From 09957ce0e4dcfd84c2de4039653059faae03065b Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Mon, 4 Nov 2019 17:09:22 +0000 Subject: [PATCH 01/47] Implement per-room message retention policies --- changelog.d/5815.feature | 1 + docs/sample_config.yaml | 63 ++++ synapse/api/constants.py | 2 + synapse/config/server.py | 172 ++++++++++ synapse/events/validator.py | 100 +++++- synapse/handlers/federation.py | 2 +- synapse/handlers/message.py | 4 +- synapse/handlers/pagination.py | 111 ++++++ synapse/storage/data_stores/main/events.py | 3 + synapse/storage/data_stores/main/room.py | 252 ++++++++++++++ .../main/schema/delta/56/room_retention.sql | 33 ++ synapse/visibility.py | 17 + tests/rest/client/test_retention.py | 320 ++++++++++++++++++ 13 files changed, 1074 insertions(+), 6 deletions(-) create mode 100644 changelog.d/5815.feature create mode 100644 synapse/storage/data_stores/main/schema/delta/56/room_retention.sql create mode 100644 tests/rest/client/test_retention.py diff --git a/changelog.d/5815.feature b/changelog.d/5815.feature new file mode 100644 index 0000000000..ca4df4e7f6 --- /dev/null +++ b/changelog.d/5815.feature @@ -0,0 +1 @@ +Implement per-room message retention policies. diff --git a/docs/sample_config.yaml b/docs/sample_config.yaml index d2f4aff826..87fba27d13 100644 --- a/docs/sample_config.yaml +++ b/docs/sample_config.yaml @@ -328,6 +328,69 @@ listeners: # #user_ips_max_age: 14d +# Message retention policy at the server level. +# +# Room admins and mods can define a retention period for their rooms using the +# 'm.room.retention' state event, and server admins can cap this period by setting +# the 'allowed_lifetime_min' and 'allowed_lifetime_max' config options. +# +# If this feature is enabled, Synapse will regularly look for and purge events +# which are older than the room's maximum retention period. Synapse will also +# filter events received over federation so that events that should have been +# purged are ignored and not stored again. +# +retention: + # The message retention policies feature is disabled by default. Uncomment the + # following line to enable it. + # + #enabled: true + + # Default retention policy. If set, Synapse will apply it to rooms that lack the + # 'm.room.retention' state event. Currently, the value of 'min_lifetime' doesn't + # matter much because Synapse doesn't take it into account yet. + # + #default_policy: + # min_lifetime: 1d + # max_lifetime: 1y + + # Retention policy limits. If set, a user won't be able to send a + # 'm.room.retention' event which features a 'min_lifetime' or a 'max_lifetime' + # that's not within this range. This is especially useful in closed federations, + # in which server admins can make sure every federating server applies the same + # rules. + # + #allowed_lifetime_min: 1d + #allowed_lifetime_max: 1y + + # Server admins can define the settings of the background jobs purging the + # events which lifetime has expired under the 'purge_jobs' section. + # + # If no configuration is provided, a single job will be set up to delete expired + # events in every room daily. + # + # Each job's configuration defines which range of message lifetimes the job + # takes care of. For example, if 'shortest_max_lifetime' is '2d' and + # 'longest_max_lifetime' is '3d', the job will handle purging expired events in + # rooms whose state defines a 'max_lifetime' that's both higher than 2 days, and + # lower than or equal to 3 days. Both the minimum and the maximum value of a + # range are optional, e.g. a job with no 'shortest_max_lifetime' and a + # 'longest_max_lifetime' of '3d' will handle every room with a retention policy + # which 'max_lifetime' is lower than or equal to three days. + # + # The rationale for this per-job configuration is that some rooms might have a + # retention policy with a low 'max_lifetime', where history needs to be purged + # of outdated messages on a very frequent basis (e.g. every 5min), but not want + # that purge to be performed by a job that's iterating over every room it knows, + # which would be quite heavy on the server. + # + #purge_jobs: + # - shortest_max_lifetime: 1d + # longest_max_lifetime: 3d + # interval: 5m: + # - shortest_max_lifetime: 3d + # longest_max_lifetime: 1y + # interval: 24h + ## TLS ## diff --git a/synapse/api/constants.py b/synapse/api/constants.py index 49c4b85054..e3f086f1c3 100644 --- a/synapse/api/constants.py +++ b/synapse/api/constants.py @@ -94,6 +94,8 @@ class EventTypes(object): ServerACL = "m.room.server_acl" Pinned = "m.room.pinned_events" + Retention = "m.room.retention" + class RejectedReason(object): AUTH_ERROR = "auth_error" diff --git a/synapse/config/server.py b/synapse/config/server.py index d556df308d..aa93a416f1 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -246,6 +246,115 @@ class ServerConfig(Config): # events with profile information that differ from the target's global profile. self.allow_per_room_profiles = config.get("allow_per_room_profiles", True) + retention_config = config.get("retention") + if retention_config is None: + retention_config = {} + + self.retention_enabled = retention_config.get("enabled", False) + + retention_default_policy = retention_config.get("default_policy") + + if retention_default_policy is not None: + self.retention_default_min_lifetime = retention_default_policy.get( + "min_lifetime" + ) + if self.retention_default_min_lifetime is not None: + self.retention_default_min_lifetime = self.parse_duration( + self.retention_default_min_lifetime + ) + + self.retention_default_max_lifetime = retention_default_policy.get( + "max_lifetime" + ) + if self.retention_default_max_lifetime is not None: + self.retention_default_max_lifetime = self.parse_duration( + self.retention_default_max_lifetime + ) + + if ( + self.retention_default_min_lifetime is not None + and self.retention_default_max_lifetime is not None + and ( + self.retention_default_min_lifetime + > self.retention_default_max_lifetime + ) + ): + raise ConfigError( + "The default retention policy's 'min_lifetime' can not be greater" + " than its 'max_lifetime'" + ) + else: + self.retention_default_min_lifetime = None + self.retention_default_max_lifetime = None + + self.retention_allowed_lifetime_min = retention_config.get("allowed_lifetime_min") + if self.retention_allowed_lifetime_min is not None: + self.retention_allowed_lifetime_min = self.parse_duration( + self.retention_allowed_lifetime_min + ) + + self.retention_allowed_lifetime_max = retention_config.get("allowed_lifetime_max") + if self.retention_allowed_lifetime_max is not None: + self.retention_allowed_lifetime_max = self.parse_duration( + self.retention_allowed_lifetime_max + ) + + if ( + self.retention_allowed_lifetime_min is not None + and self.retention_allowed_lifetime_max is not None + and self.retention_allowed_lifetime_min > self.retention_allowed_lifetime_max + ): + raise ConfigError( + "Invalid retention policy limits: 'allowed_lifetime_min' can not be" + " greater than 'allowed_lifetime_max'" + ) + + self.retention_purge_jobs = [] + for purge_job_config in retention_config.get("purge_jobs", []): + interval_config = purge_job_config.get("interval") + + if interval_config is None: + raise ConfigError( + "A retention policy's purge jobs configuration must have the" + " 'interval' key set." + ) + + interval = self.parse_duration(interval_config) + + shortest_max_lifetime = purge_job_config.get("shortest_max_lifetime") + + if shortest_max_lifetime is not None: + shortest_max_lifetime = self.parse_duration(shortest_max_lifetime) + + longest_max_lifetime = purge_job_config.get("longest_max_lifetime") + + if longest_max_lifetime is not None: + longest_max_lifetime = self.parse_duration(longest_max_lifetime) + + if ( + shortest_max_lifetime is not None + and longest_max_lifetime is not None + and shortest_max_lifetime > longest_max_lifetime + ): + raise ConfigError( + "A retention policy's purge jobs configuration's" + " 'shortest_max_lifetime' value can not be greater than its" + " 'longest_max_lifetime' value." + ) + + self.retention_purge_jobs.append({ + "interval": interval, + "shortest_max_lifetime": shortest_max_lifetime, + "longest_max_lifetime": longest_max_lifetime, + }) + + if not self.retention_purge_jobs: + self.retention_purge_jobs = [{ + "interval": self.parse_duration("1d"), + "shortest_max_lifetime": None, + "longest_max_lifetime": None, + }] + self.listeners = [] # type: List[dict] for listener in config.get("listeners", []): if not isinstance(listener.get("port", None), int): @@ -761,6 +870,69 @@ class ServerConfig(Config): # Defaults to `28d`. Set to `null` to disable clearing out of old rows. # #user_ips_max_age: 14d + + # Message retention policy at the server level. + # + # Room admins and mods can define a retention period for their rooms using the + # 'm.room.retention' state event, and server admins can cap this period by setting + # the 'allowed_lifetime_min' and 'allowed_lifetime_max' config options. + # + # If this feature is enabled, Synapse will regularly look for and purge events + # which are older than the room's maximum retention period. Synapse will also + # filter events received over federation so that events that should have been + # purged are ignored and not stored again. + # + retention: + # The message retention policies feature is disabled by default. Uncomment the + # following line to enable it. + # + #enabled: true + + # Default retention policy. If set, Synapse will apply it to rooms that lack the + # 'm.room.retention' state event. Currently, the value of 'min_lifetime' doesn't + # matter much because Synapse doesn't take it into account yet. + # + #default_policy: + # min_lifetime: 1d + # max_lifetime: 1y + + # Retention policy limits. If set, a user won't be able to send a + # 'm.room.retention' event which features a 'min_lifetime' or a 'max_lifetime' + # that's not within this range. This is especially useful in closed federations, + # in which server admins can make sure every federating server applies the same + # rules. + # + #allowed_lifetime_min: 1d + #allowed_lifetime_max: 1y + + # Server admins can define the settings of the background jobs purging the + # events which lifetime has expired under the 'purge_jobs' section. + # + # If no configuration is provided, a single job will be set up to delete expired + # events in every room daily. + # + # Each job's configuration defines which range of message lifetimes the job + # takes care of. For example, if 'shortest_max_lifetime' is '2d' and + # 'longest_max_lifetime' is '3d', the job will handle purging expired events in + # rooms whose state defines a 'max_lifetime' that's both higher than 2 days, and + # lower than or equal to 3 days. Both the minimum and the maximum value of a + # range are optional, e.g. a job with no 'shortest_max_lifetime' and a + # 'longest_max_lifetime' of '3d' will handle every room with a retention policy + # which 'max_lifetime' is lower than or equal to three days. + # + # The rationale for this per-job configuration is that some rooms might have a + # retention policy with a low 'max_lifetime', where history needs to be purged + # of outdated messages on a very frequent basis (e.g. every 5min), but not want + # that purge to be performed by a job that's iterating over every room it knows, + # which would be quite heavy on the server. + # + #purge_jobs: + # - shortest_max_lifetime: 1d + # longest_max_lifetime: 3d + # interval: 5m: + # - shortest_max_lifetime: 3d + # longest_max_lifetime: 1y + # interval: 24h """ % locals() ) diff --git a/synapse/events/validator.py b/synapse/events/validator.py index 272426e105..9b90c9ce04 100644 --- a/synapse/events/validator.py +++ b/synapse/events/validator.py @@ -13,7 +13,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -from six import string_types +from six import integer_types, string_types from synapse.api.constants import MAX_ALIAS_LENGTH, EventTypes, Membership from synapse.api.errors import Codes, SynapseError @@ -22,11 +22,12 @@ from synapse.types import EventID, RoomID, UserID class EventValidator(object): - def validate_new(self, event): + def validate_new(self, event, config): """Validates the event has roughly the right format Args: - event (FrozenEvent) + event (FrozenEvent): The event to validate. + config (Config): The homeserver's configuration. """ self.validate_builder(event) @@ -67,6 +68,99 @@ class EventValidator(object): Codes.INVALID_PARAM, ) + if event.type == EventTypes.Retention: + self._validate_retention(event, config) + + def _validate_retention(self, event, config): + """Checks that an event that defines the retention policy for a room respects the + boundaries imposed by the server's administrator. + + Args: + event (FrozenEvent): The event to validate. + config (Config): The homeserver's configuration. + """ + min_lifetime = event.content.get("min_lifetime") + max_lifetime = event.content.get("max_lifetime") + + if min_lifetime is not None: + if not isinstance(min_lifetime, integer_types): + raise SynapseError( + code=400, + msg="'min_lifetime' must be an integer", + errcode=Codes.BAD_JSON, + ) + + if ( + config.retention_allowed_lifetime_min is not None + and min_lifetime < config.retention_allowed_lifetime_min + ): + raise SynapseError( + code=400, + msg=( + "'min_lifetime' can't be lower than the minimum allowed" + " value enforced by the server's administrator" + ), + errcode=Codes.BAD_JSON, + ) + + if ( + config.retention_allowed_lifetime_max is not None + and min_lifetime > config.retention_allowed_lifetime_max + ): + raise SynapseError( + code=400, + msg=( + "'min_lifetime' can't be greater than the maximum allowed" + " value enforced by the server's administrator" + ), + errcode=Codes.BAD_JSON, + ) + + if max_lifetime is not None: + if not isinstance(max_lifetime, integer_types): + raise SynapseError( + code=400, + msg="'max_lifetime' must be an integer", + errcode=Codes.BAD_JSON, + ) + + if ( + config.retention_allowed_lifetime_min is not None + and max_lifetime < config.retention_allowed_lifetime_min + ): + raise SynapseError( + code=400, + msg=( + "'max_lifetime' can't be lower than the minimum allowed value" + " enforced by the server's administrator" + ), + errcode=Codes.BAD_JSON, + ) + + if ( + config.retention_allowed_lifetime_max is not None + and max_lifetime > config.retention_allowed_lifetime_max + ): + raise SynapseError( + code=400, + msg=( + "'max_lifetime' can't be greater than the maximum allowed" + " value enforced by the server's administrator" + ), + errcode=Codes.BAD_JSON, + ) + + if ( + min_lifetime is not None + and max_lifetime is not None + and min_lifetime > max_lifetime + ): + raise SynapseError( + code=400, + msg="'min_lifetime' can't be greater than 'max_lifetime", + errcode=Codes.BAD_JSON, + ) + def validate_builder(self, event): """Validates that the builder/event has roughly the right format. Only checks values that we expect a proto event to have, rather than all the diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 8cafcfdab0..3994137d18 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -2454,7 +2454,7 @@ class FederationHandler(BaseHandler): room_version, event_dict, event, context ) - EventValidator().validate_new(event) + EventValidator().validate_new(event, self.config) # We need to tell the transaction queue to send this out, even # though the sender isn't a local user. diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index d682dc2b7a..155ed6e06a 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -417,7 +417,7 @@ class EventCreationHandler(object): 403, "You must be in the room to create an alias for it" ) - self.validator.validate_new(event) + self.validator.validate_new(event, self.config) return (event, context) @@ -634,7 +634,7 @@ class EventCreationHandler(object): if requester: context.app_service = requester.app_service - self.validator.validate_new(event) + self.validator.validate_new(event, self.config) # If this event is an annotation then we check that that the sender # can't annotate the same way twice (e.g. stops users from liking an diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index 97f15a1c32..e1800177fa 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -15,12 +15,15 @@ # limitations under the License. import logging +from six import iteritems + from twisted.internet import defer from twisted.python.failure import Failure from synapse.api.constants import EventTypes, Membership from synapse.api.errors import SynapseError from synapse.logging.context import run_in_background +from synapse.metrics.background_process_metrics import run_as_background_process from synapse.storage.state import StateFilter from synapse.types import RoomStreamToken from synapse.util.async_helpers import ReadWriteLock @@ -80,6 +83,114 @@ class PaginationHandler(object): self._purges_by_id = {} self._event_serializer = hs.get_event_client_serializer() + self._retention_default_max_lifetime = hs.config.retention_default_max_lifetime + + if hs.config.retention_enabled: + # Run the purge jobs described in the configuration file. + for job in hs.config.retention_purge_jobs: + self.clock.looping_call( + run_as_background_process, + job["interval"], + "purge_history_for_rooms_in_range", + self.purge_history_for_rooms_in_range, + job["shortest_max_lifetime"], + job["longest_max_lifetime"], + ) + + @defer.inlineCallbacks + def purge_history_for_rooms_in_range(self, min_ms, max_ms): + """Purge outdated events from rooms within the given retention range. + + If a default retention policy is defined in the server's configuration and its + 'max_lifetime' is within this range, also targets rooms which don't have a + retention policy. + + Args: + min_ms (int|None): Duration in milliseconds that define the lower limit of + the range to handle (exclusive). If None, it means that the range has no + lower limit. + max_ms (int|None): Duration in milliseconds that define the upper limit of + the range to handle (inclusive). If None, it means that the range has no + upper limit. + """ + # We want the storage layer to to include rooms with no retention policy in its + # return value only if a default retention policy is defined in the server's + # configuration and that policy's 'max_lifetime' is either lower (or equal) than + # max_ms or higher than min_ms (or both). + if self._retention_default_max_lifetime is not None: + include_null = True + + if min_ms is not None and min_ms >= self._retention_default_max_lifetime: + # The default max_lifetime is lower than (or equal to) min_ms. + include_null = False + + if max_ms is not None and max_ms < self._retention_default_max_lifetime: + # The default max_lifetime is higher than max_ms. + include_null = False + else: + include_null = False + + rooms = yield self.store.get_rooms_for_retention_period_in_range( + min_ms, max_ms, include_null + ) + + for room_id, retention_policy in iteritems(rooms): + if room_id in self._purges_in_progress_by_room: + logger.warning( + "[purge] not purging room %s as there's an ongoing purge running" + " for this room", + room_id, + ) + continue + + max_lifetime = retention_policy["max_lifetime"] + + if max_lifetime is None: + # If max_lifetime is None, it means that include_null equals True, + # therefore we can safely assume that there is a default policy defined + # in the server's configuration. + max_lifetime = self._retention_default_max_lifetime + + # Figure out what token we should start purging at. + ts = self.clock.time_msec() - max_lifetime + + stream_ordering = ( + yield self.store.find_first_stream_ordering_after_ts(ts) + ) + + r = ( + yield self.store.get_room_event_after_stream_ordering( + room_id, stream_ordering, + ) + ) + if not r: + logger.warning( + "[purge] purging events not possible: No event found " + "(ts %i => stream_ordering %i)", + ts, stream_ordering, + ) + continue + + (stream, topo, _event_id) = r + token = "t%d-%d" % (topo, stream) + + purge_id = random_string(16) + + self._purges_by_id[purge_id] = PurgeStatus() + + logger.info( + "Starting purging events in room %s (purge_id %s)" % (room_id, purge_id) + ) + + # We want to purge everything, including local events, and to run the purge in + # the background so that it's not blocking any other operation apart from + # other purges in the same room. + run_as_background_process( + "_purge_history", + self._purge_history, + purge_id, room_id, token, True, + ) + def start_purge_history(self, room_id, token, delete_local_events=False): """Start off a history purge on a room. diff --git a/synapse/storage/data_stores/main/events.py b/synapse/storage/data_stores/main/events.py index 301f8ea128..b332a42d82 100644 --- a/synapse/storage/data_stores/main/events.py +++ b/synapse/storage/data_stores/main/events.py @@ -929,6 +929,9 @@ class EventsStore( elif event.type == EventTypes.Redaction: # Insert into the redactions table. self._store_redaction(txn, event) + elif event.type == EventTypes.Retention: + # Update the room_retention table. + self._store_retention_policy_for_room_txn(txn, event) self._handle_event_relations(txn, event) diff --git a/synapse/storage/data_stores/main/room.py b/synapse/storage/data_stores/main/room.py index 67bb1b6f60..54a7d24c73 100644 --- a/synapse/storage/data_stores/main/room.py +++ b/synapse/storage/data_stores/main/room.py @@ -19,10 +19,13 @@ import logging import re from typing import Optional, Tuple +from six import integer_types + from canonicaljson import json from twisted.internet import defer +from synapse.api.constants import EventTypes from synapse.api.errors import StoreError from synapse.storage._base import SQLBaseStore from synapse.storage.data_stores.main.search import SearchStore @@ -302,6 +305,85 @@ class RoomWorkerStore(SQLBaseStore): class RoomStore(RoomWorkerStore, SearchStore): + def __init__(self, db_conn, hs): + super(RoomStore, self).