mirror of
https://mau.dev/maunium/synapse.git
synced 2024-08-26 12:16:12 +00:00
![Eric Eastwood](/assets/img/avatar_default.png)
* Make historical messages available to federated servers Part of MSC2716: https://github.com/matrix-org/matrix-doc/pull/2716 Follow-up to https://github.com/matrix-org/synapse/pull/9247 * Debug message not available on federation * Add base starting insertion point when no chunk ID is provided * Fix messages from multiple senders in historical chunk Follow-up to https://github.com/matrix-org/synapse/pull/9247 Part of MSC2716: https://github.com/matrix-org/matrix-doc/pull/2716 --- Previously, Synapse would throw a 403, `Cannot force another user to join.`, because we were trying to use `?user_id` from a single virtual user which did not match with messages from other users in the chunk. * Remove debug lines * Messing with selecting insertion event extremeties * Move db schema change to new version * Add more better comments * Make a fake requester with just what we need See https://github.com/matrix-org/synapse/pull/10276#discussion_r660999080 * Store insertion events in table * Make base insertion event float off on its own See https://github.com/matrix-org/synapse/pull/10250#issuecomment-875711889 Conflicts: synapse/rest/client/v1/room.py * Validate that the app service can actually control the given user See https://github.com/matrix-org/synapse/pull/10276#issuecomment-876316455 Conflicts: synapse/rest/client/v1/room.py * Add some better comments on what we're trying to check for * Continue debugging * Share validation logic * Add inserted historical messages to /backfill response * Remove debug sql queries * Some marker event implemntation trials * Clean up PR * Rename insertion_event_id to just event_id * Add some better sql comments * More accurate description * Add changelog * Make it clear what MSC the change is part of * Add more detail on which insertion event came through * Address review and improve sql queries * Only use event_id as unique constraint * Fix test case where insertion event is already in the normal DAG * Remove debug changes * Add support for MSC2716 marker events * Process markers when we receive it over federation * WIP: make hs2 backfill historical messages after marker event * hs2 to better ask for insertion event extremity But running into the `sqlite3.IntegrityError: NOT NULL constraint failed: event_to_state_groups.state_group` error * Add insertion_event_extremities table * Switch to chunk events so we can auth via power_levels Previously, we were using `content.chunk_id` to connect one chunk to another. But these events can be from any `sender` and we can't tell who should be able to send historical events. We know we only want the application service to do it but these events have the sender of a real historical message, not the application service user ID as the sender. Other federated homeservers also have no indicator which senders are an application service on the originating homeserver. So we want to auth all of the MSC2716 events via power_levels and have them be sent by the application service with proper PL levels in the room. * Switch to chunk events for federation * Add unstable room version to support new historical PL * Messy: Fix undefined state_group for federated historical events ``` 2021-07-13 02:27:57,810 - synapse.handlers.federation - 1248 - ERROR - GET-4 - Failed to backfill from hs1 because NOT NULL constraint failed: event_to_state_groups.state_group Traceback (most recent call last): File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 1216, in try_backfill await self.backfill( File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 1035, in backfill await self._auth_and_persist_event(dest, event, context, backfilled=True) File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 2222, in _auth_and_persist_event await self._run_push_actions_and_persist_event(event, context, backfilled) File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 2244, in _run_push_actions_and_persist_event await self.persist_events_and_notify( File "/usr/local/lib/python3.8/site-packages/synapse/handlers/federation.py", line 3290, in persist_events_and_notify events, max_stream_token = await self.storage.persistence.persist_events( File "/usr/local/lib/python3.8/site-packages/synapse/logging/opentracing.py", line 774, in _trace_inner return await func(*args, **kwargs) File "/usr/local/lib/python3.8/site-packages/synapse/storage/persist_events.py", line 320, in persist_events ret_vals = await yieldable_gather_results(enqueue, partitioned.items()) File "/usr/local/lib/python3.8/site-packages/synapse/storage/persist_events.py", line 237, in handle_queue_loop ret = await self._per_item_callback( File "/usr/local/lib/python3.8/site-packages/synapse/storage/persist_events.py", line 577, in _persist_event_batch await self.persist_events_store._persist_events_and_state_updates( File "/usr/local/lib/python3.8/site-packages/synapse/storage/databases/main/events.py", line 176, in _persist_events_and_state_updates await self.db_pool.runInteraction( File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 681, in runInteraction result = await self.runWithConnection( File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 770, in runWithConnection return await make_deferred_yieldable( File "/usr/local/lib/python3.8/site-packages/twisted/python/threadpool.py", line 238, in inContext result = inContext.theWork() # type: ignore[attr-defined] File "/usr/local/lib/python3.8/site-packages/twisted/python/threadpool.py", line 254, in <lambda> inContext.theWork = lambda: context.call( # type: ignore[attr-defined] File "/usr/local/lib/python3.8/site-packages/twisted/python/context.py", line 118, in callWithContext return self.currentContext().callWithContext(ctx, func, *args, **kw) File "/usr/local/lib/python3.8/site-packages/twisted/python/context.py", line 83, in callWithContext return func(*args, **kw) File "/usr/local/lib/python3.8/site-packages/twisted/enterprise/adbapi.py", line 293, in _runWithConnection compat.reraise(excValue, excTraceback) File "/usr/local/lib/python3.8/site-packages/twisted/python/deprecate.py", line 298, in deprecatedFunction return function(*args, **kwargs) File "/usr/local/lib/python3.8/site-packages/twisted/python/compat.py", line 403, in reraise raise exception.with_traceback(traceback) File "/usr/local/lib/python3.8/site-packages/twisted/enterprise/adbapi.py", line 284, in _runWithConnection result = func(conn, *args, **kw) File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 765, in inner_func return func(db_conn, *args, **kwargs) File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 549, in new_transaction r = func(cursor, *args, **kwargs) File "/usr/local/lib/python3.8/site-packages/synapse/logging/utils.py", line 69, in wrapped return f(*args, **kwargs) File "/usr/local/lib/python3.8/site-packages/synapse/storage/databases/main/events.py", line 385, in _persist_events_txn self._store_event_state_mappings_txn(txn, events_and_contexts) File "/usr/local/lib/python3.8/site-packages/synapse/storage/databases/main/events.py", line 2065, in _store_event_state_mappings_txn self.db_pool.simple_insert_many_txn( File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 923, in simple_insert_many_txn txn.execute_batch(sql, vals) File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 280, in execute_batch self.executemany(sql, args) File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 300, in executemany self._do_execute(self.txn.executemany, sql, *args) File "/usr/local/lib/python3.8/site-packages/synapse/storage/database.py", line 330, in _do_execute return func(sql, *args) sqlite3.IntegrityError: NOT NULL constraint failed: event_to_state_groups.state_group ``` * Revert "Messy: Fix undefined state_group for federated historical events" This reverts commit 187ab28611546321e02770944c86f30ee2bc742a. * Fix federated events being rejected for no state_groups Add fix from https://github.com/matrix-org/synapse/pull/10439 until it merges. * Adapting to experimental room version * Some log cleanup * Add better comments around extremity fetching code and why * Rename to be more accurate to what the function returns * Add changelog * Ignore rejected events * Use simplified upsert * Add Erik's explanation of extra event checks See https://github.com/matrix-org/synapse/pull/10498#discussion_r680880332 * Clarify that the depth is not directly correlated to the backwards extremity that we return See https://github.com/matrix-org/synapse/pull/10498#discussion_r681725404 * lock only matters for sqlite See https://github.com/matrix-org/synapse/pull/10498#discussion_r681728061 * Move new SQL changes to its own delta file * Clean up upsert docstring * Bump database schema version (62)
3515 lines
134 KiB
Python
3515 lines
134 KiB
Python
# Copyright 2014-2021 The Matrix.org Foundation C.I.C.
|
|
# Copyright 2020 Sorunome
|
|
#
|
|
# 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.
|
|
|
|
"""Contains handlers for federation events."""
