[str] = None, ) -> List[JsonDict]: """ @@ -876,7 +893,7 @@ class EventsWorkerStore(SQLBaseStore): Args: context: The event context to retrieve state of the room from. - state_types_to_include: The type of state events to include. + state_keys_to_include: The state events to include, for each event type. membership_user_id: An optional user ID to include the stripped membership state events of. This is useful when generating the stripped state of a room for invites. We want to send membership events of the inviter, so that the @@ -885,21 +902,25 @@ class EventsWorkerStore(SQLBaseStore): Returns: A list of dictionaries, each representing a stripped state event from the room. """ - current_state_ids = await context.get_current_state_ids() + if membership_user_id: + types = chain( + state_keys_to_include.to_types(), + [(EventTypes.Member, membership_user_id)], + ) + filter = StateFilter.from_types(types) + else: + filter = state_keys_to_include + selected_state_ids = await context.get_current_state_ids(filter) # We know this event is not an outlier, so this must be # non-None. - assert current_state_ids is not None - - # The state to include - state_to_include_ids = [ - e_id - for k, e_id in current_state_ids.items() - if k[0] in state_types_to_include - or (membership_user_id and k == (EventTypes.Member, membership_user_id)) - ] + assert selected_state_ids is not None + + # Confusingly, get_current_state_events may return events that are discarded by + # the filter, if they're in context._state_delta_due_to_event. Strip these away. + selected_state_ids = filter.filter_state(selected_state_ids) - state_to_include = await self.get_events(state_to_include_ids) + state_to_include = await self.get_events(selected_state_ids.values()) return [ { @@ -1495,21 +1516,15 @@ class EventsWorkerStore(SQLBaseStore): Returns: a dict {event_id -> bool} """ - # if the event cache contains the event, obviously we've seen it. - - cache_results = { - event_id - for event_id in event_ids - if await self._get_event_cache.contains((event_id,)) - } - results = dict.fromkeys(cache_results, True) - remaining = [ - event_id for event_id in event_ids if event_id not in cache_results - ] - if not remaining: - return results + # TODO: We used to query the _get_event_cache here as a fast-path before + # hitting the database. For if an event were in the cache, we've presumably + # seen it before. + # + # But this is currently an invalid assumption due to the _get_event_cache + # not being invalidated when purging events from a room. The optimisation can + # be re-added after https://github.com/matrix-org/synapse/issues/13476 - def have_seen_events_txn(txn: LoggingTransaction) -> None: + def have_seen_events_txn(txn: LoggingTransaction) -> Dict[str, bool]: # we deliberately do *not* query the database for room_id, to make the # query an index-only lookup on `events_event_id_key`. # @@ -1517,16 +1532,17 @@ class EventsWorkerStore(SQLBaseStore): sql = "SELECT event_id FROM events AS e WHERE " clause, args = make_in_list_sql_clause( - txn.database_engine, "e.event_id", remaining + txn.database_engine, "e.event_id", event_ids ) txn.execute(sql + clause, args) found_events = {eid for eid, in txn} # ... and then we can update the results for each key - results.update({eid: (eid in found_events) for eid in remaining}) + return {eid: (eid in found_events) for eid in event_ids} - await self.db_pool.runInteraction("have_seen_events", have_seen_events_txn) - return results + return await self.db_pool.runInteraction( + "have_seen_events", have_seen_events_txn + ) @cached(max_entries=100000, tree=True) async def have_seen_event(self, room_id: str, event_id: str) -> bool: @@ -1571,7 +1587,7 @@ class EventsWorkerStore(SQLBaseStore): room_id: The room ID to query. Returns: - dict[str:float] of complexity version to complexity. + Map of complexity version to complexity. """ state_events = await self.get_current_state_event_counts(room_id) @@ -1969,12 +1985,17 @@ class EventsWorkerStore(SQLBaseStore): Args: room_id: room where the event lives - event_id: event to check + event: event to check (can't be an `outlier`) Returns: Boolean indicating whether it's an extremity """ + assert not event.internal_metadata.is_outlier(), ( + "is_event_next_to_backward_gap(...) can't be used with `outlier` events. " + "This function relies on `event_backward_extremities` which won't be filled in for `outliers`." + ) + def is_event_next_to_backward_gap_txn(txn: LoggingTransaction) -> bool: # If the event in question has any of its prev_events listed as a # backward extremity, it's next to a gap. @@ -2024,12 +2045,17 @@ class EventsWorkerStore(SQLBaseStore): Args: room_id: room where the event lives - event_id: event to check + event: event to check (can't be an `outlier`) Returns: Boolean indicating whether it's an extremity """ + assert not event.internal_metadata.is_outlier(), ( + "is_event_next_to_forward_gap(...) can't be used with `outlier` events. " + "This function relies on `event_edges` and `event_forward_extremities` which won't be filled in for `outliers`." + ) + def is_event_next_to_gap_txn(txn: LoggingTransaction) -> bool: # If the event in question is a forward extremity, we will just # consider any potential forward gap as not a gap since it's one of @@ -2110,13 +2136,33 @@ class EventsWorkerStore(SQLBaseStore): The closest event_id otherwise None if we can't find any event in the given direction. """ + if direction == "b": + # Find closest event *before* a given timestamp. We use descending + # (which gives values largest to smallest) because we want the + # largest possible timestamp *before* the given timestamp. + comparison_operator = "<=" + order = "DESC" + else: + # Find closest event *after* a given timestamp. We use ascending + # (which gives values smallest to largest) because we want the + # closest possible timestamp *after* the given timestamp. + comparison_operator = ">=" + order = "ASC" - sql_template = """ + sql_template = f""" SELECT event_id FROM events LEFT JOIN rejections USING (event_id) WHERE - origin_server_ts %s ? - AND room_id = ? + room_id = ? + AND origin_server_ts {comparison_operator} ? + /** + * Make sure the event isn't an `outlier` because we have no way + * to later check whether it's next to a gap. `outliers` do not + * have entries in the `event_edges`, `event_forward_extremeties`, + * and `event_backward_extremities` tables to check against + * (used by `is_event_next_to_backward_gap` and `is_event_next_to_forward_gap`). + */ + AND NOT outlier /* Make sure event is not rejected */ AND rejections.event_id IS NULL /** @@ -2126,27 +2172,14 @@ class EventsWorkerStore(SQLBaseStore): * Finally, we can tie-break based on when it was received on the server * (`stream_ordering`). */ - ORDER BY origin_server_ts %s, depth %s, stream_ordering %s + ORDER BY origin_server_ts {order}, depth {order}, stream_ordering {order} LIMIT 1; """ def get_event_id_for_timestamp_txn(txn: LoggingTransaction) -> Optional[str]: - if direction == "b": - # Find closest event *before* a given timestamp. We use descending - # (which gives values largest to smallest) because we want the - # largest possible timestamp *before* the given timestamp. - comparison_operator = "<=" - order = "DESC" - else: - # Find closest event *after* a given timestamp. We use ascending - # (which gives values smallest to largest) because we want the - # closest possible timestamp *after* the given timestamp. - comparison_operator = ">=" - order = "ASC" - txn.execute( - sql_template % (comparison_operator, order, order, order), - (timestamp, room_id), + sql_template, + (room_id, timestamp), ) row = txn.fetchone() if row: @@ -2200,7 +2233,15 @@ class EventsWorkerStore(SQLBaseStore): return result is not None async def get_partial_state_events_batch(self, room_id: str) -> List[str]: - """Get a list of events in the given room that have partial state""" + """ + Get a list of events in the given room that: + - have partial state; and + - are ready to be resynced (because they have no prev_events that are + partial-stated) + + See the docstring on `_get_partial_state_events_batch_txn` for more + information. + """ return await self.db_pool.runInteraction( "get_partial_state_events_batch", self._get_partial_state_events_batch_txn, diff --git a/synapse/storage/databases/main/filtering.py b/synapse/storage/databases/main/filtering.py index cb9ee08fa8..12f3b601f1 100644 --- a/synapse/storage/databases/main/filtering.py +++ b/synapse/storage/databases/main/filtering.py @@ -24,7 +24,7 @@ from synapse.types import JsonDict from synapse.util.caches.descriptors import cached -class FilteringStore(SQLBaseStore): +class FilteringWorkerStore(SQLBaseStore): @cached(num_args=2) async def get_user_filter( self, user_localpart: str, filter_id: Union[int, str] @@ -46,6 +46,8 @@ class FilteringStore(SQLBaseStore): return db_to_json(def_json) + +class FilteringStore(FilteringWorkerStore): async def add_user_filter(self, user_localpart: str, user_filter: JsonDict) -> int: def_json = encode_canonical_json(user_filter) diff --git a/synapse/storage/databases/main/monthly_active_users.py b/synapse/storage/databases/main/monthly_active_users.py index efd136a864..db9a24db5e 100644 --- a/synapse/storage/databases/main/monthly_active_users.py +++ b/synapse/storage/databases/main/monthly_active_users.py @@ -217,7 +217,7 @@ class MonthlyActiveUsersWorkerStore(RegistrationWorkerStore): def _reap_users(txn: LoggingTransaction, reserved_users: List[str]) -> None: """ Args: - reserved_users (tuple): reserved users to preserve + reserved_users: reserved users to preserve """ thirty_days_ago = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24 * 30) @@ -370,8 +370,8 @@ class MonthlyActiveUsersWorkerStore(RegistrationWorkerStore): should not appear in the MAU stats). Args: - txn (cursor): - user_id (str): user to add/update + txn: + user_id: user to add/update """ assert ( self._update_on_this_worker @@ -401,7 +401,7 @@ class MonthlyActiveUsersWorkerStore(RegistrationWorkerStore): add the user to the monthly active tables Args: - user_id(str): the user_id to query + user_id: the user_id to query """ assert ( self._update_on_this_worker diff --git a/synapse/storage/databases/main/push_rule.py b/synapse/storage/databases/main/push_rule.py index 8295322b0e..d4c64c46ad 100644 --- a/synapse/storage/databases/main/push_rule.py +++ b/synapse/storage/databases/main/push_rule.py @@ -12,13 +12,13 @@ # 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 abc import logging from typing import ( TYPE_CHECKING, Any, Collection, Dict, + Iterable, List, Mapping, Optional, @@ -30,7 +30,7 @@ from typing import ( from synapse.api.errors import StoreError from synapse.config.homeserver import ExperimentalConfig -from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker +from synapse.replication.tcp.streams import PushRulesStream from synapse.storage._base import SQLBaseStore from synapse.storage.database import ( DatabasePool, @@ -84,14 +84,15 @@ def _load_rules( push_rules = PushRules(ruleslist) filtered_rules = FilteredPushRules( - push_rules, enabled_map, msc3772_enabled=experimental_config.msc3772_enabled + push_rules, + enabled_map, + msc3664_enabled=experimental_config.msc3664_enabled, + msc1767_enabled=experimental_config.msc1767_enabled, ) return filtered_rules -# The ABCMeta metaclass ensures that it cannot be instantiated without -# the abstract methods being implemented. class PushRulesWorkerStore( ApplicationServiceWorkerStore, PusherWorkerStore, @@ -99,7 +100,6 @@ class PushRulesWorkerStore( ReceiptsWorkerStore, EventsWorkerStore, SQLBaseStore, - metaclass=abc.ABCMeta, ): """This is an abstract base class where subclasses must implement `get_max_push_rules_stream_id` which can be called in the initializer. @@ -113,14 +113,14 @@ class PushRulesWorkerStore( ): super().__init__(database, db_conn, hs) - if hs.config.worker.worker_app is None: - self._push_rules_stream_id_gen: AbstractStreamIdTracker = StreamIdGenerator( - db_conn, "push_rules_stream", "stream_id" - ) - else: - self._push_rules_stream_id_gen = SlavedIdTracker( - db_conn, "push_rules_stream", "stream_id" - ) + # In the worker store this is an ID tracker which we overwrite in the non-worker + # class below that is used on the main process. + self._push_rules_stream_id_gen: AbstractStreamIdTracker = StreamIdGenerator( + db_conn, + "push_rules_stream", + "stream_id", + is_writer=hs.config.worker.worker_app is None, + ) push_rules_prefill, push_rules_id = self.db_pool.get_cache_dict( db_conn, @@ -136,14 +136,23 @@ class PushRulesWorkerStore( prefilled_cache=push_rules_prefill, ) - @abc.abstractmethod def get_max_push_rules_stream_id(self) -> int: """Get the position of the push rules stream. Returns: int """ - raise NotImplementedError() + return self._push_rules_stream_id_gen.get_current_token() + + def process_replication_rows( + self, stream_name: str, instance_name: str, token: int, rows: Iterable[Any] + ) -> None: + if stream_name == PushRulesStream.NAME: + self._push_rules_stream_id_gen.advance(instance_name, token) + for row in rows: + self.get_push_rules_for_user.invalidate((row.user_id,)) + self.push_rules_stream_cache.entity_has_changed(row.user_id, token) + return super().process_replication_rows(stream_name, instance_name, token, rows) @cached(max_entries=5000) async def get_push_rules_for_user(self, user_id: str) -> FilteredPushRules: diff --git a/synapse/storage/databases/main/pusher.py b/synapse/storage/databases/main/pusher.py index 01206950a9..40fd781a6a 100644 --- a/synapse/storage/databases/main/pusher.py +++ b/synapse/storage/databases/main/pusher.py @@ -27,13 +27,18 @@ from typing import ( ) from synapse.push import PusherConfig, ThrottleParams +from synapse.replication.tcp.streams import PushersStream from synapse.storage._base import SQLBaseStore, db_to_json from synapse.storage.database import ( DatabasePool, LoggingDatabaseConnection, LoggingTransaction, ) -from synapse.storage.util.id_generators import StreamIdGenerator +from synapse.storage.util.id_generators import ( + AbstractStreamIdGenerator, + AbstractStreamIdTracker, + StreamIdGenerator, +) from synapse.types import JsonDict from synapse.util import json_encoder from synapse.util.caches.descriptors import cached @@ -52,8 +57,15 @@ class PusherWorkerStore(SQLBaseStore): hs: "HomeServer", ): super().__init__(database, db_conn, hs) - self._pushers_id_gen = StreamIdGenerator( - db_conn, "pushers", "id", extra_tables=[("deleted_pushers", "stream_id")] + + # In the worker store this is an ID tracker which we overwrite in the non-worker + # class below that is used on the main process. + self._pushers_id_gen: AbstractStreamIdTracker = StreamIdGenerator( + db_conn, + "pushers", + "id", + extra_tables=[("deleted_pushers", "stream_id")], + is_writer=hs.config.worker.worker_app is None, ) self.db_pool.updates.register_background_update_handler( @@ -96,6 +108,16 @@ class PusherWorkerStore(SQLBaseStore): yield PusherConfig(**r) + def get_pushers_stream_token(self) -> int: + return self._pushers_id_gen.get_current_token() + + def process_replication_rows( + self, stream_name: str, instance_name: str, token: int, rows: Iterable[Any] + ) -> None: + if stream_name == PushersStream.NAME: + self._pushers_id_gen.advance(instance_name, token) + return super().process_replication_rows(stream_name, instance_name, token, rows) + async def get_pushers_by_app_id_and_pushkey( self, app_id: str, pushkey: str ) -> Iterator[PusherConfig]: @@ -303,14 +325,11 @@ class PusherWorkerStore(SQLBaseStore): async def set_throttle_params( self, pusher_id: str, room_id: str, params: ThrottleParams ) -> None: - # no need to lock because `pusher_throttle` has a primary key on - # (pusher, room_id) so simple_upsert will retry await self.db_pool.simple_upsert( "pusher_throttle", {"pusher": pusher_id, "room_id": room_id}, {"last_sent_ts": params.last_sent_ts, "throttle_ms": params.throttle_ms}, desc="set_throttle_params", - lock=False, ) async def _remove_deactivated_pushers(self, progress: dict, batch_size: int) -> int: @@ -545,8 +564,9 @@ class PusherBackgroundUpdatesStore(SQLBaseStore): class PusherStore(PusherWorkerStore, PusherBackgroundUpdatesStore): - def get_pushers_stream_token(self) -> int: - return self._pushers_id_gen.get_current_token() + # Because we have write access, this will be a StreamIdGenerator + # (see PusherWorkerStore.__init__) + _pushers_id_gen: AbstractStreamIdGenerator async def add_pusher( self, @@ -566,8 +586,6 @@ class PusherStore(PusherWorkerStore, PusherBackgroundUpdatesStore): device_id: Optional[str] = None, ) -> None: async with self._pushers_id_gen.get_next() as stream_id: - # no need to lock because `pushers` has a unique key on - # (app_id, pushkey, user_name) so simple_upsert will retry await self.db_pool.simple_upsert( table="pushers", keyvalues={"app_id": app_id, "pushkey": pushkey, "user_name": user_id}, @@ -586,7 +604,6 @@ class PusherStore(PusherWorkerStore, PusherBackgroundUpdatesStore): "device_id": device_id, }, desc="add_pusher", - lock=False, ) user_has_pusher = self.get_if_user_has_pusher.cache.get_immediate( diff --git a/synapse/storage/databases/main/receipts.py b/synapse/storage/databases/main/receipts.py index 246f78ac1f..e06725f69c 100644 --- a/synapse/storage/databases/main/receipts.py +++ b/synapse/storage/databases/main/receipts.py @@ -27,7 +27,6 @@ from typing import ( ) from synapse.api.constants import EduTypes -from synapse.replication.slave.storage._slaved_id_tracker import SlavedIdTracker from synapse.replication.tcp.streams import ReceiptsStream from synapse.storage._base import SQLBaseStore, db_to_json, make_in_list_sql_clause from synapse.storage.database import ( @@ -61,6 +60,9 @@ class ReceiptsWorkerStore(SQLBaseStore): hs: "HomeServer", ): self._instance_name = hs.get_instance_name() + + # In the worker store this is an ID tracker which we overwrite in the non-worker + # class below that is used on the main process. self._receipts_id_gen: AbstractStreamIdTracker if isinstance(database.engine, PostgresEngine): @@ -87,14 +89,12 @@ class ReceiptsWorkerStore(SQLBaseStore): # `StreamIdGenerator`, otherwise we use `SlavedIdTracker` which gets # updated over replication. (Multiple writers are not supported for # SQLite). - if hs.get_instance_name() in hs.config.worker.writers.receipts: - self._receipts_id_gen = StreamIdGenerator( - db_conn, "receipts_linearized", "stream_id" - ) - else: - self._receipts_id_gen = SlavedIdTracker( - db_conn, "receipts_linearized", "stream_id" - ) + self._receipts_id_gen = StreamIdGenerator( + db_conn, + "receipts_linearized", + "stream_id", + is_writer=hs.get_instance_name() in hs.config.worker.writers.receipts, + ) super().