__init__(db_conn, hs) + + self.config = hs.config + + self.register_background_update_handler( + "insert_room_retention", self._background_insert_retention, + ) + + @defer.inlineCallbacks + def _background_insert_retention(self, progress, batch_size): + """Retrieves a list of all rooms within a range and inserts an entry for each of + them into the room_retention table. + NULLs the property's columns if missing from the retention event in the room's + state (or NULLs all of them if there's no retention event in the room's state), + so that we fall back to the server's retention policy. + """ + + last_room = progress.get("room_id", "") + + def _background_insert_retention_txn(txn): + txn.execute( + """ + SELECT state.room_id, state.event_id, events.json + FROM current_state_events as state + LEFT JOIN event_json AS events ON (state.event_id = events.event_id) + WHERE state.room_id > ? AND state.type = '%s' + ORDER BY state.room_id ASC + LIMIT ?; + """ % EventTypes.Retention, + (last_room, batch_size) + ) + + rows = self.cursor_to_dict(txn) + + if not rows: + return True + + for row in rows: + if not row["json"]: + retention_policy = {} + else: + ev = json.loads(row["json"]) + retention_policy = json.dumps(ev["content"]) + + self._simple_insert_txn( + txn=txn, + table="room_retention", + values={ + "room_id": row["room_id"], + "event_id": row["event_id"], + "min_lifetime": retention_policy.get("min_lifetime"), + "max_lifetime": retention_policy.get("max_lifetime"), + } + ) + + logger.info("Inserted %d rows into room_retention", len(rows)) + + self._background_update_progress_txn( + txn, "insert_room_retention", { + "room_id": rows[-1]["room_id"], + } + ) + + if batch_size > len(rows): + return True + else: + return False + + end = yield self.runInteraction( + "insert_room_retention", + _background_insert_retention_txn, + ) + + if end: + yield self._end_background_update("insert_room_retention") + + defer.returnValue(batch_size) + @defer.inlineCallbacks def store_room(self, room_id, room_creator_user_id, is_public): """Stores a room. @@ -502,6 +584,37 @@ class RoomStore(RoomWorkerStore, SearchStore): txn, event, "content.body", event.content["body"] ) + def _store_retention_policy_for_room_txn(self, txn, event): + if ( + hasattr(event, "content") + and ("min_lifetime" in event.content or "max_lifetime" in event.content) + ): + if ( + ("min_lifetime" in event.content and not isinstance( + event.content.get("min_lifetime"), integer_types + )) + or ("max_lifetime" in event.content and not isinstance( + event.content.get("max_lifetime"), integer_types + )) + ): + # Ignore the event if one of the value isn't an integer. + return + + self._simple_insert_txn( + txn=txn, + table="room_retention", + values={ + "room_id": event.room_id, + "event_id": event.event_id, + "min_lifetime": event.content.get("min_lifetime"), + "max_lifetime": event.content.get("max_lifetime"), + }, + ) + + self._invalidate_cache_and_stream( + txn, self.get_retention_policy_for_room, (event.room_id,) + ) + def add_event_report( self, room_id, event_id, user_id, reason, content, received_ts ): @@ -683,3 +796,142 @@ class RoomStore(RoomWorkerStore, SearchStore): remote_media_mxcs.append((hostname, media_id)) return local_media_mxcs, remote_media_mxcs + + @defer.inlineCallbacks + def get_rooms_for_retention_period_in_range(self, min_ms, max_ms, include_null=False): + """Retrieves all of the rooms within the given retention range. + + Optionally includes the rooms which don't have a retention policy. + + Args: + min_ms (int|None): Duration in milliseconds that define the lower limit of + the range to handle (exclusive). If None, doesn't set a lower limit. + max_ms (int|None): Duration in milliseconds that define the upper limit of + the range to handle (inclusive). If None, doesn't set an upper limit. + include_null (bool): Whether to include rooms which retention policy is NULL + in the returned set. + + Returns: + dict[str, dict]: The rooms within this range, along with their retention + policy. The key is "room_id", and maps to a dict describing the retention + policy associated with this room ID. The keys for this nested dict are + "min_lifetime" (int|None), and "max_lifetime" (int|None). + """ + + def get_rooms_for_retention_period_in_range_txn(txn): + range_conditions = [] + args = [] + + if min_ms is not None: + range_conditions.append("max_lifetime > ?") + args.append(min_ms) + + if max_ms is not None: + range_conditions.append("max_lifetime <= ?") + args.append(max_ms) + + # Do a first query which will retrieve the rooms that have a retention policy + # in their current state. + sql = """ + SELECT room_id, min_lifetime, max_lifetime FROM room_retention + INNER JOIN current_state_events USING (event_id, room_id) + """ + + if len(range_conditions): + sql += " WHERE (" + " AND ".join(range_conditions) + ")" + + if include_null: + sql += " OR max_lifetime IS NULL" + + txn.execute(sql, args) + + rows = self.cursor_to_dict(txn) + rooms_dict = {} + + for row in rows: + rooms_dict[row["room_id"]] = { + "min_lifetime": row["min_lifetime"], + "max_lifetime": row["max_lifetime"], + } + + if include_null: + # If required, do a second query that retrieves all of the rooms we know + # of so we can handle rooms with no retention policy. + sql = "SELECT DISTINCT room_id FROM current_state_events" + + txn.execute(sql) + + rows = self.cursor_to_dict(txn) + + # If a room isn't already in the dict (i.e. it doesn't have a retention + # policy in its state), add it with a null policy. + for row in rows: + if row["room_id"] not in rooms_dict: + rooms_dict[row["room_id"]] = { + "min_lifetime": None, + "max_lifetime": None, + } + + return rooms_dict + + rooms = yield self.runInteraction( + "get_rooms_for_retention_period_in_range", + get_rooms_for_retention_period_in_range_txn, + ) + + defer.returnValue(rooms) + + @cachedInlineCallbacks() + def get_retention_policy_for_room(self, room_id): + """Get the retention policy for a given room. + + If no retention policy has been found for this room, returns a policy defined + by the configured default policy (which has None as both the 'min_lifetime' and + the 'max_lifetime' if no default policy has been defined in the server's + configuration). + + Args: + room_id (str): The ID of the room to get the retention policy of. + + Returns: + dict[int, int]: "min_lifetime" and "max_lifetime" for this room. + """ + + def get_retention_policy_for_room_txn(txn): + txn.execute( + """ + SELECT min_lifetime, max_lifetime FROM room_retention + INNER JOIN current_state_events USING (event_id, room_id) + WHERE room_id = ?; + """, + (room_id,) + ) + + return self.cursor_to_dict(txn) + + ret = yield self.runInteraction( + "get_retention_policy_for_room", + get_retention_policy_for_room_txn, + ) + + # If we don't know this room ID, ret will be None, in this case return the default + # policy. + if not ret: + defer.returnValue({ + "min_lifetime": self.config.retention_default_min_lifetime, + "max_lifetime": self.config.retention_default_max_lifetime, + }) + + row = ret[0] + + # If one of the room's policy's attributes isn't defined, use the matching + # attribute from the default policy. + # The default values will be None if no default policy has been defined, or if one + # of the attributes is missing from the default policy. + if row["min_lifetime"] is None: + row["min_lifetime"] = self.config.retention_default_min_lifetime + + if row["max_lifetime"] is None: + row["max_lifetime"] = self.config.retention_default_max_lifetime + + defer.returnValue(row) diff --git a/synapse/storage/data_stores/main/schema/delta/56/room_retention.sql b/synapse/storage/data_stores/main/schema/delta/56/room_retention.sql new file mode 100644 index 0000000000..ee6cdf7a14 --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/room_retention.sql @@ -0,0 +1,33 @@ +/* Copyright 2019 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. + */ + +-- Tracks the retention policy of a room. +-- A NULL max_lifetime or min_lifetime means that the matching property is not defined in +-- the room's retention policy state event. +-- If a room doesn't have a retention policy state event in its state, both max_lifetime +-- and min_lifetime are NULL. +CREATE TABLE IF NOT EXISTS room_retention( + room_id TEXT, + event_id TEXT, + min_lifetime BIGINT, + max_lifetime BIGINT, + + PRIMARY KEY(room_id, event_id) +); + +CREATE INDEX room_retention_max_lifetime_idx on room_retention(max_lifetime); + +INSERT INTO background_updates (update_name, progress_json) VALUES + ('insert_room_retention', '{}'); diff --git a/synapse/visibility.py b/synapse/visibility.py index 8c843febd8..4498c156bc 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -86,6 +86,14 @@ def filter_events_for_client( erased_senders = yield storage.main.are_users_erased((e.sender for e in events)) + room_ids = set(e.room_id for e in events) + retention_policies = {} + + for room_id in room_ids: + retention_policies[room_id] = yield storage.main.get_retention_policy_for_room( + room_id + ) + def allowed(event): """ Args: @@ -103,6 +111,15 @@ def filter_events_for_client( if not event.is_state() and event.sender in ignore_list: return None + retention_policy = retention_policies[event.room_id] + max_lifetime = retention_policy.get("max_lifetime") + + if max_lifetime is not None: + oldest_allowed_ts = storage.main.clock.time_msec() - max_lifetime + + if event.origin_server_ts < oldest_allowed_ts: + return None + if event.event_id in always_include_ids: return event diff --git a/tests/rest/client/test_retention.py b/tests/rest/client/test_retention.py new file mode 100644 index 0000000000..41ea9db689 --- /dev/null +++ b/tests/rest/client/test_retention.py @@ -0,0 +1,320 @@ +# -*- coding: utf-8 -*- +# Copyright 2019 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 mock import Mock + +from synapse.api.constants import EventTypes +from synapse.rest import admin +from synapse.rest.client.v1 import login, room +from synapse.visibility import filter_events_for_client + +from tests import unittest + +one_hour_ms = 3600000 +one_day_ms = one_hour_ms * 24 + + +class RetentionTestCase(unittest.HomeserverTestCase): + servlets = [ + admin.register_servlets, + login.register_servlets, + room.register_servlets, + ] + + def make_homeserver(self, reactor, clock): + config = self.default_config() + config["default_room_version"] = "1" + config["retention"] = { + "enabled": True, + "default_policy": { + "min_lifetime": one_day_ms, + "max_lifetime": one_day_ms * 3, + }, + "allowed_lifetime_min": one_day_ms, + "allowed_lifetime_max": one_day_ms * 3, + } + + self.hs = self.setup_test_homeserver(config=config) + return self.hs + + def prepare(self, reactor, clock, homeserver): + self.user_id = self.register_user("user", "password") + self.token = self.login("user", "password") + + def test_retention_state_event(self): + """Tests that the server configuration can limit the values a user can set to the + room's retention policy. + """ + room_id = self.helper.create_room_as(self.user_id, tok=self.token) + + self.helper.send_state( + room_id=room_id, + event_type=EventTypes.Retention, + body={ + "max_lifetime": one_day_ms * 4, + }, + tok=self.token, + expect_code=400, + ) + + self.helper.send_state( + room_id=room_id, + event_type=EventTypes.Retention, + body={ + "max_lifetime": one_hour_ms, + }, + tok=self.token, + expect_code=400, + ) + + def test_retention_event_purged_with_state_event(self): + """Tests that expired events are correctly purged when the room's retention policy + is defined by a state event. + """ + room_id = self.helper.create_room_as(self.user_id, tok=self.token) + + # Set the room's retention period to 2 days. + lifetime = one_day_ms * 2 + self.helper.send_state( + room_id=room_id, + event_type=EventTypes.Retention, + body={ + "max_lifetime": lifetime, + }, + tok=self.token, + ) + + self._test_retention_event_purged(room_id, one_day_ms * 1.5) + + def test_retention_event_purged_without_state_event(self): + """Tests that expired events are correctly purged when the room's retention policy + is defined by the server's configuration's default retention policy. + """ + room_id = self.helper.create_room_as(self.user_id, tok=self.token) + + self._test_retention_event_purged(room_id, one_day_ms * 2) + + def test_visibility(self): + """Tests that synapse.visibility.filter_events_for_client correctly filters out + outdated events + """ + store = self.hs.get_datastore() + storage = self.hs.get_storage() + room_id = self.helper.create_room_as(self.user_id, tok=self.token) + events = [] + + # Send a first event, which should be filtered out at the end of the test. + resp = self.helper.send( + room_id=room_id, + body="1", + tok=self.token, + ) + + # Get the event from the store so that we end up with a FrozenEvent that we can + # give to filter_events_for_client. We need to do this now because the event won't + # be in the database anymore after it has expired. + events.append(self.get_success( + store.get_event( + resp.get("event_id") + ) + )) + + # Advance the time by 2 days. We're using the default retention policy, therefore + # after this the first event will still be valid. + self.reactor.advance(one_day_ms * 2 / 1000) + + # Send another event, which shouldn't get filtered out. + resp = self.helper.send( + room_id=room_id, + body="2", + tok=self.token, + ) + + valid_event_id = resp.get("event_id") + + events.append(self.get_success( + store.get_event( + valid_event_id + ) + )) + + # Advance the time by anothe 2 days. After this, the first event should be + # outdated but not the second one. + self.reactor.advance(one_day_ms * 2 / 1000) + + # Run filter_events_for_client with our list of FrozenEvents. + filtered_events = self.get_success(filter_events_for_client( + storage, self.user_id, events + )) + + # We should only get one event back. + self.assertEqual(len(filtered_events), 1, filtered_events) + # That event should be the second, not outdated event. + self.assertEqual(filtered_events[0].event_id, valid_event_id, filtered_events) + + def _test_retention_event_purged(self, room_id, increment): + # Send a first event to the room. This is the event we'll want to be purged at the + # end of the test. + resp = self.helper.send( + room_id=room_id, + body="1", + tok=self.token, + ) + + expired_event_id = resp.get("event_id") + + # Check that we can retrieve the event. + expired_event = self.get_event(room_id, expired_event_id) + self.assertEqual(expired_event.get("content", {}).get("body"), "1", expired_event) + + # Advance the time. + self.reactor.advance(increment / 1000) + + # Send another event. We need this because the purge job won't purge the most + # recent event in the room. + resp = self.helper.send( + room_id=room_id, + body="2", + tok=self.token, + ) + + valid_event_id = resp.get("event_id") + + # Advance the time again. Now our first event should have expired but our second + # one should still be kept. + self.reactor.advance(increment / 1000) + + # Check that the event has been purged from the database. + self.get_event(room_id, expired_event_id, expected_code=404) + + # Check that the event that hasn't been purged can still be retrieved. + valid_event = self.get_event(room_id, valid_event_id) + self.assertEqual(valid_event.get("content", {}).get("body"), "2", valid_event) + + def get_event(self, room_id, event_id, expected_code=200): + url = "/_matrix/client/r0/rooms/%s/event/%s" % (room_id, event_id) + + request, channel = self.make_request("GET", url, access_token=self.token) + self.render(request) + + self.assertEqual(channel.code, expected_code, channel.result) + + return channel.json_body + + +class RetentionNoDefaultPolicyTestCase(unittest.HomeserverTestCase): + servlets = [ + admin.register_servlets, + login.register_servlets, + room.register_servlets, + ] + + def make_homeserver(self, reactor, clock): + config = self.default_config() + config["default_room_version"] = "1" + config["retention"] = { + "enabled": True, + } + + mock_federation_client = Mock(spec=["backfill"]) + + self.hs = self.setup_test_homeserver( + config=config, + federation_client=mock_federation_client, + ) + return self.hs + + def prepare(self, reactor, clock, homeserver): + self.user_id = self.register_user("user", "password") + self.token = self.login("user", "password") + + def test_no_default_policy(self): + """Tests that an event doesn't get expired if there is neither a default retention + policy nor a policy specific to the room. + """ + room_id = self.helper.create_room_as(self.user_id, tok=self.token) + + self._test_retention(room_id) + + def test_state_policy(self): + """Tests that an event gets correctly expired if there is no default retention + policy but there's a policy specific to the room. + """ + room_id = self.helper.create_room_as(self.user_id, tok=self.token) + + # Set the maximum lifetime to 35 days so that the first event gets expired but not + # the second one. + self.helper.send_state( + room_id=room_id, + event_type=EventTypes.Retention, + body={ + "max_lifetime": one_day_ms * 35, + }, + tok=self.token, + ) + + self._test_retention(room_id, expected_code_for_first_event=404) + + def _test_retention(self, room_id, expected_code_for_first_event=200): + # Send a first event to the room. This is the event we'll want to be purged at the + # end of the test. + resp = self.helper.send( + room_id=room_id, + body="1", + tok=self.token, + ) + + first_event_id = resp.get("event_id") + + # Check that we can retrieve the event. + expired_event = self.get_event(room_id, first_event_id) + self.assertEqual(expired_event.get("content", {}).get("body"), "1", expired_event) + + # Advance the time by a month. + self.reactor.advance(one_day_ms * 30 / 1000) + + # Send another event. We need this because the purge job won't purge the most + # recent event in the room. + resp = self.helper.send( + room_id=room_id, + body="2", + tok=self.token, + ) + + second_event_id = resp.get("event_id") + + # Advance the time by another month. + self.reactor.advance(one_day_ms * 30 / 1000) + + # Check if the event has been purged from the database. + first_event = self.get_event( + room_id, first_event_id, expected_code=expected_code_for_first_event + ) + + if expected_code_for_first_event == 200: + self.assertEqual(first_event.get("content", {}).get("body"), "1", first_event) + + # Check that the event that hasn't been purged can still be retrieved. + second_event = self.get_event(room_id, second_event_id) + self.assertEqual(second_event.get("content", {}).get("body"), "2", second_event) + + def get_event(self, room_id, event_id, expected_code=200): + url = "/_matrix/client/r0/rooms/%s/event/%s" % (room_id, event_id) + + request, channel = self.make_request("GET", url, access_token=self.token) + self.render(request) + + self.assertEqual(channel.code, expected_code, channel.result) + + return channel.json_body From c16e192e2f9970cc62adfd758034244631968102 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 5 Nov 2019 15:49:43 +0000 Subject: [PATCH 02/47] Fix caching devices for remote servers in worker. When the `/keys/query` API is hit on client_reader worker Synapse may decide that it needs to resync some remote deivces. Usually this happens on master, and then gets cached. However, that fails on workers and so it falls back to fetching devices from remotes directly, which may in turn fail if the remote is down. --- synapse/handlers/e2e_keys.py | 19 ++++++-- synapse/replication/http/__init__.py | 10 +++- synapse/replication/http/devices.py | 69 ++++++++++++++++++++++++++++ 3 files changed, 94 insertions(+), 4 deletions(-) create mode 100644 synapse/replication/http/devices.py diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index f09a0b73c8..28c12753c1 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -30,6 +30,7 @@ from twisted.internet import defer from synapse.api.errors import CodeMessageException, Codes, NotFoundError, SynapseError 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 ( UserID, get_domain_from_id, @@ -53,6 +54,12 @@ class E2eKeysHandler(object): self._edu_updater = SigningKeyEduUpdater(hs, self) + self._is_master = hs.config.worker_app is None + if not self._is_master: + self._user_device_resync_client = ReplicationUserDevicesResyncRestServlet.make_client( + hs + ) + federation_registry = hs.get_federation_registry() # FIXME: switch to m.signing_key_update when MSC1756 is merged into the spec @@ -191,9 +198,15 @@ class E2eKeysHandler(object): # probably be tracking their device lists. However, we haven't # done an initial sync on the device list so we do it now. try: - user_devices = yield self.device_handler.device_list_updater.user_device_resync( - user_id - ) + if self._is_master: + user_devices = yield self.device_handler.device_list_updater.user_device_resync( + user_id + ) + else: + user_devices = yield self._user_device_resync_client( + user_id=user_id + ) + user_devices = user_devices["devices"] for device in user_devices: results[user_id] = {device["device_id"]: device["keys"]} diff --git a/synapse/replication/http/__init__.py b/synapse/replication/http/__init__.py index 81b85352b1..28dbc6fcba 100644 --- a/synapse/replication/http/__init__.py +++ b/synapse/replication/http/__init__.py @@ -14,7 +14,14 @@ # limitations under the License. from synapse.http.server import JsonResource -from synapse.replication.http import federation, login, membership, register, send_event +from synapse.replication.http import ( + devices, + federation, + login, + membership, + register, + send_event, +) REPLICATION_PREFIX = "/_synapse/replication" @@ -30,3 +37,4 @@ class ReplicationRestResource(JsonResource): federation.register_servlets(hs, self) login.register_servlets(hs, self) register.register_servlets(hs, self) + devices.register_servlets(hs, self) diff --git a/synapse/replication/http/devices.py b/synapse/replication/http/devices.py new file mode 100644 index 0000000000..795ca7b65e --- /dev/null +++ b/synapse/replication/http/devices.py @@ -0,0 +1,69 @@ +# -*- coding: utf-8 -*- +# Copyright 2018 New Vector Ltd +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging + +from synapse.replication.http._base import ReplicationEndpoint + +logger = logging.getLogger(__name__) + + +class ReplicationUserDevicesResyncRestServlet(ReplicationEndpoint): + """Notifies that a user has joined or left the room + + Request format: + + POST /_synapse/replication/user_device_resync/:user_id + + {} + + Response is equivalent to ` /_matrix/federation/v1/user/devices/:user_id` + response, e.g.: + + { + "user_id": "@alice:example.org", + "devices": [ + { + "device_id": "JLAFKJWSCS", + "keys": { ... }, + "device_display_name": "Alice's Mobile Phone" + } + ] + } + """ + + NAME = "user_device_resync" + PATH_ARGS = ("user_id",) + CACHE = False + + def __init__(self, hs): + super(ReplicationUserDevicesResyncRestServlet, self).