|
|
|
|
import itertools
|
|
import logging
|
|
from collections.abc import Container
|
|
from http import HTTPStatus
|
|
from typing import (
|
|
TYPE_CHECKING,
|
|
Collection,
|
|
Dict,
|
|
Iterable,
|
|
List,
|
|
Optional,
|
|
Sequence,
|
|
Set,
|
|
Tuple,
|
|
Union,
|
|
)
|
|
|
|
import attr
|
|
from prometheus_client import Counter
|
|
from signedjson.key import decode_verify_key_bytes
|
|
from signedjson.sign import verify_signed_json
|
|
from unpaddedbase64 import decode_base64
|
|
|
|
from twisted.internet import defer
|
|
|
|
from synapse import event_auth
|
|
from synapse.api.constants import (
|
|
EventContentFields,
|
|
EventTypes,
|
|
Membership,
|
|
RejectedReason,
|
|
RoomEncryptionAlgorithms,
|
|
)
|
|
from synapse.api.errors import (
|
|
AuthError,
|
|
CodeMessageException,
|
|
Codes,
|
|
FederationDeniedError,
|
|
FederationError,
|
|
HttpResponseException,
|
|
NotFoundError,
|
|
RequestSendFailed,
|
|
SynapseError,
|
|
)
|
|
from synapse.api.room_versions import KNOWN_ROOM_VERSIONS, RoomVersion, RoomVersions
|
|
from synapse.crypto.event_signing import compute_event_signature
|
|
from synapse.event_auth import auth_types_for_event
|
|
from synapse.events import EventBase
|
|
from synapse.events.snapshot import EventContext
|
|
from synapse.events.validator import EventValidator
|
|
from synapse.handlers._base import BaseHandler
|
|
from synapse.http.servlet import assert_params_in_dict
|
|
from synapse.logging.context import (
|
|
make_deferred_yieldable,
|
|
nested_logging_context,
|
|
preserve_fn,
|
|
run_in_background,
|
|
)
|
|
from synapse.logging.utils import log_function
|
|
from synapse.metrics.background_process_metrics import run_as_background_process
|
|
from synapse.replication.http.devices import ReplicationUserDevicesResyncRestServlet
|
|
from synapse.replication.http.federation import (
|
|
ReplicationCleanRoomRestServlet,
|
|
ReplicationFederationSendEventsRestServlet,
|
|
ReplicationStoreRoomOnOutlierMembershipRestServlet,
|
|
)
|
|
from synapse.state import StateResolutionStore
|
|
from synapse.storage.databases.main.events_worker import EventRedactBehaviour
|
|
from synapse.types import (
|
|
JsonDict,
|
|
MutableStateMap,
|
|
PersistedEventPosition,
|
|
RoomStreamToken,
|
|
StateMap,
|
|
UserID,
|
|
get_domain_from_id,
|
|
)
|
|
from synapse.util.async_helpers import Linearizer, concurrently_execute
|
|
from synapse.util.iterutils import batch_iter
|
|
from synapse.util.retryutils import NotRetryingDestination
|
|
from synapse.util.stringutils import shortstr
|
|
from synapse.visibility import filter_events_for_server
|
|
|
|
if TYPE_CHECKING:
|
|
from synapse.server import HomeServer
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
soft_failed_event_counter = Counter(
|
|
"synapse_federation_soft_failed_events_total",
|
|
"Events received over federation that we marked as soft_failed",
|
|
)
|
|
|
|
|
|
@attr.s(slots=True)
|
|
class _NewEventInfo:
|
|
"""Holds information about a received event, ready for passing to _auth_and_persist_events
|
|
|
|
Attributes:
|
|
event: the received event
|
|
|
|
state: the state at that event
|
|
|
|
auth_events: the auth_event map for that event
|
|
"""
|
|
|
|
event = attr.ib(type=EventBase)
|
|
state = attr.ib(type=Optional[Sequence[EventBase]], default=None)
|
|
auth_events = attr.ib(type=Optional[MutableStateMap[EventBase]], default=None)
|
|
|
|
|
|
class FederationHandler(BaseHandler):
|
|
"""Handles events that originated from federation.