__init__(database, db_conn, hs) @@ -113,24 +113,6 @@ class ReceiptsWorkerStore(SQLBaseStore): prefilled_cache=receipts_stream_prefill, ) - self.db_pool.updates.register_background_index_update( - "receipts_linearized_unique_index", - index_name="receipts_linearized_unique_index", - table="receipts_linearized", - columns=["room_id", "receipt_type", "user_id"], - where_clause="thread_id IS NULL", - unique=True, - ) - - self.db_pool.updates.register_background_index_update( - "receipts_graph_unique_index", - index_name="receipts_graph_unique_index", - table="receipts_graph", - columns=["room_id", "receipt_type", "user_id"], - where_clause="thread_id IS NULL", - unique=True, - ) - def get_max_receipt_stream_id(self) -> int: """Get the current max stream ID for receipts stream""" return self._receipts_id_gen.get_current_token() @@ -418,6 +400,8 @@ class ReceiptsWorkerStore(SQLBaseStore): receipt_type = event_entry.setdefault(row["receipt_type"], {}) receipt_type[row["user_id"]] = db_to_json(row["data"]) + if row["thread_id"]: + receipt_type[row["user_id"]]["thread_id"] = row["thread_id"] results = { room_id: [results[room_id]] if room_id in results else [] @@ -700,9 +684,6 @@ class ReceiptsWorkerStore(SQLBaseStore): "data": json_encoder.encode(data), }, where_clause=where_clause, - # receipts_linearized has a unique constraint on - # (user_id, room_id, receipt_type), so no need to lock - lock=False, ) return rx_ts @@ -860,14 +841,13 @@ class ReceiptsWorkerStore(SQLBaseStore): "data": json_encoder.encode(data), }, where_clause=where_clause, - # receipts_graph has a unique constraint on - # (user_id, room_id, receipt_type), so no need to lock - lock=False, ) class ReceiptsBackgroundUpdateStore(SQLBaseStore): POPULATE_RECEIPT_EVENT_STREAM_ORDERING = "populate_event_stream_ordering" + RECEIPTS_LINEARIZED_UNIQUE_INDEX_UPDATE_NAME = "receipts_linearized_unique_index" + RECEIPTS_GRAPH_UNIQUE_INDEX_UPDATE_NAME = "receipts_graph_unique_index" def __init__( self, @@ -881,6 +861,14 @@ class ReceiptsBackgroundUpdateStore(SQLBaseStore): self.POPULATE_RECEIPT_EVENT_STREAM_ORDERING, self._populate_receipt_event_stream_ordering, ) + self.db_pool.updates.register_background_update_handler( + self.RECEIPTS_LINEARIZED_UNIQUE_INDEX_UPDATE_NAME, + self._background_receipts_linearized_unique_index, + ) + self.db_pool.updates.register_background_update_handler( + self.RECEIPTS_GRAPH_UNIQUE_INDEX_UPDATE_NAME, + self._background_receipts_graph_unique_index, + ) async def _populate_receipt_event_stream_ordering( self, progress: JsonDict, batch_size: int @@ -936,6 +924,116 @@ class ReceiptsBackgroundUpdateStore(SQLBaseStore): return batch_size + async def _background_receipts_linearized_unique_index( + self, progress: dict, batch_size: int + ) -> int: + """Removes duplicate receipts and adds a unique index on + `(room_id, receipt_type, user_id)` to `receipts_linearized`, for non-thread + receipts.""" + + def _remote_duplicate_receipts_txn(txn: LoggingTransaction) -> None: + # Identify any duplicate receipts arising from + # https://github.com/matrix-org/synapse/issues/14406. + # We expect the following query to use the per-thread receipt index and take + # less than a minute. + sql = """ + SELECT MAX(stream_id), room_id, receipt_type, user_id + FROM receipts_linearized + WHERE thread_id IS NULL + GROUP BY room_id, receipt_type, user_id + HAVING COUNT(*) > 1 + """ + txn.execute(sql) + duplicate_keys = cast(List[Tuple[int, str, str, str]], list(txn)) + + # Then remove duplicate receipts, keeping the one with the highest + # `stream_id`. There should only be a single receipt with any given + # `stream_id`. + for max_stream_id, room_id, receipt_type, user_id in duplicate_keys: + sql = """ + DELETE FROM receipts_linearized + WHERE + room_id = ? AND + receipt_type = ? AND + user_id = ? AND + thread_id IS NULL AND + stream_id < ? + """ + txn.execute(sql, (room_id, receipt_type, user_id, max_stream_id)) + + await self.db_pool.runInteraction( + self.RECEIPTS_LINEARIZED_UNIQUE_INDEX_UPDATE_NAME, + _remote_duplicate_receipts_txn, + ) + + await self.db_pool.updates.create_index_in_background( + index_name="receipts_linearized_unique_index", + table="receipts_linearized", + columns=["room_id", "receipt_type", "user_id"], + where_clause="thread_id IS NULL", + unique=True, + ) + + await self.db_pool.updates._end_background_update( + self.RECEIPTS_LINEARIZED_UNIQUE_INDEX_UPDATE_NAME + ) + + return 1 + + async def _background_receipts_graph_unique_index( + self, progress: dict, batch_size: int + ) -> int: + """Removes duplicate receipts and adds a unique index on + `(room_id, receipt_type, user_id)` to `receipts_graph`, for non-thread + receipts.""" + + def _remote_duplicate_receipts_txn(txn: LoggingTransaction) -> None: + # Identify any duplicate receipts arising from + # https://github.com/matrix-org/synapse/issues/14406. + # We expect the following query to use the per-thread receipt index and take + # less than a minute. + sql = """ + SELECT room_id, receipt_type, user_id FROM receipts_graph + WHERE thread_id IS NULL + GROUP BY room_id, receipt_type, user_id + HAVING COUNT(*) > 1 + """ + txn.execute(sql) + duplicate_keys = cast(List[Tuple[str, str, str]], list(txn)) + + # Then remove all duplicate receipts. + # We could be clever and try to keep the latest receipt out of every set of + # duplicates, but it's far simpler to remove them all. + for room_id, receipt_type, user_id in duplicate_keys: + sql = """ + DELETE FROM receipts_graph + WHERE + room_id = ? AND + receipt_type = ? AND + user_id = ? AND + thread_id IS NULL + """ + txn.execute(sql, (room_id, receipt_type, user_id)) + + await self.db_pool.runInteraction( + self.RECEIPTS_GRAPH_UNIQUE_INDEX_UPDATE_NAME, + _remote_duplicate_receipts_txn, + ) + + await self.db_pool.updates.create_index_in_background( + index_name="receipts_graph_unique_index", + table="receipts_graph", + columns=["room_id", "receipt_type", "user_id"], + where_clause="thread_id IS NULL", + unique=True, + ) + + await self.db_pool.updates._end_background_update( + self.RECEIPTS_GRAPH_UNIQUE_INDEX_UPDATE_NAME + ) + + return 1 + class ReceiptsStore(ReceiptsWorkerStore, ReceiptsBackgroundUpdateStore): pass diff --git a/synapse/storage/databases/main/registration.py b/synapse/storage/databases/main/registration.py index 2996d6bb4d..31f0f2bd3d 100644 --- a/synapse/storage/databases/main/registration.py +++ b/synapse/storage/databases/main/registration.py @@ -21,7 +21,13 @@ from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union, cast import attr from synapse.api.constants import UserTypes -from synapse.api.errors import Codes, StoreError, SynapseError, ThreepidValidationError +from synapse.api.errors import ( + Codes, + NotFoundError, + StoreError, + SynapseError, + ThreepidValidationError, +) from synapse.config.homeserver import HomeServerConfig from synapse.metrics.background_process_metrics import wrap_as_background_process from synapse.storage.database import ( @@ -50,6 +56,14 @@ class ExternalIDReuseException(Exception): because this external id is given to an other user.""" +class LoginTokenExpired(Exception): + """Exception if the login token sent expired""" + + +class LoginTokenReused(Exception): + """Exception if the login token sent was already used""" + + @attr.s(frozen=True, slots=True, auto_attribs=True) class TokenLookupResult: """Result of looking up an access token. @@ -115,6 +129,20 @@ class RefreshTokenLookupResult: If None, the session can be refreshed indefinitely.""" +@attr.s(auto_attribs=True, frozen=True, slots=True) +class LoginTokenLookupResult: + """Result of looking up a login token.""" + + user_id: str + """The user this token belongs to.""" + + auth_provider_id: Optional[str] + """The SSO Identity Provider that the user authenticated with, to get this token.""" + + auth_provider_session_id: Optional[str] + """The session ID advertised by the SSO Identity Provider.""" + + class RegistrationWorkerStore(CacheInvalidationWorkerStore): def __init__( self, @@ -925,7 +953,7 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): """Returns user id from threepid Args: - txn (cursor): + txn: medium: threepid medium e.g. email address: threepid address e.g. me@example.com @@ -1255,8 +1283,8 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): """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 + user_id: User ID to set an expiration date for. + use_delta: 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. @@ -1789,6 +1817,130 @@ class RegistrationWorkerStore(CacheInvalidationWorkerStore): "replace_refresh_token", _replace_refresh_token_txn ) + async def add_login_token_to_user( + self, + user_id: str, + token: str, + expiry_ts: int, + auth_provider_id: Optional[str], + auth_provider_session_id: Optional[str], + ) -> None: + """Adds a short-term login token for the given user. + + Args: + user_id: The user ID. + token: The new login token to add. + expiry_ts (milliseconds since the epoch): Time after which the login token + cannot be used. + auth_provider_id: The SSO Identity Provider that the user authenticated with + to get this token, if any + auth_provider_session_id: The session ID advertised by the SSO Identity + Provider, if any. + """ + await self.db_pool.simple_insert( + "login_tokens", + { + "token": token, + "user_id": user_id, + "expiry_ts": expiry_ts, + "auth_provider_id": auth_provider_id, + "auth_provider_session_id": auth_provider_session_id, + }, + desc="add_login_token_to_user", + ) + + def _consume_login_token( + self, + txn: LoggingTransaction, + token: str, + ts: int, + ) -> LoginTokenLookupResult: + values = self.db_pool.simple_select_one_txn( + txn, + "login_tokens", + keyvalues={"token": token}, + retcols=( + "user_id", + "expiry_ts", + "used_ts", + "auth_provider_id", + "auth_provider_session_id", + ), + allow_none=True, + ) + + if values is None: + raise NotFoundError() + + self.db_pool.simple_update_one_txn( + txn, + "login_tokens", + keyvalues={"token": token}, + updatevalues={"used_ts": ts}, + ) + user_id = values["user_id"] + expiry_ts = values["expiry_ts"] + used_ts = values["used_ts"] + auth_provider_id = values["auth_provider_id"] + auth_provider_session_id = values["auth_provider_session_id"] + + # Token was already used + if used_ts is not None: + raise LoginTokenReused() + + # Token expired + if ts > int(expiry_ts): + raise LoginTokenExpired() + + return LoginTokenLookupResult( + user_id=user_id, + auth_provider_id=auth_provider_id, + auth_provider_session_id=auth_provider_session_id, + ) + + async def consume_login_token(self, token: str) -> LoginTokenLookupResult: + """Lookup a login token and consume it. + + Args: + token: The login token. + + Returns: + The data stored with that token, including the `user_id`. Returns `None` if + the token does not exist or if it expired. + + Raises: + NotFound if the login token was not found in database + LoginTokenExpired if the login token expired + LoginTokenReused if the login token was already used + """ + return await self.db_pool.runInteraction( + "consume_login_token", + self._consume_login_token, + token, + self._clock.time_msec(), + ) + + async def invalidate_login_tokens_by_session_id( + self, auth_provider_id: str, auth_provider_session_id: str + ) -> None: + """Invalidate login tokens with the given IdP session ID. + + Args: + auth_provider_id: The SSO Identity Provider that the user authenticated with + to get this token + auth_provider_session_id: The session ID advertised by the SSO Identity + Provider + """ + await self.db_pool.simple_update( + table="login_tokens", + keyvalues={ + "auth_provider_id": auth_provider_id, + "auth_provider_session_id": auth_provider_session_id, + }, + updatevalues={"used_ts": self._clock.time_msec()}, + desc="invalidate_login_tokens_by_session_id", + ) + @cached() async def is_guest(self, user_id: str) -> bool: res = await self.db_pool.simple_select_one_onecol( @@ -2019,6 +2171,12 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore): and hs.config.experimental.msc3866.require_approval_for_new_accounts ) + # Create a background job for removing expired login tokens + if hs.config.worker.run_background_tasks: + self._clock.looping_call( + self._delete_expired_login_tokens, THIRTY_MINUTES_IN_MS + ) + async def add_access_token_to_user( self, user_id: str, @@ -2617,6 +2775,23 @@ class RegistrationStore(StatsStore, RegistrationBackgroundUpdateStore): approved, ) + @wrap_as_background_process("delete_expired_login_tokens") + async def _delete_expired_login_tokens(self) -> None: + """Remove login tokens with expiry dates that have passed.""" + + def _delete_expired_login_tokens_txn(txn: LoggingTransaction, ts: int) -> None: + sql = "DELETE FROM login_tokens WHERE expiry_ts <= ?" + txn.execute(sql, (ts,)) + + # We keep the expired tokens for an extra 5 minutes so we can measure how many + # times a token is being used after its expiry + now = self._clock.time_msec() + await self.db_pool.runInteraction( + "delete_expired_login_tokens", + _delete_expired_login_tokens_txn, + now - (5 * 60 * 1000), + ) + def find_max_generated_user_id_localpart(cur: Cursor) -> int: """ diff --git a/synapse/storage/databases/main/relations.py b/synapse/storage/databases/main/relations.py index 154385b1e8..aea96e9d24 100644 --- a/synapse/storage/databases/main/relations.py +++ b/synapse/storage/databases/main/relations.py @@ -14,11 +14,13 @@ import logging from typing import ( + TYPE_CHECKING, Collection, Dict, FrozenSet, Iterable, List, + Mapping, Optional, Set, Tuple, @@ -28,19 +30,48 @@ from typing import ( import attr -from synapse.api.constants import RelationTypes +from synapse.api.constants import MAIN_TIMELINE, RelationTypes +from synapse.api.errors import SynapseError from synapse.events import EventBase from synapse.storage._base import SQLBaseStore -from synapse.storage.database import LoggingTransaction, make_in_list_sql_clause +from synapse.storage.database import ( + DatabasePool, + LoggingDatabaseConnection, + LoggingTransaction, + make_in_list_sql_clause, +) from synapse.storage.databases.main.stream import generate_pagination_where_clause from synapse.storage.engines import PostgresEngine from synapse.types import JsonDict, RoomStreamToken, StreamKeyType, StreamToken from synapse.util.caches.descriptors import cached, cachedList +if TYPE_CHECKING: + from synapse.server import HomeServer + logger = logging.getLogger(__name__) @attr.s(slots=True, frozen=True, auto_attribs=True) +class ThreadsNextBatch: + topological_ordering: int + stream_ordering: int + + def __str__(self) -> str: + return f"{self.topological_ordering}_{self.stream_ordering}" + + @classmethod + def from_string(cls, string: str) -> "ThreadsNextBatch": + """ + Creates a ThreadsNextBatch from its textual representation. + """ + try: + keys = (int(s) for s in string.split("_")) + return cls(*keys) + except Exception: + raise SynapseError(400, "Invalid threads token") + + +@attr.s(slots=True, frozen=True, auto_attribs=True) class _RelatedEvent: """ Contains enough information about a related event in order to properly filter @@ -51,11 +82,79 @@ class _RelatedEvent: event_id: str # The sender of the related event. sender: str - topological_ordering: Optional[int] - stream_ordering: int class RelationsWorkerStore(SQLBaseStore): + def __init__( + self, + database: DatabasePool, + db_conn: LoggingDatabaseConnection, + hs: "HomeServer", + ): + super().__init__(database, db_conn, hs) + + self.db_pool.updates.register_background_update_handler( + "threads_backfill", self._backfill_threads + ) + + async def _backfill_threads(self, progress: JsonDict, batch_size: int) -> int: + """Backfill the threads table.""" + + def threads_backfill_txn(txn: LoggingTransaction) -> int: + last_thread_id = progress.get("last_thread_id", "") + + # Get the latest event in each thread by topo ordering / stream ordering. + # + # Note that the MAX(event_id) is needed to abide by the rules of group by, + # but doesn't actually do anything since there should only be a single event + # ID per topo/stream ordering pair. + sql = f""" + SELECT room_id, relates_to_id, MAX(topological_ordering), MAX(stream_ordering), MAX(event_id) + FROM event_relations + INNER JOIN events USING (event_id) + WHERE + relates_to_id > ? AND + relation_type = '{RelationTypes.THREAD}' + GROUP BY room_id, relates_to_id + ORDER BY relates_to_id + LIMIT ? + """ + txn.execute(sql, (last_thread_id, batch_size)) + + # No more rows to process. + rows = txn.fetchall() + if not rows: + return 0 + + # Insert the rows into the threads table. If a matching thread already exists, + # assume it is from a newer event. + sql = """ + INSERT INTO threads (room_id, thread_id, topological_ordering, stream_ordering, latest_event_id) + VALUES %s + ON CONFLICT (room_id, thread_id) + DO NOTHING + """ + if isinstance(txn.database_engine, PostgresEngine): + txn.execute_values(sql % ("?",), rows, fetch=False) + else: + txn.execute_batch(sql % ("(?, ?, ?, ?, ?)",), rows) + + # Mark the progress. + self.db_pool.updates._background_update_progress_txn( + txn, "threads_backfill", {"last_thread_id": rows[-1][1]} + ) + + return txn.rowcount + + result = await self.db_pool.runInteraction( + "threads_backfill", threads_backfill_txn + ) + + if not result: + await self.db_pool.updates._end_background_update("threads_backfill") + + return result + @cached(uncached_args=("event",), tree=True) async def get_relations_for_event( self, @@ -145,13 +244,17 @@ class RelationsWorkerStore(SQLBaseStore): txn.execute(sql, where_args + [limit + 1]) events = [] - for event_id, relation_type, sender, topo_ordering, stream_ordering in txn: + topo_orderings: List[int] = [] + stream_orderings: List[int] = [] + for event_id, relation_type, sender, topo_ordering, stream_ordering in cast( + List[Tuple[str, str, str, int, int]], txn + ): # Do not include edits for redacted events as they leak event # content. if not is_redacted or relation_type != RelationTypes.REPLACE: - events.append( - _RelatedEvent(event_id, sender, topo_ordering, stream_ordering) - ) + events.append(_RelatedEvent(event_id, sender)) + topo_orderings.append(topo_ordering) + stream_orderings.append(stream_ordering) # If there are more events, generate the next pagination key from the # last event returned. @@ -160,9 +263,11 @@ class RelationsWorkerStore(SQLBaseStore): # Instead of using the last row (which tells us there is more # data), use the last row to be returned. events = events[:limit] + topo_orderings = topo_orderings[:limit] + stream_orderings = stream_orderings[:limit] - topo = events[-1].topological_ordering - token = events[-1].stream_ordering + topo = topo_orderings[-1] + token = stream_orderings[-1] if direction == "b": # Tokens are positions between events. # This token points *after* the last event in the chunk. @@ -195,6 +300,42 @@ class RelationsWorkerStore(SQLBaseStore): "get_recent_references_for_event", _get_recent_references_for_event_txn ) + async def get_all_relations_for_event_with_types( + self, + event_id: str, + relation_types: List[str], + ) -> List[str]: + """Get the event IDs of all events that have a relation to the given event with + one of the given relation types. + + Args: + event_id: The event for which to look for related events. + relation_types: The types of relations to look for. + + Returns: + A list of the IDs of the events that relate to the given event with one of + the given relation types. + """ + + def get_all_relation_ids_for_event_with_types_txn( + txn: LoggingTransaction, + ) -> List[str]: + rows = self.db_pool.simple_select_many_txn( + txn=txn, + table="event_relations", + column="relation_type", + iterable=relation_types, + keyvalues={"relates_to_id": event_id}, + retcols=["event_id"], + ) + + return [row["event_id"] for row in rows] + + return await self.db_pool.runInteraction( + desc="get_all_relation_ids_for_event_with_types", + func=get_all_relation_ids_for_event_with_types_txn, + ) + async def event_includes_relation(self, event_id: str) -> bool: """Check if the given event relates to another event. @@ -258,112 +399,196 @@ class RelationsWorkerStore(SQLBaseStore): ) return result is not None - @cached(tree=True) - async def get_aggregation_groups_for_event( - self, event_id: str, room_id: str, limit: int = 5 - ) -> List[JsonDict]: - """Get a list of annotations on the event, grouped by event type and + @cached() + async def get_aggregation_groups_for_event(self, event_id: str) -> List[JsonDict]: + raise NotImplementedError() + + @cachedList( + cached_method_name="get_aggregation_groups_for_event", list_name="event_ids" + ) + async def get_aggregation_groups_for_events( + self, event_ids: Collection[str] + ) -> Mapping[str, Optional[List[JsonDict]]]: + """Get a list of annotations on the given events, grouped by event type and aggregation key, sorted by count. This is used e.g. to get the what and how many reactions have happend on an event. Args: - event_id: Fetch events that relate to this event ID. - room_id: The room the event belongs to. - limit: Only fetch the `limit` groups. + event_ids: Fetch events that relate to these event IDs. Returns: - List of groups of annotations that match. Each row is a dict with - `type`, `key` and `count` fields. + A map of event IDs to a list of groups of annotations that match. + Each entry is a dict with `type`, `key` and `count` fields. """ + # The number of entries to return per event ID. + limit = 5 - args = [ - event_id, - room_id, - RelationTypes.ANNOTATION, - limit, - ] + clause, args = make_in_list_sql_clause( + self.database_engine, "relates_to_id", event_ids + ) + args.append(RelationTypes.ANNOTATION) - sql = """ - SELECT type, aggregation_key, COUNT(DISTINCT sender) - FROM event_relations - INNER JOIN events USING (event_id) - WHERE relates_to_id = ? AND room_id = ? AND relation_type = ? - GROUP BY relation_type, type, aggregation_key - ORDER BY COUNT(*) DESC - LIMIT ? + sql = f""" + SELECT + relates_to_id, + annotation.type, + aggregation_key, + COUNT(DISTINCT annotation.sender) + FROM events AS annotation + INNER JOIN event_relations USING (event_id) + INNER JOIN events AS parent ON + parent.event_id = relates_to_id + AND parent.room_id = annotation.room_id + WHERE + {clause} + AND relation_type = ? + GROUP BY relates_to_id, annotation.type, aggregation_key + ORDER BY relates_to_id, COUNT(*) DESC """ - def _get_aggregation_groups_for_event_txn( + def _get_aggregation_groups_for_events_txn( txn: LoggingTransaction, - ) -> List[JsonDict]: + ) -> Mapping[str, List[JsonDict]]: txn.execute(sql, args) - return [{"type": row[0], "key": row[1], "count": row[2]} for row in txn] + result: Dict[str, List[JsonDict]] = {} + for event_id, type, key, count in cast( + List[Tuple[str, str, str, int]], txn + ): + event_results = result.setdefault(event_id, []) + + # Limit the number of results per event ID. + if len(event_results) == limit: + continue + + event_results.append({"type": type, "key": key, "count": count}) + + return result return await self.db_pool.runInteraction( - "get_aggregation_groups_for_event", _get_aggregation_groups_for_event_txn + "get_aggregation_groups_for_events", _get_aggregation_groups_for_events_txn ) async def get_aggregation_groups_for_users( - self, - event_id: str, - room_id: str, - limit: int, - users: FrozenSet[str] = frozenset(), - ) -> Dict[Tuple[str, str], int]: + self, event_ids: Collection[str], users: FrozenSet[str] + ) -> Dict[str, Dict[Tuple[str, str], int]]: """Fetch the partial aggregations for an event for specific users. This is used, in conjunction with get_aggregation_groups_for_event, to remove information from the results for ignored users. Args: - event_id: Fetch events that relate to this event ID. - room_id: The room the event belongs to. - limit: Only fetch the `limit` groups. + event_ids: Fetch events that relate to these event IDs. users: The users to fetch information for. Returns: - A map of (event type, aggregation key) to a count of users. + A map of event ID to a map of (event type, aggregation key) to a + count of users. """ if not users: return {} - args: List[Union[str, int]] = [ - event_id, - room_id, - RelationTypes.ANNOTATION, - ] + events_sql, args = make_in_list_sql_clause( + self.database_engine, "relates_to_id", event_ids + ) users_sql, users_args = make_in_list_sql_clause( - self.database_engine, "sender", users + self.database_engine, "annotation.sender", users ) args.extend(users_args) + args.append(RelationTypes.ANNOTATION) sql = f""" - SELECT type, aggregation_key, COUNT(DISTINCT sender) - FROM event_relations - INNER JOIN events USING (event_id) - WHERE relates_to_id = ? AND room_id = ? AND relation_type = ? AND {users_sql} - GROUP BY relation_type, type, aggregation_key - ORDER BY COUNT(*) DESC - LIMIT ? + SELECT + relates_to_id, + annotation.type, + aggregation_key, + COUNT(DISTINCT annotation.sender) + FROM events AS annotation + INNER JOIN event_relations USING (event_id) + INNER JOIN events AS parent ON + parent.event_id = relates_to_id + AND parent.room_id = annotation.room_id + WHERE {events_sql} AND {users_sql} AND relation_type = ? + GROUP BY relates_to_id, annotation.type, aggregation_key + ORDER BY relates_to_id, COUNT(*) DESC """ def _get_aggregation_groups_for_users_txn( txn: LoggingTransaction, - ) -> Dict[Tuple[str, str], int]: - txn.execute(sql, args + [limit]) + ) -> Dict[str, Dict[Tuple[str, str], int]]: + txn.execute(sql, args) - return {(row[0], row[1]): row[2] for row in txn} + result: Dict[str, Dict[Tuple[str, str], int]] = {} + for event_id, type, key, count in cast( + List[Tuple[str, str, str, int]], txn + ): + result.setdefault(event_id, {})[(type, key)] = count + + return result return await self.db_pool.runInteraction( "get_aggregation_groups_for_users", _get_aggregation_groups_for_users_txn ) @cached() + async def get_references_for_event(self, event_id: str) -> List[JsonDict]: + raise NotImplementedError() + + @cachedList(cached_method_name="get_references_for_event", list_name="event_ids") + async def get_references_for_events( + self, event_ids: Collection[str] + ) -> Mapping[str, Optional[List[_RelatedEvent]]]: + """Get a list of references to the given events. + + Args: + event_ids: Fetch events that relate to these event IDs. + + Returns: + A map of event IDs to a list of related event IDs (and their senders). + """ + + clause, args = make_in_list_sql_clause( + self.database_engine, "relates_to_id", event_ids + ) + args.append(RelationTypes.REFERENCE) + + sql = f""" + SELECT relates_to_id, ref.event_id, ref.sender + FROM events AS ref + INNER JOIN event_relations USING (event_id) + INNER JOIN events AS parent ON + parent.event_id = relates_to_id + AND parent.room_id = ref.room_id + WHERE + {clause} + AND relation_type = ? + ORDER BY ref.topological_ordering, ref.stream_ordering + """ + + def _get_references_for_events_txn( + txn: LoggingTransaction, + ) -> Mapping[str, List[_RelatedEvent]]: + txn.execute(sql, args) + + result: Dict[str, List[_RelatedEvent]] = {} + for relates_to_id, event_id, sender in cast( + List[Tuple[str, str, str]], txn + ): + result.setdefault(relates_to_id, []).append( + _RelatedEvent(event_id, sender) + ) + + return result + + return await self.db_pool.runInteraction( + "_get_references_for_events_txn", _get_references_for_events_txn + ) + + @cached() def get_applicable_edit(self, event_id: str) -> Optional[EventBase]: raise NotImplementedError() @@ -779,95 +1004,194 @@ class RelationsWorkerStore(SQLBaseStore): "get_if_user_has_annotated_event", _get_if_user_has_annotated_event ) - @cached(iterable=True) - async def get_mutual_event_relations_for_rel_type( - self, event_id: str, relation_type: str - ) -> Set[Tuple[str, str]]: - raise NotImplementedError() - - @cachedList( - cached_method_name="get_mutual_event_relations_for_rel_type", - list_name="relation_types", - ) - async def get_mutual_event_relations( - self, event_id: str, relation_types: Collection[str] - ) -> Dict[str, Set[Tuple[str, str]]]: - """ - Fetch event metadata for events which related to the same event as the given event. - - If the given event has no relation information, returns an empty dictionary. + @cached(tree=True) + async def get_threads( + self, + room_id: str, + limit: int = 5, + from_token: Optional[ThreadsNextBatch] = None, + ) -> Tuple[List[str], Optional[ThreadsNextBatch]]: + """Get a list of thread IDs, ordered by topological ordering of their + latest reply. Args: - event_id: The event ID which is targeted by relations. - relation_types: The relation types to check for mutual relations. + room_id: The room the event belongs to. + limit: Only fetch the most recent `limit` threads. + from_token: Fetch rows from a previous next_batch, or from the start if None. Returns: - A dictionary of relation type to: - A set of tuples of: - The sender - The event type + A tuple of: + A list of thread root event IDs. + + The next_batch, if one exists. """ - rel_type_sql, rel_type_args = make_in_list_sql_clause( - self.database_engine, "relation_type", relation_types - ) + # Generate the pagination clause, if necessary. + # + # Find any threads where the latest reply is equal / before the last + # thread's topo ordering and earlier in stream ordering. + pagination_clause = "" + pagination_args: tuple = () + if from_token: + pagination_clause = "AND topological_ordering <= ? AND stream_ordering < ?" + pagination_args = ( + from_token.topological_ordering, + from_token.stream_ordering, + ) sql = f""" - SELECT DISTINCT relation_type, sender, type FROM event_relations - INNER JOIN events USING (event_id) - WHERE relates_to_id = ? AND {rel_type_sql} + SELECT thread_id, topological_ordering, stream_ordering + FROM threads + WHERE + room_id = ? + {pagination_clause} + ORDER BY topological_ordering DESC, stream_ordering DESC + LIMIT ? """ - def _get_event_relations( + def _get_threads_txn( txn: LoggingTransaction, - ) -> Dict[str, Set[Tuple[str, str]]]: - txn.execute(sql, [event_id] + rel_type_args) - result: Dict[str, Set[Tuple[str, str]]] = { - rel_type: set() for rel_type in relation_types - } - for rel_type, sender, type in txn.fetchall(): - result[rel_type].add((sender, type)) - return result + ) -> Tuple[List[str], Optional[ThreadsNextBatch]]: + txn.execute(sql, (room_id, *pagination_args, limit + 1)) - return await self.db_pool.runInteraction( - "get_event_relations", _get_event_relations - ) + rows = cast(List[Tuple[str, int, int]], txn.fetchall()) + thread_ids = [r[0] for r in rows] + + # If there are more events, generate the next pagination key from the + # last thread which will be returned. + next_token = None + if len(thread_ids) > limit: + last_topo_id = rows[-2][1] + last_stream_id = rows[-2][2] + next_token = ThreadsNextBatch(last_topo_id, last_stream_id) + + return thread_ids[:limit], next_token + + return await self.db_pool.runInteraction("get_threads", _get_threads_txn) @cached() - async def get_thread_id(self, event_id: str) -> Optional[str]: + async def get_thread_id(self, event_id: str) -> str: """ Get the thread ID for an event. This considers multi-level relations, e.g. an annotation to an event which is part of a thread. + It only searches up the relations tree, i.e. it only searches for events + which the given event is related to (and which those events are related + to, etc.) + + Given the following DAG: + + A <---[m.thread]-- B <--[m.annotation]-- C + ^ + |--[m.reference]-- D <--[m.annotation]-- E + + get_thread_id(X) considers events B and C as part of thread A. + + See also get_thread_id_for_receipts. + Args: event_id: The event ID to fetch the thread ID for. Returns: The event ID of the root event in the thread, if this event is part - of a thread. None, otherwise. + of a thread. "main", otherwise. """ - # Since event relations form a tree, we should only ever find 0 or 1 - # results from the below query. + + # Recurse event relations up to the *root* event, then search that chain + # of relations for a thread relation. If one is found, the root event is + # returned. + # + # Note that this should only ever find 0 or 1 entries since it is invalid + # for an event to have a thread relation to an event which also has a + # relation. sql = """ WITH RECURSIVE related_events AS ( - SELECT event_id, relates_to_id, relation_type + SELECT event_id, relates_to_id, relation_type, 0 depth FROM event_relations WHERE event_id = ? - UNION SELECT e.event_id, e.relates_to_id, e.relation_type + UNION SELECT e.event_id, e.relates_to_id, e.relation_type, depth + 1 FROM event_relations e INNER JOIN related_events r ON r.relates_to_id = e.event_id - ) SELECT relates_to_id FROM related_events WHERE relation_type = 'm.thread'; + WHERE depth <= 3 + ) + SELECT relates_to_id FROM related_events + WHERE relation_type = 'm.thread' + ORDER BY depth DESC + LIMIT 1; """ - def _get_thread_id(txn: LoggingTransaction) -> Optional[str]: + def _get_thread_id(txn: LoggingTransaction) -> str: txn.execute(sql, (event_id,)) - # TODO Should we ensure there's only a single result here? row = txn.fetchone() if row: return row[0] - return None + + # If no thread was found, it is part of the main timeline. + return MAIN_TIMELINE return await self.db_pool.runInteraction("get_thread_id", _get_thread_id) + @cached() + async def get_thread_id_for_receipts(self, event_id: str) -> str: + """ + Get the thread ID for an event by traversing to the top-most related event + and confirming any children events form a thread. + + Given the following DAG: + + A <---[m.thread]-- B <--[m.annotation]-- C + ^ + |--[m.reference]-- D <--[m.annotation]-- E + + get_thread_id_for_receipts(X) considers events A, B, C, D, and E as part + of thread A. + + See also get_thread_id. + + Args: + event_id: The event ID to fetch the thread ID for. + + Returns: + The event ID of the root event in the thread, if this event is part + of a thread. "main", otherwise. + """ + + # Recurse event relations up to the *root* event, then search for any events + # related to that root node for a thread relation. If one is found, the + # root event is returned. + # + # Note that there cannot be thread relations in the middle of the chain since + # it is invalid for an event to have a thread relation to an event which also + # has a relation. + sql = """ + SELECT relates_to_id FROM event_relations WHERE relates_to_id = COALESCE(( + WITH RECURSIVE related_events AS ( + SELECT event_id, relates_to_id, relation_type, 0 depth + FROM event_relations + WHERE event_id = ? + UNION SELECT e.event_id, e.relates_to_id, e.relation_type, depth + 1 + FROM event_relations e + INNER JOIN related_events r ON r.relates_to_id = e.event_id + WHERE depth <= 3 + ) + SELECT relates_to_id FROM related_events + ORDER BY depth DESC + LIMIT 1 + ), ?) AND relation_type = 'm.thread' LIMIT 1; + """ + + def _get_related_thread_id(txn: LoggingTransaction) -> str: + txn.execute(sql, (event_id, event_id)) + row = txn.fetchone() + if row: + return row[0] + + # If no thread was found, it is part of the main timeline. + return MAIN_TIMELINE + + return await self.db_pool.runInteraction( + "get_related_thread_id", _get_related_thread_id + ) + class RelationsStore(RelationsWorkerStore): pass diff --git a/synapse/storage/databases/main/room.py b/synapse/storage/databases/main/room.py index e41c99027a..78906a5e1d 100644 --- a/synapse/storage/databases/main/room.py +++ b/synapse/storage/databases/main/room.py @@ -1,5 +1,5 @@ # Copyright 2014-2016 OpenMarket Ltd -# Copyright 2019 The Matrix.org Foundation C.I.C. +# Copyright 2019, 2022 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. @@ -50,8 +50,14 @@ from synapse.storage.database import ( LoggingTransaction, ) from synapse.storage.databases.main.cache import CacheInvalidationWorkerStore +from synapse.storage.engines import PostgresEngine from synapse.storage.types import Cursor -from synapse.storage.util.id_generators import IdGenerator +from synapse.storage.util.id_generators import ( + AbstractStreamIdGenerator, + IdGenerator, + MultiWriterIdGenerator, + StreamIdGenerator, +) from synapse.types import JsonDict, RetentionPolicy, ThirdPartyInstanceID from synapse.util import json_encoder from synapse.util.caches.descriptors import cached @@ -97,6 +103,12 @@ class RoomSortOrder(Enum): STATE_EVENTS = "state_events" +@attr.s(slots=True, frozen=True, auto_attribs=True) +class PartialStateResyncInfo: + joined_via: Optional[str] + servers_in_room: List[str] = attr.ib(factory=list) + + class RoomWorkerStore(CacheInvalidationWorkerStore): def __init__( self, @@ -108,6 +120,26 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): self.config: HomeServerConfig = hs.config + self._un_partial_stated_rooms_stream_id_gen: AbstractStreamIdGenerator + + if isinstance(database.engine, PostgresEngine): + self._un_partial_stated_rooms_stream_id_gen = MultiWriterIdGenerator( + db_conn=db_conn, + db=database, + stream_name="un_partial_stated_room_stream", + instance_name=self._instance_name, + tables=[ + ("un_partial_stated_room_stream", "instance_name", "stream_id") + ], + sequence_name="un_partial_stated_room_stream_sequence", + # TODO(faster_joins, multiple writers) Support multiple writers. + writers=["master"], + ) + else: + self._un_partial_stated_rooms_stream_id_gen = StreamIdGenerator( + db_conn, "un_partial_stated_room_stream", "stream_id" + ) + async def store_room( self, room_id: str, @@ -906,7 +938,11 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): event_json = db_to_json(content_json) content = event_json["content"] content_url = content.get("url") - thumbnail_url = content.get("info", {}).get("thumbnail_url") + info = content.get("info") + if isinstance(info, dict): + thumbnail_url = info.get("thumbnail_url") + else: + thumbnail_url = None for url in (content_url, thumbnail_url): if not url: @@ -1160,17 +1196,29 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): desc="get_partial_state_servers_at_join", ) - async def get_partial_state_rooms_and_servers( + async def get_partial_state_room_resync_info( self, - ) -> Mapping[str, Collection[str]]: - """Get all rooms containing events with partial state, and the servers known - to be in the room. + ) -> Mapping[str, PartialStateResyncInfo]: + """Get all rooms containing events with partial state, and the information + needed to restart a "resync" of those rooms. Returns: A dictionary of rooms with partial state, with room IDs as keys and lists of servers in rooms as values. """ - room_servers: Dict[str, List[str]] = {} + room_servers: Dict[str, PartialStateResyncInfo] = {} + + rows = await self.db_pool.simple_select_list( + table="partial_state_rooms", + keyvalues={}, + retcols=("room_id", "joined_via"), + desc="get_server_which_served_partial_join", + ) + + for row in rows: + room_id = row["room_id"] + joined_via = row["joined_via"] + room_servers[room_id] = PartialStateResyncInfo(joined_via=joined_via) rows = await self.db_pool.simple_select_list( "partial_state_rooms_servers", @@ -1182,74 +1230,18 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): for row in rows: room_id = row["room_id"] server_name = row["server_name"] - room_servers.setdefault(room_id, []).append(server_name) + entry = room_servers.get(room_id) + if entry is None: + # There is a foreign key constraint which enforces that every room_id in + # partial_state_rooms_servers appears in partial_state_rooms. So we + # expect `entry` to be non-null. (This reasoning fails if we've + # partial-joined between the two SELECTs, but this is unlikely to happen + # in practice.) + continue + entry.servers_in_room.append(server_name) return room_servers - async def clear_partial_state_room(self, room_id: str) -> bool: - """Clears the partial state flag for a room. - - Args: - room_id: The room whose partial state flag is to be cleared. - - Returns: - `True` if the partial state flag has been cleared successfully. - - `False` if the partial state flag could not be cleared because the room - still contains events with partial state. - """ - try: - await self.db_pool.runInteraction( - "clear_partial_state_room", self._clear_partial_state_room_txn, room_id - ) - return True - except self.db_pool.engine.module.IntegrityError as e: - # Assume that any `IntegrityError`s are due to partial state events. - logger.info( - "Exception while clearing lazy partial-state-room %s, retrying: %s", - room_id, - e, - ) - return False - - def _clear_partial_state_room_txn( - self, txn: LoggingTransaction, room_id: str - ) -> None: - DatabasePool.simple_delete_txn( - txn, - table="partial_state_rooms_servers", - keyvalues={"room_id": room_id}, - ) - DatabasePool.simple_delete_one_txn( - txn, - table="partial_state_rooms", - keyvalues={"room_id": room_id}, - ) - self._invalidate_cache_and_stream(txn, self.is_partial_state_room, (room_id,)) - self._invalidate_cache_and_stream( - txn, self.get_partial_state_servers_at_join, (room_id,) - ) - - # We now delete anything from `device_lists_remote_pending` with a - # stream ID less than the minimum - # `partial_state_rooms.device_lists_stream_id`, as we no longer need them. - device_lists_stream_id = DatabasePool.simple_select_one_onecol_txn( - txn, - table="partial_state_rooms", - keyvalues={}, - retcol="MIN(device_lists_stream_id)", - allow_none=True, - ) - if device_lists_stream_id is None: - # There are no rooms being currently partially joined, so we delete everything. - txn.execute("DELETE FROM device_lists_remote_pending") - else: - sql = """ - DELETE FROM device_lists_remote_pending - WHERE stream_id <= ? - """ - txn.execute(sql, (device_lists_stream_id,)) - @cached() async def is_partial_state_room(self, room_id: str) -> bool: """Checks if this room has partial state. @@ -1285,6 +1277,66 @@ class RoomWorkerStore(CacheInvalidationWorkerStore): ) return result["join_event_id"], result["device_lists_stream_id"] + def get_un_partial_stated_rooms_token(self) -> int: + # TODO(faster_joins, multiple writers): This is inappropriate if there + # are multiple writers because workers that don't write often will + # hold all readers up. + # (See `MultiWriterIdGenerator.get_persisted_upto_position` for an + # explanation.) + return self._un_partial_stated_rooms_stream_id_gen.get_current_token() + + async def get_un_partial_stated_rooms_from_stream( + self, instance_name: str, last_id: int, current_id: int, limit: int + ) -> Tuple[List[Tuple[int, Tuple[str]]], int, bool]: + """Get updates for caches replication stream. + + Args: + instance_name: The writer we want to fetch updates from. Unused + here since there is only ever one writer. + last_id: The token to fetch updates from. Exclusive. + current_id: The token to fetch updates up to. Inclusive. + limit: The requested limit for the number of rows to return. The + function may return more or fewer rows. + + Returns: + A tuple consisting of: the updates, a token to use to fetch + subsequent updates, and whether we returned fewer rows than exists + between the requested tokens due to the limit. + + The token returned can be used in a subsequent call to this + function to get further updatees. + + The updates are a list of 2-tuples of stream ID and the row data + """ + + if last_id == current_id: + return [], current_id, False + + def get_un_partial_stated_rooms_from_stream_txn( + txn: LoggingTransaction, + ) -> Tuple[List[Tuple[int, Tuple[str]]], int, bool]: + sql = """ + SELECT stream_id, room_id + FROM un_partial_stated_room_stream + WHERE ? < stream_id AND stream_id <= ? AND instance_name = ? + ORDER BY stream_id ASC + LIMIT ? + """ + txn.execute(sql, (last_id, current_id, instance_name, limit)) + updates = [(row[0], (row[1],)) for row in txn] + limited = False + upto_token = current_id + if len(updates) >= limit: + upto_token = updates[-1][0] + limited = True + + return updates, upto_token, limited + + return await self.db_pool.runInteraction( + "get_un_partial_stated_rooms_from_stream", + get_un_partial_stated_rooms_from_stream_txn, + ) + class _BackgroundUpdates: REMOVE_TOMESTONED_ROOMS_BG_UPDATE = "remove_tombstoned_rooms_from_directory" @@ -1776,6 +1828,8 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): self._event_reports_id_gen = IdGenerator(db_conn, "event_reports", "id") + self._instance_name = hs.get_instance_name() + async def upsert_room_on_join( self, room_id: str, room_version: RoomVersion, state_events: List[EventBase] ) -> None: @@ -1817,9 +1871,6 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): "creator": room_creator, "has_auth_chain_index": has_auth_chain_index, }, - # rooms has a unique constraint on room_id, so no need to lock when doing an - # emulated upsert. - lock=False, ) async def store_partial_state_room( @@ -1827,6 +1878,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): room_id: str, servers: Collection[str], device_lists_stream_id: int, + joined_via: str, ) -> None: """Mark the given room as containing events with partial state. @@ -1842,6 +1894,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): servers: other servers known to be in the room device_lists_stream_id: the device_lists stream ID at the time when we first joined the room. + joined_via: the server name we requested a partial join from. """ await self.db_pool.runInteraction( "store_partial_state_room", @@ -1849,6 +1902,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): room_id, servers, device_lists_stream_id, + joined_via, ) def _store_partial_state_room_txn( @@ -1857,6 +1911,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): room_id: str, servers: Collection[str], device_lists_stream_id: int, + joined_via: str, ) -> None: DatabasePool.simple_insert_txn( txn, @@ -1866,6 +1921,7 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): "device_lists_stream_id": device_lists_stream_id, # To be updated later once the join event is persisted. "join_event_id": None, + "joined_via": joined_via, }, ) DatabasePool.simple_insert_many_txn( @@ -1935,9 +1991,6 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): "creator": "", "has_auth_chain_index": has_auth_chain_index, }, - # rooms has a unique constraint on room_id, so no need to lock when doing an - # emulated upsert. - lock=False, ) async def set_room_is_public(self, room_id: str, is_public: bool) -> None: @@ -2026,7 +2079,8 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): Args: report_id: ID of reported event in database Returns: - event_report: json list of information from event report + JSON dict of information from an event report or None if the + report does not exist. """ def _get_event_report_txn( @@ -2099,8 +2153,9 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): user_id: search for user_id. Ignored if user_id is None room_id: search for room_id. Ignored if room_id is None Returns: - event_reports: json list of event reports - count: total number of event reports matching the filter criteria + Tuple of: + json list of event reports + total number of event reports matching the filter criteria """ def _get_event_reports_paginate_txn( @@ -2239,3 +2294,84 @@ class RoomStore(RoomBackgroundUpdateStore, RoomWorkerStore): self.is_room_blocked, (room_id,), ) + + async def clear_partial_state_room(self, room_id: str) -> bool: + """Clears the partial state flag for a room. + + Args: + room_id: The room whose partial state flag is to be cleared. + + Returns: + `True` if the partial state flag has been cleared successfully. + + `False` if the partial state flag could not be cleared because the room + still contains events with partial state. + """ + try: + async with self._un_partial_stated_rooms_stream_id_gen.get_next() as un_partial_state_room_stream_id: + await self.db_pool.runInteraction( + "clear_partial_state_room", + self._clear_partial_state_room_txn, + room_id, + un_partial_state_room_stream_id, + ) + return True + except self.db_pool.engine.module.IntegrityError as e: + # Assume that any `IntegrityError`s are due to partial state events. + logger.info( + "Exception while clearing lazy partial-state-room %s, retrying: %s", + room_id, + e, + ) + return False + + def _clear_partial_state_room_txn( + self, + txn: LoggingTransaction, + room_id: str, + un_partial_state_room_stream_id: int, + ) -> None: + DatabasePool.simple_delete_txn( + txn, + table="partial_state_rooms_servers", + keyvalues={"room_id": room_id}, + ) + DatabasePool.simple_delete_one_txn( + txn, + table="partial_state_rooms", + keyvalues={"room_id": room_id}, + ) + self._invalidate_cache_and_stream(txn, self.is_partial_state_room, (room_id,)) + self._invalidate_cache_and_stream( + txn, self.get_partial_state_servers_at_join, (room_id,) + ) + + DatabasePool.simple_insert_txn( + txn, + "un_partial_stated_room_stream", + { + "stream_id": un_partial_state_room_stream_id, + "instance_name": self._instance_name, + "room_id": room_id, + }, + ) + + # We now delete anything from `device_lists_remote_pending` with a + # stream ID less than the minimum + # `partial_state_rooms.device_lists_stream_id`, as we no longer need them. + device_lists_stream_id = DatabasePool.simple_select_one_onecol_txn( + txn, + table="partial_state_rooms", + keyvalues={}, + retcol="MIN(device_lists_stream_id)", + allow_none=True, + ) + if device_lists_stream_id is None: + # There are no rooms being currently partially joined, so we delete everything. + txn.execute("DELETE FROM device_lists_remote_pending") + else: + sql = """ + DELETE FROM device_lists_remote_pending + WHERE stream_id <= ? + """ + txn.execute(sql, (device_lists_stream_id,)) diff --git a/synapse/storage/databases/main/room_batch.py b/synapse/storage/databases/main/room_batch.py index 39e80f6f5b..131f357d04 100644 --- a/synapse/storage/databases/main/room_batch.py +++ b/synapse/storage/databases/main/room_batch.py @@ -44,6 +44,4 @@ class RoomBatchStore(SQLBaseStore): table="event_to_state_groups", keyvalues={"event_id": event_id}, values={"state_group": state_group_id, "event_id": event_id}, - # Unique constraint on event_id so we don't have to lock - lock=False, ) diff --git a/synapse/storage/databases/main/roommember.py b/synapse/storage/databases/main/roommember.py index 2337289d88..f02c1d7ea7 100644 --- a/synapse/storage/databases/main/roommember.py +++ b/synapse/storage/databases/main/roommember.py @@ -152,6 +152,9 @@ class RoomMemberWorkerStore(EventsWorkerStore): the forward extremities of those rooms will exclude most members. We may also calculate room state incorrectly for such rooms and believe that a member is or is not in the room when the opposite is true. + + Note: If you only care about users in the room local to the homeserver, use + `get_local_users_in_room(...)` instead which will be more performant. """ return await self.db_pool.simple_select_onecol( table="current_state_events", @@ -666,7 +669,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): cached_method_name="get_rooms_for_user", list_name="user_ids", ) - async def get_rooms_for_users( + async def _get_rooms_for_users( self, user_ids: Collection[str] ) -> Dict[str, FrozenSet[str]]: """A batched version of `get_rooms_for_user`. @@ -697,6 +700,21 @@ class RoomMemberWorkerStore(EventsWorkerStore): return {key: frozenset(rooms) for key, rooms in user_rooms.items()} + async def get_rooms_for_users( + self, user_ids: Collection[str] + ) -> Dict[str, FrozenSet[str]]: + """A batched wrapper around `_get_rooms_for_users`, to prevent locking + other calls to `get_rooms_for_user` for large user lists. + """ + all_user_rooms: Dict[str, FrozenSet[str]] = {} + + # 250 users is pretty arbitrary but the data can be quite large if users + # are in many rooms. + for batch_user_ids in batch_iter(user_ids, 250): + all_user_rooms.update(await self._get_rooms_for_users(batch_user_ids)) + + return all_user_rooms + @cached(max_entries=10000) async def does_pair_of_users_share_a_room( self, user_id: str, other_user_id: str @@ -727,7 +745,7 @@ class RoomMemberWorkerStore(EventsWorkerStore): # user and the set of other users, and then checking if there is any # overlap. sql = f""" - SELECT b.state_key + SELECT DISTINCT b.state_key FROM ( SELECT room_id FROM current_state_events WHERE type = 'm.room.member' AND membership = 'join' AND state_key = ? @@ -736,7 +754,6 @@ class RoomMemberWorkerStore(EventsWorkerStore): SELECT room_id, state_key FROM current_state_events WHERE type = 'm.room.member' AND membership = 'join' AND {clause} ) AS b using (room_id) - LIMIT 1 """ txn.execute(sql, (user_id, *args)) @@ -1500,6 +1517,36 @@ class RoomMemberStore( await self.db_pool.runInteraction("forget_membership", f) +def extract_heroes_from_room_summary( + details: Mapping[str, MemberSummary], me: str +) -> List[str]: + """Determine the users that represent a room, from the perspective of the `me` user. + + The rules which say which users we select are specified in the "Room Summary" + section of + https://spec.matrix.org/v1.4/client-server-api/#get_matrixclientv3sync + + Returns a list (possibly empty) of heroes' mxids. + """ + empty_ms = MemberSummary([], 0) + + joined_user_ids = [ + r[0] for r in details.get(Membership.JOIN, empty_ms).members if r[0] != me + ] + invited_user_ids = [ + r[0] for r in details.get(Membership.INVITE, empty_ms).members if r[0] != me + ] + gone_user_ids = [ + r[0] for r in details.get(Membership.LEAVE, empty_ms).members if r[0] != me + ] + [r[0] for r in details.get(Membership.BAN, empty_ms).members if r[0] != me] + + # FIXME: order by stream ordering rather than as returned by SQL + if joined_user_ids or invited_user_ids: + return sorted(joined_user_ids + invited_user_ids)[0:5] + else: + return sorted(gone_user_ids)[0:5] + + @attr.s(slots=True, auto_attribs=True) class _JoinedHostsCache: """The cached data used by the `_get_joined_hosts_cache`.""" diff --git a/synapse/storage/databases/main/search.py b/synapse/storage/databases/main/search.py index 1b79acf955..3fe433f66c 100644 --- a/synapse/storage/databases/main/search.py +++ b/synapse/storage/databases/main/search.py @@ -11,10 +11,22 @@ # 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 enum import logging import re -from typing import TYPE_CHECKING, Any, Collection, Iterable, List, Optional, Set, Tuple +from collections import deque +from dataclasses import dataclass +from typing import ( + TYPE_CHECKING, + Any, + Collection, + Iterable, + List, + Optional, + Set, + Tuple, + Union, +) import attr @@ -27,7 +39,7 @@ from synapse.storage.database import ( LoggingTransaction, ) from synapse.storage.databases.main.events_worker import EventRedactBehaviour -from synapse.storage.engines import BaseDatabaseEngine, PostgresEngine, Sqlite3Engine +from synapse.storage.engines import PostgresEngine, Sqlite3Engine from synapse.types import JsonDict if TYPE_CHECKING: @@ -68,11 +80,11 @@ class SearchWorkerStore(SQLBaseStore): if not self.hs.config.server.enable_search: return if isinstance(self.database_engine, PostgresEngine): - sql = ( - "INSERT INTO event_search" - " (event_id, room_id, key, vector, stream_ordering, origin_server_ts)" - " VALUES (?,?,?,to_tsvector('english', ?),?,?)" - ) + sql = """ + INSERT INTO event_search + (event_id, room_id, key, vector, stream_ordering, origin_server_ts) + VALUES (?,?,?,to_tsvector('english', ?),?,?) + """ args1 = ( ( @@ -89,20 +101,20 @@ class SearchWorkerStore(SQLBaseStore): txn.execute_batch(sql, args1) elif isinstance(self.database_engine, Sqlite3Engine): - sql = ( - "INSERT INTO event_search (event_id, room_id, key, value)" - " VALUES (?,?,?,?)" - ) - args2 = ( - ( - entry.event_id, - entry.room_id, - entry.key, - _clean_value_for_search(entry.value), - ) - for entry in entries + self.db_pool.simple_insert_many_txn( + txn, + table="event_search", + keys=("event_id", "room_id", "key", "value"), + values=( + ( + entry.event_id, + entry.room_id, + entry.key, + _clean_value_for_search(entry.value), + ) + for entry in entries + ), ) - txn.execute_batch(sql, args2) else: # This should be unreachable. @@ -150,15 +162,17 @@ class SearchBackgroundUpdateStore(SearchWorkerStore): TYPES = ["m.room.name", "m.room.message", "m.room.topic"] def reindex_search_txn(txn: LoggingTransaction) -> int: - sql = ( - "SELECT stream_ordering, event_id, room_id, type, json, " - " origin_server_ts FROM events" - " JOIN event_json USING (room_id, event_id)" - " WHERE ? <= stream_ordering AND stream_ordering < ?" - " AND (%s)" - " ORDER BY stream_ordering DESC" - " LIMIT ?" - ) % (" OR ".join("type = '%s'" % (t,) for t in TYPES),) + sql = """ + SELECT stream_ordering, event_id, room_id, type, json, origin_server_ts + FROM events + JOIN event_json USING (room_id, event_id) + WHERE ? <= stream_ordering AND stream_ordering < ? + AND (%s) + ORDER BY stream_ordering DESC + LIMIT ? + """ % ( + " OR ".join("type = '%s'" % (t,) for t in TYPES), + ) txn.execute(sql, (target_min_stream_id, max_stream_id, batch_size)) @@ -272,8 +286,10 @@ class SearchBackgroundUpdateStore(SearchWorkerStore): try: c.execute( - "CREATE INDEX CONCURRENTLY event_search_fts_idx" - " ON event_search USING GIN (vector)" + """ + CREATE INDEX CONCURRENTLY event_search_fts_idx + ON event_search USING GIN (vector) + """ ) except psycopg2.ProgrammingError as e: logger.warning( @@ -311,12 +327,16 @@ class SearchBackgroundUpdateStore(SearchWorkerStore): # We create with NULLS FIRST so that when we search *backwards* # we get the ones with non null origin_server_ts *first* c.execute( - "CREATE INDEX CONCURRENTLY event_search_room_order ON event_search(" - "room_id, origin_server_ts NULLS FIRST, stream_ordering NULLS FIRST)" + """ + CREATE INDEX CONCURRENTLY event_search_room_order + ON event_search(room_id, origin_server_ts NULLS FIRST, stream_ordering NULLS FIRST) + """ ) c.execute( - "CREATE INDEX CONCURRENTLY event_search_order ON event_search(" - "origin_server_ts NULLS FIRST, stream_ordering NULLS FIRST)" + """ + CREATE INDEX CONCURRENTLY event_search_order + ON event_search(origin_server_ts NULLS FIRST, stream_ordering NULLS FIRST) + """ ) conn.set_session(autocommit=False) @@ -333,14 +353,14 @@ class SearchBackgroundUpdateStore(SearchWorkerStore): ) def reindex_search_txn(txn: LoggingTransaction) -> Tuple[int, bool]: - sql = ( - "UPDATE event_search AS es SET stream_ordering = e.stream_ordering," - " origin_server_ts = e.origin_server_ts" - " FROM events AS e" - " WHERE e.event_id = es.event_id" - " AND ? <= e.stream_ordering AND e.stream_ordering < ?" - " RETURNING es.stream_ordering" - ) + sql = """ + UPDATE event_search AS es + SET stream_ordering = e.stream_ordering, origin_server_ts = e.origin_server_ts + FROM events AS e + WHERE e.event_id = es.event_id + AND ? <= e.stream_ordering AND e.stream_ordering < ? + RETURNING es.stream_ordering + """ min_stream_id = max_stream_id - batch_size txn.execute(sql, (min_stream_id, max_stream_id)) @@ -421,8 +441,6 @@ class SearchStore(SearchBackgroundUpdateStore): """ clauses = [] - search_query = _parse_query(self.database_engine, search_term) - args: List[Any] = [] # Make sure we don't explode because the person is in too many rooms. @@ -444,32 +462,35 @@ class SearchStore(SearchBackgroundUpdateStore): count_clauses = clauses if isinstance(self.database_engine, PostgresEngine): - sql = ( - "SELECT ts_rank_cd(vector, to_tsquery('english', ?)) AS rank," - " room_id, event_id" - " FROM event_search" - " WHERE vector @@ to_tsquery('english', ?)" - ) + search_query = search_term + sql = """ + SELECT ts_rank_cd(vector, websearch_to_tsquery('english', ?)) AS rank, + room_id, event_id + FROM event_search + WHERE vector @@ websearch_to_tsquery('english', ?) + """ args = [search_query, search_query] + args - count_sql = ( - "SELECT room_id, count(*) as count FROM event_search" - " WHERE vector @@ to_tsquery('english', ?)" - ) + count_sql = """ + SELECT room_id, count(*) as count FROM event_search + WHERE vector @@ websearch_to_tsquery('english', ?) + """ count_args = [search_query] + count_args elif isinstance(self.database_engine, Sqlite3Engine): - sql = ( - "SELECT rank(matchinfo(event_search)) as rank, room_id, event_id" - " FROM event_search" - " WHERE value MATCH ?" - ) + search_query = _parse_query_for_sqlite(search_term) + + sql = """ + SELECT rank(matchinfo(event_search)) as rank, room_id, event_id + FROM event_search + WHERE value MATCH ? + """ args = [search_query] + args - count_sql = ( - "SELECT room_id, count(*) as count FROM event_search" - " WHERE value MATCH ?" - ) - count_args = [search_term] + count_args + count_sql = """ + SELECT room_id, count(*) as count FROM event_search + WHERE value MATCH ? + """ + count_args = [search_query] + count_args else: # This should be unreachable. raise Exception("Unrecognized database engine") @@ -510,7 +531,6 @@ class SearchStore(SearchBackgroundUpdateStore): ) count = sum(row["count"] for row in count_results if row["room_id"] in room_ids) - return { "results": [ {"event": event_map[r["event_id"]], "rank": r["rank"]} @@ -542,9 +562,6 @@ class SearchStore(SearchBackgroundUpdateStore): Each match as a dictionary. """ clauses = [] - - search_query = _parse_query(self.database_engine, search_term) - args: List[Any] = [] # Make sure we don't explode because the person is in too many rooms. @@ -576,26 +593,29 @@ class SearchStore(SearchBackgroundUpdateStore): raise SynapseError(400, "Invalid pagination token") clauses.append( - "(origin_server_ts < ?" - " OR (origin_server_ts = ? AND stream_ordering < ?))" + """ + (origin_server_ts < ? OR (origin_server_ts = ? AND stream_ordering < ?)) + """ ) args.extend([origin_server_ts, origin_server_ts, stream]) if isinstance(self.database_engine, PostgresEngine): - sql = ( - "SELECT ts_rank_cd(vector, to_tsquery('english', ?)) as rank," - " origin_server_ts, stream_ordering, room_id, event_id" - " FROM event_search" - " WHERE vector @@ to_tsquery('english', ?) AND " - ) + search_query = search_term + sql = """ + SELECT ts_rank_cd(vector, websearch_to_tsquery('english', ?)) as rank, + origin_server_ts, stream_ordering, room_id, event_id + FROM event_search + WHERE vector @@ websearch_to_tsquery('english', ?) AND + """ args = [search_query, search_query] + args - count_sql = ( - "SELECT room_id, count(*) as count FROM event_search" - " WHERE vector @@ to_tsquery('english', ?) AND " - ) + count_sql = """ + SELECT room_id, count(*) as count FROM event_search + WHERE vector @@ websearch_to_tsquery('english', ?) AND + """ count_args = [search_query] + count_args elif isinstance(self.database_engine, Sqlite3Engine): + # We use CROSS JOIN here to ensure we use the right indexes. # https://sqlite.org/optoverview.html#crossjoin # @@ -604,23 +624,25 @@ class SearchStore(SearchBackgroundUpdateStore): # in the events table to get the topological ordering. We need # to use the indexes in this order because sqlite refuses to # MATCH unless it uses the