__init__(hs) + + self.device_list_updater = hs.get_device_handler().device_list_updater + self.store = hs.get_datastore() + self.clock = hs.get_clock() + + @staticmethod + def _serialize_payload(user_id): + return {} + + async def _handle_request(self, request, user_id): + user_devices = await self.device_list_updater.user_device_resync(user_id) + + return 200, user_devices + + +def register_servlets(hs, http_server): + ReplicationUserDevicesResyncRestServlet(hs).register(http_server) From 248111bae8600a3e2d4da49c6e64b72c76219850 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 5 Nov 2019 15:54:23 +0000 Subject: [PATCH 03/47] Newsfile --- changelog.d/6332.bugfix | 1 + 1 file changed, 1 insertion(+) create mode 100644 changelog.d/6332.bugfix diff --git a/changelog.d/6332.bugfix b/changelog.d/6332.bugfix new file mode 100644 index 0000000000..b14bd7e43c --- /dev/null +++ b/changelog.d/6332.bugfix @@ -0,0 +1 @@ +Fix caching devices for remote users when using workers. From f03c9d34442368c8f2c26d2ac16b770bc451c76d Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Wed, 6 Nov 2019 15:47:40 +0000 Subject: [PATCH 04/47] Don't apply retention policy based filtering on state events As per MSC1763, 'Retention is only considered for non-state events.', so don't filter out state events based on the room's retention policy. --- synapse/visibility.py | 15 +++++++++------ tests/rest/client/test_retention.py | 10 ++++++++++ 2 files changed, 19 insertions(+), 6 deletions(-) diff --git a/synapse/visibility.py b/synapse/visibility.py index 4498c156bc..4d4141dacc 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -111,14 +111,17 @@ def filter_events_for_client( if not event.is_state() and event.sender in ignore_list: return None - retention_policy = retention_policies[event.room_id] - max_lifetime = retention_policy.get("max_lifetime") + # Don't try to apply the room's retention policy if the event is a state event, as + # MSC1763 states that retention is only considered for non-state events. + if not event.is_state(): + retention_policy = retention_policies[event.room_id] + max_lifetime = retention_policy.get("max_lifetime") - if max_lifetime is not None: - oldest_allowed_ts = storage.main.clock.time_msec() - max_lifetime + if max_lifetime is not None: + oldest_allowed_ts = storage.main.clock.time_msec() - max_lifetime - if event.origin_server_ts < oldest_allowed_ts: - return None + if event.origin_server_ts < oldest_allowed_ts: + return None if event.event_id in always_include_ids: return event diff --git a/tests/rest/client/test_retention.py b/tests/rest/client/test_retention.py index 41ea9db689..7b6f25a838 100644 --- a/tests/rest/client/test_retention.py +++ b/tests/rest/client/test_retention.py @@ -164,6 +164,12 @@ class RetentionTestCase(unittest.HomeserverTestCase): self.assertEqual(filtered_events[0].event_id, valid_event_id, filtered_events) def _test_retention_event_purged(self, room_id, increment): + # Get the create event to, later, check that we can still access it. + message_handler = self.hs.get_message_handler() + create_event = self.get_success( + message_handler.get_room_data(self.user_id, room_id, EventTypes.Create) + ) + # Send a first event to the room. This is the event we'll want to be purged at the # end of the test. resp = self.helper.send( @@ -202,6 +208,10 @@ class RetentionTestCase(unittest.HomeserverTestCase): valid_event = self.get_event(room_id, valid_event_id) self.assertEqual(valid_event.get("content", {}).get("body"), "2", valid_event) + # Check that we can still access state events that were sent before the event that + # has been purged. + self.get_event(room_id, create_event.event_id) + def get_event(self, room_id, event_id, expected_code=200): url = "/_matrix/client/r0/rooms/%s/event/%s" % (room_id, event_id) From 772d414975608c03d5690e2d8f65c7f382403a99 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Fri, 11 Oct 2019 16:05:21 +0100 Subject: [PATCH 05/47] Simplify _update_auth_events_and_context_for_auth move event_key calculation into _update_context_for_auth_events, since it's only used there. --- synapse/handlers/federation.py | 20 +++++++++----------- 1 file changed, 9 insertions(+), 11 deletions(-) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 05dd8d2671..ab152e8dcd 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -2098,11 +2098,6 @@ class FederationHandler(BaseHandler): """ event_auth_events = set(event.auth_event_ids()) - if event.is_state(): - event_key = (event.type, event.state_key) - else: - event_key = None - # if the event's auth_events refers to events which are not in our # calculated auth_events, we need to fetch those events from somewhere. # @@ -2231,13 +2226,13 @@ class FederationHandler(BaseHandler): auth_events.update(new_state) context = yield self._update_context_for_auth_events( - event, context, auth_events, event_key + event, context, auth_events ) return context @defer.inlineCallbacks - def _update_context_for_auth_events(self, event, context, auth_events, event_key): + def _update_context_for_auth_events(self, event, context, auth_events): """Update the state_ids in an event context after auth event resolution, storing the changes as a new state group. @@ -2246,18 +2241,21 @@ class FederationHandler(BaseHandler): context (synapse.events.snapshot.EventContext): initial event context - auth_events (dict[(str, str)->str]): Events to update in the event + auth_events (dict[(str, str)->EventBase]): Events to update in the event context. - event_key ((str, str)): (type, state_key) for the current event. - this will not be included in the current_state in the context. - Returns: Deferred[EventContext]: new event context """ + # exclude the state key of the new event from the current_state in the context. + if event.is_state(): + event_key = (event.type, event.state_key) + else: + event_key = None state_updates = { k: a.event_id for k, a in iteritems(auth_events) if k != event_key } + current_state_ids = yield context.get_current_state_ids(self.store) current_state_ids = dict(current_state_ids) From f8407975e7ad080d04a028771ae5a84590a19da1 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Fri, 8 Nov 2019 12:18:20 +0000 Subject: [PATCH 06/47] Update some docstrings and comments --- synapse/handlers/federation.py | 39 +++++++++++++++++++++++++++------- 1 file changed, 31 insertions(+), 8 deletions(-) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index ab152e8dcd..4bc4d57efb 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -2040,8 +2040,10 @@ class FederationHandler(BaseHandler): auth_events (dict[(str, str)->synapse.events.EventBase]): Map from (event_type, state_key) to event - What we expect the event's auth_events to be, based on the event's - position in the dag. I think? maybe?? + Normally, our calculated auth_events based on the state of the room + at the event's position in the DAG, though occasionally (eg if the + event is an outlier), may be the auth events claimed by the remote + server. Also NB that this function adds entries to it. Returns: @@ -2091,25 +2093,35 @@ class FederationHandler(BaseHandler): origin (str): event (synapse.events.EventBase): context (synapse.events.snapshot.EventContext): + auth_events (dict[(str, str)->synapse.events.EventBase]): + Map from (event_type, state_key) to event + + Normally, our calculated auth_events based on the state of the room + at the event's position in the DAG, though occasionally (eg if the + event is an outlier), may be the auth events claimed by the remote + server. + + Also NB that this function adds entries to it. Returns: defer.Deferred[EventContext]: updated context """ event_auth_events = set(event.auth_event_ids()) - # if the event's auth_events refers to events which are not in our - # calculated auth_events, we need to fetch those events from somewhere. - # - # we start by fetching them from the store, and then try calling /event_auth/. + # missing_auth is the set of the event's auth_events which we don't yet have + # in auth_events. missing_auth = event_auth_events.difference( e.event_id for e in auth_events.values() ) + # if we have missing events, we need to fetch those events from somewhere. + # + # we start by checking if they are in the store, and then try calling /event_auth/. if missing_auth: # TODO: can we use store.have_seen_events here instead? have_events = yield self.store.get_seen_events_with_rejections(missing_auth) - logger.debug("Got events %s from store", have_events) + logger.debug("Found events %s in the store", have_events) missing_auth.difference_update(have_events.keys()) else: have_events = {} @@ -2164,15 +2176,23 @@ class FederationHandler(BaseHandler): event.auth_event_ids() ) except Exception: - # FIXME: logger.exception("Failed to get auth chain") if event.internal_metadata.is_outlier(): + # XXX: given that, for an outlier, we'll be working with the + # event's *claimed* auth events rather than those we calculated: + # (a) is there any point in this test, since different_auth below will + # obviously be empty + # (b) alternatively, why don't we do it earlier? logger.info("Skipping auth_event fetch for outlier") return context # FIXME: Assumes we have and stored all the state for all the # prev_events + # + # FIXME: what does the fixme above mean? where do prev_events come into + # it, why do we care about the state for those events, and what does "have and + # stored" mean? Seems erik wrote it in c1d860870b different_auth = event_auth_events.difference( e.event_id for e in auth_events.values() ) @@ -2186,6 +2206,9 @@ class FederationHandler(BaseHandler): different_auth, ) + # now we state-resolve between our own idea of the auth events, and the remote's + # idea of them. + room_version = yield self.store.get_room_version(event.room_id) different_events = yield make_deferred_yieldable( From f41027f74678f35ad9e9eb2531c416dd58a65127 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Fri, 8 Nov 2019 12:21:28 +0000 Subject: [PATCH 07/47] Use get_events_as_list rather than lots of calls to get_event It's more efficient and clearer. --- synapse/handlers/federation.py | 24 ++++++++---------------- 1 file changed, 8 insertions(+), 16 deletions(-) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 4bc4d57efb..3d4197ed69 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -2210,26 +2210,18 @@ class FederationHandler(BaseHandler): # idea of them. room_version = yield self.store.get_room_version(event.room_id) + different_event_ids = [ + d for d in different_auth if d in have_events and not have_events[d] + ] - different_events = yield make_deferred_yieldable( - defer.gatherResults( - [ - run_in_background( - self.store.get_event, d, allow_none=True, allow_rejected=False - ) - for d in different_auth - if d in have_events and not have_events[d] - ], - consumeErrors=True, - ) - ).addErrback(unwrapFirstError) + if different_event_ids: + # XXX: currently this checks for redactions but I'm not convinced that is + # necessary? + different_events = yield self.store.get_events_as_list(different_event_ids) - if different_events: local_view = dict(auth_events) remote_view = dict(auth_events) - remote_view.update( - {(d.type, d.state_key): d for d in different_events if d} - ) + remote_view.update({(d.type, d.state_key): d for d in different_events}) new_state = yield self.state_handler.resolve_events( room_version, From c4bdf2d7855dcddb7e294e9dba458884db2be7e0 Mon Sep 17 00:00:00 2001 From: Andrew Morgan Date: Fri, 8 Nov 2019 11:42:55 +0000 Subject: [PATCH 08/47] Remove content from being sent for account data rdata stream --- synapse/replication/tcp/streams/_base.py | 6 +++--- synapse/storage/data_stores/main/account_data.py | 6 +++--- 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/synapse/replication/tcp/streams/_base.py b/synapse/replication/tcp/streams/_base.py index 9e45429d49..dd87733842 100644 --- a/synapse/replication/tcp/streams/_base.py +++ b/synapse/replication/tcp/streams/_base.py @@ -89,7 +89,7 @@ TagAccountDataStreamRow = namedtuple( ) AccountDataStreamRow = namedtuple( "AccountDataStream", - ("user_id", "room_id", "data_type", "data"), # str # str # str # dict + ("user_id", "room_id", "data_type"), # str # str # str ) GroupsStreamRow = namedtuple( "GroupsStreamRow", @@ -421,8 +421,8 @@ class AccountDataStream(Stream): results = list(room_results) results.extend( - (stream_id, user_id, None, account_data_type, content) - for stream_id, user_id, account_data_type, content in global_results + (stream_id, user_id, None, account_data_type) + for stream_id, user_id, account_data_type in global_results ) return results diff --git a/synapse/storage/data_stores/main/account_data.py b/synapse/storage/data_stores/main/account_data.py index 6afbfc0d74..22093484ed 100644 --- a/synapse/storage/data_stores/main/account_data.py +++ b/synapse/storage/data_stores/main/account_data.py @@ -184,14 +184,14 @@ class AccountDataWorkerStore(SQLBaseStore): current_id(int): The position to fetch up to. Returns: A deferred pair of lists of tuples of stream_id int, user_id string, - room_id string, type string, and content string. + room_id string, and type string. """ if last_room_id == current_id and last_global_id == current_id: return defer.succeed(([], [])) def get_updated_account_data_txn(txn): sql = ( - "SELECT stream_id, user_id, account_data_type, content" + "SELECT stream_id, user_id, account_data_type" " FROM account_data WHERE ? < stream_id AND stream_id <= ?" " ORDER BY stream_id ASC LIMIT ?" ) @@ -199,7 +199,7 @@ class AccountDataWorkerStore(SQLBaseStore): global_results = txn.fetchall() sql = ( - "SELECT stream_id, user_id, room_id, account_data_type, content" + "SELECT stream_id, user_id, room_id, account_data_type" " FROM room_account_data WHERE ? < stream_id AND stream_id <= ?" " ORDER BY stream_id ASC LIMIT ?" ) From 318dd21b4767039014f73917a4dddb9fc885bc56 Mon Sep 17 00:00:00 2001 From: Andrew Morgan Date: Fri, 8 Nov 2019 15:45:08 +0000 Subject: [PATCH 09/47] Add changelog --- changelog.d/6333.bugfix | 1 + 1 file changed, 1 insertion(+) create mode 100644 changelog.d/6333.bugfix diff --git a/changelog.d/6333.bugfix b/changelog.d/6333.bugfix new file mode 100644 index 0000000000..a25d6ef3cb --- /dev/null +++ b/changelog.d/6333.bugfix @@ -0,0 +1 @@ +Prevent account data syncs getting lost across TCP replication. \ No newline at end of file From cd96b4586f8dfa8a6857e1bee7de2bd9660238d5 Mon Sep 17 00:00:00 2001 From: Andrew Morgan Date: Fri, 8 Nov 2019 15:45:45 +0000 Subject: [PATCH 10/47] lint --- synapse/replication/tcp/streams/_base.py | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/synapse/replication/tcp/streams/_base.py b/synapse/replication/tcp/streams/_base.py index dd87733842..8512923eae 100644 --- a/synapse/replication/tcp/streams/_base.py +++ b/synapse/replication/tcp/streams/_base.py @@ -88,8 +88,7 @@ TagAccountDataStreamRow = namedtuple( "TagAccountDataStreamRow", ("user_id", "room_id", "data") # str # str # dict ) AccountDataStreamRow = namedtuple( - "AccountDataStream", - ("user_id", "room_id", "data_type"), # str # str # str + "AccountDataStream", ("user_id", "room_id", "data_type") # str # str # str ) GroupsStreamRow = namedtuple( "GroupsStreamRow", From 20d687516fdc34a4be19dcbbadd8a9a9726203e4 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Fri, 8 Nov 2019 16:17:02 +0000 Subject: [PATCH 11/47] newsfile --- changelog.d/6343.misc | 1 + 1 file changed, 1 insertion(+) create mode 100644 changelog.d/6343.misc diff --git a/changelog.d/6343.misc b/changelog.d/6343.misc new file mode 100644 index 0000000000..d9a44389b9 --- /dev/null +++ b/changelog.d/6343.misc @@ -0,0 +1 @@ +Refactor some code in the event authentication path for clarity. From 7c24d0f443724082376c89f9f75954d81f524a8e Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Tue, 19 Nov 2019 13:22:37 +0000 Subject: [PATCH 12/47] Lint --- synapse/config/server.py | 39 ++++++++----- synapse/handlers/pagination.py | 17 ++---- synapse/storage/data_stores/main/room.py | 49 ++++++++-------- tests/rest/client/test_retention.py | 73 ++++++++---------------- 4 files changed, 77 insertions(+), 101 deletions(-) diff --git a/synapse/config/server.py b/synapse/config/server.py index aa93a416f1..8a55ffac4f 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -19,7 +19,7 @@ import logging import os.path import re from textwrap import indent -from typing import List +from typing import List, Dict, Optional import attr import yaml @@ -287,13 +287,17 @@ class ServerConfig(Config): self.retention_default_min_lifetime = None self.retention_default_max_lifetime = None - self.retention_allowed_lifetime_min = retention_config.get("allowed_lifetime_min") + self.retention_allowed_lifetime_min = retention_config.get( + "allowed_lifetime_min" + ) if self.retention_allowed_lifetime_min is not None: self.retention_allowed_lifetime_min = self.parse_duration( self.retention_allowed_lifetime_min ) - self.retention_allowed_lifetime_max = retention_config.get("allowed_lifetime_max") + self.retention_allowed_lifetime_max = retention_config.get( + "allowed_lifetime_max" + ) if self.retention_allowed_lifetime_max is not None: self.retention_allowed_lifetime_max = self.parse_duration( self.retention_allowed_lifetime_max @@ -302,14 +306,15 @@ class ServerConfig(Config): if ( self.retention_allowed_lifetime_min is not None and self.retention_allowed_lifetime_max is not None - and self.retention_allowed_lifetime_min > self.retention_allowed_lifetime_max + and self.retention_allowed_lifetime_min + > self.retention_allowed_lifetime_max ): raise ConfigError( "Invalid retention policy limits: 'allowed_lifetime_min' can not be" " greater than 'allowed_lifetime_max'" ) - self.retention_purge_jobs = [] + self.retention_purge_jobs = [] # type: List[Dict[str, Optional[int]]] for purge_job_config in retention_config.get("purge_jobs", []): interval_config = purge_job_config.get("interval") @@ -342,18 +347,22 @@ class ServerConfig(Config): " 'longest_max_lifetime' value." ) - self.retention_purge_jobs.append({ - "interval": interval, - "shortest_max_lifetime": shortest_max_lifetime, - "longest_max_lifetime": longest_max_lifetime, - }) + self.retention_purge_jobs.append( + { + "interval": interval, + "shortest_max_lifetime": shortest_max_lifetime, + "longest_max_lifetime": longest_max_lifetime, + } + ) if not self.retention_purge_jobs: - self.retention_purge_jobs = [{ - "interval": self.parse_duration("1d"), - "shortest_max_lifetime": None, - "longest_max_lifetime": None, - }] + self.retention_purge_jobs = [ + { + "interval": self.parse_duration("1d"), + "shortest_max_lifetime": None, + "longest_max_lifetime": None, + } + ] self.listeners = [] # type: List[dict] for listener in config.get("listeners", []): diff --git a/synapse/handlers/pagination.py b/synapse/handlers/pagination.py index e1800177fa..d122c11a4d 100644 --- a/synapse/handlers/pagination.py +++ b/synapse/handlers/pagination.py @@ -154,20 +154,17 @@ class PaginationHandler(object): # Figure out what token we should start purging at. ts = self.clock.time_msec() - max_lifetime - stream_ordering = ( - yield self.store.find_first_stream_ordering_after_ts(ts) - ) + stream_ordering = yield self.store.find_first_stream_ordering_after_ts(ts) - r = ( - yield self.store.get_room_event_after_stream_ordering( - room_id, stream_ordering, - ) + r = yield self.store.get_room_event_after_stream_ordering( + room_id, stream_ordering, ) if not r: logger.warning( "[purge] purging events not possible: No event found " "(ts %i => stream_ordering %i)", - ts, stream_ordering, + ts, + stream_ordering, ) continue @@ -186,9 +183,7 @@ class PaginationHandler(object): # the background so that it's not blocking any other operation apart from # other purges in the same room. run_as_background_process( - "_purge_history", - self._purge_history, - purge_id, room_id, token, True, + "_purge_history", self._purge_history, purge_id, room_id, token, True, ) def start_purge_history(self, room_id, token, delete_local_events=False): diff --git a/synapse/storage/data_stores/main/room.py b/synapse/storage/data_stores/main/room.py index 54a7d24c73..7fceae59ca 100644 --- a/synapse/storage/data_stores/main/room.py +++ b/synapse/storage/data_stores/main/room.py @@ -334,8 +334,9 @@ class RoomStore(RoomWorkerStore, SearchStore): WHERE state.room_id > ? AND state.type = '%s' ORDER BY state.room_id ASC LIMIT ?; - """ % EventTypes.Retention, - (last_room, batch_size) + """ + % EventTypes.Retention, + (last_room, batch_size), ) rows = self.cursor_to_dict(txn) @@ -358,15 +359,13 @@ class RoomStore(RoomWorkerStore, SearchStore): "event_id": row["event_id"], "min_lifetime": retention_policy.get("min_lifetime"), "max_lifetime": retention_policy.get("max_lifetime"), - } + }, ) logger.info("Inserted %d rows into