|
|
Responsible for:
|
|
a) handling received Pdus before handing them on as Events to the rest
|
|
of the homeserver (including auth and state conflict resolutions)
|
|
b) converting events that were produced by local clients that may need
|
|
to be sent to remote homeservers.
|
|
c) doing the necessary dances to invite remote users and join remote
|
|
rooms.
|
|
"""
|
|
|
|
def __init__(self, hs: "HomeServer"):
|
|
super().__init__(hs)
|
|
|
|
self.hs = hs
|
|
|
|
self.store = hs.get_datastore()
|
|
self.storage = hs.get_storage()
|
|
self.state_store = self.storage.state
|
|
self.federation_client = hs.get_federation_client()
|
|
self.state_handler = hs.get_state_handler()
|
|
self._state_resolution_handler = hs.get_state_resolution_handler()
|
|
self.server_name = hs.hostname
|
|
self.keyring = hs.get_keyring()
|
|
self.action_generator = hs.get_action_generator()
|
|
self.is_mine_id = hs.is_mine_id
|
|
self.spam_checker = hs.get_spam_checker()
|
|
self.event_creation_handler = hs.get_event_creation_handler()
|
|
self._event_auth_handler = hs.get_event_auth_handler()
|
|
self._message_handler = hs.get_message_handler()
|
|
self._server_notices_mxid = hs.config.server_notices_mxid
|
|
self.config = hs.config
|
|
self.http_client = hs.get_proxied_blacklisted_http_client()
|
|
self._instance_name = hs.get_instance_name()
|
|
self._replication = hs.get_replication_data_handler()
|
|
|
|
self._send_events = ReplicationFederationSendEventsRestServlet.make_client(hs)
|
|
self._clean_room_for_join_client = ReplicationCleanRoomRestServlet.make_client(
|
|
hs
|
|
)
|
|
|
|
if hs.config.worker_app:
|
|
self._user_device_resync = (
|
|
ReplicationUserDevicesResyncRestServlet.make_client(hs)
|
|
)
|
|
self._maybe_store_room_on_outlier_membership = (
|
|
ReplicationStoreRoomOnOutlierMembershipRestServlet.make_client(hs)
|
|
)
|
|
else:
|
|
self._device_list_updater = hs.get_device_handler().device_list_updater
|
|
self._maybe_store_room_on_outlier_membership = (
|
|
self.store.maybe_store_room_on_outlier_membership
|
|
)
|
|
|
|
# When joining a room we need to queue any events for that room up.
|
|
# For each room, a list of (pdu, origin) tuples.
|
|
self.room_queues: Dict[str, List[Tuple[EventBase, str]]] = {}
|
|
self._room_pdu_linearizer = Linearizer("fed_room_pdu")
|
|
|
|
self._room_backfill = Linearizer("room_backfill")
|
|
|
|
self.third_party_event_rules = hs.get_third_party_event_rules()
|
|
|
|
self._ephemeral_messages_enabled = hs.config.enable_ephemeral_messages
|
|
|
|
async def on_receive_pdu(
|
|
self, origin: str, pdu: EventBase, sent_to_us_directly: bool = False
|
|
) -> None:
|
|
"""Process a PDU received via a federation /send/ transaction, or
|
|
via backfill of missing prev_events
|
|
|
|
Args:
|
|
origin: server which initiated the /send/ transaction. Will
|
|
be used to fetch missing events or state.
|
|
pdu: received PDU
|
|
sent_to_us_directly: True if this event was pushed to us; False if
|
|
we pulled it as the result of a missing prev_event.