full text search index - sql = ( - "SELECT rank(matchinfo) as rank, room_id, event_id," - " origin_server_ts, stream_ordering" - " FROM (SELECT key, event_id, matchinfo(event_search) as matchinfo" - " FROM event_search" - " WHERE value MATCH ?" - " )" - " CROSS JOIN events USING (event_id)" - " WHERE " + sql = """ + SELECT + rank(matchinfo) as rank, room_id, event_id, origin_server_ts, stream_ordering + FROM ( + SELECT key, event_id, matchinfo(event_search) as matchinfo + FROM event_search + WHERE value MATCH ? ) + CROSS JOIN events USING (event_id) + WHERE + """ + search_query = _parse_query_for_sqlite(search_term) args = [search_query] + args - count_sql = ( - "SELECT room_id, count(*) as count FROM event_search" - " WHERE value MATCH ? AND " - ) - count_args = [search_term] + count_args + count_sql = """ + SELECT room_id, count(*) as count FROM event_search + WHERE value MATCH ? AND + """ + count_args = [search_query] + count_args else: # This should be unreachable. raise Exception("Unrecognized database engine") @@ -631,10 +653,10 @@ class SearchStore(SearchBackgroundUpdateStore): # We add an arbitrary limit here to ensure we don't try to pull the # entire table from the database. if isinstance(self.database_engine, PostgresEngine): - sql += ( - " ORDER BY origin_server_ts DESC NULLS LAST," - " stream_ordering DESC NULLS LAST LIMIT ?" - ) + sql += """ + ORDER BY origin_server_ts DESC NULLS LAST, stream_ordering DESC NULLS LAST + LIMIT ? + """ elif isinstance(self.database_engine, Sqlite3Engine): sql += " ORDER BY origin_server_ts DESC, stream_ordering DESC LIMIT ?" else: @@ -729,13 +751,16 @@ class SearchStore(SearchBackgroundUpdateStore): while stop_sel in value: stop_sel += ">" - query = "SELECT ts_headline(?, to_tsquery('english', ?), %s)" % ( - _to_postgres_options( - { - "StartSel": start_sel, - "StopSel": stop_sel, - "MaxFragments": "50", - } + query = ( + "SELECT ts_headline(?, websearch_to_tsquery('english', ?), %s)" + % ( + _to_postgres_options( + { + "StartSel": start_sel, + "StopSel": stop_sel, + "MaxFragments": "50", + } + ) ) ) txn.execute(query, (value, search_query)) @@ -760,20 +785,127 @@ def _to_postgres_options(options_dict: JsonDict) -> str: return "'%s'" % (",".join("%s=%s" % (k, v) for k, v in options_dict.items()),) -def _parse_query(database_engine: BaseDatabaseEngine, search_term: str) -> str: - """Takes a plain unicode string from the user and converts it into a form - that can be passed to database. - We use this so that we can add prefix matching, which isn't something - that is supported by default. +@dataclass +class Phrase: + phrase: List[str] + + +class SearchToken(enum.Enum): + Not = enum.auto() + Or = enum.auto() + And = enum.auto() + + +Token = Union[str, Phrase, SearchToken] +TokenList = List[Token] + + +def _is_stop_word(word: str) -> bool: + # TODO Pull these out of the dictionary: + # https://github.com/postgres/postgres/blob/master/src/backend/snowball/stopwords/english.stop + return word in {"the", "a", "you", "me", "and", "but"} + + +def _tokenize_query(query: str) -> TokenList: """ + Convert the user-supplied `query` into a TokenList, which can be translated into + some DB-specific syntax. + + The following constructs are supported: + + - phrase queries using "double quotes" + - case-insensitive `or` and `and` operators + - negation of a keyword via unary `-` + - unary hyphen to denote NOT e.g. 'include -exclude' - # Pull out the individual words, discarding any non-word characters. - results = re.findall(r"([\w\-]+)", search_term, re.UNICODE) + The following differs from websearch_to_tsquery: - if isinstance(database_engine, PostgresEngine): - return " & ".join(result + ":*" for result in results) - elif isinstance(database_engine, Sqlite3Engine): - return " & ".join(result + "*" for result in results) - else: - # This should be unreachable. - raise Exception("Unrecognized database engine") + - Stop words are not removed. + - Unclosed phrases are treated differently. + + """ + tokens: TokenList = [] + + # Find phrases. + in_phrase = False + parts = deque(query.split('"')) + for i, part in enumerate(parts): + # The contents inside double quotes is treated as a phrase. + in_phrase = bool(i % 2) + + # Pull out the individual words, discarding any non-word characters. + words = deque(re.findall(r"([\w\-]+)", part, re.UNICODE)) + + # Phrases have simplified handling of words. + if in_phrase: + # Skip stop words. + phrase = [word for word in words if not _is_stop_word(word)] + + # Consecutive words are implicitly ANDed together. + if tokens and tokens[-1] not in (SearchToken.Not, SearchToken.Or): + tokens.append(SearchToken.And) + + # Add the phrase. + tokens.append(Phrase(phrase)) + continue + + # Otherwise, not in a phrase. + while words: + word = words.popleft() + + if word.startswith("-"): + tokens.append(SearchToken.Not) + + # If there's more word, put it back to be processed again. + word = word[1:] + if word: + words.appendleft(word) + elif word.lower() == "or": + tokens.append(SearchToken.Or) + else: + # Skip stop words. + if _is_stop_word(word): + continue + + # Consecutive words are implicitly ANDed together. + if tokens and tokens[-1] not in (SearchToken.Not, SearchToken.Or): + tokens.append(SearchToken.And) + + # Add the search term. + tokens.append(word) + + return tokens + + +def _tokens_to_sqlite_match_query(tokens: TokenList) -> str: + """ + Convert the list of tokens to a string suitable for passing to sqlite's MATCH. + Assume sqlite was compiled with enhanced query syntax. + + Ref: https://www.sqlite.org/fts3.html#full_text_index_queries + """ + match_query = [] + for token in tokens: + if isinstance(token, str): + match_query.append(token) + elif isinstance(token, Phrase): + match_query.append('"' + " ".join(token.phrase) + '"') + elif token == SearchToken.Not: + # TODO: SQLite treats NOT as a *binary* operator. Hopefully a search + # term has already been added before this. + match_query.append(" NOT ") + elif token == SearchToken.Or: + match_query.append(" OR ") + elif token == SearchToken.And: + match_query.append(" AND ") + else: + raise ValueError(f"unknown token {token}") + + return "".join(match_query) + + +def _parse_query_for_sqlite(search_term: str) -> str: + """Takes a plain unicode string from the user and converts it into a form + that can be passed to sqllite's matchinfo(). + """ + return _tokens_to_sqlite_match_query(_tokenize_query(search_term)) diff --git a/synapse/storage/databases/main/state.py b/synapse/storage/databases/main/state.py index af7bebee80..c801a93b5b 100644 --- a/synapse/storage/databases/main/state.py +++ b/synapse/storage/databases/main/state.py @@ -33,8 +33,8 @@ from synapse.storage.database import ( ) from synapse.storage.databases.main.events_worker import EventsWorkerStore from synapse.storage.databases.main.roommember import RoomMemberWorkerStore -from synapse.storage.state import StateFilter from synapse.types import JsonDict, JsonMapping, StateMap +from synapse.types.state import StateFilter from synapse.util.caches import intern_string from synapse.util.caches.descriptors import cached, cachedList from synapse.util.cancellation import cancellable diff --git a/synapse/storage/databases/main/stream.py b/synapse/storage/databases/main/stream.py index 530f04e149..cc27ec3804 100644 --- a/synapse/storage/databases/main/stream.py +++ b/synapse/storage/databases/main/stream.py @@ -357,6 +357,24 @@ def filter_to_clause(event_filter: Optional[Filter]) -> Tuple[str, List[str]]: ) args.extend(event_filter.related_by_rel_types) + if event_filter.rel_types: + clauses.append( + "(%s)" + % " OR ".join( + "event_relation.relation_type = ?" for _ in event_filter.rel_types + ) + ) + args.extend(event_filter.rel_types) + + if event_filter.not_rel_types: + clauses.append( + "((%s) OR event_relation.relation_type IS NULL)" + % " AND ".join( + "event_relation.relation_type != ?" for _ in event_filter.not_rel_types + ) + ) + args.extend(event_filter.not_rel_types) + return " AND ".join(clauses), args @@ -397,6 +415,7 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): ) self._stream_order_on_start = self.get_room_max_stream_ordering() + self._min_stream_order_on_start = self.get_room_min_stream_ordering() def get_room_max_stream_ordering(self) -> int: """Get the stream_ordering of regular events that we have committed up to @@ -1024,28 +1043,31 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): "after": {"event_ids": events_after, "token": end_token}, } - async def get_all_new_events_stream( - self, from_id: int, current_id: int, limit: int, get_prev_content: bool = False - ) -> Tuple[int, List[EventBase], Dict[str, Optional[int]]]: + async def get_all_new_event_ids_stream( + self, + from_id: int, + current_id: int, + limit: int, + ) -> Tuple[int, Dict[str, Optional[int]]]: """Get all new events - Returns all events with from_id < stream_ordering <= current_id. + Returns all event ids with from_id < stream_ordering <= current_id. Args: from_id: the stream_ordering of the last event we processed current_id: the stream_ordering of the most recently processed event limit: the maximum number of events to return - get_prev_content: whether to fetch previous event content Returns: - A tuple of (next_id, events, event_to_received_ts), where `next_id` + A tuple of (next_id, event_to_received_ts), where `next_id` is the next value to pass as `from_id` (it will either be the stream_ordering of the last returned event, or, if fewer than `limit` events were found, the `current_id`). The `event_to_received_ts` is - a dictionary mapping event ID to the event `received_ts`. + a dictionary mapping event ID to the event `received_ts`, sorted by ascending + stream_ordering. """ - def get_all_new_events_stream_txn( + def get_all_new_event_ids_stream_txn( txn: LoggingTransaction, ) -> Tuple[int, Dict[str, Optional[int]]]: sql = ( @@ -1070,15 +1092,10 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): return upper_bound, event_to_received_ts upper_bound, event_to_received_ts = await self.db_pool.runInteraction( - "get_all_new_events_stream", get_all_new_events_stream_txn - ) - - events = await self.get_events_as_list( - event_to_received_ts.keys(), - get_prev_content=get_prev_content, + "get_all_new_event_ids_stream", get_all_new_event_ids_stream_txn ) - return upper_bound, events, event_to_received_ts + return upper_bound, event_to_received_ts async def get_federation_out_pos(self, typ: str) -> int: if self._need_to_reset_federation_stream_positions: @@ -1202,8 +1219,6 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): `to_token`), or `limit` is zero. """ - assert int(limit) >= 0 - # Tokens really represent positions between elements, but we use # the convention of pointing to the event before the gap. Hence # we have a bit of asymmetry when it comes to equalities. @@ -1282,8 +1297,8 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): # 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 there is a relation_senders and relation_types filter join to the + # relations table to get events related to the current event. if event_filter and ( event_filter.related_by_senders or event_filter.related_by_rel_types ): @@ -1298,6 +1313,13 @@ class StreamWorkerStore(EventsWorkerStore, SQLBaseStore): LEFT JOIN events AS related_event ON (relation.event_id = related_event.event_id) """ + # If there is a not_rel_types filter join to the relations table to get + # the event's relation information. + if event_filter and (event_filter.rel_types or event_filter.not_rel_types): + join_clause += """ + LEFT JOIN event_relations AS event_relation USING (event_id) + """ + if needs_distinct: select_keywords += " DISTINCT" diff --git a/synapse/storage/databases/main/user_directory.py b/synapse/storage/databases/main/user_directory.py index ddb25b5cea..14ef5b040d 100644 --- a/synapse/storage/databases/main/user_directory.py +++ b/synapse/storage/databases/main/user_directory.py @@ -26,6 +26,14 @@ from typing import ( cast, ) +try: + # Figure out if ICU support is available for searching users. + import icu + + USE_ICU = True +except ModuleNotFoundError: + USE_ICU = False + from typing_extensions import TypedDict from synapse.api.errors import StoreError @@ -185,9 +193,8 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore): - who should be in the user_directory. Args: - progress (dict) - batch_size (int): Maximum number of state events to process - per cycle. + progress + batch_size: Maximum number of state events to process per cycle. Returns: number of events processed. @@ -482,7 +489,6 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore): table="user_directory", keyvalues={"user_id": user_id}, values={"display_name": display_name, "avatar_url": avatar_url}, - lock=False, # We're only inserter ) if isinstance(self.database_engine, PostgresEngine): @@ -512,7 +518,6 @@ class UserDirectoryBackgroundUpdateStore(StateDeltasStore): table="user_directory_search", keyvalues={"user_id": user_id}, values={"value": value}, - lock=False, # We're only inserter ) else: # This should be unreachable. @@ -708,10 +713,10 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): Returns the rooms that a user is in. Args: - user_id(str): Must be a local user + user_id: Must be a local user Returns: - list: user_id + List of room IDs """ rows = await self.db_pool.simple_select_onecol( table="users_who_share_private_rooms", @@ -889,7 +894,7 @@ class UserDirectoryStore(UserDirectoryBackgroundUpdateStore): limited = len(results) > limit - return {"limited": limited, "results": results} + return {"limited": limited, "results": results[0:limit]} def _parse_query_sqlite(search_term: str) -> str: @@ -903,7 +908,7 @@ def _parse_query_sqlite(search_term: str) -> str: """ # Pull out the individual words, discarding any non-word characters. - results = re.findall(r"([\w\-]+)", search_term, re.UNICODE) + results = _parse_words(search_term) return " & ".join("(%s* OR %s)" % (result, result) for result in results) @@ -913,12 +918,63 @@ def _parse_query_postgres(search_term: str) -> Tuple[str, str, str]: We use this so that we can add prefix matching, which isn't something that is supported by default. """ - - # Pull out the individual words, discarding any non-word characters. - results = re.findall(r"([\w\-]+)", search_term, re.UNICODE) + results = _parse_words(search_term) both = " & ".join("(%s:* | %s)" % (result, result) for result in results) exact = " & ".join("%s" % (result,) for result in results) prefix = " & ".join("%s:*" % (result,) for result in results) return both, exact, prefix + + +def _parse_words(search_term: str) -> List[str]: + """Split the provided search string into a list of its words. + + If support for ICU (International Components for Unicode) is available, use it. + Otherwise, fall back to using a regex to detect word boundaries. This latter + solution works well enough for most latin-based languages, but doesn't work as well + with other languages. + + Args: + search_term: The search string. + + Returns: + A list of the words in the search string. + """ + if USE_ICU: + return _parse_words_with_icu(search_term) + + return re.findall(r"([\w\-]+)", search_term, re.UNICODE) + + +def _parse_words_with_icu(search_term: str) -> List[str]: + """Break down the provided search string into its individual words using ICU + (International Components for Unicode). + + Args: + search_term: The search string. + + Returns: + A list of the words in the search string. + """ + results = [] + breaker = icu.BreakIterator.createWordInstance(icu.Locale.getDefault()) + breaker.setText(search_term) + i = 0 + while True: + j = breaker.nextBoundary() + if j < 0: + break + + result = search_term[i:j] + + # libicu considers spaces and punctuation between words as words, but we don't + # want to include those in results as they would result in syntax errors in SQL + # queries (e.g. "foo bar" would result in the search query including "foo & & + # bar"). + if len(re.findall(r"([\w\-]+)", result, re.UNICODE)): + results.append(result) + + i = j + + return results diff --git a/synapse/storage/databases/state/bg_updates.py b/synapse/storage/databases/state/bg_updates.py index a7fcc564a9..d743282f13 100644 --- a/synapse/storage/databases/state/bg_updates.py +++ b/synapse/storage/databases/state/bg_updates.py @@ -22,8 +22,8 @@ from synapse.storage.database import ( LoggingTransaction, ) from synapse.storage.engines import PostgresEngine -from synapse.storage.state import StateFilter from synapse.types import MutableStateMap, StateMap +from synapse.types.state import StateFilter from synapse.util.caches import intern_string if TYPE_CHECKING: @@ -93,13 +93,6 @@ class StateGroupBackgroundUpdateStore(SQLBaseStore): results: Dict[int, MutableStateMap[str]] = {group: {} for group in groups} - where_clause, where_args = state_filter.make_sql_filter_clause() - - # Unless the filter clause is empty, we're going to append it after an - # existing where clause - if where_clause: - where_clause = " AND (%s)" % (where_clause,) - if isinstance(self.database_engine, PostgresEngine): # Temporarily disable sequential scans in this transaction. This is # a temporary hack until we can add the right indices in @@ -110,31 +103,91 @@ class StateGroupBackgroundUpdateStore(SQLBaseStore): # against `state_groups_state` to fetch the latest state. # It assumes that previous state groups are always numerically # lesser. - # The PARTITION is used to get the event_id in the greatest state - # group for the given type, state_key. # This may return multiple rows per (type, state_key), but last_value # should be the same. sql = """ - WITH RECURSIVE state(state_group) AS ( + WITH RECURSIVE sgs(state_group) AS ( VALUES(?::bigint) UNION ALL - SELECT prev_state_group FROM state_group_edges e, state s + SELECT prev_state_group FROM state_group_edges e, sgs s WHERE s.state_group = e.state_group ) - SELECT DISTINCT ON (type, state_key) - type, state_key, event_id - FROM state_groups_state - WHERE state_group IN ( - SELECT state_group FROM state - ) %s - ORDER BY type, state_key, state_group DESC + %s """ + overall_select_query_args: List[Union[int, str]] = [] + + # This is an optimization to create a select clause per-condition. This + # makes the query planner a lot smarter on what rows should pull out in the + # first place and we end up with something that takes 10x less time to get a + # result. + use_condition_optimization = ( + not state_filter.include_others and not state_filter.is_full() + ) + state_filter_condition_combos: List[Tuple[str, Optional[str]]] = [] + # We don't need to caclculate this list if we're not using the condition + # optimization + if use_condition_optimization: + for etype, state_keys in state_filter.types.items(): + if state_keys is None: + state_filter_condition_combos.append((etype, None)) + else: + for state_key in state_keys: + state_filter_condition_combos.append((etype, state_key)) + # And here is the optimization itself. We don't want to do the optimization + # if there are too many individual conditions. 