room_retention", len(rows)) self._background_update_progress_txn( - txn, "insert_room_retention", { - "room_id": rows[-1]["room_id"], - } + txn, "insert_room_retention", {"room_id": rows[-1]["room_id"]} ) if batch_size > len(rows): @@ -375,8 +374,7 @@ class RoomStore(RoomWorkerStore, SearchStore): return False end = yield self.runInteraction( - "insert_room_retention", - _background_insert_retention_txn, + "insert_room_retention", _background_insert_retention_txn, ) if end: @@ -585,17 +583,15 @@ class RoomStore(RoomWorkerStore, SearchStore): ) def _store_retention_policy_for_room_txn(self, txn, event): - if ( - hasattr(event, "content") - and ("min_lifetime" in event.content or "max_lifetime" in event.content) + if hasattr(event, "content") and ( + "min_lifetime" in event.content or "max_lifetime" in event.content ): if ( - ("min_lifetime" in event.content and not isinstance( - event.content.get("min_lifetime"), integer_types - )) - or ("max_lifetime" in event.content and not isinstance( - event.content.get("max_lifetime"), integer_types - )) + "min_lifetime" in event.content + and not isinstance(event.content.get("min_lifetime"), integer_types) + ) or ( + "max_lifetime" in event.content + and not isinstance(event.content.get("max_lifetime"), integer_types) ): # Ignore the event if one of the value isn't an integer. return @@ -798,7 +794,9 @@ class RoomStore(RoomWorkerStore, SearchStore): return local_media_mxcs, remote_media_mxcs @defer.inlineCallbacks - def get_rooms_for_retention_period_in_range(self, min_ms, max_ms, include_null=False): + def get_rooms_for_retention_period_in_range( + self, min_ms, max_ms, include_null=False + ): """Retrieves all of the rooms within the given retention range. Optionally includes the rooms which don't have a retention policy. @@ -904,23 +902,24 @@ class RoomStore(RoomWorkerStore, SearchStore): INNER JOIN current_state_events USING (event_id, room_id) WHERE room_id = ?; """, - (room_id,) + (room_id,), ) return self.cursor_to_dict(txn) ret = yield self.runInteraction( - "get_retention_policy_for_room", - get_retention_policy_for_room_txn, + "get_retention_policy_for_room", get_retention_policy_for_room_txn, ) # If we don't know this room ID, ret will be None, in this case return the default # policy. if not ret: - defer.returnValue({ - "min_lifetime": self.config.retention_default_min_lifetime, - "max_lifetime": self.config.retention_default_max_lifetime, - }) + defer.returnValue( + { + "min_lifetime": self.config.retention_default_min_lifetime, + "max_lifetime": self.config.retention_default_max_lifetime, + } + ) row = ret[0] diff --git a/tests/rest/client/test_retention.py b/tests/rest/client/test_retention.py index 7b6f25a838..6bf485c239 100644 --- a/tests/rest/client/test_retention.py +++ b/tests/rest/client/test_retention.py @@ -61,9 +61,7 @@ class RetentionTestCase(unittest.HomeserverTestCase): self.helper.send_state( room_id=room_id, event_type=EventTypes.Retention, - body={ - "max_lifetime": one_day_ms * 4, - }, + body={"max_lifetime": one_day_ms * 4}, tok=self.token, expect_code=400, ) @@ -71,9 +69,7 @@ class RetentionTestCase(unittest.HomeserverTestCase): self.helper.send_state( room_id=room_id, event_type=EventTypes.Retention, - body={ - "max_lifetime": one_hour_ms, - }, + body={"max_lifetime": one_hour_ms}, tok=self.token, expect_code=400, ) @@ -89,9 +85,7 @@ class RetentionTestCase(unittest.HomeserverTestCase): self.helper.send_state( room_id=room_id, event_type=EventTypes.Retention, - body={ - "max_lifetime": lifetime, - }, + body={"max_lifetime": lifetime}, tok=self.token, ) @@ -115,20 +109,12 @@ class RetentionTestCase(unittest.HomeserverTestCase): events = [] # Send a first event, which should be filtered out at the end of the test. - resp = self.helper.send( - room_id=room_id, - body="1", - tok=self.token, - ) + resp = self.helper.send(room_id=room_id, body="1", tok=self.token) # Get the event from the store so that we end up with a FrozenEvent that we can # give to filter_events_for_client. We need to do this now because the event won't # be in the database anymore after it has expired. - events.append(self.get_success( - store.get_event( - resp.get("event_id") - ) - )) + events.append(self.get_success(store.get_event(resp.get("event_id")))) # Advance the time by 2 days. We're using the default retention policy, therefore # after this the first event will still be valid. @@ -143,20 +129,16 @@ class RetentionTestCase(unittest.HomeserverTestCase): valid_event_id = resp.get("event_id") - events.append(self.get_success( - store.get_event( - valid_event_id - ) - )) + events.append(self.get_success(store.get_event(valid_event_id))) # Advance the time by anothe 2 days. After this, the first event should be # outdated but not the second one. self.reactor.advance(one_day_ms * 2 / 1000) # Run filter_events_for_client with our list of FrozenEvents. - filtered_events = self.get_success(filter_events_for_client( - storage, self.user_id, events - )) + filtered_events = self.get_success( + filter_events_for_client(storage, self.user_id, events) + ) # We should only get one event back. self.assertEqual(len(filtered_events), 1, filtered_events) @@ -172,28 +154,22 @@ class RetentionTestCase(unittest.HomeserverTestCase): # Send a first event to the room. This is the event we'll want to be purged at the # end of the test. - resp = self.helper.send( - room_id=room_id, - body="1", - tok=self.token, - ) + resp = self.helper.send(room_id=room_id, body="1", tok=self.token) expired_event_id = resp.get("event_id") # Check that we can retrieve the event. expired_event = self.get_event(room_id, expired_event_id) - self.assertEqual(expired_event.get("content", {}).get("body"), "1", expired_event) + self.assertEqual( + expired_event.get("content", {}).get("body"), "1", expired_event + ) # Advance the time. self.reactor.advance(increment / 1000) # Send another event. We need this because the purge job won't purge the most # recent event in the room. - resp = self.helper.send( - room_id=room_id, - body="2", - tok=self.token, - ) + resp = self.helper.send(room_id=room_id, body="2", tok=self.token) valid_event_id = resp.get("event_id") @@ -240,8 +216,7 @@ class RetentionNoDefaultPolicyTestCase(unittest.HomeserverTestCase): mock_federation_client = Mock(spec=["backfill"]) self.hs = self.setup_test_homeserver( - config=config, - federation_client=mock_federation_client, + config=config, federation_client=mock_federation_client, ) return self.hs @@ -268,9 +243,7 @@ class RetentionNoDefaultPolicyTestCase(unittest.HomeserverTestCase): self.helper.send_state( room_id=room_id, event_type=EventTypes.Retention, - body={ - "max_lifetime": one_day_ms * 35, - }, + body={"max_lifetime": one_day_ms * 35}, tok=self.token, ) @@ -289,18 +262,16 @@ class RetentionNoDefaultPolicyTestCase(unittest.HomeserverTestCase): # Check that we can retrieve the event. expired_event = self.get_event(room_id, first_event_id) - self.assertEqual(expired_event.get("content", {}).get("body"), "1", expired_event) + self.assertEqual( + expired_event.get("content", {}).get("body"), "1", expired_event + ) # Advance the time by a month. self.reactor.advance(one_day_ms * 30 / 1000) # Send another event. We need this because the purge job won't purge the most # recent event in the room. - resp = self.helper.send( - room_id=room_id, - body="2", - tok=self.token, - ) + resp = self.helper.send(room_id=room_id, body="2", tok=self.token) second_event_id = resp.get("event_id") @@ -313,7 +284,9 @@ class RetentionNoDefaultPolicyTestCase(unittest.HomeserverTestCase): ) if expected_code_for_first_event == 200: - self.assertEqual(first_event.get("content", {}).get("body"), "1", first_event) + self.assertEqual( + first_event.get("content", {}).get("body"), "1", first_event + ) # Check that the event that hasn't been purged can still be retrieved. second_event = self.get_event(room_id, second_event_id) From bf9a11c54d3c9ca1e4fa9420b567efceb1e46d5b Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Tue, 19 Nov 2019 13:30:04 +0000 Subject: [PATCH 13/47] Lint again --- synapse/config/server.py | 2 +- tests/rest/client/test_retention.py | 12 ++---------- 2 files changed, 3 insertions(+), 11 deletions(-) diff --git a/synapse/config/server.py b/synapse/config/server.py index 8a55ffac4f..ed916e1400 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -19,7 +19,7 @@ import logging import os.path import re from textwrap import indent -from typing import List, Dict, Optional +from typing import Dict, List, Optional import attr import yaml diff --git a/tests/rest/client/test_retention.py b/tests/rest/client/test_retention.py index 6bf485c239..9e549d8a91 100644 --- a/tests/rest/client/test_retention.py +++ b/tests/rest/client/test_retention.py @@ -121,11 +121,7 @@ class RetentionTestCase(unittest.HomeserverTestCase): self.reactor.advance(one_day_ms * 2 / 1000) # Send another event, which shouldn't get filtered out. - resp = self.helper.send( - room_id=room_id, - body="2", - tok=self.token, - ) + resp = self.helper.send(room_id=room_id, body="2", tok=self.token) valid_event_id = resp.get("event_id") @@ -252,11 +248,7 @@ class RetentionNoDefaultPolicyTestCase(unittest.HomeserverTestCase): def _test_retention(self, room_id, expected_code_for_first_event=200): # Send a first event to the room. This is the event we'll want to be purged at the # end of the test. - resp = self.helper.send( - room_id=room_id, - body="1", - tok=self.token, - ) + resp = self.helper.send(room_id=room_id, body="1", tok=self.token) first_event_id = resp.get("event_id") From 97b863fe32c03ce30c4cd5bda1479c1f9a03da83 Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Tue, 19 Nov 2019 13:33:58 +0000 Subject: [PATCH 14/47] Lint again --- synapse/config/server.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/synapse/config/server.py b/synapse/config/server.py index ed916e1400..4404953e27 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -879,7 +879,7 @@ class ServerConfig(Config): # Defaults to `28d`. Set to `null` to disable clearing out of old rows. # #user_ips_max_age: 14d - + # Message retention policy at the server level. # # Room admins and mods can define a retention period for their rooms using the From a6fc6754f8a680aa31e2be2b10d02e953d4ff368 Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Tue, 19 Nov 2019 14:07:39 +0000 Subject: [PATCH 15/47] Fix 3PID invite exchange --- synapse/handlers/federation.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 3994137d18..ab82f83625 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -2569,7 +2569,7 @@ class FederationHandler(BaseHandler): event, context = yield self.event_creation_handler.create_new_client_event( builder=builder ) - EventValidator().validate_new(event) + EventValidator().validate_new(event, self.config) return (event, context) @defer.inlineCallbacks From cdd3cb870d2e81fc28073abade2f6eaf4ec54fdd Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Tue, 19 Nov 2019 14:40:21 +0000 Subject: [PATCH 16/47] Fix worker mode --- synapse/storage/data_stores/main/room.py | 112 +++++++++++------------ 1 file changed, 56 insertions(+), 56 deletions(-) diff --git a/synapse/storage/data_stores/main/room.py b/synapse/storage/data_stores/main/room.py index 7fceae59ca..b7f9024811 100644 --- a/synapse/storage/data_stores/main/room.py +++ b/synapse/storage/data_stores/main/room.py @@ -303,6 +303,62 @@ class RoomWorkerStore(SQLBaseStore): else: return None + @cachedInlineCallbacks() + def get_retention_policy_for_room(self, room_id): + """Get the retention policy for a given room. + + If no retention policy has been found for this room, returns a policy defined + by the configured default policy (which has None as both the 'min_lifetime' and + the 'max_lifetime' if no default policy has been defined in the server's + configuration). + + Args: + room_id (str): The ID of the room to get the retention policy of. + + Returns: + dict[int, int]: "min_lifetime" and "max_lifetime" for this room. + """ + + def get_retention_policy_for_room_txn(txn): + txn.execute( + """ + SELECT min_lifetime, max_lifetime FROM room_retention + INNER JOIN current_state_events USING (event_id, room_id) + WHERE room_id = ?; + """, + (room_id,), + ) + + return self.cursor_to_dict(txn) + + ret = yield self.runInteraction( + "get_retention_policy_for_room", get_retention_policy_for_room_txn, + ) + + # If we don't know this room ID, ret will be None, in this case return the default + # policy. + if not ret: + defer.returnValue( + { + "min_lifetime": self.config.retention_default_min_lifetime, + "max_lifetime": self.config.retention_default_max_lifetime, + } + ) + + row = ret[0] + + # If one of the room's policy's attributes isn't defined, use the matching + # attribute from the default policy. + # The default values will be None if no default policy has been defined, or if one + # of the attributes is missing from the default policy. + if row["min_lifetime"] is None: + row["min_lifetime"] = self.config.retention_default_min_lifetime + + if row["max_lifetime"] is None: + row["max_lifetime"] = self.config.retention_default_max_lifetime + + defer.returnValue(row) + class RoomStore(RoomWorkerStore, SearchStore): def __init__(self, db_conn, hs): @@ -878,59 +934,3 @@ class RoomStore(RoomWorkerStore, SearchStore): ) defer.returnValue(rooms) - - @cachedInlineCallbacks() - def get_retention_policy_for_room(self, room_id): - """Get the retention policy for a given room. - - If no retention policy has been found for this room, returns a policy defined - by the configured default policy (which has None as both the 'min_lifetime' and - the 'max_lifetime' if no default policy has been defined in the server's - configuration). - - Args: - room_id (str): The ID of the room to get the retention policy of. - - Returns: - dict[int, int]: "min_lifetime" and "max_lifetime" for this room. - """ - - def get_retention_policy_for_room_txn(txn): - txn.execute( - """ - SELECT min_lifetime, max_lifetime FROM room_retention - INNER JOIN current_state_events USING (event_id, room_id) - WHERE room_id = ?; - """, - (room_id,), - ) - - return self.cursor_to_dict(txn) - - ret = yield self.runInteraction( - "get_retention_policy_for_room", get_retention_policy_for_room_txn, - ) - - # If we don't know this room ID, ret will be None, in this case return the default - # policy. - if not ret: - defer.returnValue( - { - "min_lifetime": self.config.retention_default_min_lifetime, - "max_lifetime": self.config.retention_default_max_lifetime, - } - ) - - row = ret[0] - - # If one of the room's policy's attributes isn't defined, use the matching - # attribute from the default policy. - # The default values will be None if no default policy has been defined, or if one - # of the attributes is missing from the default policy. - if row["min_lifetime"] is None: - row["min_lifetime"] = self.config.retention_default_min_lifetime - - if row["max_lifetime"] is None: - row["max_lifetime"] = self.config.retention_default_max_lifetime - - defer.returnValue(row) From b7367c339db153824fb47728d3eebe2f944530e6 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Mon, 25 Nov 2019 13:26:59 +0000 Subject: [PATCH 17/47] Fix exceptions from background database update for event labels. (#6407) Add some exception handling here so that events whose json cannot be parsed are ignored rather than getting us stuck in a loop. Fixes #6404. --- changelog.d/6407.bugfix | 1 + .../data_stores/main/events_bg_updates.py | 41 +++++++++++-------- 2 files changed, 25 insertions(+), 17 deletions(-) create mode 100644 changelog.d/6407.bugfix diff --git a/changelog.d/6407.bugfix b/changelog.d/6407.bugfix new file mode 100644 index 0000000000..0fdbf2a781 --- /dev/null +++ b/changelog.d/6407.bugfix @@ -0,0 +1 @@ +Fix a bug which could cause the background database update hander for event labels to get stuck in a loop raising exceptions. diff --git a/synapse/storage/data_stores/main/events_bg_updates.py b/synapse/storage/data_stores/main/events_bg_updates.py index 0ed59ef48e..aa87f9abc5 100644 --- a/synapse/storage/data_stores/main/events_bg_updates.py +++ b/synapse/storage/data_stores/main/events_bg_updates.py @@ -530,24 +530,31 @@ class EventsBackgroundUpdatesStore(BackgroundUpdateStore): nbrows = 0 last_row_event_id = "" for (event_id, event_json_raw) in results: - event_json = json.loads(event_json_raw) + try: + event_json = json.loads(event_json_raw) - self._simple_insert_many_txn( - txn=txn, - table="event_labels", - values=[ - { - "event_id": event_id, - "label": label, - "room_id": event_json["room_id"], - "topological_ordering": event_json["depth"], - } - for label in event_json["content"].get( - EventContentFields.LABELS, [] - ) - if isinstance(label, str) - ], - ) + self._simple_insert_many_txn( + txn=txn, + table="event_labels", + values=[ + { + "event_id": event_id, + "label": label, + "room_id": event_json["room_id"], + "topological_ordering": event_json["depth"], + } + for label in event_json["content"].get( + EventContentFields.LABELS, [] + ) + if isinstance(label, str) + ], + ) + except Exception as e: + logger.warning( + "Unable to load event %s (no labels will be imported): %s", + event_id, + e, + ) nbrows += 1 last_row_event_id = event_id From f9c9e1f07646262cb064782d4bf427dd1634617f Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Mon, 25 Nov 2019 13:28:12 +0000 Subject: [PATCH 18/47] 1.6.0rc2 --- CHANGES.md | 9 +++++++++ changelog.d/6407.bugfix | 1 - synapse/__init__.py | 2 +- 3 files changed, 10 insertions(+), 2 deletions(-) delete mode 100644 changelog.d/6407.bugfix diff --git a/CHANGES.md b/CHANGES.md index f4f61db5d4..d26bc7a86f 100644 --- a/CHANGES.md +++ b/CHANGES.md @@ -1,3 +1,12 @@ +Synapse 1.6.0rc2 (2019-11-25) +============================= + +Bugfixes +-------- + +- Fix a bug which could cause the background database update hander for event labels to get stuck in a loop raising exceptions. ([\#6407](https://github.com/matrix-org/synapse/issues/6407)) + + Synapse 1.6.0rc1 (2019-11-20) ============================= diff --git a/changelog.d/6407.bugfix b/changelog.d/6407.bugfix deleted file mode 100644 index 0fdbf2a781..0000000000 --- a/changelog.d/6407.bugfix +++ /dev/null @@ -1 +0,0 @@ -Fix a bug which could cause the background database update hander for event labels to get stuck in a loop raising exceptions. diff --git a/synapse/__init__.py b/synapse/__init__.py index 1d962f5dc8..051c83774e 100644 --- a/synapse/__init__.py +++ b/synapse/__init__.py @@ -36,7 +36,7 @@ try: except ImportError: pass -__version__ = "1.6.0rc1" +__version__ = "1.6.0rc2" if bool(os.environ.get("SYNAPSE_TEST_PATCH_LOG_CONTEXTS", False)): # We import here so that we don't have to install a bunch of deps when From c01d5435843ad4af3d520851e86d9938b47b2d12 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Mon, 25 Nov 2019 21:03:17 +0000 Subject: [PATCH 19/47] Make sure that we close cursors before returning from a query (#6408) There are lots of words in the comment as to why this is a good idea. Fixes #6403. --- changelog.d/6408.bugfix | 1 + synapse/storage/_base.py | 51 ++++++++++++++++---- synapse/storage/data_stores/main/receipts.py | 2 +- 3 files changed, 44 insertions(+), 10 deletions(-) create mode 100644 changelog.d/6408.bugfix diff --git a/changelog.d/6408.bugfix b/changelog.d/6408.bugfix new file mode 100644 index 0000000000..c9babe599b --- /dev/null +++ b/changelog.d/6408.bugfix @@ -0,0 +1 @@ +Fix an intermittent exception when handling read-receipts. diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py index 6b8a9cd89a..459901ac60 100644 --- a/synapse/storage/_base.py +++ b/synapse/storage/_base.py @@ -409,16 +409,15 @@ class SQLBaseStore(object): i = 0 N = 5 while True: + cursor = LoggingTransaction( + conn.cursor(), + name, + self.database_engine, + after_callbacks, + exception_callbacks, + ) try: - txn = conn.cursor() - txn = LoggingTransaction( - txn, - name, - self.database_engine, - after_callbacks, - exception_callbacks, - ) - r = func(txn, *args, **kwargs) + r = func(cursor, *args, **kwargs) conn.commit() return r except self.database_engine.module.OperationalError as e: @@ -456,6 +455,40 @@ class SQLBaseStore(object): ) continue raise + finally: + # we're either about to retry with a new cursor, or we're about to + # release the connection. Once we release the connection, it could + # get used for another query, which might do a conn.rollback(). + # + # In the latter case, even though that probably wouldn't affect the + # results of this transaction, python's sqlite will reset all + # statements on the connection [1], which will make our cursor + # invalid [2]. + # + # In any case, continuing to read rows after commit()ing seems + # dubious from the PoV of ACID transactional semantics + # (sqlite explicitly says that once you commit, you may see rows + # from subsequent updates.) + # + # In psycopg2, cursors are essentially a client-side fabrication - + # all the data is transferred to the client side when the statement + # finishes executing - so in theory we could go on streaming results + # from the cursor, but attempting to do so would make us + # incompatible with sqlite, so let's make sure we're not doing that + # by closing the cursor. + # + # (*named* cursors in psycopg2 are different and are proper server- + # side things, but (a) we don't use them and (b) they are implicitly + # closed by ending the transaction anyway.) + # + # In short, if we haven't finished with the cursor yet, that's a + # problem waiting to bite us. + # + # TL;DR: we're done with the cursor, so we can close it. + # + # [1]: https://github.com/python/cpython/blob/v3.8.0/Modules/_sqlite/connection.c#L465 + # [2]: https://github.com/python/cpython/blob/v3.8.0/Modules/_sqlite/cursor.c#L236 + cursor.close() except Exception as e: logger.debug("[TXN FAIL] {%s} %s", name, e) raise diff --git