|
|
"""
|
|
|
|
room_id = pdu.room_id
|
|
event_id = pdu.event_id
|
|
|
|
logger.info("handling received PDU: %s", pdu)
|
|
|
|
# We reprocess pdus when we have seen them only as outliers
|
|
existing = await self.store.get_event(
|
|
event_id, allow_none=True, allow_rejected=True
|
|
)
|
|
|
|
# FIXME: Currently we fetch an event again when we already have it
|
|
# if it has been marked as an outlier.
|
|
|
|
already_seen = existing and (
|
|
not existing.internal_metadata.is_outlier()
|
|
or pdu.internal_metadata.is_outlier()
|
|
)
|
|
if already_seen:
|
|
logger.debug("Already seen pdu")
|
|
return
|
|
|
|
# do some initial sanity-checking of the event. In particular, make
|
|
# sure it doesn't have hundreds of prev_events or auth_events, which
|
|
# could cause a huge state resolution or cascade of event fetches.
|
|
try:
|
|
self._sanity_check_event(pdu)
|
|
except SynapseError as err:
|
|
logger.warning("Received event failed sanity checks")
|
|
raise FederationError("ERROR", err.code, err.msg, affected=pdu.event_id)
|
|
|
|
# If we are currently in the process of joining this room, then we
|
|
# queue up events for later processing.
|
|
if room_id in self.room_queues:
|
|
logger.info(
|
|
"Queuing PDU from %s for now: join in progress",
|
|
origin,
|
|
)
|
|
self.room_queues[room_id].append((pdu, origin))
|
|
return
|
|
|
|
# If we're not in the room just ditch the event entirely. This is
|
|
# probably an old server that has come back and thinks we're still in
|
|
# the room (or we've been rejoined to the room by a state reset).
|
|
#
|
|
# Note that if we were never in the room then we would have already
|
|
# dropped the event, since we wouldn't know the room version.
|
|
is_in_room = await self._event_auth_handler.check_host_in_room(
|
|
room_id, self.server_name
|
|
)
|
|
if not is_in_room:
|
|
logger.info(
|
|
"Ignoring PDU from %s as we're not in the room",
|
|
origin,
|
|
)
|
|
return None
|
|
|
|
state = None
|
|
|
|
# Check that the event passes auth based on the state at the event. This is
|
|
# done for events that are to be added to the timeline (non-outliers).
|
|
#
|
|
# Get missing pdus if necessary:
|
|
# - Fetching any missing prev events to fill in gaps in the graph
|
|
# - Fetching state if we have a hole in the graph
|
|
if not pdu.internal_metadata.is_outlier():
|
|
# We only backfill backwards to the min depth.
|
|
min_depth = await self.get_min_depth_for_context(pdu.room_id)
|
|
|
|
logger.debug("min_depth: %d", min_depth)
|
|
|
|
prevs = set(pdu.prev_event_ids())
|
|
seen = await self.store.have_events_in_timeline(prevs)
|
|
|
|
if min_depth is not None and pdu.depth < min_depth:
|
|
# This is so that we don't notify the user about this
|
|
# message, to work around the fact that some events will
|
|
# reference really really old events we really don't want to
|
|
# send to the clients.
|
|
pdu.internal_metadata.outlier = True
|
|
elif min_depth is not None and pdu.depth > min_depth:
|
|
missing_prevs = prevs - seen
|
|
if sent_to_us_directly and missing_prevs:
|
|
# If we're missing stuff, ensure we only fetch stuff one
|
|
# at a time.
|
|
logger.info(
|
|
"Acquiring room lock to fetch %d missing prev_events: %s",
|
|
len(missing_prevs),
|
|
shortstr(missing_prevs),
|
|
)
|
|
with (await self._room_pdu_linearizer.queue(pdu.room_id)):
|
|
logger.info(
|
|
"Acquired room lock to fetch %d missing prev_events",
|
|
len(missing_prevs),
|
|
)
|
|
|
|
try:
|
|
await self._get_missing_events_for_pdu(
|
|
origin, pdu, prevs, min_depth
|
|
)
|
|
except Exception as e:
|
|
raise Exception(
|
|
"Error fetching missing prev_events for %s: %s"
|
|
% (event_id, e)
|
|
) from e
|
|
|
|
# Update the set of things we've seen after trying to
|
|
# fetch the missing stuff
|
|
seen = await self.store.have_events_in_timeline(prevs)
|
|
|
|
if not prevs - seen:
|
|
logger.info(
|
|
"Found all missing prev_events",
|
|
)
|
|
|
|
if prevs - seen:
|
|
# We've still not been able to get all of the prev_events for this event.
|
|
#
|
|
# In this case, we need to fall back to asking another server in the
|
|
# federation for the state at this event. That's ok provided we then
|
|
# resolve the state against other bits of the DAG before using it (which
|
|
# will ensure that you can't just take over a room by sending an event,
|
|
# withholding its prev_events, and declaring yourself to be an admin in
|
|
# the subsequent state request).
|
|
#
|
|
# Now, if we're pulling this event as a missing prev_event, then clearly
|
|
# this event is not going to become the only forward-extremity and we are
|
|
# guaranteed to resolve its state against our existing forward
|
|
# extremities, so that should be fine.
|
|
#
|
|
# On the other hand, if this event was pushed to us, it is possible for
|
|
# it to become the only forward-extremity in the room, and we would then
|
|
# trust its state to be the state for the whole room. This is very bad.
|
|
# Further, if the event was pushed to us, there is no excuse for us not to
|
|
# have all the prev_events. We therefore reject any such events.
|
|
#
|
|
# XXX this really feels like it could/should be merged with the above,
|
|
# but there is an interaction with min_depth that I'm not really
|
|
# following.
|
|
|
|
if sent_to_us_directly:
|
|
logger.warning(
|
|
"Rejecting: failed to fetch %d prev events: %s",
|
|
len(prevs - seen),
|
|
shortstr(prevs - seen),
|
|
)
|
|
raise FederationError(
|
|
"ERROR",
|
|
403,
|
|
(
|
|
"Your server isn't divulging details about prev_events "
|
|
"referenced in this event."
|
|
),
|
|
affected=pdu.event_id,
|
|
)
|
|
|
|
logger.info(
|
|
"Event %s is missing prev_events: calculating state for a "
|
|
"backwards extremity",
|
|
event_id,
|
|
)
|
|
|
|
# Calculate the state after each of the previous events, and
|
|
# resolve them to find the correct state at the current event.
|
|
event_map = {event_id: pdu}
|
|
try:
|
|
# Get the state of the events we know about
|
|
ours = await self.state_store.get_state_groups_ids(room_id, seen)
|
|
|
|
# state_maps is a list of mappings from (type, state_key) to event_id
|
|
state_maps: List[StateMap[str]] = list(ours.values())
|
|
|
|
# we don't need this any more, let's delete it.
|
|
del ours
|
|
|
|
# Ask the remote server for the states we don't
|
|
# know about
|
|
for p in prevs - seen:
|
|
logger.info("Requesting state after missing prev_event %s", p)
|
|
|
|
with nested_logging_context(p):
|
|
# note that if any of the missing prevs share missing state or
|
|
# auth events, the requests to fetch those events are deduped
|
|
# by the get_pdu_cache in federation_client.
|
|
remote_state = (
|
|
await self._get_state_after_missing_prev_event(
|
|
origin, room_id, p
|
|
)
|
|
)
|
|
|
|
remote_state_map = {
|
|
(x.type, x.state_key): x.event_id for x in remote_state
|
|
}
|
|
state_maps.append(remote_state_map)
|
|
|
|
for x in remote_state:
|
|
event_map[x.event_id] = x
|
|
|
|
room_version = await self.store.get_room_version_id(room_id)
|
|
state_map = (
|
|
await self._state_resolution_handler.resolve_events_with_store(
|
|
room_id,
|
|
room_version,
|
|
state_maps,
|
|
event_map,
|
|
state_res_store=StateResolutionStore(self.store),
|
|
)
|
|
)