10 is an arbitrary number + # with no testing behind it but we do know that we specifically made this + # optimization for when we grab the necessary state out for + # `filter_events_for_client` which just uses 2 conditions + # (`EventTypes.RoomHistoryVisibility` and `EventTypes.Member`). + if use_condition_optimization and len(state_filter_condition_combos) < 10: + select_clause_list: List[str] = [] + for etype, skey in state_filter_condition_combos: + if skey is None: + where_clause = "(type = ?)" + overall_select_query_args.extend([etype]) + else: + where_clause = "(type = ? AND state_key = ?)" + overall_select_query_args.extend([etype, skey]) + + select_clause_list.append( + f""" + ( + SELECT DISTINCT ON (type, state_key) + type, state_key, event_id + FROM state_groups_state + INNER JOIN sgs USING (state_group) + WHERE {where_clause} + ORDER BY type, state_key, state_group DESC + ) + """ + ) + + overall_select_clause = " UNION ".join(select_clause_list) + else: + where_clause, where_args = state_filter.make_sql_filter_clause() + # Unless the filter clause is empty, we're going to append it after an + # existing where clause + if where_clause: + where_clause = " AND (%s)" % (where_clause,) + + overall_select_query_args.extend(where_args) + + overall_select_clause = f""" + SELECT DISTINCT ON (type, state_key) + type, state_key, event_id + FROM state_groups_state + WHERE state_group IN ( + SELECT state_group FROM sgs + ) {where_clause} + ORDER BY type, state_key, state_group DESC + """ + for group in groups: args: List[Union[int, str]] = [group] - args.extend(where_args) + args.extend(overall_select_query_args) - txn.execute(sql % (where_clause,), args) + txn.execute(sql % (overall_select_clause,), args) for row in txn: typ, state_key, event_id = row key = (intern_string(typ), intern_string(state_key)) @@ -142,6 +195,12 @@ class StateGroupBackgroundUpdateStore(SQLBaseStore): else: max_entries_returned = state_filter.max_entries_returned() + where_clause, where_args = state_filter.make_sql_filter_clause() + # Unless the filter clause is empty, we're going to append it after an + # existing where clause + if where_clause: + where_clause = " AND (%s)" % (where_clause,) + # We don't use WITH RECURSIVE on sqlite3 as there are distributions # that ship with an sqlite3 version that doesn't support it (e.g. wheezy) for group in groups: diff --git a/synapse/storage/databases/state/store.py b/synapse/storage/databases/state/store.py index f8cfcaca83..1a7232b276 100644 --- a/synapse/storage/databases/state/store.py +++ b/synapse/storage/databases/state/store.py @@ -25,10 +25,10 @@ from synapse.storage.database import ( LoggingTransaction, ) from synapse.storage.databases.state.bg_updates import StateBackgroundUpdateStore -from synapse.storage.state import StateFilter from synapse.storage.types import Cursor from synapse.storage.util.sequence import build_sequence_generator from synapse.types import MutableStateMap, StateKey, StateMap +from synapse.types.state import StateFilter from synapse.util.caches.descriptors import cached from synapse.util.caches.dictionary_cache import DictionaryCache from synapse.util.cancellation import cancellable diff --git a/synapse/storage/engines/postgres.py b/synapse/storage/engines/postgres.py index d8c0f64d9a..f9f562ea45 100644 --- a/synapse/storage/engines/postgres.py +++ b/synapse/storage/engines/postgres.py @@ -77,12 +77,12 @@ class PostgresEngine( # docs: The number is formed by converting the major, minor, and # revision numbers into two-decimal-digit numbers and appending them # together. For example, version 8.1.5 will be returned as 80105 - self._version = cast(int, db_conn.server_version) + self._version = db_conn.server_version allow_unsafe_locale = self.config.get("allow_unsafe_locale", False) # Are we on a supported PostgreSQL version? - if not allow_outdated_version and self._version < 100000: - raise RuntimeError("Synapse requires PostgreSQL 10 or above.") + if not allow_outdated_version and self._version < 110000: + raise RuntimeError("Synapse requires PostgreSQL 11 or above.") with db_conn.cursor() as txn: txn.execute("SHOW SERVER_ENCODING") diff --git a/synapse/storage/engines/sqlite.py b/synapse/storage/engines/sqlite.py index faa574dbfd..14260442b6 100644 --- a/synapse/storage/engines/sqlite.py +++ b/synapse/storage/engines/sqlite.py @@ -88,6 +88,10 @@ class Sqlite3Engine(BaseDatabaseEngine[sqlite3.Connection, sqlite3.Cursor]): db_conn.create_function("rank", 1, _rank) db_conn.execute("PRAGMA foreign_keys = ON;") + + # Enable WAL. + # see https://www.sqlite.org/wal.html + db_conn.execute("PRAGMA journal_mode = WAL;") db_conn.commit() def is_deadlock(self, error: Exception) -> bool: diff --git a/synapse/storage/schema/main/delta/73/06thread_notifications_backfill.sql b/synapse/storage/schema/main/delta/73/06thread_notifications_backfill.sql deleted file mode 100644 index 0ffde9bbeb..0000000000 --- a/synapse/storage/schema/main/delta/73/06thread_notifications_backfill.sql +++ /dev/null @@ -1,29 +0,0 @@ -/* Copyright 2022 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. - */ - --- Forces the background updates from 06thread_notifications.sql to run in the --- foreground as code will now require those to be "done". - -DELETE FROM background_updates WHERE update_name = 'event_push_backfill_thread_id'; - --- Overwrite any null thread_id columns. -UPDATE event_push_actions_staging SET thread_id = 'main' WHERE thread_id IS NULL; -UPDATE event_push_actions SET thread_id = 'main' WHERE thread_id IS NULL; -UPDATE event_push_summary SET thread_id = 'main' WHERE thread_id IS NULL; - --- Do not run the event_push_summary_unique_index job if it is pending; the --- thread_id field will be made required. -DELETE FROM background_updates WHERE update_name = 'event_push_summary_unique_index'; -DROP INDEX IF EXISTS event_push_summary_unique_index; diff --git a/synapse/storage/schema/main/delta/73/06thread_notifications_thread_id_idx.sql b/synapse/storage/schema/main/delta/73/06thread_notifications_thread_id_idx.sql new file mode 100644 index 0000000000..8b3c636594 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/06thread_notifications_thread_id_idx.sql @@ -0,0 +1,23 @@ +/* Copyright 2022 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. + */ + +-- Allow there to be multiple summaries per user/room. +DROP INDEX IF EXISTS event_push_summary_unique_index; + +INSERT INTO background_updates (ordering, update_name, progress_json, depends_on) VALUES + (7306, 'event_push_actions_thread_id_null', '{}', 'event_push_backfill_thread_id'); + +INSERT INTO background_updates (ordering, update_name, progress_json, depends_on) VALUES + (7306, 'event_push_summary_thread_id_null', '{}', 'event_push_backfill_thread_id'); diff --git a/synapse/storage/schema/main/delta/73/07thread_notifications_not_null.sql.sqlite b/synapse/storage/schema/main/delta/73/07thread_notifications_not_null.sql.sqlite deleted file mode 100644 index 5322ad77a4..0000000000 --- a/synapse/storage/schema/main/delta/73/07thread_notifications_not_null.sql.sqlite +++ /dev/null @@ -1,101 +0,0 @@ -/* Copyright 2022 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. - */ - --- SQLite doesn't support modifying columns to an existing table, so it must --- be recreated. - --- Create the new tables. -CREATE TABLE event_push_actions_staging_new ( - event_id TEXT NOT NULL, - user_id TEXT NOT NULL, - actions TEXT NOT NULL, - notif SMALLINT NOT NULL, - highlight SMALLINT NOT NULL, - unread SMALLINT, - thread_id TEXT NOT NULL, - inserted_ts BIGINT -); - -CREATE TABLE event_push_actions_new ( - room_id TEXT NOT NULL, - event_id TEXT NOT NULL, - user_id TEXT NOT NULL, - profile_tag VARCHAR(32), - actions TEXT NOT NULL, - topological_ordering BIGINT, - stream_ordering BIGINT, - notif SMALLINT, - highlight SMALLINT, - unread SMALLINT, - thread_id TEXT NOT NULL, - CONSTRAINT event_id_user_id_profile_tag_uniqueness UNIQUE (room_id, event_id, user_id, profile_tag) -); - -CREATE TABLE event_push_summary_new ( - user_id TEXT NOT NULL, - room_id TEXT NOT NULL, - notif_count BIGINT NOT NULL, - stream_ordering BIGINT NOT NULL, - unread_count BIGINT, - last_receipt_stream_ordering BIGINT, - thread_id TEXT NOT NULL -); - --- Swap the indexes. -DROP INDEX IF EXISTS event_push_actions_staging_id; -CREATE INDEX event_push_actions_staging_id ON event_push_actions_staging_new(event_id); - -DROP INDEX IF EXISTS event_push_actions_room_id_user_id; -DROP INDEX IF EXISTS event_push_actions_rm_tokens; -DROP INDEX IF EXISTS event_push_actions_stream_ordering; -DROP INDEX IF EXISTS event_push_actions_u_highlight; -DROP INDEX IF EXISTS event_push_actions_highlights_index; -CREATE INDEX event_push_actions_room_id_user_id on event_push_actions_new(room_id, user_id); -CREATE INDEX event_push_actions_rm_tokens on event_push_actions_new( user_id, room_id, topological_ordering, stream_ordering ); -CREATE INDEX event_push_actions_stream_ordering on event_push_actions_new( stream_ordering, user_id ); -CREATE INDEX event_push_actions_u_highlight ON event_push_actions_new (user_id, stream_ordering); -CREATE INDEX event_push_actions_highlights_index ON event_push_actions_new (user_id, room_id, topological_ordering, stream_ordering); - --- Copy the data. -INSERT INTO event_push_actions_staging_new (event_id, user_id, actions, notif, highlight, unread, thread_id, inserted_ts) - SELECT event_id, user_id, actions, notif, highlight, unread, thread_id, inserted_ts - FROM event_push_actions_staging; - -INSERT INTO event_push_actions_new (room_id, event_id, user_id, profile_tag, actions, topological_ordering, stream_ordering, notif, highlight, unread, thread_id) - SELECT room_id, event_id, user_id, profile_tag, actions, topological_ordering, stream_ordering, notif, highlight, unread, thread_id - FROM event_push_actions; - -INSERT INTO event_push_summary_new (user_id, room_id, notif_count, stream_ordering, unread_count, last_receipt_stream_ordering, thread_id) - SELECT user_id, room_id, notif_count, stream_ordering, unread_count, last_receipt_stream_ordering, thread_id - FROM event_push_summary; - --- Drop the old tables. -DROP TABLE event_push_actions_staging; -DROP TABLE event_push_actions; -DROP TABLE event_push_summary; - --- Rename the tables. -ALTER TABLE event_push_actions_staging_new RENAME TO event_push_actions_staging; -ALTER TABLE event_push_actions_new RENAME TO event_push_actions; -ALTER TABLE event_push_summary_new RENAME TO event_push_summary; - --- Re-run background updates from 72/02event_push_actions_index.sql and --- 72/06thread_notifications.sql. -INSERT INTO background_updates (ordering, update_name, progress_json) VALUES - (7307, 'event_push_summary_unique_index2', '{}') - ON CONFLICT (update_name) DO NOTHING; -INSERT INTO background_updates (ordering, update_name, progress_json) VALUES - (7307, 'event_push_actions_stream_highlight_index', '{}') - ON CONFLICT (update_name) DO NOTHING; diff --git a/synapse/storage/schema/main/delta/73/07thread_notifications_not_null.sql.postgres b/synapse/storage/schema/main/delta/73/09partial_joined_via_destination.sql index 33674f8c62..066d602b18 100644 --- a/synapse/storage/schema/main/delta/73/07thread_notifications_not_null.sql.postgres +++ b/synapse/storage/schema/main/delta/73/09partial_joined_via_destination.sql @@ -13,7 +13,6 @@ * limitations under the License. */ --- The columns can now be made non-nullable. -ALTER TABLE event_push_actions_staging ALTER COLUMN thread_id SET NOT NULL; -ALTER TABLE event_push_actions ALTER COLUMN thread_id SET NOT NULL; -ALTER TABLE event_push_summary ALTER COLUMN thread_id SET NOT NULL; +-- When we resync partial state, we prioritise doing so using the server we +-- partial-joined from. To do this we need to record which server that was! +ALTER TABLE partial_state_rooms ADD COLUMN joined_via TEXT; diff --git a/synapse/storage/schema/main/delta/73/09threads_table.sql b/synapse/storage/schema/main/delta/73/09threads_table.sql new file mode 100644 index 0000000000..aa7c5e9a2e --- /dev/null +++ b/synapse/storage/schema/main/delta/73/09threads_table.sql @@ -0,0 +1,30 @@ +/* Copyright 2022 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 threads ( + room_id TEXT NOT NULL, + -- The event ID of the root event in the thread. + thread_id TEXT NOT NULL, + -- The latest event ID and corresponding topo / stream ordering. + latest_event_id TEXT NOT NULL, + topological_ordering BIGINT NOT NULL, + stream_ordering BIGINT NOT NULL, + CONSTRAINT threads_uniqueness UNIQUE (room_id, thread_id) +); + +CREATE INDEX threads_ordering_idx ON threads(room_id, topological_ordering, stream_ordering); + +INSERT INTO background_updates (ordering, update_name, progress_json) VALUES + (7309, 'threads_backfill', '{}'); diff --git a/synapse/storage/schema/main/delta/73/10_update_sqlite_fts4_tokenizer.py b/synapse/storage/schema/main/delta/73/10_update_sqlite_fts4_tokenizer.py new file mode 100644 index 0000000000..3de0a709eb --- /dev/null +++ b/synapse/storage/schema/main/delta/73/10_update_sqlite_fts4_tokenizer.py @@ -0,0 +1,62 @@ +# Copyright 2022 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 json + +from synapse.storage.engines import BaseDatabaseEngine, Sqlite3Engine +from synapse.storage.types import Cursor + + +def run_create(cur: Cursor, database_engine: BaseDatabaseEngine) -> None: + """ + Upgrade the event_search table to use the porter tokenizer if it isn't already + + Applies only for sqlite. + """ + if not isinstance(database_engine, Sqlite3Engine): + return + + # Rebuild the table event_search table with tokenize=porter configured. + cur.execute("DROP TABLE event_search") + cur.execute( + """ + CREATE VIRTUAL TABLE event_search + USING fts4 (tokenize=porter, event_id, room_id, sender, key, value ) + """ + ) + + # Re-run the background job to re-populate the event_search table. + cur.execute("SELECT MIN(stream_ordering) FROM events") + row = cur.fetchone() + min_stream_id = row[0] + + # If there are not any events, nothing to do. + if min_stream_id is None: + return + + cur.execute("SELECT MAX(stream_ordering) FROM events") + row = cur.fetchone() + max_stream_id = row[0] + + progress = { + "target_min_stream_id_inclusive": min_stream_id, + "max_stream_id_exclusive": max_stream_id + 1, + } + progress_json = json.dumps(progress) + + sql = """ + INSERT into background_updates (ordering, update_name, progress_json) + VALUES (?, ?, ?) + """ + + cur.execute(sql, (7310, "event_search", progress_json)) diff --git a/synapse/storage/schema/main/delta/73/10login_tokens.sql b/synapse/storage/schema/main/delta/73/10login_tokens.sql new file mode 100644 index 0000000000..a39b7bcece --- /dev/null +++ b/synapse/storage/schema/main/delta/73/10login_tokens.sql @@ -0,0 +1,35 @@ +/* + * Copyright 2022 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. + */ + +-- Login tokens are short-lived tokens that are used for the m.login.token +-- login method, mainly during SSO logins +CREATE TABLE login_tokens ( + token TEXT PRIMARY KEY, + user_id TEXT NOT NULL, + expiry_ts BIGINT NOT NULL, + used_ts BIGINT, + auth_provider_id TEXT, + auth_provider_session_id TEXT +); + +-- We're sometimes querying them by their session ID we got from their IDP +CREATE INDEX login_tokens_auth_provider_idx + ON login_tokens (auth_provider_id, auth_provider_session_id); + +-- We're deleting them by their expiration time +CREATE INDEX login_tokens_expiry_time_idx + ON login_tokens (expiry_ts); + diff --git a/synapse/storage/schema/main/delta/73/11event_search_room_id_n_distinct.sql.postgres b/synapse/storage/schema/main/delta/73/11event_search_room_id_n_distinct.sql.postgres new file mode 100644 index 0000000000..93cdaefca1 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/11event_search_room_id_n_distinct.sql.postgres @@ -0,0 +1,33 @@ +/* Copyright 2022 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. + */ + + +-- By default the postgres statistics collector massively underestimates the +-- number of distinct rooms in `event_search`, which can cause postgres to use +-- table scans for queries for multiple rooms. +-- +-- To work around this we can manually tell postgres the number of distinct rooms +-- by setting `n_distinct` (a negative value here is the number of distinct values +-- divided by the number of rows, so -0.01 means on average there are 100 rows per +-- distinct value). We don't need a particularly accurate number here, as a) we just +-- want it to always use index scans and b) our estimate is going to be better than the +-- one made by the statistics collector. + +ALTER TABLE event_search ALTER COLUMN room_id SET (n_distinct = -0.01); + +-- Ideally we'd do an `ANALYZE event_search (room_id)` here so that +-- the above gets picked up immediately, but that can take a bit of time so we +-- rely on the autovacuum eventually getting run and doing that in the +-- background for us. diff --git a/synapse/storage/schema/main/delta/73/12refactor_device_list_outbound_pokes.sql b/synapse/storage/schema/main/delta/73/12refactor_device_list_outbound_pokes.sql new file mode 100644 index 0000000000..93d7fcb79b --- /dev/null +++ b/synapse/storage/schema/main/delta/73/12refactor_device_list_outbound_pokes.sql @@ -0,0 +1,53 @@ +/* Copyright 2022 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. + */ + +-- Prior to this schema delta, we tracked the set of unconverted rows in +-- `device_lists_changes_in_room` using the `converted_to_destinations` flag. When rows +-- were converted to `device_lists_outbound_pokes`, the `converted_to_destinations` flag +-- would be set. +-- +-- After this schema delta, the `converted_to_destinations` is still populated like +-- before, but the set of unconverted rows is determined by the `stream_id` in the new +-- `device_lists_changes_converted_stream_position` table. +-- +-- If rolled back, Synapse will re-send all device list changes that happened since the +-- schema delta. + +CREATE TABLE IF NOT EXISTS device_lists_changes_converted_stream_position( + Lock CHAR(1) NOT NULL DEFAULT 'X' UNIQUE, -- Makes sure this table only has one row. + -- The (stream id, room id) of the last row in `device_lists_changes_in_room` that + -- has been converted to `device_lists_outbound_pokes`. Rows with a strictly larger + -- (stream id, room id) where `converted_to_destinations` is `FALSE` have not been + -- converted. + stream_id BIGINT NOT NULL, + -- `room_id` may be an empty string, which compares less than all valid room IDs. + room_id TEXT NOT NULL, + CHECK (Lock='X') +); + +INSERT INTO device_lists_changes_converted_stream_position (stream_id, room_id) VALUES ( + ( + SELECT COALESCE( + -- The last converted stream id is the smallest unconverted stream id minus + -- one. + MIN(stream_id) - 1, + -- If there is no unconverted stream id, the last converted stream id is the + -- largest stream id. + -- Otherwise, pick 1, since stream ids start at 2. + (SELECT COALESCE(MAX(stream_id), 1) FROM device_lists_changes_in_room) + ) FROM device_lists_changes_in_room WHERE NOT converted_to_destinations + ), + '' +); diff --git a/synapse/storage/schema/main/delta/73/13add_device_lists_index.sql b/synapse/storage/schema/main/delta/73/13add_device_lists_index.sql new file mode 100644 index 0000000000..3725022a13 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/13add_device_lists_index.sql @@ -0,0 +1,20 @@ +/* Copyright 2022 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. + */ + + +-- Adds an index on `device_lists_changes_in_room (room_id, stream_id)`, which +-- speeds up `/sync` queries. +INSERT INTO background_updates (ordering, update_name, progress_json) VALUES + (7313, 'device_lists_changes_in_room_by_room_index', '{}'); diff --git a/synapse/storage/schema/main/delta/73/20_un_partial_stated_room_stream.sql b/synapse/storage/schema/main/delta/73/20_un_partial_stated_room_stream.sql new file mode 100644 index 0000000000..743196cfe3 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/20_un_partial_stated_room_stream.sql @@ -0,0 +1,32 @@ +/* Copyright 2022 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. + */ + +-- Stream for notifying that a room has become un-partial-stated. +CREATE TABLE un_partial_stated_room_stream( + -- Position in the stream + stream_id BIGINT PRIMARY KEY NOT NULL, + + -- Which instance wrote this entry. + instance_name TEXT NOT NULL, + + -- Which room has been un-partial-stated. + room_id TEXT NOT NULL REFERENCES rooms(room_id) ON DELETE CASCADE +); + +-- We want an index here because of the foreign key constraint: +-- upon deleting a room, the database needs to be able to check here. +-- This index is not unique because we can join a room multiple times in a server's lifetime, +-- so the same room could be un-partial-stated multiple times! +CREATE INDEX un_partial_stated_room_stream_room_id ON un_partial_stated_room_stream (room_id); diff --git a/synapse/storage/schema/main/delta/73/21_un_partial_stated_room_stream_seq.sql.postgres b/synapse/storage/schema/main/delta/73/21_un_partial_stated_room_stream_seq.sql.postgres new file mode 100644 index 0000000000..c1aac0b385 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/21_un_partial_stated_room_stream_seq.sql.postgres @@ -0,0 +1,20 @@ +/* Copyright 2022 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 SEQUENCE IF NOT EXISTS un_partial_stated_room_stream_sequence; + +SELECT setval('un_partial_stated_room_stream_sequence', ( + SELECT COALESCE(MAX(stream_id), 1) FROM un_partial_stated_room_stream +)); diff --git a/synapse/storage/schema/main/delta/73/22_rebuild_user_dir_stats.sql b/synapse/storage/schema/main/delta/73/22_rebuild_user_dir_stats.sql