a/synapse/storage/data_stores/main/receipts.py b/synapse/storage/data_stores/main/receipts.py index 0c24430f28..8b17334ff4 100644 --- a/synapse/storage/data_stores/main/receipts.py +++ b/synapse/storage/data_stores/main/receipts.py @@ -280,7 +280,7 @@ class ReceiptsWorkerStore(SQLBaseStore): args.append(limit) txn.execute(sql, args) - return (r[0:5] + (json.loads(r[5]),) for r in txn) + return list(r[0:5] + (json.loads(r[5]),) for r in txn) return self.runInteraction( "get_all_updated_receipts", get_all_updated_receipts_txn From 35f9165e96f6261e15aadb439a5d2199bede3c99 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 26 Nov 2019 12:04:48 +0000 Subject: [PATCH 20/47] Fixup docs --- changelog.d/6332.bugfix | 2 +- synapse/replication/http/devices.py | 6 +++++- 2 files changed, 6 insertions(+), 2 deletions(-) diff --git a/changelog.d/6332.bugfix b/changelog.d/6332.bugfix index b14bd7e43c..67d5170ba0 100644 --- a/changelog.d/6332.bugfix +++ b/changelog.d/6332.bugfix @@ -1 +1 @@ -Fix caching devices for remote users when using workers. +Fix caching devices for remote users when using workers, so that we don't attempt to refetch (and potentially fail) each time a user requests devices. diff --git a/synapse/replication/http/devices.py b/synapse/replication/http/devices.py index 795ca7b65e..e32aac0a25 100644 --- a/synapse/replication/http/devices.py +++ b/synapse/replication/http/devices.py @@ -21,7 +21,11 @@ logger = logging.getLogger(__name__) class ReplicationUserDevicesResyncRestServlet(ReplicationEndpoint): - """Notifies that a user has joined or left the room + """Ask master to resync the device list for a user by contacting their + server. + + This must happen on master so that the results can be correctly cached in + the database and streamed to workers. Request format: From 4d394d6415e3e4b64a577a1c83ee8acf147ce0af Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Tue, 26 Nov 2019 12:32:37 +0000 Subject: [PATCH 21/47] remove confusing fixme --- synapse/handlers/federation.py | 6 ------ 1 file changed, 6 deletions(-) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index 66852153c4..97d045db10 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -2187,12 +2187,6 @@ class FederationHandler(BaseHandler): logger.info("Skipping auth_event fetch for outlier") return context - # FIXME: Assumes we have and stored all the state for all the - # prev_events - # - # FIXME: what does the fixme above mean? where do prev_events come into - # it, why do we care about the state for those events, and what does "have and - # stored" mean? Seems erik wrote it in c1d860870b different_auth = event_auth_events.difference( e.event_id for e in auth_events.values() ) From 65d54c5e8c1434eada5d1b670d7db2e37d68d4f1 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 26 Nov 2019 13:10:09 +0000 Subject: [PATCH 22/47] Fix phone home stats (#6418) Fix phone home stats --- changelog.d/6418.bugfix | 1 + synapse/app/homeserver.py | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) create mode 100644 changelog.d/6418.bugfix diff --git a/changelog.d/6418.bugfix b/changelog.d/6418.bugfix new file mode 100644 index 0000000000..a1f488d3a2 --- /dev/null +++ b/changelog.d/6418.bugfix @@ -0,0 +1 @@ +Fix phone home stats reporting. diff --git a/synapse/app/homeserver.py b/synapse/app/homeserver.py index 73e2c29d06..883b3fb70b 100644 --- a/synapse/app/homeserver.py +++ b/synapse/app/homeserver.py @@ -585,7 +585,7 @@ def run(hs): def performance_stats_init(): _stats_process.clear() _stats_process.append( - (int(hs.get_clock().time(), resource.getrusage(resource.RUSAGE_SELF))) + (int(hs.get_clock().time()), resource.getrusage(resource.RUSAGE_SELF)) ) def start_phone_stats_home(): From b98971e8a437eb3903506eadbefdf6cb2e0853d6 Mon Sep 17 00:00:00 2001 From: Andrew Morgan Date: Tue, 26 Nov 2019 12:15:46 +0000 Subject: [PATCH 23/47] 1.6.0 --- CHANGES.md | 9 +++++++++ changelog.d/6418.bugfix | 1 - debian/changelog | 6 ++++++ synapse/__init__.py | 2 +- 4 files changed, 16 insertions(+), 2 deletions(-) delete mode 100644 changelog.d/6418.bugfix diff --git a/CHANGES.md b/CHANGES.md index d26bc7a86f..42281483b3 100644 --- a/CHANGES.md +++ b/CHANGES.md @@ -1,3 +1,12 @@ +Synapse 1.6.0 (2019-11-26) +========================== + +Bugfixes +-------- + +- Fix phone home stats reporting. ([\#6418](https://github.com/matrix-org/synapse/issues/6418)) + + Synapse 1.6.0rc2 (2019-11-25) ============================= diff --git a/changelog.d/6418.bugfix b/changelog.d/6418.bugfix deleted file mode 100644 index a1f488d3a2..0000000000 --- a/changelog.d/6418.bugfix +++ /dev/null @@ -1 +0,0 @@ -Fix phone home stats reporting. diff --git a/debian/changelog b/debian/changelog index c4415f460a..82dae017f1 100644 --- a/debian/changelog +++ b/debian/changelog @@ -1,3 +1,9 @@ +matrix-synapse-py3 (1.6.0) stable; urgency=medium + + * New synapse release 1.6.0. + + -- Synapse Packaging team Tue, 26 Nov 2019 12:15:40 +0000 + matrix-synapse-py3 (1.5.1) stable; urgency=medium * New synapse release 1.5.1. diff --git a/synapse/__init__.py b/synapse/__init__.py index 051c83774e..53eedc0048 100644 --- a/synapse/__init__.py +++ b/synapse/__init__.py @@ -36,7 +36,7 @@ try: except ImportError: pass -__version__ = "1.6.0rc2" +__version__ = "1.6.0" if bool(os.environ.get("SYNAPSE_TEST_PATCH_LOG_CONTEXTS", False)): # We import here so that we don't have to install a bunch of deps when From 8bb7b15894462c6a0ba81f7198f23a140100331d Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 26 Nov 2019 15:50:17 +0000 Subject: [PATCH 24/47] Fix find_next_generated_user_id_localpart --- .../storage/data_stores/main/registration.py | 17 +++++++---------- 1 file changed, 7 insertions(+), 10 deletions(-) diff --git a/synapse/storage/data_stores/main/registration.py b/synapse/storage/data_stores/main/registration.py index 6a594c160c..c124bbb88b 100644 --- a/synapse/storage/data_stores/main/registration.py +++ b/synapse/storage/data_stores/main/registration.py @@ -19,7 +19,6 @@ import logging import re from six import iterkeys -from six.moves import range from twisted.internet import defer from twisted.internet.defer import Deferred @@ -482,12 +481,8 @@ class RegistrationWorkerStore(SQLBaseStore): """ Gets the localpart of the next generated user ID. - Generated user IDs are integers, and we aim for them to be as small as - we can. Unfortunately, it's possible some of them are already taken by - existing users, and there may be gaps in the already taken range. This - function returns the start of the first allocatable gap. This is to - avoid the case of ID 1000 being pre-allocated and starting at 1001 while - 0-999 are available. + Generated user IDs are integers, so we find the largest integer user ID + already taken and return that plus one. """ def _find_next_generated_user_id(txn): @@ -503,9 +498,11 @@ class RegistrationWorkerStore(SQLBaseStore): match = regex.search(user_id) if match: found.add(int(match.group(1))) - for i in range(len(found) + 1): - if i not in found: - return i + + if not found: + return 1 + + return max(found) + 1 return ( ( From ba110a2030274dc785b37b1d5ad20acc4de9c612 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 26 Nov 2019 15:54:10 +0000 Subject: [PATCH 25/47] Newsfile --- changelog.d/6420.bugfix | 1 + 1 file changed, 1 insertion(+) create mode 100644 changelog.d/6420.bugfix diff --git a/changelog.d/6420.bugfix b/changelog.d/6420.bugfix new file mode 100644 index 0000000000..aef47cccaa --- /dev/null +++ b/changelog.d/6420.bugfix @@ -0,0 +1 @@ +Fix broken guest registration when there are existing blocks of numeric user IDs. From f8f14ba466a18ee38827629b8e9ab2d5931e5713 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 26 Nov 2019 16:06:41 +0000 Subject: [PATCH 26/47] Don't construct a set --- synapse/storage/data_stores/main/registration.py | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/synapse/storage/data_stores/main/registration.py b/synapse/storage/data_stores/main/registration.py index c124bbb88b..0a3c1f0510 100644 --- a/synapse/storage/data_stores/main/registration.py +++ b/synapse/storage/data_stores/main/registration.py @@ -492,17 +492,14 @@ class RegistrationWorkerStore(SQLBaseStore): regex = re.compile(r"^@(\d+):") - found = set() + max_found = 0 for (user_id,) in txn: match = regex.search(user_id) if match: - found.add(int(match.group(1))) + max_found = max(int(match.group(1)), max_found) - if not found: - return 1 - - return max(found) + 1 + return max_found + 1 return ( ( From f0ef9708241ec65fe6f32f1ad36f719ab4ab2b53 Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Tue, 26 Nov 2019 17:49:12 +0000 Subject: [PATCH 27/47] Don't restrict the tests to v1 rooms --- tests/rest/client/test_retention.py | 2 -- 1 file changed, 2 deletions(-) diff --git a/tests/rest/client/test_retention.py b/tests/rest/client/test_retention.py index 9e549d8a91..95475bb651 100644 --- a/tests/rest/client/test_retention.py +++ b/tests/rest/client/test_retention.py @@ -34,7 +34,6 @@ class RetentionTestCase(unittest.HomeserverTestCase): def make_homeserver(self, reactor, clock): config = self.default_config() - config["default_room_version"] = "1" config["retention"] = { "enabled": True, "default_policy": { @@ -204,7 +203,6 @@ class RetentionNoDefaultPolicyTestCase(unittest.HomeserverTestCase): def make_homeserver(self, reactor, clock): config = self.default_config() - config["default_room_version"] = "1" config["retention"] = { "enabled": True, } From ef1a85e7733bc1979f48357dd59b638110285075 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Tue, 26 Nov 2019 18:10:50 +0000 Subject: [PATCH 28/47] Fix startup error when http proxy is defined. (#6421) Guess I only tested this on python 2 :/ Fixes #6419. --- changelog.d/6421.bugfix | 1 + synapse/rest/media/v1/preview_url_resource.py | 4 ++-- synapse/server.py | 4 ++-- 3 files changed, 5 insertions(+), 4 deletions(-) create mode 100644 changelog.d/6421.bugfix diff --git a/changelog.d/6421.bugfix b/changelog.d/6421.bugfix new file mode 100644 index 0000000000..7969f7f71d --- /dev/null +++ b/changelog.d/6421.bugfix @@ -0,0 +1 @@ +Fix startup error when http proxy is defined. diff --git a/synapse/rest/media/v1/preview_url_resource.py b/synapse/rest/media/v1/preview_url_resource.py index a23d6f5c75..fb0d02aa83 100644 --- a/synapse/rest/media/v1/preview_url_resource.py +++ b/synapse/rest/media/v1/preview_url_resource.py @@ -77,8 +77,8 @@ class PreviewUrlResource(DirectServeResource): treq_args={"browser_like_redirects": True}, ip_whitelist=hs.config.url_preview_ip_range_whitelist, ip_blacklist=hs.config.url_preview_ip_range_blacklist, - http_proxy=os.getenv("http_proxy"), - https_proxy=os.getenv("HTTPS_PROXY"), + http_proxy=os.getenvb(b"http_proxy"), + https_proxy=os.getenvb(b"HTTPS_PROXY"), ) self.media_repo = media_repo self.primary_base_path = media_repo.primary_base_path diff --git a/synapse/server.py b/synapse/server.py index 90c3b072e8..be9af7f986 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -318,8 +318,8 @@ class HomeServer(object): def build_proxied_http_client(self): return SimpleHttpClient( self, - http_proxy=os.getenv("http_proxy"), - https_proxy=os.getenv("HTTPS_PROXY"), + http_proxy=os.getenvb(b"http_proxy"), + https_proxy=os.getenvb(b"HTTPS_PROXY"), ) def build_room_creation_handler(self): From ce578031f4d0fe6f1eb26de4cb3d30a4175468db Mon Sep 17 00:00:00 2001 From: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com> Date: Tue, 26 Nov 2019 18:42:27 +0000 Subject: [PATCH 29/47] Remove assertion and provide a clear warning on startup for missing public_baseurl (#6379) --- changelog.d/6379.misc | 1 + synapse/config/emailconfig.py | 2 ++ synapse/config/registration.py | 7 +++++++ tests/rest/client/v2_alpha/test_register.py | 1 + 4 files changed, 11 insertions(+) create mode 100644 changelog.d/6379.misc diff --git a/changelog.d/6379.misc b/changelog.d/6379.misc new file mode 100644 index 0000000000..725c2e7d87 --- /dev/null +++ b/changelog.d/6379.misc @@ -0,0 +1 @@ +Complain on startup instead of 500'ing during runtime when `public_baseurl` isn't set when necessary. \ No newline at end of file diff --git a/synapse/config/emailconfig.py b/synapse/config/emailconfig.py index 43fad0bf8b..ac1724045f 100644 --- a/synapse/config/emailconfig.py +++ b/synapse/config/emailconfig.py @@ -146,6 +146,8 @@ class EmailConfig(Config): if k not in email_config: missing.append("email." + k) + # public_baseurl is required to build password reset and validation links that + # will be emailed to users if config.get("public_baseurl") is None: missing.append("public_baseurl") diff --git a/synapse/config/registration.py b/synapse/config/registration.py index 1f6dac69da..ee9614c5f7 100644 --- a/synapse/config/registration.py +++ b/synapse/config/registration.py @@ -106,6 +106,13 @@ class RegistrationConfig(Config): account_threepid_delegates = config.get("account_threepid_delegates") or {} self.account_threepid_delegate_email = account_threepid_delegates.get("email") self.account_threepid_delegate_msisdn = account_threepid_delegates.get("msisdn") + if self.account_threepid_delegate_msisdn and not self.public_baseurl: + raise ConfigError( + "The configuration option `public_baseurl` is required if " + "`account_threepid_delegate.msisdn` is set, such that " + "clients know where to submit validation tokens to. Please " + "configure `public_baseurl`." + ) self.default_identity_server = config.get("default_identity_server") self.allow_guest_access = config.get("allow_guest_access", False) diff --git a/tests/rest/client/v2_alpha/test_register.py b/tests/rest/client/v2_alpha/test_register.py index dab87e5edf..c0d0d2b44e 100644 --- a/tests/rest/client/v2_alpha/test_register.py +++ b/tests/rest/client/v2_alpha/test_register.py @@ -203,6 +203,7 @@ class RegisterRestServletTestCase(unittest.HomeserverTestCase): @unittest.override_config( { + "public_baseurl": "https://test_server", "enable_registration_captcha": True, "user_consent": { "version": "1", From 70c0da4d82e30f8e84d0424666ea4797e437d4ca Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Tue, 26 Nov 2019 19:00:24 +0000 Subject: [PATCH 30/47] clean up buildkite output --- .buildkite/merge_base_branch.sh | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/.buildkite/merge_base_branch.sh b/.buildkite/merge_base_branch.sh index eb7219a56d..361440fd1a 100755 --- a/.buildkite/merge_base_branch.sh +++ b/.buildkite/merge_base_branch.sh @@ -1,6 +1,6 @@ #!/usr/bin/env bash -set -ex +set -e if [[ "$BUILDKITE_BRANCH" =~ ^(develop|master|dinsic|shhs|release-.*)$ ]]; then echo "Not merging forward, as this is a release branch" @@ -18,6 +18,8 @@ else GITBASE=$BUILDKITE_PULL_REQUEST_BASE_BRANCH fi +echo "--- merge_base_branch $GITBASE" + # Show what we are before git --no-pager show -s From 9b9ee75666ffca8e14b44efec6c360c2ccbcf615 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Tue, 26 Nov 2019 18:10:50 +0000 Subject: [PATCH 31/47] Fix startup error when http proxy is defined. (#6421) Guess I only tested this on python 2 :/ Fixes #6419. --- changelog.d/6421.bugfix | 1 + synapse/rest/media/v1/preview_url_resource.py | 4 ++-- synapse/server.py | 4 ++-- 3 files changed, 5 insertions(+), 4 deletions(-) create mode 100644 changelog.d/6421.bugfix diff --git a/changelog.d/6421.bugfix b/changelog.d/6421.bugfix new file mode 100644 index 0000000000..7969f7f71d --- /dev/null +++ b/changelog.d/6421.bugfix @@ -0,0 +1 @@ +Fix startup error when http proxy is defined. diff --git a/synapse/rest/media/v1/preview_url_resource.py b/synapse/rest/media/v1/preview_url_resource.py index 15c15a12f5..87343d9db9 100644 --- a/synapse/rest/media/v1/preview_url_resource.py +++ b/synapse/rest/media/v1/preview_url_resource.py @@ -77,8 +77,8 @@ class PreviewUrlResource(DirectServeResource): treq_args={"browser_like_redirects": True}, ip_whitelist=hs.config.url_preview_ip_range_whitelist, ip_blacklist=hs.config.url_preview_ip_range_blacklist, - http_proxy=os.getenv("http_proxy"), - https_proxy=os.getenv("HTTPS_PROXY"), + http_proxy=os.getenvb(b"http_proxy"), + https_proxy=os.getenvb(b"HTTPS_PROXY"), ) self.media_repo = media_repo self.primary_base_path = media_repo.primary_base_path diff --git a/synapse/server.py b/synapse/server.py index 90c3b072e8..be9af7f986 100644 --- a/synapse/server.py +++ b/synapse/server.py @@ -318,8 +318,8 @@ class HomeServer(object): def build_proxied_http_client(self): return SimpleHttpClient( self, - http_proxy=os.getenv("http_proxy"), - https_proxy=os.getenv("HTTPS_PROXY"), + http_proxy=os.getenvb(b"http_proxy"), + https_proxy=os.getenvb(b"HTTPS_PROXY"), ) def build_room_creation_handler(self): From 6f4a63df0044a304109f7d6958be94262558a7cf Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Wed, 27 Nov 2019 18:18:33 +0000 Subject: [PATCH 32/47] Add more tests to the worker blacklist (#6429) --- .buildkite/worker-blacklist | 29 +++++++++++++++++++++++++++++ changelog.d/6429.misc | 1 + 2 files changed, 30 insertions(+) create mode 100644 changelog.d/6429.misc diff --git a/.buildkite/worker-blacklist b/.buildkite/worker-blacklist index cda5c84e94..d7908af177 100644 --- a/.buildkite/worker-blacklist +++ b/.buildkite/worker-blacklist @@ -28,3 +28,32 @@ User sees updates to presence from other users in the incremental sync. Gapped incremental syncs include all state changes Old members are included in gappy incr LL sync if they start speaking + +# new failures as of https://github.com/matrix-org/sytest/pull/732 +Device list doesn't change if remote server is down +Remote servers cannot set power levels in rooms without existing powerlevels +Remote servers should reject attempts by non-creators to set the power levels + +# new failures as of https://github.com/matrix-org/sytest/pull/753 +GET /rooms/:room_id/messages returns a message +GET /rooms/:room_id/messages lazy loads members correctly +Read receipts are sent as events +Only original members of the room can see messages from erased users +Device deletion propagates over federation +If user leaves room, remote user changes device and rejoins we see update in /sync and /keys/changes +Changing user-signing key notifies local users +Newly updated tags appear in an incremental v2 /sync +Server correctly handles incoming m.device_list_update +Local device key changes get to remote servers with correct prev_id +AS-ghosted users can use rooms via AS +Ghost user must register before joining room +Test that a message is pushed +Invites are pushed +Rooms with aliases are correctly named in pushed +Rooms with names are correctly named in pushed +Rooms with canonical alias are correctly named in pushed +Rooms with many users are correctly pushed +Don't get pushed for rooms you've muted +Rejected events are not pushed +Test that rejected pushers are removed. +Events come down the correct room diff --git a/changelog.d/6429.misc b/changelog.d/6429.misc new file mode 100644 index 0000000000..4b32cdeac6 --- /dev/null +++ b/changelog.d/6429.misc @@ -0,0 +1 @@ +Add more tests to the blacklist when running in worker mode. From 0d27aba900136514a8801b902f9a8ac69150e2c0 Mon Sep 17 00:00:00 2001 From: Hubert Chathi Date: Wed, 27 Nov 2019 16:14:44 -0500 Subject: [PATCH 33/47] add etag and count to key backup endpoints (#5858) --- changelog.d/5858.feature | 1 + synapse/handlers/e2e_room_keys.py | 126 ++++++---- synapse/rest/client/v2_alpha/room_keys.py | 8 +- .../storage/data_stores/main/e2e_room_keys.py | 226 +++++++++++++----- .../main/schema/delta/56/room_key_etag.sql | 17 ++ tests/handlers/test_e2e_room_keys.py | 31 +++ tests/storage/test_e2e_room_keys.py | 8 +- 7 files changed, 295 insertions(+), 122 deletions(-) create mode 100644 changelog.d/5858.feature create mode 100644 synapse/storage/data_stores/main/schema/delta/56/room_key_etag.sql diff --git a/changelog.d/5858.feature b/changelog.d/5858.feature new file mode 100644 index 0000000000..55ee93051e --- /dev/null +++ b/changelog.d/5858.feature @@ -0,0 +1 @@ +Add etag and count fields to key backup endpoints to help clients guess if there are new keys. diff --git a/synapse/handlers/e2e_room_keys.py b/synapse/handlers/e2e_room_keys.py index 0cea445f0d..f1b4424a02 100644 --- a/synapse/handlers/e2e_room_keys.py +++ b/synapse/handlers/e2e_room_keys.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2017, 2018 New Vector Ltd +# 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. @@ -103,14 +104,35 @@ class E2eRoomKeysHandler(object): rooms session_id(string): session ID to delete keys for, for None to delete keys for all sessions + Raises: + NotFoundError: if the backup version does not exist Returns: - A deferred of the deletion transaction + A dict containing the count and etag for the backup version """ # lock for consistency with uploading with (yield self._upload_linearizer.queue(user_id)): + # make sure the backup version exists + try: + version_info = yield self.store.get_e2e_room_keys_version_info( + user_id, version + ) + except StoreError as e: + if e.code == 404: + raise NotFoundError("Unknown backup version") + else: + raise + yield self.store.delete_e2e_room_keys(user_id, version, room_id, session_id) + version_etag = version_info["etag"] + 1 + yield self.store.update_e2e_room_keys_version( + user_id, version, None, version_etag + ) + + count = yield self.store.count_e2e_room_keys(user_id, version) + return {"etag": str(version_etag), "count": count} + @trace @defer.inlineCallbacks def upload_room_keys(self, user_id, version, room_keys): @@ -138,6 +160,9 @@ class E2eRoomKeysHandler(object): } } + Returns: + A dict containing the count and etag for the backup version + Raises: NotFoundError: if there are no versions defined RoomKeysVersionError: if the uploaded version is not the current version @@ -171,59 +196,62 @@ class E2eRoomKeysHandler(object): else: raise - # go through the room_keys. - # XXX: this should/could be done concurrently, given we're in a lock. + # Fetch any existing room keys for the sessions that have been + # submitted. Then compare them with the submitted keys. If the + # key is new, insert it; if the key should be updated, then update + # it; otherwise, drop it. + existing_keys = yield self.store.get_e2e_room_keys_multi( + user_id, version, room_keys["rooms"] + ) + to_insert = [] # batch the inserts together + changed = False # if anything has changed, we need to update the etag for room_id, room in iteritems(room_keys["rooms"]): - for session_id, session in iteritems(room["sessions"]): - yield self._upload_room_key( - user_id, version, room_id, session_id, session + for session_id, room_key in iteritems(room["sessions"]): + log_kv( + { + "message": "Trying to upload room key", + "room_id": room_id, + "session_id": session_id, + "user_id": user_id, + } ) + current_room_key = existing_keys.get(room_id, {}).get(session_id) + if current_room_key: + if self._should_replace_room_key(current_room_key, room_key): + log_kv({"message": "Replacing room key."