|
|
|
|
# We need to give _process_received_pdu the actual state events
|
|
# rather than event ids, so generate that now.
|
|
|
|
# First though we need to fetch all the events that are in
|
|
# state_map, so we can build up the state below.
|
|
evs = await self.store.get_events(
|
|
list(state_map.values()),
|
|
get_prev_content=False,
|
|
redact_behaviour=EventRedactBehaviour.AS_IS,
|
|
)
|
|
event_map.update(evs)
|
|
|
|
state = [event_map[e] for e in state_map.values()]
|
|
except Exception:
|
|
logger.warning(
|
|
"Error attempting to resolve state at missing " "prev_events",
|
|
exc_info=True,
|
|
)
|
|
raise FederationError(
|
|
"ERROR",
|
|
403,
|
|
"We can't get valid state history.",
|
|
affected=event_id,
|
|
)
|
|
|
|
# A second round of checks for all events. Check that the event passes auth
|
|
# based on `auth_events`, this allows us to assert that the event would
|
|
# have been allowed at some point. If an event passes this check its OK
|
|
# for it to be used as part of a returned `/state` request, as either
|
|
# a) we received the event as part of the original join and so trust it, or
|
|
# b) we'll do a state resolution with existing state before it becomes
|
|
# part of the "current state", which adds more protection.
|
|
await self._process_received_pdu(origin, pdu, state=state)
|
|
|
|
async def _get_missing_events_for_pdu(
|
|
self, origin: str, pdu: EventBase, prevs: Set[str], min_depth: int
|
|
) -> None:
|
|
"""
|
|
Args:
|
|
origin: Origin of the pdu. Will be called to get the missing events
|
|
pdu: received pdu
|
|
prevs: List of event ids which we are missing
|
|
min_depth: Minimum depth of events to return.
|
|
"""
|
|
|
|
room_id = pdu.room_id
|
|
event_id = pdu.event_id
|
|
|
|
seen = await self.store.have_events_in_timeline(prevs)
|
|
|
|
if not prevs - seen:
|
|
return
|
|
|
|
latest_list = await self.store.get_latest_event_ids_in_room(room_id)
|
|
|
|
# We add the prev events that we have seen to the latest
|
|
# list to ensure the remote server doesn't give them to us
|
|
latest = set(latest_list)
|
|
latest |= seen
|
|
|
|
logger.info(
|
|
"Requesting missing events between %s and %s",
|
|
shortstr(latest),
|
|
event_id,
|
|
)