new file mode 100644 index 0000000000..afab1e4bb7 --- /dev/null +++ b/synapse/storage/schema/main/delta/73/22_rebuild_user_dir_stats.sql @@ -0,0 +1,29 @@ +/* Copyright 2022 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. + */ + +INSERT INTO background_updates (ordering, update_name, progress_json, depends_on) VALUES + -- Set up user directory staging tables. + (7322, 'populate_user_directory_createtables', '{}', NULL), + -- Run through each room and update the user directory according to who is in it. + (7322, 'populate_user_directory_process_rooms', '{}', 'populate_user_directory_createtables'), + -- Insert all users into the user directory, if search_all_users is on. + (7322, 'populate_user_directory_process_users', '{}', 'populate_user_directory_process_rooms'), + -- Clean up user directory staging tables. + (7322, 'populate_user_directory_cleanup', '{}', 'populate_user_directory_process_users'), + -- Rebuild the room_stats_current and room_stats_state tables. + (7322, 'populate_stats_process_rooms', '{}', NULL), + -- Update the user_stats_current table. + (7322, 'populate_stats_process_users', '{}', NULL) +ON CONFLICT (update_name) DO NOTHING; diff --git a/synapse/storage/util/id_generators.py b/synapse/storage/util/id_generators.py index 2dfe4c0b66..0d7108f01b 100644 --- a/synapse/storage/util/id_generators.py +++ b/synapse/storage/util/id_generators.py @@ -186,11 +186,13 @@ class StreamIdGenerator(AbstractStreamIdGenerator): column: str, extra_tables: Iterable[Tuple[str, str]] = (), step: int = 1, + is_writer: bool = True, ) -> None: assert step != 0 self._lock = threading.Lock() self._step: int = step self._current: int = _load_current_id(db_conn, table, column, step) + self._is_writer = is_writer for table, column in extra_tables: self._current = (max if step > 0 else min)( self._current, _load_current_id(db_conn, table, column, step) @@ -204,9 +206,11 @@ class StreamIdGenerator(AbstractStreamIdGenerator): self._unfinished_ids: OrderedDict[int, int] = OrderedDict() def advance(self, instance_name: str, new_id: int) -> None: - # `StreamIdGenerator` should only be used when there is a single writer, - # so replication should never happen. - raise Exception("Replication is not supported by StreamIdGenerator") + # Advance should never be called on a writer instance, only over replication + if self._is_writer: + raise Exception("Replication is not supported by writer StreamIdGenerator") + + self._current = (max if self._step > 0 else min)(self._current, new_id) def get_next(self) -> AsyncContextManager[int]: with self._lock: @@ -249,6 +253,9 @@ class StreamIdGenerator(AbstractStreamIdGenerator): return _AsyncCtxManagerWrapper(manager()) def get_current_token(self) -> int: + if not self._is_writer: + return self._current + with self._lock: if self._unfinished_ids: return next(iter(self._unfinished_ids)) - self._step diff --git a/synapse/streams/__init__.py b/synapse/streams/__init__.py index 806b671305..2dcd43d0a2 100644 --- a/synapse/streams/__init__.py +++ b/synapse/streams/__init__.py @@ -27,7 +27,7 @@ class EventSource(Generic[K, R]): self, user: UserID, from_key: K, - limit: Optional[int], + limit: int, room_ids: Collection[str], is_guest: bool, explicit_room_id: Optional[str] = None, diff --git a/synapse/streams/config.py b/synapse/streams/config.py index b52723e2b8..6df2de919c 100644 --- a/synapse/streams/config.py +++ b/synapse/streams/config.py @@ -35,17 +35,19 @@ class PaginationConfig: from_token: Optional[StreamToken] to_token: Optional[StreamToken] direction: str - limit: Optional[int] + limit: int @classmethod async def from_request( cls, store: "DataStore", request: SynapseRequest, - raise_invalid_params: bool = True, - default_limit: Optional[int] = None, + default_limit: int, + default_dir: str = "f", ) -> "PaginationConfig": - direction = parse_string(request, "dir", default="f", allowed_values=["f", "b"]) + direction = parse_string( + request, "dir", default=default_dir, allowed_values=["f", "b"] + ) from_tok_str = parse_string(request, "from") to_tok_str = parse_string(request, "to") @@ -67,12 +69,10 @@ class PaginationConfig: raise SynapseError(400, "'to' parameter is invalid") limit = parse_integer(request, "limit", default=default_limit) + if limit < 0: + raise SynapseError(400, "Limit must be 0 or above") - if limit: - if limit < 0: - raise SynapseError(400, "Limit must be 0 or above") - - limit = min(int(limit), MAX_LIMIT) + limit = min(limit, MAX_LIMIT) try: return PaginationConfig(from_tok, to_tok, direction, limit) diff --git a/synapse/streams/events.py b/synapse/streams/events.py index bcd840bd88..619eb7f601 100644 --- a/synapse/streams/events.py +++ b/synapse/streams/events.py @@ -45,9 +45,12 @@ class _EventSourcesInner: class EventSources: def __init__(self, hs: "HomeServer"): self.sources = _EventSourcesInner( - # mypy thinks attribute.type is `Optional`, but we know it's never `None` here since - # all the attributes of `_EventSourcesInner` are annotated. - *(attribute.type(hs) for attribute in attr.fields(_EventSourcesInner)) # type: ignore[misc] + # mypy previously warned that attribute.type is `Optional`, but we know it's + # never `None` here since all the attributes of `_EventSourcesInner` are + # annotated. + # As of the stubs in attrs 22.1.0, `attr.fields()` now returns Any, + # so the call to `attribute.type` is not checked. + *(attribute.type(hs) for attribute in attr.fields(_EventSourcesInner)) ) self.store = hs.get_datastores().main @@ -71,6 +74,19 @@ class EventSources: return token @trace + async def get_start_token_for_pagination(self, room_id: str) -> StreamToken: + """Get the start token for a given room to be used to paginate + events. + + The returned token does not have the current values for fields other + than `room`, since they are not used during pagination. + + Returns: + The start token for pagination. + """ + return StreamToken.START + + @trace async def get_current_token_for_pagination(self, room_id: str) -> StreamToken: """Get the current token for a given room to be used to paginate events. diff --git a/synapse/types.py b/synapse/types/__init__.py index 773f0438d5..f2d436ddc3 100644 --- a/synapse/types.py +++ b/synapse/types/__init__.py @@ -143,8 +143,8 @@ class Requester: Requester. Args: - store (DataStore): Used to convert AS ID to AS object - input (dict): A dict produced by `serialize` + store: Used to convert AS ID to AS object + input: A dict produced by `serialize` Returns: Requester diff --git a/synapse/storage/state.py b/synapse/types/state.py index 0004d955b4..743a4f9217 100644 --- a/synapse/storage/state.py +++ b/synapse/types/state.py @@ -118,6 +118,15 @@ class StateFilter: ) ) + def to_types(self) -> Iterable[Tuple[str, Optional[str]]]: + """The inverse to `from_types`.""" + for (event_type, state_keys) in self.types.items(): + if state_keys is None: + yield event_type, None + else: + for state_key in state_keys: + yield event_type, state_key + @staticmethod def from_lazy_load_member_list(members: Iterable[str]) -> "StateFilter": """Creates a filter that returns all non-member events, plus the member @@ -343,6 +352,15 @@ class StateFilter: for s in state_keys ] + def wildcard_types(self) -> List[str]: + """Returns a list of event types which require us to fetch all state keys. + This will be empty unless `has_wildcards` returns True. + + Returns: + A list of event types. + """ + return [t for t, state_keys in self.types.items() if state_keys is None] + def get_member_split(self) -> Tuple["StateFilter", "StateFilter"]: """Return the filter split into two: one which assumes it's exclusively matching against member state, and one which assumes it's matching diff --git a/synapse/util/async_helpers.py b/synapse/util/async_helpers.py index 7f1d41eb3c..d24c4f68c4 100644 --- a/synapse/util/async_helpers.py +++ b/synapse/util/async_helpers.py @@ -217,7 +217,8 @@ async def concurrently_execute( limit: Maximum number of conccurent executions. Returns: - Deferred: Resolved when all function invocations have finished. + None, when all function invocations have finished. The return values + from those functions are discarded. """ it = iter(args) diff --git a/synapse/util/caches/__init__.py b/synapse/util/caches/__init__.py index f7c3a6794e..9387632d0d 100644 --- a/synapse/util/caches/__init__.py +++ b/synapse/util/caches/__init__.py @@ -197,7 +197,7 @@ def register_cache( resize_callback: A function which can be called to resize the cache. Returns: - CacheMetric: an object which provides inc_{hits,misses,evictions} methods + an object which provides inc_{hits,misses,evictions} methods """ if resizable: if not resize_callback: diff --git a/synapse/util/caches/deferred_cache.py b/synapse/util/caches/deferred_cache.py index 6425f851ea..bf7bd351e0 100644 --- a/synapse/util/caches/deferred_cache.py +++ b/synapse/util/caches/deferred_cache.py @@ -153,7 +153,7 @@ class DeferredCache(Generic[KT, VT]): Args: key: callback: Gets called when the entry in the cache is invalidated - update_metrics (bool): whether to update the cache hit rate metrics + update_metrics: whether to update the cache hit rate metrics Returns: A Deferred which completes with the result. Note that this may later fail @@ -395,8 +395,8 @@ class DeferredCache(Generic[KT, VT]): # _pending_deferred_cache.pop should either return a CacheEntry, or, in the # case of a TreeCache, a dict of keys to cache entries. Either way calling # iterate_tree_cache_entry on it will do the right thing. - for entry in iterate_tree_cache_entry(entry): - for cb in entry.get_invalidation_callbacks(key): + for iter_entry in iterate_tree_cache_entry(entry): + for cb in iter_entry.get_invalidation_callbacks(key): cb() def invalidate_all(self) -> None: diff --git a/synapse/util/caches/descriptors.py b/synapse/util/caches/descriptors.py index 0391966462..72227359b9 100644 --- a/synapse/util/caches/descriptors.py +++ b/synapse/util/caches/descriptors.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 enum import functools import inspect import logging @@ -146,109 +145,6 @@ class _CacheDescriptorBase: ) -class _LruCachedFunction(Generic[F]): - cache: LruCache[CacheKey, Any] - __call__: F - - -def lru_cache( - *, max_entries: int = 1000, cache_context: bool = False -) -> Callable[[F], _LruCachedFunction[F]]: - """A method decorator that applies a memoizing cache around the function. - - This is more-or-less a drop-in equivalent to functools.lru_cache, although note - that the signature is slightly different. - - The main differences with functools.lru_cache are: - (a) the size of the cache can be controlled via the cache_factor mechanism - (b) the wrapped function can request a "cache_context" which provides a - callback mechanism to indicate that the result is no longer valid - (c) prometheus metrics are exposed automatically. - - The function should take zero or more arguments, which are used as the key for the - cache. Single-argument functions use that argument as the cache key; otherwise the - arguments are built into a tuple. - - Cached functions can be "chained" (i.e. a cached function can call other cached - functions and get appropriately invalidated when they called caches are - invalidated) by adding a special "cache_context" argument to the function - and passing that as a kwarg to all caches called. For example: - - @lru_cache(cache_context=True) - def foo(self, key, cache_context): - r1 = self.bar1(key, on_invalidate=cache_context.invalidate) - r2 = self.bar2(key, on_invalidate=cache_context.invalidate) - return r1 + r2 - - The wrapped function also has a 'cache' property which offers direct access to the - underlying LruCache. - """ - - def func(orig: F) -> _LruCachedFunction[F]: - desc = LruCacheDescriptor( - orig, - max_entries=max_entries, - cache_context=cache_context, - ) - return cast(_LruCachedFunction[F], desc) - - return func - - -class LruCacheDescriptor(_CacheDescriptorBase): - """Helper for @lru_cache""" - - class _Sentinel(enum.Enum): - sentinel = object() - - def __init__( - self, - orig: Callable[..., Any], - max_entries: int = 1000, - cache_context: bool = False, - ): - super().__init__( - orig, num_args=None, uncached_args=None, cache_context=cache_context - ) - self.max_entries = max_entries - - def __get__(self, obj: Optional[Any], owner: Optional[Type]) -> Callable[..., Any]: - cache: LruCache[CacheKey, Any] = LruCache( - cache_name=self.name, - max_size=self.max_entries, - ) - - get_cache_key = self.cache_key_builder - sentinel = LruCacheDescriptor._Sentinel.sentinel - - @functools.wraps(self.orig) - def _wrapped(*args: Any, **kwargs: Any) -> Any: - invalidate_callback = kwargs.pop("on_invalidate", None) - callbacks = (invalidate_callback,) if invalidate_callback else () - - cache_key = get_cache_key(args, kwargs) - - ret = cache.get(cache_key, default=sentinel, callbacks=callbacks) - if ret != sentinel: - return ret - - # Add our own `cache_context` to argument list if the wrapped function - # has asked for one - if self.add_cache_context: - kwargs["cache_context"] = _CacheContext.get_instance(cache, cache_key) - - ret2 = self.orig(obj, *args, **kwargs) - cache.set(cache_key, ret2, callbacks=callbacks) - - return ret2 - - wrapped = cast(CachedFunction, _wrapped) - wrapped.cache = cache - obj.__dict__[self.name] = wrapped - - return wrapped - - class DeferredCacheDescriptor(_CacheDescriptorBase): """A method decorator that applies a memoizing cache around the function. @@ -432,7 +328,7 @@ class DeferredCacheListDescriptor(_CacheDescriptorBase): num_args = cached_method.num_args if num_args != self.num_args: - raise Exception( + raise TypeError( "Number of args (%s) does not match underlying cache_method_name=%s (%s)." % (self.num_args, self.cached_method_name, num_args) ) @@ -607,7 +503,7 @@ def cachedList( is specified as a list that is iterated through to lookup keys in the original cache. A new tuple consisting of the (deduplicated) keys that weren't in the cache gets passed to the original function, which is expected to results - in a map of key to value for each passed value. THe new results are stored in the + in a map of key to value for each passed value. The new results are stored in the original cache. Note that any missing values are cached as None. Args: diff --git a/synapse/util/caches/dictionary_cache.py b/synapse/util/caches/dictionary_cache.py index fa91479c97..5eaf70c7ab 100644 --- a/synapse/util/caches/dictionary_cache.py +++ b/synapse/util/caches/dictionary_cache.py @@ -169,10 +169,11 @@ class DictionaryCache(Generic[KT, DKT, DV]): if it is in the cache. Returns: - DictionaryEntry: If `dict_keys` is not None then `DictionaryEntry` - will contain include the keys that are in the cache. If None then - will either return the full dict if in the cache, or the empty - dict (with `full` set to False) if it isn't. + If `dict_keys` is not None then `DictionaryEntry` will contain include + the keys that are in the cache. + + If None then will either return the full dict if in the cache, or the + empty dict (with `full` set to False) if it isn't. """ if dict_keys is None: # The caller wants the full set of dictionary keys for this cache key diff --git a/synapse/util/caches/expiringcache.py b/synapse/util/caches/expiringcache.py index c6a5d0dfc0..01ad02af67 100644 --- a/synapse/util/caches/expiringcache.py +++ b/synapse/util/caches/expiringcache.py @@ -207,7 +207,7 @@ class ExpiringCache(Generic[KT, VT]): items from the cache. Returns: - bool: Whether the cache changed size or not. + Whether the cache changed size or not. """ new_size = int(self._original_max_size * factor) if new_size != self._max_size: diff --git a/synapse/util/caches/lrucache.py b/synapse/util/caches/lrucache.py index aa93109d13..dcf0eac3bf 100644 --- a/synapse/util/caches/lrucache.py +++ b/synapse/util/caches/lrucache.py @@ -389,11 +389,11 @@ class LruCache(Generic[KT, VT]): cache_name: The name of this cache, for the prometheus metrics. If unset, no metrics will be reported on this cache. - cache_type (type): + cache_type: type of underlying cache to be used. Typically one of dict or TreeCache. - size_callback (func(V) -> int | None): + size_callback: metrics_collection_callback: metrics collection callback. This is called early in the metrics @@ -403,7 +403,7 @@ class LruCache(Generic[KT, VT]): Ignored if cache_name is None. - apply_cache_factor_from_config (bool): If true, `max_size` will be + apply_cache_factor_from_config: If true, `max_size` will be multiplied by a cache factor derived from the homeserver config clock: @@ -796,7 +796,7 @@ class LruCache(Generic[KT, VT]): items from the cache. Returns: - bool: Whether the cache changed size or not. + Whether the cache changed size or not. """ if not self.apply_cache_factor_from_config: return False diff --git a/synapse/util/caches/stream_change_cache.py b/synapse/util/caches/stream_change_cache.py index 330709b8b7..1657459549 100644 --- a/synapse/util/caches/stream_change_cache.py +++ b/synapse/util/caches/stream_change_cache.py @@ -16,6 +16,7 @@ import logging import math from typing import Collection, Dict, FrozenSet, List, Mapping, Optional, Set, Union +import attr from sortedcontainers import SortedDict from synapse.util import caches @@ -26,14 +27,41 @@ logger = logging.getLogger(__name__) EntityType = str +@attr.s(auto_attribs=True, frozen=True, slots=True) +class AllEntitiesChangedResult: + """Return type of `get_all_entities_changed`. + + Callers must check that there was a cache hit, via `result.hit`, before + using the entities in `result.entities`. + + This specifically does *not* implement helpers such as `__bool__` to ensure + that callers do the correct checks. + """ + + _entities: Optional[List[EntityType]] + + @property + def hit(self) -> bool: + return self._entities is not None + + @property + def entities(self) -> List[EntityType]: + assert self._entities is not None + return self._entities + + class StreamChangeCache: - """Keeps track of the stream positions of the latest change in a set of entities. + """ + Keeps track of the stream positions of the latest change in a set of entities. + + The entity will is typically a room ID or user ID, but can be any string. - Typically the entity will be a room or user id. + Can be queried for whether a specific entity has changed after a stream position + or for a list of changed entities after a stream position. See the individual + methods for more information. - Given a list of entities and a stream position, it will give a subset of - entities that may have changed since that position. If position key is too - old then the cache will simply return all given entities. + Only tracks to a maximum cache size, any position earlier than the earliest + known stream position must be treated as unknown. """ def __init__( @@ -45,16 +73,20 @@ class StreamChangeCache: ) -> None: self._original_max_size: int = max_size self._max_size = math.floor(max_size) - self._entity_to_key: Dict[EntityType, int] = {} - # map from stream id to the a set of entities which changed at that stream id. + # map from stream id to the set of entities which changed at that stream id. self._cache: SortedDict[int, Set[EntityType]] = SortedDict() + # map from entity to the stream ID of the latest change for that entity. + # + # Must be kept in sync with _cache. + self._entity_to_key: Dict[EntityType, int] = {} # the earliest stream_pos for which we can reliably answer # get_all_entities_changed. In other words, one less than the earliest # stream_pos for which we know _cache is valid. # self._earliest_known_stream_pos = current_stream_pos + self.name = name self.metrics = caches.register_cache( "cache", self.name, self._cache, resize_callback=self.set_cache_factor @@ -72,7 +104,7 @@ class StreamChangeCache: items from the cache. Returns: - bool: Whether the cache changed size or not. + Whether the cache changed size or not. """ new_size = math.floor(self._original_max_size * factor) if new_size != self._max_size: @@ -82,22 +114,46 @@ class StreamChangeCache: return False def has_entity_changed(self, entity: EntityType, stream_pos: int) -> bool: - """Returns True if the entity may have been updated since stream_pos""" + """ + Returns True if the entity may have been updated after stream_pos. + + Args: + entity: The entity to check for changes. + stream_pos: The stream position to check for changes after. + + Return: + True if the entity may have been updated, this happens if: + * The given stream position is at or earlier than the earliest + known stream position. + * The given stream position is earlier than the latest change for + the entity. + + False otherwise: + * The entity is unknown. + * The given stream position is at or later than the latest change + for the entity. + """ assert isinstance(stream_pos, int) - if stream_pos < self._earliest_known_stream_pos: + # _cache is not valid at or before the earliest known stream position, so + # return that the entity has changed. + if stream_pos <= self._earliest_known_stream_pos: self.metrics.inc_misses() return True + # If the entity is unknown, it hasn't changed. latest_entity_change_pos = self._entity_to_key.get(entity, None) if latest_entity_change_pos is None: self.metrics.inc_hits() return False + # This is a known entity, return true if the stream position is earlier + # than the last change. if stream_pos < latest_entity_change_pos: self.metrics.inc_misses() return True + # Otherwise, the stream position is after the latest change: return false. self.metrics.inc_hits() return False @@ -105,23 +161,35 @@ class StreamChangeCache: self, entities: Collection[EntityType], stream_pos: int ) -> Union[Set[EntityType], FrozenSet[EntityType]]: """ - Returns subset of entities that have had new things since the given - position. Entities unknown to the cache will be returned. If the - position is too old it will just return the given list. + Returns the subset of the given entities that have had changes after the given position. + + Entities unknown to the cache will be returned. + + If the position is too old it will just return the given list. + + Args: + entities: Entities to check for changes. + stream_pos: The stream position to check for changes after. + + Return: + A subset of entities which have changed after the given stream position. + + This will be all entities if the given stream position is at or earlier + than the earliest known stream position. """ - changed_entities = self.get_all_entities_changed(stream_pos) - if changed_entities is not None: + cache_result = self.get_all_entities_changed(stream_pos) + if cache_result.hit: # We now do an intersection, trying to do so in the most efficient # way possible (some of these sets are *large*). First check in the - # given iterable is already set that we can reuse, otherwise we + # given iterable is already a set that we can reuse, otherwise we # create a set of the *smallest* of the two iterables and call # `intersection(..)