}) + # updates are done one at a time in the DB, so send + # updates right away rather than batching them up, + # like we do with the inserts + yield self.store.update_e2e_room_key( + user_id, version, room_id, session_id, room_key + ) + changed = True + else: + log_kv({"message": "Not replacing room_key."}) + else: + log_kv( + { + "message": "Room key not found.", + "room_id": room_id, + "user_id": user_id, + } + ) + log_kv({"message": "Replacing room key."}) + to_insert.append((room_id, session_id, room_key)) + changed = True - @defer.inlineCallbacks - def _upload_room_key(self, user_id, version, room_id, session_id, room_key): - """Upload a given room_key for a given room and session into a given - version of the backup. Merges the key with any which might already exist. + if len(to_insert): + yield self.store.add_e2e_room_keys(user_id, version, to_insert) - Args: - user_id(str): the user whose backup we're setting - version(str): the version ID of the backup we're updating - room_id(str): the ID of the room whose keys we're setting - session_id(str): the session whose room_key we're setting - room_key(dict): the room_key being set - """ - log_kv( - { - "message": "Trying to upload room key", - "room_id": room_id, - "session_id": session_id, - "user_id": user_id, - } - ) - # get the room_key for this particular row - current_room_key = None - try: - current_room_key = yield self.store.get_e2e_room_key( - user_id, version, room_id, session_id - ) - except StoreError as e: - if e.code == 404: - log_kv( - { - "message": "Room key not found.", - "room_id": room_id, - "user_id": user_id, - } + version_etag = version_info["etag"] + if changed: + version_etag = version_etag + 1 + yield self.store.update_e2e_room_keys_version( + user_id, version, None, version_etag ) - else: - raise - if self._should_replace_room_key(current_room_key, room_key): - log_kv({"message": "Replacing room key."}) - yield self.store.set_e2e_room_key( - user_id, version, room_id, session_id, room_key - ) - else: - log_kv({"message": "Not replacing room_key."}) + count = yield self.store.count_e2e_room_keys(user_id, version) + return {"etag": str(version_etag), "count": count} @staticmethod def _should_replace_room_key(current_room_key, room_key): @@ -314,6 +342,8 @@ class E2eRoomKeysHandler(object): raise NotFoundError("Unknown backup version") else: raise + + res["count"] = yield self.store.count_e2e_room_keys(user_id, res["version"]) return res @trace diff --git a/synapse/rest/client/v2_alpha/room_keys.py b/synapse/rest/client/v2_alpha/room_keys.py index d596786430..d83ac8e3c5 100644 --- a/synapse/rest/client/v2_alpha/room_keys.py +++ b/synapse/rest/client/v2_alpha/room_keys.py @@ -134,8 +134,8 @@ class RoomKeysServlet(RestServlet): if room_id: body = {"rooms": {room_id: body}} - yield self.e2e_room_keys_handler.upload_room_keys(user_id, version, body) - return 200, {} + ret = yield self.e2e_room_keys_handler.upload_room_keys(user_id, version, body) + return 200, ret @defer.inlineCallbacks def on_GET(self, request, room_id, session_id): @@ -239,10 +239,10 @@ class RoomKeysServlet(RestServlet): user_id = requester.user.to_string() version = parse_string(request, "version") - yield self.e2e_room_keys_handler.delete_room_keys( + ret = yield self.e2e_room_keys_handler.delete_room_keys( user_id, version, room_id, session_id ) - return 200, {} + return 200, ret class RoomKeysNewVersionServlet(RestServlet): diff --git a/synapse/storage/data_stores/main/e2e_room_keys.py b/synapse/storage/data_stores/main/e2e_room_keys.py index 1cbbae5b63..113224fd7c 100644 --- a/synapse/storage/data_stores/main/e2e_room_keys.py +++ b/synapse/storage/data_stores/main/e2e_room_keys.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2017 New Vector Ltd +# 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. @@ -24,49 +25,8 @@ from synapse.storage._base import SQLBaseStore class EndToEndRoomKeyStore(SQLBaseStore): @defer.inlineCallbacks - def get_e2e_room_key(self, user_id, version, room_id, session_id): - """Get the encrypted E2E room key for a given session from a given - backup version of room_keys. We only store the 'best' room key for a given - session at a given time, as determined by the handler. - - Args: - user_id(str): the user whose backup we're querying - version(str): the version ID of the backup for the set of keys we're querying - room_id(str): the ID of the room whose keys we're querying. - This is a bit redundant as it's implied by the session_id, but - we include for consistency with the rest of the API. - session_id(str): the session whose room_key we're querying. - - Returns: - A deferred dict giving the session_data and message metadata for - this room key. - """ - - row = yield self._simple_select_one( - table="e2e_room_keys", - keyvalues={ - "user_id": user_id, - "version": version, - "room_id": room_id, - "session_id": session_id, - }, - retcols=( - "first_message_index", - "forwarded_count", - "is_verified", - "session_data", - ), - desc="get_e2e_room_key", - ) - - row["session_data"] = json.loads(row["session_data"]) - - return row - - @defer.inlineCallbacks - def set_e2e_room_key(self, user_id, version, room_id, session_id, room_key): - """Replaces or inserts the encrypted E2E room key for a given session in - a given backup + def update_e2e_room_key(self, user_id, version, room_id, session_id, room_key): + """Replaces the encrypted E2E room key for a given session in a given backup Args: user_id(str): the user whose backup we're setting @@ -78,7 +38,7 @@ class EndToEndRoomKeyStore(SQLBaseStore): StoreError """ - yield self._simple_upsert( + yield self._simple_update_one( table="e2e_room_keys", keyvalues={ "user_id": user_id, @@ -86,21 +46,51 @@ class EndToEndRoomKeyStore(SQLBaseStore): "room_id": room_id, "session_id": session_id, }, - values={ + updatevalues={ "first_message_index": room_key["first_message_index"], "forwarded_count": room_key["forwarded_count"], "is_verified": room_key["is_verified"], "session_data": json.dumps(room_key["session_data"]), }, - lock=False, + desc="update_e2e_room_key", ) - log_kv( - { - "message": "Set room key", - "room_id": room_id, - "session_id": session_id, - "room_key": room_key, - } + + @defer.inlineCallbacks + def add_e2e_room_keys(self, user_id, version, room_keys): + """Bulk add room keys to a given backup. + + Args: + user_id (str): the user whose backup we're adding to + version (str): the version ID of the backup for the set of keys we're adding to + room_keys (iterable[(str, str, dict)]): the keys to add, in the form + (roomID, sessionID, keyData) + """ + + values = [] + for (room_id, session_id, room_key) in room_keys: + values.append( + { + "user_id": user_id, + "version": version, + "room_id": room_id, + "session_id": session_id, + "first_message_index": room_key["first_message_index"], + "forwarded_count": room_key["forwarded_count"], + "is_verified": room_key["is_verified"], + "session_data": json.dumps(room_key["session_data"]), + } + ) + log_kv( + { + "message": "Set room key", + "room_id": room_id, + "session_id": session_id, + "room_key": room_key, + } + ) + + yield self._simple_insert_many( + table="e2e_room_keys", values=values, desc="add_e2e_room_keys" ) @trace @@ -110,11 +100,11 @@ class EndToEndRoomKeyStore(SQLBaseStore): room, or a given session. Args: - user_id(str): the user whose backup we're querying - version(str): the version ID of the backup for the set of keys we're querying - room_id(str): Optional. the ID of the room whose keys we're querying, if any. + user_id (str): the user whose backup we're querying + version (str): the version ID of the backup for the set of keys we're querying + room_id (str): Optional. the ID of the room whose keys we're querying, if any. If not specified, we return the keys for all the rooms in the backup. - session_id(str): Optional. the session whose room_key we're querying, if any. + session_id (str): Optional. the session whose room_key we're querying, if any. If specified, we also require the room_id to be specified. If not specified, we return all the keys in this version of the backup (or for the specified room) @@ -162,6 +152,95 @@ class EndToEndRoomKeyStore(SQLBaseStore): return sessions + def get_e2e_room_keys_multi(self, user_id, version, room_keys): + """Get multiple room keys at a time. The difference between this function and + get_e2e_room_keys is that this function can be used to retrieve + multiple specific keys at a time, whereas get_e2e_room_keys is used for + getting all the keys in a backup version, all the keys for a room, or a + specific key. + + Args: + user_id (str): the user whose backup we're querying + version (str): the version ID of the backup we're querying about + room_keys (dict[str, dict[str, iterable[str]]]): a map from + room ID -> {"session": [session ids]} indicating the session IDs + that we want to query + + Returns: + Deferred[dict[str, dict[str, dict]]]: a map of room IDs to session IDs to room key + """ + + return self.runInteraction( + "get_e2e_room_keys_multi", + self._get_e2e_room_keys_multi_txn, + user_id, + version, + room_keys, + ) + + @staticmethod + def _get_e2e_room_keys_multi_txn(txn, user_id, version, room_keys): + if not room_keys: + return {} + + where_clauses = [] + params = [user_id, version] + for room_id, room in room_keys.items(): + sessions = list(room["sessions"]) + if not sessions: + continue + params.append(room_id) + params.extend(sessions) + where_clauses.append( + "(room_id = ? AND session_id IN (%s))" + % (",".join(["?" for _ in sessions]),) + ) + + # check if we're actually querying something + if not where_clauses: + return {} + + sql = """ + SELECT room_id, session_id, first_message_index, forwarded_count, + is_verified, session_data + FROM e2e_room_keys + WHERE user_id = ? AND version = ? AND (%s) + """ % ( + " OR ".join(where_clauses) + ) + + txn.execute(sql, params) + + ret = {} + + for row in txn: + room_id = row[0] + session_id = row[1] + ret.setdefault(room_id, {}) + ret[room_id][session_id] = { + "first_message_index": row[2], + "forwarded_count": row[3], + "is_verified": row[4], + "session_data": json.loads(row[5]), + } + + return ret + + def count_e2e_room_keys(self, user_id, version): + """Get the number of keys in a backup version. + + Args: + user_id (str): the user whose backup we're querying + version (str): the version ID of the backup we're querying about + """ + + return self._simple_select_one_onecol( + table="e2e_room_keys", + keyvalues={"user_id": user_id, "version": version}, + retcol="COUNT(*)", + desc="count_e2e_room_keys", + ) + @trace @defer.inlineCallbacks def delete_e2e_room_keys(self, user_id, version, room_id=None, session_id=None): @@ -219,6 +298,7 @@ class EndToEndRoomKeyStore(SQLBaseStore): 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): @@ -236,10 +316,12 @@ class EndToEndRoomKeyStore(SQLBaseStore): txn, table="e2e_room_keys_versions", keyvalues={"user_id": user_id, "version": this_version, "deleted": 0}, - retcols=("version", "algorithm", "auth_data"), + retcols=("version", "algorithm", "auth_data", "etag"), ) result["auth_data"] = json.loads(result["auth_data"]) result["version"] = str(result["version"]) + if result["etag"] is None: + result["etag"] = 0 return result return self.runInteraction( @@ -288,21 +370,33 @@ class EndToEndRoomKeyStore(SQLBaseStore): ) @trace - def update_e2e_room_keys_version(self, user_id, version, info): + def update_e2e_room_keys_version( + self, user_id, version, info=None, version_etag=None + ): """Update a given backup version Args: user_id(str): the user whose backup version we're updating version(str): the version ID of the backup version we're updating - info(dict): the new backup version info to store + info (dict): the new backup version info to store. If None, then + the backup version info is not updated + version_etag (Optional[int]): etag of the keys in the backup. If + None, then the etag is not updated """ + updatevalues = {} - return self._simple_update( - table="e2e_room_keys_versions", - keyvalues={"user_id": user_id, "version": version}, - updatevalues={"auth_data": json.dumps(info["auth_data"])}, - desc="update_e2e_room_keys_version", - ) + if info is not None and "auth_data" in info: + updatevalues["auth_data"] = json.dumps(info["auth_data"]) + if version_etag is not None: + updatevalues["etag"] = version_etag + + if updatevalues: + return self._simple_update( + table="e2e_room_keys_versions", + keyvalues={"user_id": user_id, "version": version}, + updatevalues=updatevalues, + desc="update_e2e_room_keys_version", + ) @trace def delete_e2e_room_keys_version(self, user_id, version=None): diff --git a/synapse/storage/data_stores/main/schema/delta/56/room_key_etag.sql b/synapse/storage/data_stores/main/schema/delta/56/room_key_etag.sql new file mode 100644 index 0000000000..7d70dd071e --- /dev/null +++ b/synapse/storage/data_stores/main/schema/delta/56/room_key_etag.sql @@ -0,0 +1,17 @@ +/* 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. + */ + +-- store the current etag of backup version +ALTER TABLE e2e_room_keys_versions ADD COLUMN etag BIGINT; diff --git a/tests/handlers/test_e2e_room_keys.py b/tests/handlers/test_e2e_room_keys.py index 0bb96674a2..70f172eb02 100644 --- a/tests/handlers/test_e2e_room_keys.py +++ b/tests/handlers/test_e2e_room_keys.py @@ -1,6 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2016 OpenMarket Ltd # Copyright 2017 New Vector Ltd +# 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. @@ -94,23 +95,29 @@ class E2eRoomKeysHandlerTestCase(unittest.TestCase): # check we can retrieve it as the current version res = yield self.handler.get_version_info(self.local_user) + version_etag = res["etag"] + del res["etag"] self.assertDictEqual( res, { "version": "1", "algorithm": "m.megolm_backup.v1", "auth_data": "first_version_auth_data", + "count": 0, }, ) # check we can retrieve it as a specific version res = yield self.handler.get_version_info(self.local_user, "1") + self.assertEqual(res["etag"], version_etag) + del res["etag"] self.assertDictEqual( res, { "version": "1", "algorithm": "m.megolm_backup.v1", "auth_data": "first_version_auth_data", + "count": 0, }, ) @@ -126,12 +133,14 @@ class E2eRoomKeysHandlerTestCase(unittest.TestCase): # check we can retrieve it as the current version res = yield self.handler.get_version_info(self.local_user) + del res["etag"] self.assertDictEqual( res, { "version": "2", "algorithm": "m.megolm_backup.v1", "auth_data": "second_version_auth_data", + "count": 0, }, ) @@ -158,12 +167,14 @@ class E2eRoomKeysHandlerTestCase(unittest.TestCase): # check we can retrieve it as the current version res = yield self.handler.get_version_info(self.local_user) + del res["etag"] self.assertDictEqual( res, { "algorithm": "m.megolm_backup.v1", "auth_data": "revised_first_version_auth_data", "version": version, + "count": 0, }, ) @@ -207,12 +218,14 @@ class E2eRoomKeysHandlerTestCase(unittest.TestCase): # check we can retrieve it as the current version res = yield self.handler.get_version_info(self.local_user) + del res["etag"] # etag is opaque, so don't test its contents self.assertDictEqual( res, { "algorithm": "m.megolm_backup.v1", "auth_data": "revised_first_version_auth_data", "version": version, + "count": 0, }, ) @@ -409,6 +422,11 @@ class E2eRoomKeysHandlerTestCase(unittest.TestCase): yield self.handler.upload_room_keys(self.local_user, version, room_keys) + # get the etag to compare to future versions + res = yield self.handler.get_version_info(self.local_user) + backup_etag = res["etag"] + self.assertEqual(res["count"], 1) + new_room_keys = copy.deepcopy(room_keys) new_room_key = new_room_keys["rooms"]["!abc:matrix.org"]["sessions"]["c0ff33"] @@ -423,6 +441,10 @@ class E2eRoomKeysHandlerTestCase(unittest.TestCase): "SSBBTSBBIEZJU0gK", ) + # the etag should be the same since the session did not change + res = yield self.handler.get_version_info(self.local_user) + self.assertEqual(res["etag"], backup_etag) + # test that marking the session as verified however /does/ replace it new_room_key["is_verified"] = True yield self.handler.upload_room_keys(self.local_user, version, new_room_keys) @@ -432,6 +454,11 @@ class E2eRoomKeysHandlerTestCase(unittest.TestCase): res["rooms"]["!abc:matrix.org"]["sessions"]["c0ff33"]["session_data"], "new" ) + # the etag should NOT be equal now, since the key changed + res = yield self.handler.get_version_info(self.local_user) + self.assertNotEqual(res["etag"], backup_etag) + backup_etag = res["etag"] + # test that a session with a higher forwarded_count doesn't replace one # with a lower forwarding count new_room_key["forwarded_count"] = 2 @@ -443,6 +470,10 @@ class E2eRoomKeysHandlerTestCase(unittest.TestCase): res["rooms"]["!abc:matrix.org"]["sessions"]["c0ff33"]["session_data"], "new" ) + # the etag should be the same since the session did not change + res = yield self.handler.get_version_info(self.local_user) + self.assertEqual(res["etag"], backup_etag) + # TODO: check edge cases as well as the common variations here @defer.inlineCallbacks diff --git a/tests/storage/test_e2e_room_keys.py b/tests/storage/test_e2e_room_keys.py index d128fde441..35dafbb904 100644 --- a/tests/storage/test_e2e_room_keys.py +++ b/tests/storage/test_e2e_room_keys.py @@ -39,8 +39,8 @@ class E2eRoomKeysHandlerTestCase(unittest.HomeserverTestCase): ) self.get_success( - self.store.set_e2e_room_key( - "user_id", version1, "room", "session", room_key + self.store.add_e2e_room_keys( + "user_id", version1, [("room", "session", room_key)] ) ) @@ -51,8 +51,8 @@ class E2eRoomKeysHandlerTestCase(unittest.HomeserverTestCase): ) self.get_success( - self.store.set_e2e_room_key( - "user_id", version2, "room", "session", room_key + self.store.add_e2e_room_keys( + "user_id", version2, [("room", "session", room_key)] ) ) From 0f87b912aba7e678041632bc9a6d1f7c2d24342c Mon Sep 17 00:00:00 2001 From: Amber Brown Date: Thu, 28 Nov 2019 08:54:07 +1100 Subject: [PATCH 34/47] Implementation of MSC2314 (#6176) --- changelog.d/6176.feature | 1 + synapse/federation/federation_server.py | 26 ++++++--- synapse/federation/transport/server.py | 6 +- sytest-blacklist | 6 +- tests/federation/test_complexity.py | 28 +-------- tests/federation/test_federation_sender.py | 4 +- tests/federation/test_federation_server.py | 63 ++++++++++++++++++++ tests/handlers/test_typing.py | 3 + tests/replication/slave/storage/_base.py | 3 + tests/replication/tcp/streams/_base.py | 4 ++ tests/storage/test_roommember.py | 26 +-------- tests/unittest.py | 68 +++++++++++++++++++++- tests/utils.py | 1 + 13 files changed, 174 insertions(+), 65 deletions(-) create mode 100644 changelog.d/6176.feature diff --git a/changelog.d/6176.feature b/changelog.d/6176.feature new file mode 100644 index 0000000000..3c66d689d4 --- /dev/null +++ b/changelog.d/6176.feature @@ -0,0 +1 @@ +Implement the `/_matrix/federation/unstable/net.atleastfornow/state/` API as drafted in MSC2314. diff --git a/synapse/federation/federation_server.py b/synapse/federation/federation_server.py index d942d77a72..84d4eca041 100644 --- a/synapse/federation/federation_server.py +++ b/synapse/federation/federation_server.py @@ -1,6 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2015, 2016 OpenMarket Ltd # Copyright 2018 New Vector Ltd +# Copyright 2019 Matrix.org Federation 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. @@ -73,6 +74,7 @@ class FederationServer(FederationBase): self.auth = hs.get_auth() self.handler = hs.get_handlers().federation_handler + self.state = hs.get_state_handler() self._server_linearizer = Linearizer("fed_server") self._transaction_linearizer = Linearizer("fed_txn_handler") @@ -264,9 +266,6 @@ class FederationServer(FederationBase): await self.registry.on_edu(edu_type, origin, content) async def on_context_state_request(self, origin, room_id, event_id): - if not event_id: - raise NotImplementedError("Specify an event") - origin_host, _ = parse_server_name(origin) await self.check_server_matches_acl(origin_host, room_id) @@ -280,13 +279,18 @@ class FederationServer(FederationBase): # - but that's non-trivial to get right, and anyway somewhat defeats # the point of the linearizer. with (await self._server_linearizer.queue((origin, room_id))): - resp = await self._state_resp_cache.wrap( - (room_id, event_id), - self._on_context_state_request_compute, - room_id, - event_id, + resp = dict( + await self._state_resp_cache.wrap( + (room_id, event_id), + self._on_context_state_request_compute, + room_id, + event_id, + ) ) + room_version = await self.store.get_room_version(room_id) + resp["room_version"] = room_version + return 200, resp async def on_state_ids_request(self, origin, room_id, event_id): @@ -306,7 +310,11 @@ class FederationServer(FederationBase): return 