` on it (this can be twice as fast as the reverse). if isinstance(entities, (set, frozenset)): - result = entities.intersection(changed_entities) - elif len(changed_entities) < len(entities): - result = set(changed_entities).intersection(entities) + result = entities.intersection(cache_result.entities) + elif len(cache_result.entities) < len(entities): + result = set(cache_result.entities).intersection(entities) else: - result = set(entities).intersection(changed_entities) + result = set(entities).intersection(cache_result.entities) self.metrics.inc_hits() else: result = set(entities) @@ -130,43 +198,76 @@ class StreamChangeCache: return result def has_any_entity_changed(self, stream_pos: int) -> bool: - """Returns if any entity has changed""" - assert type(stream_pos) is int + """ + Returns true if any entity has changed after the given stream position. - if not self._cache: - # If the cache is empty, nothing can have changed. - return False + Args: + stream_pos: The stream position to check for changes after. - if stream_pos >= self._earliest_known_stream_pos: - self.metrics.inc_hits() - return self._cache.bisect_right(stream_pos) < len(self._cache) - else: + Return: + True if any entity has changed after the given stream position or + if the given stream position is at or earlier than the earliest + known stream position. + + False otherwise. + """ + assert isinstance(stream_pos, int) + + # _cache is not valid at or before the earliest known stream position, so + # return that an entity has changed. + if stream_pos <= self._earliest_known_stream_pos: self.metrics.inc_misses() return True - def get_all_entities_changed(self, stream_pos: int) -> Optional[List[EntityType]]: - """Returns all entities that have had new things since the given - position. If the position is too old it will return None. + # If the cache is empty, nothing can have changed. + if not self._cache: + self.metrics.inc_misses() + return False + + self.metrics.inc_hits() + return stream_pos < self._cache.peekitem()[0] + + def get_all_entities_changed(self, stream_pos: int) -> AllEntitiesChangedResult: + """ + Returns all entities that have had changes after the given position. + + If the stream change cache does not go far enough back, i.e. the + position is too old, it will return None. Returns the entities in the order that they were changed. + + Args: + stream_pos: The stream position to check for changes after. + + Return: + A class indicating if we have the requested data cached, and if so + includes the entities in the order they were changed. """ - assert type(stream_pos) is int + assert isinstance(stream_pos, int) - if stream_pos < self._earliest_known_stream_pos: - return None + # _cache is not valid at or before the earliest known stream position, so + # return None to mark that it is unknown if an entity has changed. + if stream_pos <= self._earliest_known_stream_pos: + return AllEntitiesChangedResult(None) changed_entities: List[EntityType] = [] for k in self._cache.islice(start=self._cache.bisect_right(stream_pos)): changed_entities.extend(self._cache[k]) - return changed_entities + return AllEntitiesChangedResult(changed_entities) def entity_has_changed(self, entity: EntityType, stream_pos: int) -> None: - """Informs the cache that the entity has been changed at the given - position. """ - assert type(stream_pos) is int + Informs the cache that the entity has been changed at the given position. + + Args: + entity: The entity to mark as changed. + stream_pos: The stream position to update the entity to. + """ + assert isinstance(stream_pos, int) + # For a change before _cache is valid (e.g. at or before the earliest known + # stream position) there's nothing to do. if stream_pos <= self._earliest_known_stream_pos: return @@ -188,14 +289,13 @@ class StreamChangeCache: self._entity_to_key[entity] = stream_pos self._evict() - # if the cache is too big, remove entries - while len(self._cache) > self._max_size: - k, r = self._cache.popitem(0) - self._earliest_known_stream_pos = max(k, self._earliest_known_stream_pos) - for entity in r: - del self._entity_to_key[entity] - def _evict(self) -> None: + """ + Ensure the cache has not exceeded the maximum size. + + Evicts entries until it is at the maximum size. + """ + # if the cache is too big, remove entries while len(self._cache) > self._max_size: k, r = self._cache.popitem(0) self._earliest_known_stream_pos = max(k, self._earliest_known_stream_pos) @@ -203,8 +303,14 @@ class StreamChangeCache: self._entity_to_key.pop(entity, None) def get_max_pos_of_last_change(self, entity: EntityType) -> int: - """Returns an upper bound of the stream id of the last change to an entity. + + Args: + entity: The entity to check. + + Return: + The stream position of the latest change for the given entity or + the earliest known stream position if the entitiy is unknown. """ return self._entity_to_key.get(entity, self._earliest_known_stream_pos) diff --git a/synapse/util/httpresourcetree.py b/synapse/util/httpresourcetree.py index a0606851f7..39fab4fe06 100644 --- a/synapse/util/httpresourcetree.py +++ b/synapse/util/httpresourcetree.py @@ -15,7 +15,9 @@ import logging from typing import Dict -from twisted.web.resource import NoResource, Resource +from twisted.web.resource import Resource + +from synapse.http.server import UnrecognizedRequestResource logger = logging.getLogger(__name__) @@ -49,7 +51,7 @@ def create_resource_tree( for path_seg in full_path.split(b"/")[1:-1]: if path_seg not in last_resource.listNames(): # resource doesn't exist, so make a "dummy resource" - child_resource: Resource = NoResource() + child_resource: Resource = UnrecognizedRequestResource() last_resource.putChild(path_seg, child_resource) res_id = _resource_id(last_resource, path_seg) resource_mappings[res_id] = child_resource diff --git a/synapse/util/macaroons.py b/synapse/util/macaroons.py index df77edcce2..5df03d3ddc 100644 --- a/synapse/util/macaroons.py +++ b/synapse/util/macaroons.py @@ -24,7 +24,7 @@ from typing_extensions import Literal from synapse.util import Clock, stringutils -MacaroonType = Literal["access", "delete_pusher", "session", "login"] +MacaroonType = Literal["access", "delete_pusher", "session"] def get_value_from_macaroon(macaroon: pymacaroons.Macaroon, key: str) -> str: @@ -111,19 +111,6 @@ class OidcSessionData: """The session ID of the ongoing UI Auth ("" if this is a login)""" -@attr.s(slots=True, frozen=True, auto_attribs=True) -class LoginTokenAttributes: - """Data we store in a short-term login token""" - - user_id: str - - auth_provider_id: str - """The SSO Identity Provider that the user authenticated with, to get this token.""" - - auth_provider_session_id: Optional[str] - """The session ID advertised by the SSO Identity Provider.""" - - class MacaroonGenerator: def __init__(self, clock: Clock, location: str, secret_key: bytes): self._clock = clock @@ -165,35 +152,6 @@ class MacaroonGenerator: macaroon.add_first_party_caveat(f"pushkey = {pushkey}") return macaroon.serialize() - def generate_short_term_login_token( - self, - user_id: str, - auth_provider_id: str, - auth_provider_session_id: Optional[str] = None, - duration_in_ms: int = (2 * 60 * 1000), - ) -> str: - """Generate a short-term login token used during SSO logins - - Args: - user_id: The user for which the token is valid. - auth_provider_id: The SSO IdP the user used. - auth_provider_session_id: The session ID got during login from the SSO IdP. - - Returns: - A signed token valid for using as a ``m.login.token`` token. - """ - now = self._clock.time_msec() - expiry = now + duration_in_ms - macaroon = self._generate_base_macaroon("login") - macaroon.add_first_party_caveat(f"user_id = {user_id}") - macaroon.add_first_party_caveat(f"time < {expiry}") - macaroon.add_first_party_caveat(f"auth_provider_id = {auth_provider_id}") - if auth_provider_session_id is not None: - macaroon.add_first_party_caveat( - f"auth_provider_session_id = {auth_provider_session_id}" - ) - return macaroon.serialize() - def generate_oidc_session_token( self, state: str, @@ -233,49 +191,6 @@ class MacaroonGenerator: return macaroon.serialize() - def verify_short_term_login_token(self, token: str) -> LoginTokenAttributes: - """Verify a short-term-login macaroon - - Checks that the given token is a valid, unexpired short-term-login token - minted by this server. - - Args: - token: The login token to verify. - - Returns: - A set of attributes carried by this token, including the - ``user_id`` and informations about the SSO IDP used during that - login. - - Raises: - MacaroonVerificationFailedException if the verification failed - """ - macaroon = pymacaroons.Macaroon.deserialize(token) - - v = self._base_verifier("login") - v.satisfy_general(lambda c: c.startswith("user_id = ")) - v.satisfy_general(lambda c: c.startswith("auth_provider_id = ")) - v.satisfy_general(lambda c: c.startswith("auth_provider_session_id = ")) - satisfy_expiry(v, self._clock.time_msec) - v.verify(macaroon, self._secret_key) - - user_id = get_value_from_macaroon(macaroon, "user_id") - auth_provider_id = get_value_from_macaroon(macaroon, "auth_provider_id") - - auth_provider_session_id: Optional[str] = None - try: - auth_provider_session_id = get_value_from_macaroon( - macaroon, "auth_provider_session_id" - ) - except MacaroonVerificationFailedException: - pass - - return LoginTokenAttributes( - user_id=user_id, - auth_provider_id=auth_provider_id, - auth_provider_session_id=auth_provider_session_id, - ) - def verify_guest_token(self, token: str) -> str: """Verify a guest access token macaroon diff --git a/synapse/util/ratelimitutils.py b/synapse/util/ratelimitutils.py index 9f64fed0d7..2aceb1a47f 100644 --- a/synapse/util/ratelimitutils.py +++ b/synapse/util/ratelimitutils.py @@ -183,7 +183,7 @@ class FederationRateLimiter: # Handle request ... Args: - host (str): Origin of incoming request. + host: Origin of incoming request. Returns: context manager which returns a deferred. diff --git a/synapse/util/retryutils.py b/synapse/util/retryutils.py index d0a69ff843..dcc037b982 100644 --- a/synapse/util/retryutils.py +++ b/synapse/util/retryutils.py @@ -51,7 +51,7 @@ class NotRetryingDestination(Exception): destination: the domain in question """ - msg = "Not retrying server %s." % (destination,) + msg = f"Not retrying server {destination} because we tried it recently retry_last_ts={retry_last_ts} and we won't check for another retry_interval={retry_interval}ms." super().__init__(msg) self.retry_last_ts = retry_last_ts diff --git a/synapse/util/stringutils.py b/synapse/util/stringutils.py index 27a363d7e5..4961fe9313 100644 --- a/synapse/util/stringutils.py +++ b/synapse/util/stringutils.py @@ -86,7 +86,7 @@ def parse_server_name(server_name: str) -> Tuple[str, Optional[int]]: ValueError if the server name could not be parsed. """ try: - if server_name[-1] == "]": + if server_name and server_name[-1] == "]": # ipv6 literal, hopefully return server_name, None @@ -123,7 +123,7 @@ def parse_and_validate_server_name(server_name: str) -> Tuple[str, Optional[int] # that nobody is sneaking IP literals in that look like hostnames, etc. # look for ipv6 literals - if host[0] == "[": + if host and host[0] == "[": if host[-1] != "]": raise ValueError("Mismatched [...] in server name '%s'" % (server_name,)) diff --git a/synapse/util/threepids.py b/synapse/util/threepids.py index 1e9c2faa64..54bc7589fd 100644 --- a/synapse/util/threepids.py +++ b/synapse/util/threepids.py @@ -48,7 +48,7 @@ async def check_3pid_allowed( registration: whether we want to bind the 3PID as part of registering a new user. Returns: - bool: whether the 3PID medium/address is allowed to be added to this HS + whether the 3PID medium/address is allowed to be added to this HS """ if not await hs.get_password_auth_provider().is_3pid_allowed( medium, address, registration diff --git a/synapse/util/wheel_timer.py b/synapse/util/wheel_timer.py index 177e198e7e..b1ec7f4bd8 100644 --- a/synapse/util/wheel_timer.py +++ b/synapse/util/wheel_timer.py @@ -90,10 +90,10 @@ class WheelTimer(Generic[T]): """Fetch any objects that have timed out Args: - now (ms): Current time in msec + now: Current time in msec Returns: - list: List of objects that have timed out + List of objects that have timed out """ now_key = int(now / self.bucket_size) diff --git a/synapse/visibility.py b/synapse/visibility.py index c4048d2477..e442de3173 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -26,8 +26,8 @@ from synapse.events.utils import prune_event from synapse.logging.opentracing import trace from synapse.storage.controllers import StorageControllers from synapse.storage.databases.main import DataStore -from synapse.storage.state import StateFilter from synapse.types import RetentionPolicy, StateMap, get_domain_from_id +from synapse.types.state import StateFilter from synapse.util import Clock logger = logging.getLogger(__name__) @@ -84,7 +84,15 @@ async def filter_events_for_client( """ # Filter out events that have been soft failed so that we don't relay them # to clients. + events_before_filtering = events events = [e for e in events if not e.internal_metadata.is_soft_failed()] + if len(events_before_filtering) != len(events): + if logger.isEnabledFor(logging.DEBUG): + logger.debug( + "filter_events_for_client: Filtered out soft-failed events: Before=%s, After=%s", + [event.event_id for event in events_before_filtering], + [event.event_id for event in events], + ) types = (_HISTORY_VIS_KEY, (EventTypes.Member, user_id)) @@ -301,6 +309,10 @@ def _check_client_allowed_to_see_event( _check_filter_send_to_client(event, clock, retention_policy, sender_ignored) == _CheckFilter.DENIED ): + logger.debug( + "_check_client_allowed_to_see_event(event=%s): Filtered out event because `_check_filter_send_to_client` returned `_CheckFilter.DENIED`", + event.event_id, + ) return None if event.event_id in always_include_ids: @@ -312,9 +324,17 @@ def _check_client_allowed_to_see_event( # for out-of-band membership events (eg, incoming invites, or rejections of # said invite) for the user themselves. if event.type == EventTypes.Member and event.state_key == user_id: - logger.debug("Returning out-of-band-membership event %s", event) + logger.debug( + "_check_client_allowed_to_see_event(event=%s): Returning out-of-band-membership event %s", + event.event_id, + event, + ) return event + logger.debug( + "_check_client_allowed_to_see_event(event=%s): Filtered out event because it's an outlier", + event.event_id, + ) return None if state is None: @@ -337,11 +357,21 @@ def _check_client_allowed_to_see_event( membership_result = _check_membership(user_id, event, visibility, state, is_peeking) if not membership_result.allowed: + logger.debug( + "_check_client_allowed_to_see_event(event=%s): Filtered out event because the user can't see the event because of their membership, membership_result.allowed=%s membership_result.joined=%s", + event.event_id, + membership_result.allowed, + membership_result.joined, + ) return None # If the sender has been erased and the user was not joined at the time, we # must only return the redacted form. if sender_erased and not membership_result.joined: + logger.debug( + "_check_client_allowed_to_see_event(event=%s): Returning pruned event because `sender_erased` and the user was not joined at the time", + event.event_id, + ) event = prune_event(event) return event @@ -533,7 +563,8 @@ def get_effective_room_visibility_from_state(state: StateMap[EventBase]) -> str: async def filter_events_for_server( storage: StorageControllers, - server_name: str, + target_server_name: str, + local_server_name: str, events: List[EventBase], redact: bool = True, check_history_visibility_only: bool = False, @@ -573,7 +604,7 @@ async def filter_events_for_server( # if the server is either in the room or has been invited # into the room. for ev in memberships.values(): - assert get_domain_from_id(ev.state_key) == server_name + assert get_domain_from_id(ev.state_key) == target_server_name memtype = ev.membership if memtype == Membership.JOIN: @@ -592,6 +623,24 @@ async def filter_events_for_server( # to no users having been erased. erased_senders = {} + # Filter out non-local events when we are in the middle of a partial join, since our servers + # list can be out of date and we could leak events to servers not in the room anymore. + # This can also be true for local events but we consider it to be an acceptable risk. + + # We do this check as a first step and before retrieving membership events because + # otherwise a room could be fully joined after we retrieve those, which would then bypass + # this check but would base the filtering on an outdated view of the membership events. + + partial_state_invisible_events = set() + if not check_history_visibility_only: + for e in events: + sender_domain = get_domain_from_id(e.sender) + if ( + sender_domain != local_server_name + and await storage.main.is_partial_state_room(e.room_id) + ): + partial_state_invisible_events.add(e) + # Let's check to see if all the events have a history visibility # of "shared" or "world_readable". If that's the case then we don't # need to check membership (as we know the server is in the room). @@ -606,7 +655,7 @@ async def filter_events_for_server( if event_to_history_vis[e.event_id] not in (HistoryVisibility.SHARED, HistoryVisibility.WORLD_READABLE) ], - server_name, + target_server_name, ) to_return = [] @@ -615,6 +664,10 @@ async def filter_events_for_server( visible = check_event_is_visible( event_to_history_vis[e.event_id], event_to_memberships.get(e.event_id, {}) ) + + if e in partial_state_invisible_events: + visible = False + if visible and not erased: to_return.append(e) elif redact: |