200, {"pdu_ids": state_ids, "auth_chain_ids": auth_chain_ids} async def _on_context_state_request_compute(self, room_id, event_id): - pdus = await self.handler.get_state_for_pdu(room_id, event_id) + if event_id: + pdus = await self.handler.get_state_for_pdu(room_id, event_id) + else: + pdus = (await self.state.get_current_state(room_id)).values() + auth_chain = await self.store.get_auth_chain([pdu.event_id for pdu in pdus]) return { diff --git a/synapse/federation/transport/server.py b/synapse/federation/transport/server.py index 09baa9c57d..fefc789c85 100644 --- a/synapse/federation/transport/server.py +++ b/synapse/federation/transport/server.py @@ -421,7 +421,7 @@ class FederationEventServlet(BaseFederationServlet): return await self.handler.on_pdu_request(origin, event_id) -class FederationStateServlet(BaseFederationServlet): +class FederationStateV1Servlet(BaseFederationServlet): PATH = "/state/(?P[^/]*)/?" # This is when someone asks for all data for a given context. @@ -429,7 +429,7 @@ class FederationStateServlet(BaseFederationServlet): return await self.handler.on_context_state_request( origin, context, - parse_string_from_args(query, "event_id", None, required=True), + parse_string_from_args(query, "event_id", None, required=False), ) @@ -1360,7 +1360,7 @@ class RoomComplexityServlet(BaseFederationServlet): FEDERATION_SERVLET_CLASSES = ( FederationSendServlet, FederationEventServlet, - FederationStateServlet, + FederationStateV1Servlet, FederationStateIdsServlet, FederationBackfillServlet, FederationQueryServlet, diff --git a/sytest-blacklist b/sytest-blacklist index 11785fd43f..411cce0692 100644 --- a/sytest-blacklist +++ b/sytest-blacklist @@ -1,6 +1,6 @@ # This file serves as a blacklist for SyTest tests that we expect will fail in # Synapse. -# +# # Each line of this file is scanned by sytest during a run and if the line # exactly matches the name of a test, it will be marked as "expected fail", # meaning the test will still run, but failure will not mark the entire test @@ -29,3 +29,7 @@ Enabling an unknown default rule fails with 404 # Blacklisted due to https://github.com/matrix-org/synapse/issues/1663 New federated private chats get full presence information (SYN-115) + +# Blacklisted due to https://github.com/matrix-org/matrix-doc/pull/2314 removing +# this requirement from the spec +Inbound federation of state requires event_id as a mandatory paramater diff --git a/tests/federation/test_complexity.py b/tests/federation/test_complexity.py index 51714a2b06..24fa8dbb45 100644 --- a/tests/federation/test_complexity.py +++ b/tests/federation/test_complexity.py @@ -18,17 +18,14 @@ from mock import Mock from twisted.internet import defer from synapse.api.errors import Codes, SynapseError -from synapse.config.ratelimiting import FederationRateLimitConfig -from synapse.federation.transport import server from synapse.rest import admin from synapse.rest.client.v1 import login, room from synapse.types import UserID -from synapse.util.ratelimitutils import FederationRateLimiter from tests import unittest -class RoomComplexityTests(unittest.HomeserverTestCase): +class RoomComplexityTests(unittest.FederatingHomeserverTestCase): servlets = [ admin.register_servlets, @@ -41,25 +38,6 @@ class RoomComplexityTests(unittest.HomeserverTestCase): config["limit_remote_rooms"] = {"enabled": True, "complexity": 0.05} return config - def prepare(self, reactor, clock, homeserver): - class Authenticator(object): - def authenticate_request(self, request, content): - return defer.succeed("otherserver.nottld") - - ratelimiter = FederationRateLimiter( - clock, - FederationRateLimitConfig( - window_size=1, - sleep_limit=1, - sleep_msec=1, - reject_limit=1000, - concurrent_requests=1000, - ), - ) - server.register_servlets( - homeserver, self.resource, Authenticator(), ratelimiter - ) - def test_complexity_simple(self): u1 = self.register_user("u1", "pass") @@ -105,7 +83,7 @@ class RoomComplexityTests(unittest.HomeserverTestCase): d = handler._remote_join( None, - ["otherserver.example"], + ["other.example.com"], "roomid", UserID.from_string(u1), {"membership": "join"}, @@ -146,7 +124,7 @@ class RoomComplexityTests(unittest.HomeserverTestCase): d = handler._remote_join( None, - ["otherserver.example"], + ["other.example.com"], room_1, UserID.from_string(u1), {"membership": "join"}, diff --git a/tests/federation/test_federation_sender.py b/tests/federation/test_federation_sender.py index cce8d8c6de..d456267b87 100644 --- a/tests/federation/test_federation_sender.py +++ b/tests/federation/test_federation_sender.py @@ -19,7 +19,7 @@ from twisted.internet import defer from synapse.types import ReadReceipt -from tests.unittest import HomeserverTestCase +from tests.unittest import HomeserverTestCase, override_config class FederationSenderTestCases(HomeserverTestCase): @@ -29,6 +29,7 @@ class FederationSenderTestCases(HomeserverTestCase): federation_transport_client=Mock(spec=["send_transaction"]), ) + @override_config({"send_federation": True}) def test_send_receipts(self): mock_state_handler = self.hs.get_state_handler() mock_state_handler.get_current_hosts_in_room.return_value = ["test", "host2"] @@ -69,6 +70,7 @@ class FederationSenderTestCases(HomeserverTestCase): ], ) + @override_config({"send_federation": True}) def test_send_receipts_with_backoff(self): """Send two receipts in quick succession; the second should be flushed, but only after 20ms""" diff --git a/tests/federation/test_federation_server.py b/tests/federation/test_federation_server.py index b08be451aa..1ec8c40901 100644 --- a/tests/federation/test_federation_server.py +++ b/tests/federation/test_federation_server.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2018 New Vector Ltd +# Copyright 2019 Matrix.org Federation 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. @@ -16,6 +17,8 @@ import logging from synapse.events import FrozenEvent from synapse.federation.federation_server import server_matches_acl_event +from synapse.rest import admin +from synapse.rest.client.v1 import login, room from tests import unittest @@ -41,6 +44,66 @@ class ServerACLsTestCase(unittest.TestCase): self.assertTrue(server_matches_acl_event("1:2:3:4", e)) +class StateQueryTests(unittest.FederatingHomeserverTestCase): + + servlets = [ + admin.register_servlets, + room.register_servlets, + login.register_servlets, + ] + + def test_without_event_id(self): + """ + Querying v1/state/ without an event ID will return the current + known state. + """ + u1 = self.register_user("u1", "pass") + u1_token = self.login("u1", "pass") + + room_1 = self.helper.create_room_as(u1, tok=u1_token) + self.inject_room_member(room_1, "@user:other.example.com", "join") + + request, channel = self.make_request( + "GET", "/_matrix/federation/v1/state/%s" % (room_1,) + ) + self.render(request) + self.assertEquals(200, channel.code, channel.result) + + self.assertEqual( + channel.json_body["room_version"], + self.hs.config.default_room_version.identifier, + ) + + members = set( + map( + lambda x: x["state_key"], + filter( + lambda x: x["type"] == "m.room.member", channel.json_body["pdus"] + ), + ) + ) + + self.assertEqual(members, set(["@user:other.example.com", u1])) + self.assertEqual(len(channel.json_body["pdus"]), 6) + + def test_needs_to_be_in_room(self): + """ + Querying v1/state/ requires the server + be in the room to provide data. + """ + u1 = self.register_user("u1", "pass") + u1_token = self.login("u1", "pass") + + room_1 = self.helper.create_room_as(u1, tok=u1_token) + + request, channel = self.make_request( + "GET", "/_matrix/federation/v1/state/%s" % (room_1,) + ) + self.render(request) + self.assertEquals(403, channel.code, channel.result) + self.assertEqual(channel.json_body["errcode"], "M_FORBIDDEN") + + def _create_acl_event(content): return FrozenEvent( { diff --git a/tests/handlers/test_typing.py b/tests/handlers/test_typing.py index 5ec568f4e6..f6d8660285 100644 --- a/tests/handlers/test_typing.py +++ b/tests/handlers/test_typing.py @@ -24,6 +24,7 @@ from synapse.api.errors import AuthError from synapse.types import UserID from tests import unittest +from tests.unittest import override_config from tests.utils import register_federation_servlets # Some local users to test with @@ -174,6 +175,7 @@ class TypingNotificationsTestCase(unittest.HomeserverTestCase): ], ) + @override_config({"send_federation": True}) def test_started_typing_remote_send(self): self.room_members = [U_APPLE, U_ONION] @@ -237,6 +239,7 @@ class TypingNotificationsTestCase(unittest.HomeserverTestCase): ], ) + @override_config({"send_federation": True}) def test_stopped_typing(self): self.room_members = [U_APPLE, U_BANANA, U_ONION] diff --git a/tests/replication/slave/storage/_base.py b/tests/replication/slave/storage/_base.py index 4f924ce451..e7472e3a93 100644 --- a/tests/replication/slave/storage/_base.py +++ b/tests/replication/slave/storage/_base.py @@ -48,7 +48,10 @@ class BaseSlavedStoreTestCase(unittest.HomeserverTestCase): server_factory = ReplicationStreamProtocolFactory(self.hs) self.streamer = server_factory.streamer + handler_factory = Mock() self.replication_handler = ReplicationClientHandler(self.slaved_store) + self.replication_handler.factory = handler_factory + client_factory = ReplicationClientFactory( self.hs, "client_name", self.replication_handler ) diff --git a/tests/replication/tcp/streams/_base.py b/tests/replication/tcp/streams/_base.py index ce3835ae6a..1d14e77255 100644 --- a/tests/replication/tcp/streams/_base.py +++ b/tests/replication/tcp/streams/_base.py @@ -12,6 +12,8 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +from mock import Mock + from synapse.replication.tcp.commands import ReplicateCommand from synapse.replication.tcp.protocol import ClientReplicationStreamProtocol from synapse.replication.tcp.resource import ReplicationStreamProtocolFactory @@ -30,7 +32,9 @@ class BaseStreamTestCase(unittest.HomeserverTestCase): server = server_factory.buildProtocol(None) # build a replication client, with a dummy handler + handler_factory = Mock() self.test_handler = TestReplicationClientHandler() + self.test_handler.factory = handler_factory self.client = ClientReplicationStreamProtocol( "client", "test", clock, self.test_handler ) diff --git a/tests/storage/test_roommember.py b/tests/storage/test_roommember.py index 9ddd17f73d..105a0c2b02 100644 --- a/tests/storage/test_roommember.py +++ b/tests/storage/test_roommember.py @@ -16,8 +16,7 @@ from unittest.mock import Mock -from synapse.api.constants import EventTypes, Membership -from synapse.api.room_versions import RoomVersions +from synapse.api.constants import Membership from synapse.rest.admin import register_servlets_for_client_rest_resource from synapse.rest.client.v1 import login, room from synapse.types import Requester, UserID @@ -44,9 +43,6 @@ class RoomMemberStoreTestCase(unittest.HomeserverTestCase): # We can't test the RoomMemberStore on its own without the other event # storage logic self.store = hs.get_datastore() - self.storage = hs.get_storage() - self.event_builder_factory = hs.get_event_builder_factory() - self.event_creation_handler = hs.get_event_creation_handler() self.u_alice = self.register_user("alice", "pass") self.t_alice = self.login("alice", "pass") @@ -55,26 +51,6 @@ class RoomMemberStoreTestCase(unittest.HomeserverTestCase): # User elsewhere on another host self.u_charlie = UserID.from_string("@charlie:elsewhere") - def inject_room_member(self, room, user, membership, replaces_state=None): - builder = self.event_builder_factory.for_room_version( - RoomVersions.V1, - { - "type": EventTypes.Member, - "sender": user, - "state_key": user, - "room_id": room, - "content": {"membership": membership}, - }, - ) - - event, context = self.get_success( - self.event_creation_handler.create_new_client_event(builder) - ) - - self.get_success(self.storage.persistence.persist_event(event, context)) - - return event - def test_one_member(self): # Alice creates the room, and is automatically joined diff --git a/tests/unittest.py b/tests/unittest.py index 561cebc223..31997a0f31 100644 --- a/tests/unittest.py +++ b/tests/unittest.py @@ -1,6 +1,7 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 OpenMarket Ltd # Copyright 2018 New Vector +# Copyright 2019 Matrix.org Federation 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. @@ -13,6 +14,7 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. + import gc import hashlib import hmac @@ -27,13 +29,17 @@ from twisted.internet.defer import Deferred, succeed from twisted.python.threadpool import ThreadPool from twisted.trial import unittest -from synapse.api.constants import EventTypes +from synapse.api.constants import EventTypes, Membership +from synapse.api.room_versions import KNOWN_ROOM_VERSIONS from synapse.config.homeserver import HomeServerConfig +from synapse.config.ratelimiting import FederationRateLimitConfig +from synapse.federation.transport import server as federation_server from synapse.http.server import JsonResource from synapse.http.site import SynapseRequest from synapse.logging.context import LoggingContext from synapse.server import HomeServer from synapse.types import Requester, UserID, create_requester +from synapse.util.ratelimitutils import FederationRateLimiter from tests.server import get_clock, make_request, render, setup_test_homeserver from tests.test_utils.logging_setup import setup_logging @@ -559,6 +565,66 @@ class HomeserverTestCase(TestCase): self.render(request) self.assertEqual(channel.code, 403, channel.result) + def inject_room_member(self, room: str, user: str, membership: Membership) -> None: + """ + Inject a membership event into a room. + + Args: + room: Room ID to inject the event into. + user: MXID of the user to inject the membership for. + membership: The membership type. + """ + event_builder_factory = self.hs.get_event_builder_factory() + event_creation_handler = self.hs.get_event_creation_handler() + + room_version = self.get_success(self.hs.get_datastore().get_room_version(room)) + + builder = event_builder_factory.for_room_version( + KNOWN_ROOM_VERSIONS[room_version], + { + "type": EventTypes.Member, + "sender": user, + "state_key": user, + "room_id": room, + "content": {"membership": membership}, + }, + ) + + event, context = self.get_success( + event_creation_handler.create_new_client_event(builder) + ) + + self.get_success( + self.hs.get_storage().persistence.persist_event(event, context) + ) + + +class FederatingHomeserverTestCase(HomeserverTestCase): + """ + A federating homeserver that authenticates incoming requests as `other.example.com`. + """ + + def prepare(self, reactor, clock, homeserver): + class Authenticator(object): + def authenticate_request(self, request, content): + return succeed("other.example.com") + + ratelimiter = FederationRateLimiter( + clock, + FederationRateLimitConfig( + window_size=1, + sleep_limit=1, + sleep_msec=1, + reject_limit=1000, + concurrent_requests=1000, + ), + ) + federation_server.register_servlets( + homeserver, self.resource, Authenticator(), ratelimiter + ) + + return super().prepare(reactor, clock, homeserver) + def override_config(extra_config): """A decorator which can be applied to test functions to give additional HS config diff --git a/tests/utils.py b/tests/utils.py index 7dc9bdc505..de2ac1ed33 100644 --- a/tests/utils.py +++ b/tests/utils.py @@ -109,6 +109,7 @@ def default_config(name, parse=False): """ config_dict = { "server_name": name, + "send_federation": False, "media_store_path": "media", "uploads_path": "uploads", # the test signing key is just an arbitrary ed25519 key to keep the config From c48ea9800769c22d763cd97ecb137141050739e1 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff <1389908+richvdh@users.noreply.github.com> Date: Thu, 28 Nov 2019 09:29:18 +0000 Subject: [PATCH 35/47] Clarifications for the email configuration settings. (#6423) Cf #6422 --- changelog.d/6423.misc | 1 + docs/sample_config.yaml | 17 ++++++++++++++++- synapse/config/emailconfig.py | 17 ++++++++++++++++- 3 files changed, 33 insertions(+), 2 deletions(-) create mode 100644 changelog.d/6423.misc diff --git a/changelog.d/6423.misc b/changelog.d/6423.misc new file mode 100644 index 0000000000..9bcd5d36c1 --- /dev/null +++ b/changelog.d/6423.misc @@ -0,0 +1 @@ +Clarifications for the email configuration settings. diff --git a/docs/sample_config.yaml b/docs/sample_config.yaml index 09dd21352f..c7391f0c48 100644 --- a/docs/sample_config.yaml +++ b/docs/sample_config.yaml @@ -1333,8 +1333,23 @@ password_config: # smtp_user: "exampleusername" # smtp_pass: "examplepassword" # require_transport_security: false +# +# # notif_from defines the "From" address to use when sending emails. +# # It must be set if email sending is enabled. +# # +# # The placeholder '%(app)s' will be replaced by the application name, +# # which is normally 'app_name' (below), but may be overridden by the +# # Matrix client application. +# # +# # Note that the placeholder must be written '%(app)s', including the +# # trailing 's'. +# # # notif_from: "Your Friendly %(app)s homeserver " -# app_name: Matrix +# +# # app_name defines the default value for '%(app)s' in notif_from. It +# # defaults to 'Matrix'. +# # +# #app_name: my_branded_matrix_server # # # Enable email notifications by default # # diff --git a/synapse/config/emailconfig.py b/synapse/config/emailconfig.py index ac1724045f..18f42a87f9 100644 --- a/synapse/config/emailconfig.py +++ b/synapse/config/emailconfig.py @@ -307,8 +307,23 @@ class EmailConfig(Config): # smtp_user: "exampleusername" # smtp_pass: "examplepassword" # require_transport_security: false + # + # # notif_from defines the "From" address to use when sending emails. + # # It must be set if email sending is enabled. + # # + # # The placeholder '%(app)s' will be replaced by the application name, + # # which is normally 'app_name' (below), but may be overridden by the + # # Matrix client application. + # # + # # Note that the placeholder must be written '%(app)s', including the + # # trailing 's'. + # # # notif_from: "Your Friendly %(app)s homeserver " - # app_name: Matrix + # + # # app_name defines the default value for '%(app)s' in notif_from. It + # # defaults to 'Matrix'. + # # + # #app_name: my_branded_matrix_server # # # Enable email notifications by default # # From a9c44d4008deb29503e2de00e5aae1a56a72d630 Mon Sep 17 00:00:00 2001 From: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com> Date: Thu, 28 Nov 2019 10:40:42 +0000 Subject: [PATCH 36/47] Remove local threepids on account deactivation (#6426) --- changelog.d/6426.bugfix | 1 + synapse/handlers/deactivate_account.py | 3 +++ synapse/storage/data_stores/main/registration.py | 13 +++++++++++++ 3 files changed, 17 insertions(+) create mode 100644 changelog.d/6426.bugfix diff --git a/changelog.d/6426.bugfix b/changelog.d/6426.bugfix new file mode 100644 index 0000000000..3acfde4211 --- /dev/null +++ b/changelog.d/6426.bugfix @@ -0,0 +1 @@ +Clean up local threepids from user on account deactivation. \ No newline at end of file diff --git a/synapse/handlers/deactivate_account.py b/synapse/handlers/deactivate_account.py index 63267a0a4c..6dedaaff8d 100644 --- a/synapse/handlers/deactivate_account.py +++ b/synapse/handlers/deactivate_account.py @@ -95,6 +95,9 @@ class DeactivateAccountHandler(BaseHandler): user_id, threepid["medium"], threepid["address"] ) + # Remove all 3PIDs this user has bound to the homeserver + yield self.store.user_delete_threepids(user_id) + # delete any devices belonging to the user, which will also # delete corresponding access tokens. yield self._device_handler.delete_all_devices_for_user(user_id) diff --git a/synapse/storage/data_stores/main/registration.py b/synapse/storage/data_stores/main/registration.py index 0a3c1f0510..98cf6427c3 100644 --- a/synapse/storage/data_stores/main/registration.py +++ b/synapse/storage/data_stores/main/registration.py @@ -569,6 +569,19 @@ class RegistrationWorkerStore(SQLBaseStore): return self._simple_delete( "user_threepids", keyvalues={"user_id": user_id, "medium": medium, "address": address}, + desc="user_delete_threepid", + ) + + def user_delete_threepids(self, user_id: str): + """Delete all threepid this user has bound + + Args: + user_id: The user id to delete all threepids of + + """ + return self._simple_delete( + "user_threepids", + keyvalues={"user_id": user_id}, desc="user_delete_threepids", ) From 69d8fb83c6da35d7e1f04fa3afba0fd5406bd9d9 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Thu, 28 Nov 2019 11:02:04 +0000 Subject: [PATCH 37/47] MSC2367 Allow reason field on all member events --- synapse/rest/client/v1/room.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/synapse/rest/client/v1/room.py b/synapse/rest/client/v1/room.py index 86bbcc0eea..711d4ad304 100644 --- a/synapse/rest/client/v1/room.py +++ b/synapse/rest/client/v1/room.py @@ -714,7 +714,7 @@ class RoomMembershipRestServlet(TransactionRestServlet): target = UserID.from_string(content["user_id"]) event_content = None - if "reason" in content and membership_action in ["kick", "ban"]: + if "reason" in content: event_content = {"reason": content["reason"]} await self.room_member_handler.update_membership( From 2030193e5523810c4fd6158f97b7a223cee4cb72 Mon Sep 17 00:00:00 2001 From: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com> Date: Thu, 28 Nov 2019 10:40:42 +0000 Subject: [PATCH 38/47] Remove local threepids on account deactivation (#6426) --- changelog.d/6426.bugfix | 1 + synapse/handlers/deactivate_account.py | 3 +++ synapse/storage/data_stores/main/registration.py | 13 +++++++++++++ 3 files changed, 17 insertions(+) create mode 100644 changelog.d/6426.bugfix diff --git a/changelog.d/6426.bugfix b/changelog.d/6426.bugfix new file mode 100644 index 0000000000..3acfde4211 --- /dev/null +++ b/changelog.d/6426.bugfix @@ -0,0 +1 @@ +Clean up local threepids from user on account deactivation. \ No newline at end of file diff --git a/synapse/handlers/deactivate_account.py b/synapse/handlers/deactivate_account.py index 63267a0a4c..6dedaaff8d 100644 --- a/synapse/handlers/deactivate_account.py +++ b/synapse/handlers/deactivate_account.py @@ -95,6 +95,9 @@ class DeactivateAccountHandler(BaseHandler): user_id, threepid["medium"], threepid["address"] ) + # Remove all 3PIDs this user has bound to the homeserver + yield self.store.user_delete_threepids(user_id) + # delete any devices belonging to the user, which will also # delete corresponding access tokens. yield self._device_handler.delete_all_devices_for_user(user_id) diff --git a/synapse/storage/data_stores/main/registration.py b/synapse/storage/data_stores/main/registration.py index ee1b2b2bbf..89147ad511 100644 --- a/synapse/storage/data_stores/main/registration.py +++ b/synapse/storage/data_stores/main/registration.py @@ -577,6 +577,19 @@ class RegistrationWorkerStore(SQLBaseStore): return self._simple_delete( "user_threepids", keyvalues={"user_id": user_id, "medium": medium, "address": address}, + desc="user_delete_threepid", + ) + + def user_delete_threepids(self, user_id: str): + """Delete all threepid this user has bound + + Args: + user_id: The user id to delete all threepids of + + """ + return self._simple_delete( + "user_threepids", + keyvalues={"user_id": user_id}, desc="user_delete_threepids", ) From e7777f3668d09c87335830f785f42c851827b497 Mon Sep 17 00:00:00 2001 From: Andrew Morgan Date: Thu, 28 Nov 2019 11:24:11 +0000 Subject: [PATCH 39/47] 1.6.1 --- CHANGES.md | 15 +++++++++++++++ changelog.d/6421.bugfix | 1 - changelog.d/6426.bugfix | 1 - debian/changelog | 6 ++++++ synapse/__init__.py | 2 +- 5 files changed, 22 insertions(+), 3 deletions(-) delete mode 100644 changelog.d/6421.bugfix delete mode 100644 changelog.d/6426.bugfix diff --git a/CHANGES.md b/CHANGES.md index 42281483b3..a9afd36d2c 100644 --- a/CHANGES.md +++ b/CHANGES.md @@ -1,3 +1,18 @@ +Synapse 1.6.1 (2019-11-28) +========================== + +Security updates +---------------- + +This release includes a security fix ([\#6426](https://github.com/matrix-org/synapse/issues/6426), below). Administrators are encouraged to upgrade as soon as possible. + +Bugfixes +-------- + +- Clean up local threepids from user on account deactivation. ([\#6426](https://github.com/matrix-org/synapse/issues/6426)) +- Fix startup error when http proxy is defined. ([\#6421](https://github.com/matrix-org/synapse/issues/6421)) + + Synapse 1.6.0 (2019-11-26) ========================== diff --git a/changelog.d/6421.bugfix b/changelog.d/6421.bugfix deleted file mode 100644 index 7969f7f71d..0000000000 --- a/changelog.d/6421.bugfix +++ /dev/null @@ -1 +0,0 @@ -Fix startup error when http proxy is defined. diff --git a/changelog.d/6426.bugfix b/changelog.d/6426.bugfix deleted file mode 100644 index 3acfde4211..0000000000 --- a/changelog.d/6426.bugfix +++ /dev/null @@ -1 +0,0 @@ -Clean up local threepids from user on account deactivation. \ No newline at end of file diff --git a/debian/changelog b/debian/changelog index 82dae017f1..b8a43788ef 100644 --- a/debian/changelog +++ b/debian/changelog @@ -1,3 +1,9 @@ +matrix-synapse-py3 (1.6.1) stable; urgency=medium + + * New synapse release 1.6.1. + + -- Synapse Packaging team Thu, 28 Nov 2019 11:10:40 +0000 + matrix-synapse-py3 (1.6.0) stable; urgency=medium * New synapse release 1.6.0. diff --git a/synapse/__init__.py b/synapse/__init__.py index 53eedc0048..f99de2f3f3 100644 --- a/synapse/__init__.py +++ b/synapse/__init__.py @@ -36,7 +36,7 @@ try: except ImportError: pass -__version__ = "1.6.0" +__version__ = "1.6.1" if bool(os.environ.get("SYNAPSE_TEST_PATCH_LOG_CONTEXTS", False)): # We import here so that we don't have to install a bunch of deps when From 2173785f0d9124037ca841b568349ad0424b39cd Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Thu, 28 Nov 2019 11:31:56 +0000 Subject: [PATCH 40/47] Propagate reason in remotely rejected invites --- synapse/handlers/federation.py | 4 ++-- synapse/handlers/room_member.py | 13 +++++++++---- synapse/handlers/room_member_worker.py | 5 ++++- synapse/replication/http/membership.py | 7 +++++-- 4 files changed, 20 insertions(+), 9 deletions(-) diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index a5ae7b77d1..d3267734f7 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -1428,9 +1428,9 @@ class FederationHandler(BaseHandler): return event @defer.inlineCallbacks - def do_remotely_reject_invite(self, target_hosts, room_id, user_id): + def do_remotely_reject_invite(self, target_hosts, room_id, user_id, content): origin, event, event_format_version = yield self._make_and_verify_event( - target_hosts, room_id, user_id, "leave" + target_hosts, room_id, user_id, "leave", content=content, ) # Mark as outlier as we don't have any state for this event; we're not # even in the room. diff --git a/synapse/handlers/room_member.py b/synapse/handlers/room_member.py index 6cfee4b361..7b7270fc61 100644 --- a/synapse/handlers/room_member.py +++ b/synapse/handlers/room_member.py @@ -94,7 +94,9 @@ class RoomMemberHandler(object): raise NotImplementedError() @abc.abstractmethod - def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target): + def _remote_reject_invite( + self, requester, remote_room_hosts, room_id, target, content + ): """Attempt to reject an invite for a room this server is not in. If we fail to do so we locally mark the invite as rejected. @@ -104,6 +106,7 @@ class RoomMemberHandler(object): reject invite room_id (str) target (UserID): The user rejecting the invite + content (dict): The content for the rejection event Returns: Deferred[dict]: A dictionary to be returned to the client, may @@ -471,7 +474,7 @@ class RoomMemberHandler(object): # send the rejection to the inviter's HS. remote_room_hosts = remote_room_hosts + [inviter.domain] res = yield self._remote_reject_invite( - requester, remote_room_hosts, room_id, target + requester, remote_room_hosts, room_id, target, content, ) return res @@ -971,13 +974,15 @@ class RoomMemberMasterHandler(RoomMemberHandler): ) @defer.inlineCallbacks - def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target): + def _remote_reject_invite( + self, requester, remote_room_hosts, room_id, target, content + ): """Implements RoomMemberHandler._remote_reject_invite """ fed_handler = self.federation_handler try: ret = yield fed_handler.do_remotely_reject_invite( - remote_room_hosts, room_id, target.to_string() + remote_room_hosts, room_id, target.to_string(), content=content, ) return ret except Exception as e: diff --git a/synapse/handlers/room_member_worker.py b/synapse/handlers/room_member_worker.py index 75e96ae1a2..69be86893b 100644 --- a/synapse/handlers/room_member_worker.py +++ b/synapse/handlers/room_member_worker.py @@ -55,7 +55,9 @@ class RoomMemberWorkerHandler(RoomMemberHandler): return ret - def _remote_reject_invite(self, requester, remote_room_hosts, room_id, target): + def _remote_reject_invite( + self, requester, remote_room_hosts, room_id, target, content + ): """Implements RoomMemberHandler._remote_reject_invite """ return self._remote_reject_client( @@ -63,6 +65,7 @@ class RoomMemberWorkerHandler(RoomMemberHandler): remote_room_hosts=remote_room_hosts, room_id=room_id, user_id=target.to_string(), + content=content, ) def _user_joined_room(self, target, room_id): diff --git a/synapse/replication/http/membership.py b/synapse/replication/http/membership.py index cc1f249740..3577611fd7 100644 --- a/synapse/replication/http/membership.py +++ b/synapse/replication/http/membership.py @@ -93,6 +93,7 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint): { "requester": ..., "remote_room_hosts": [...], + "content": { ... } } """ @@ -107,7 +108,7 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint): self.clock = hs.get_clock() @staticmethod - def _serialize_payload(requester, room_id, user_id, remote_room_hosts): + def _serialize_payload(requester, room_id, user_id, remote_room_hosts, content): """ Args: requester(Requester) @@ -118,12 +119,14 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint): return { "requester": requester.serialize(), "remote_room_hosts": remote_room_hosts, + "content": content, } async def _handle_request(self, request, room_id, user_id): content = parse_json_object_from_request(request) remote_room_hosts = content["remote_room_hosts"] + event_content = content["content"] requester = Requester.deserialize(self.store, content["requester"]) @@ -134,7 +137,7 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint): try: event = await self.federation_handler.do_remotely_reject_invite( - remote_room_hosts, room_id, user_id + remote_room_hosts, room_id, user_id, event_content, ) ret = event.get_pdu_json() except Exception as e: From 8c9a713f8db1d6fcc1f876ac6fbd0e54b5e5819c Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Thu, 28 Nov 2019 11:32:06 +0000 Subject: [PATCH 41/47] Add tests --- tests/rest/client/v1/test_rooms.py | 140 +++++++++++++++++++++++++++++ 1 file changed, 140 insertions(+) diff --git a/tests/rest/client/v1/test_rooms.py b/tests/rest/client/v1/test_rooms.py index e84e578f99..eda2fabc71 100644 --- a/tests/rest/client/v1/test_rooms.py +++ b/tests/rest/client/v1/test_rooms.py @@ -1180,3 +1180,143 @@ class PerRoomProfilesForbiddenTestCase(unittest.HomeserverTestCase): res_displayname = channel.json_body["content"]["displayname"] self.assertEqual(res_displayname, self.displayname, channel.result) + + +class RoomMembershipReasonTestCase(unittest.HomeserverTestCase): + """Tests that clients can add a "reason" field to membership events and + that they get correctly added to the generated events and propagated. + """ + + servlets = [ + synapse.rest.admin.register_servlets_for_client_rest_resource, + room.register_servlets, + login.register_servlets, + ] + + def prepare(self, reactor, clock, homeserver): + self.creator = self.register_user("creator", "test") + self.creator_tok = self.login("creator", "test") + + self.second_user_id = self.register_user("second", "test") + self.second_tok = self.login("second", "test") + + self.room_id = self.helper.create_room_as(self.creator, tok=self.creator_tok) + + def test_join_reason(self): + reason = "hello" + request, channel = self.make_request( + "POST", + "/_matrix/client/r0/rooms/{}/join".format(self.room_id), + content={"reason": reason}, + access_token=self.second_tok, + ) + self.render(request) + self.assertEqual(channel.code, 200, channel.result) + + self._check_for_reason(reason) + + def test_leave_reason(self): + self.helper.join(self.room_id, user=self.second_user_id, tok=self.second_tok) + + reason = "hello" + request, channel = self.make_request( + "POST", + "/_matrix/client/r0/rooms/{}/leave".format(self.room_id), + content={"reason": reason}, + access_token=self.second_tok, + ) + self.render(request) + self.assertEqual(channel.code, 200, channel.result) + + self._check_for_reason(reason) + + def test_kick_reason(self): + self.helper.join(self.room_id, user=self.second_user_id, tok=self.second_tok) + + reason = "hello" + request, channel = self.make_request( + "POST", + "/_matrix/client/r0/rooms/{}/kick".format(self.room_id), + content={"reason": reason, "user_id": self.second_user_id}, + access_token=self.second_tok, + ) + self.render(request) + self.assertEqual(channel.code, 200, channel.result) + + self._check_for_reason(reason) + + def test_ban_reason(self): + self.helper.join(self.room_id, user=self.second_user_id, tok=self.second_tok) + + reason = "hello" + request, channel = self.make_request( + "POST", + "/_matrix/client/r0/rooms/{}/ban".format(self.room_id), + content={"reason": reason, "user_id": self.second_user_id}, + access_token=self.creator_tok, + ) + self.render(request) + self.assertEqual(channel.code, 200, channel.result) + + self._check_for_reason(reason) + + def test_unban_reason(self): + reason = "hello" + request, channel = self.make_request( + "POST", + "/_matrix/client/r0/rooms/{}/unban".format(self.room_id), + content={"reason": reason, "user_id": self.second_user_id}, + access_token=self.creator_tok, + ) + self.render(request) + self.assertEqual(channel.code, 200, channel.result) + + self._check_for_reason(reason) + + def test_invite_reason(self): + reason = "hello" + request, channel = self.make_request( + "POST", + "/_matrix/client/r0/rooms/{}/invite".format(self.room_id), + content={"reason": reason, "user_id": self.second_user_id}, + access_token=self.creator_tok, + ) + self.render(request) + self.assertEqual(channel.code, 200, channel.result) + + self._check_for_reason(reason) + + def test_reject_invite_reason(self): + self.helper.invite( + self.room_id, + src=self.creator, + targ=self.second_user_id, + tok=self.creator_tok, + ) + + reason = "hello" + request, channel = self.make_request( + "POST", + "/_matrix/client/r0/rooms/{}/leave".format(self.room_id), + content={"reason": reason}, + access_token=self.second_tok, + ) + self.render(request) + self.assertEqual(channel.code, 200, channel.result) + + self._check_for_reason(reason) + + def _check_for_reason(self, reason): + request, channel = self.make_request( + "GET", + "/_matrix/client/r0/rooms/{}/state/m.room.member/{}".format( + self.room_id, self.second_user_id + ), + access_token=self.creator_tok, + ) + self.render(request) + self.assertEqual(channel.code, 200, channel.result) + + event_content = channel.json_body + + self.assertEqual(event_content.get("reason"), reason, channel.result) From 19ba7c142eaced06110c1cb2d22a489dae2ac155 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Thu, 28 Nov 2019 13:59:32 +0000 Subject: [PATCH 42/47] Newsfile --- changelog.d/6434.feature | 1 + 1 file changed, 1 insertion(+) create mode 100644 changelog.d/6434.feature diff --git a/changelog.d/6434.feature b/changelog.d/6434.feature new file mode 100644 index 0000000000..affa5d50c1 --- /dev/null +++ b/changelog.d/6434.feature @@ -0,0 +1 @@ +Add support for MSC 2367, which allows specifying a reason on all membership events. From 7baeea9f37f1cb7bf9305f4991b1e1b357f161cf Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Thu, 28 Nov 2019 14:55:19 +0000 Subject: [PATCH 43/47] blacklist more tests --- .buildkite/worker-blacklist | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/.buildkite/worker-blacklist b/.buildkite/worker-blacklist index d7908af177..7950d19db3 100644 --- a/.buildkite/worker-blacklist +++ b/.buildkite/worker-blacklist @@ -57,3 +57,10 @@ Don't get pushed for rooms you've muted Rejected events are not pushed Test that rejected pushers are removed. Events come down the correct room + +# https://buildkite.com/matrix-dot-org/sytest/builds/326#cca62404-a88a-4fcb-ad41-175fd3377603 +Presence changes to UNAVAILABLE are reported to remote room members +If remote user leaves room, changes device and rejoins we see update in sync +uploading self-signing key notifies over federation +Inbound federation can receive redacted events +Outbound federation can request missing events From 708cef88cfbf8dd6df44d2da4ab4dbc7eb584f74 Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Thu, 28 Nov 2019 19:26:13 +0000 Subject: [PATCH 44/47] Discard retention policies when retrieving state Purge jobs don't delete the latest event in a room in order to keep the forward extremity and not break the room. On the other hand, get_state_events, when given an at_token argument calls filter_events_for_client to know if the user can see the event that matches that (sync) token. That function uses the retention policies of the events it's given to filter out those that are too old from a client's view. Some clients, such as Riot, when loading a room, request the list of members for the latest sync token it knows about, and get confused to the point of refusing to send any message if the server tells it that it can't get that information. This can happen very easily with the message retention feature turned on and a room with low activity so that the last event sent becomes too old according to the room's retention policy. An easy and clean fix for that issue is to discard the room's retention policies when retrieving state. --- synapse/handlers/message.py | 2 +- synapse/visibility.py | 22 ++++++++++++++-------- 2 files changed, 15 insertions(+), 9 deletions(-) diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 155ed6e06a..3b0156f516 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -138,7 +138,7 @@ class MessageHandler(object): raise NotFoundError("Can't find event for token %s" % (at_token,)) visible_events = yield filter_events_for_client( - self.storage, user_id, last_events + self.storage, user_id, last_events, apply_retention_policies=False ) event = last_events[0] diff --git a/synapse/visibility.py b/synapse/visibility.py index 4d4141dacc..7b037eeb0c 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -44,7 +44,8 @@ MEMBERSHIP_PRIORITY = ( @defer.inlineCallbacks def filter_events_for_client( - storage: Storage, user_id, events, is_peeking=False, always_include_ids=frozenset() + storage: Storage, user_id, events, is_peeking=False, always_include_ids=frozenset(), + apply_retention_policies=True, ): """ Check which events a user is allowed to see @@ -59,6 +60,10 @@ def filter_events_for_client( events always_include_ids (set(event_id)): set of event ids to specifically include (unless sender is ignored) + apply_retention_policies (bool): Whether to filter out events that's older than + allowed by the room's retention policy. Useful when this function is called + to e.g. check whether a user should be allowed to see the state at a given + event rather than to know if it should send an event to a user's client(s). Returns: Deferred[list[synapse.events.EventBase]] @@ -86,13 +91,14 @@ def filter_events_for_client( erased_senders = yield storage.main.are_users_erased((e.sender for e in events)) - room_ids = set(e.room_id for e in events) - retention_policies = {} + if apply_retention_policies: + room_ids = set(e.room_id for e in events) + retention_policies = {} - for room_id in room_ids: - retention_policies[room_id] = yield storage.main.get_retention_policy_for_room( - room_id - ) + for room_id in room_ids: + retention_policies[room_id] = ( + yield storage.main.get_retention_policy_for_room(room_id) + ) def allowed(event): """ @@ -113,7 +119,7 @@ def filter_events_for_client( # Don't try to apply the room's retention policy if the event is a state event, as # MSC1763 states that retention is only considered for non-state events. - if not event.is_state(): + if apply_retention_policies and not event.is_state(): retention_policy = retention_policies[event.room_id] max_lifetime = retention_policy.get("max_lifetime") From 5ee2beeddbbcbf09ac054679de71db0e0bf9df31 Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Thu, 28 Nov 2019 19:32:49 +0000 Subject: [PATCH 45/47] Changelog --- changelog.d/6436.bugfix | 1 + 1 file changed, 1 insertion(+) create mode 100644 changelog.d/6436.bugfix diff --git a/changelog.d/6436.bugfix b/changelog.d/6436.bugfix new file mode 100644 index 0000000000..954a4e1d84 --- /dev/null +++ b/changelog.d/6436.bugfix @@ -0,0 +1 @@ +Fix a bug where a room could become unusable with a low retention policy and a low activity. From 78ec11c08562bfd635497621da238c7197e69b6f Mon Sep 17 00:00:00 2001 From: Brendan Abolivier Date: Thu, 28 Nov 2019 20:35:22 +0000 Subject: [PATCH 46/47] Lint --- synapse/visibility.py | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) diff --git a/synapse/visibility.py b/synapse/visibility.py index 7b037eeb0c..dffe943b28 100644 --- a/synapse/visibility.py +++ b/synapse/visibility.py @@ -44,7 +44,11 @@ MEMBERSHIP_PRIORITY = ( @defer.inlineCallbacks def filter_events_for_client( - storage: Storage, user_id, events, is_peeking=False, always_include_ids=frozenset(), + storage: Storage, + user_id, + events, + is_peeking=False, + always_include_ids=frozenset(), apply_retention_policies=True, ): """ @@ -96,9 +100,9 @@ def filter_events_for_client( retention_policies = {} for room_id in room_ids: - retention_policies[room_id] = ( - yield storage.main.get_retention_policy_for_room(room_id) - ) + retention_policies[ + room_id + ] = yield storage.main.get_retention_policy_for_room(room_id) def allowed(event): """ From 23ea5721259059d50b80083bb7240a8cb56cf297 Mon Sep 17 00:00:00 2001 From: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com> Date: Fri, 29 Nov 2019 13:51:14 +0000 Subject: [PATCH 47/47] Add User-Interactive Auth to /account/3pid/add (#6119) --- changelog.d/6119.feature | 1 + synapse/rest/client/v2_alpha/account.py | 5 +++++ 2 files changed, 6 insertions(+) create mode 100644 changelog.d/6119.feature diff --git a/changelog.d/6119.feature b/changelog.d/6119.feature new file mode 100644 index 0000000000..1492e83c5a --- /dev/null +++ b/changelog.d/6119.feature @@ -0,0 +1 @@ +Require User-Interactive Authentication for `/account/3pid/add`, meaning the user's password will be required to add a third-party ID to their account. \ No newline at end of file diff --git a/synapse/rest/client/v2_alpha/account.py b/synapse/rest/client/v2_alpha/account.py index f26eae794c..ad674239ab 100644 --- a/synapse/rest/client/v2_alpha/account.py +++ b/synapse/rest/client/v2_alpha/account.py @@ -642,6 +642,7 @@ class ThreepidAddRestServlet(RestServlet): self.auth = hs.get_auth() self.auth_handler = hs.get_auth_handler() + @interactive_auth_handler @defer.inlineCallbacks def on_POST(self, request): requester = yield self.auth.get_user_by_req(request) @@ -652,6 +653,10 @@ class ThreepidAddRestServlet(RestServlet): client_secret = body["client_secret"] sid = body["sid"] + yield self.auth_handler.validate_user_via_ui_auth( + requester, body, self.hs.get_ip_from_request(request) + ) + validation_session = yield self.identity_handler.validate_threepid_session( client_secret, sid )