|
|
|
|
# XXX: we set timeout to 10s to help workaround
|
|
# https://github.com/matrix-org/synapse/issues/1733.
|
|
# The reason is to avoid holding the linearizer lock
|
|
# whilst processing inbound /send transactions, causing
|
|
# FDs to stack up and block other inbound transactions
|
|
# which empirically can currently take up to 30 minutes.
|
|
#
|
|
# N.B. this explicitly disables retry attempts.
|
|
#
|
|
# N.B. this also increases our chances of falling back to
|
|
# fetching fresh state for the room if the missing event
|
|
# can't be found, which slightly reduces our security.
|
|
# it may also increase our DAG extremity count for the room,
|
|
# causing additional state resolution? See #1760.
|
|
# However, fetching state doesn't hold the linearizer lock
|
|
# apparently.
|
|
#
|
|
# see https://github.com/matrix-org/synapse/pull/1744
|
|
#
|
|
# ----
|
|
#
|
|
# Update richvdh 2018/09/18: There are a number of problems with timing this
|
|
# request out aggressively on the client side:
|
|
#
|
|
# - it plays badly with the server-side rate-limiter, which starts tarpitting you
|
|
# if you send too many requests at once, so you end up with the server carefully
|
|
# working through the backlog of your requests, which you have already timed
|
|
# out.
|
|
#
|
|
# - for this request in particular, we now (as of
|
|
# https://github.com/matrix-org/synapse/pull/3456) reject any PDUs where the
|
|
# server can't produce a plausible-looking set of prev_events - so we becone
|
|
# much more likely to reject the event.
|
|
#
|
|
# - contrary to what it says above, we do *not* fall back to fetching fresh state
|
|
# for the room if get_missing_events times out. Rather, we give up processing
|
|
# the PDU whose prevs we are missing, which then makes it much more likely that
|
|
# we'll end up back here for the *next* PDU in the list, which exacerbates the
|
|
# problem.
|
|
#
|
|
# - the aggressive 10s timeout was introduced to deal with incoming federation
|
|
# requests taking 8 hours to process. It's not entirely clear why that was going
|
|
# on; certainly there were other issues causing traffic storms which are now
|
|
# resolved, and I think in any case we may be more sensible about our locking
|
|
# now. We're *certainly* more sensible about our logging.
|
|
#
|
|
# All that said: Let's try increasing the timeout to 60s and see what happens.
|
|
|
|
try:
|
|
missing_events = await self.federation_client.get_missing_events(
|
|
origin,
|
|
room_id,
|
|
earliest_events_ids=list(latest),
|
|
latest_events=[pdu],
|
|
limit=10,
|
|
min_depth=min_depth,
|
|
timeout=60000,
|
|
)
|
|
except (RequestSendFailed, HttpResponseException, NotRetryingDestination) as e:
|
|
# We failed to get the missing events, but since we need to handle
|
|
# the case of `get_missing_events` not returning the necessary
|
|
# events anyway, it is safe to simply log the error and continue.
|
|
logger.warning("Failed to get prev_events: %s", e)
|
|
return
|
|
|
|
logger.info(
|
|
"Got %d prev_events: %s",
|
|
len(missing_events),
|
|
shortstr(missing_events),
|
|
)
|
|
|
|
# We want to sort these by depth so we process them and
|
|
# tell clients about them in order.
|
|
missing_events.sort(key=lambda x: x.depth)
|
|
|
|
for ev in missing_events:
|
|
logger.info(
|
|
"Handling received prev_event %s",
|
|
ev.event_id,
|
|
)
|
|
with nested_logging_context(ev.event_id):
|
|
try:
|
|
await self.on_receive_pdu(origin, ev, sent_to_us_directly=False)
|
|
except FederationError as e:
|
|
if e.code == 403:
|
|
logger.warning(
|
|
"Received prev_event %s failed history check.",
|
|
ev.event_id,
|
|
)
|
|
else:
|
|
raise
|
|
|
|
async def _get_state_for_room(
|
|
self,
|
|
destination: str,
|
|
room_id: str,
|
|
event_id: str,
|
|
) -> List[EventBase]:
|
|
"""Requests all of the room state at a given event from a remote
|
|
homeserver.
|
|
|
|
Will also fetch any missing events reported in the `auth_chain_ids`
|
|
section of `/state_ids`.
|
|
|
|
Args:
|
|
destination: The remote homeserver to query for the state.
|
|
room_id: The id of the room we're interested in.
|
|
event_id: The id of the event we want the state at.
|
|
|
|
Returns:
|
|
A list of events in the state, not including the event itself.
|
|
"""
|
|
(
|
|
state_event_ids,
|
|
auth_event_ids,
|
|
) = await self.federation_client.get_room_state_ids(
|
|
destination, room_id, event_id=event_id
|
|
)
|
|
|
|
# Fetch the state events from the DB, and check we have the auth events.
|
|
event_map = await self.store.get_events(state_event_ids, allow_rejected=True)
|
|
auth_events_in_store = await self.store.have_seen_events(
|
|
room_id, auth_event_ids
|
|
)
|
|
|
|
# Check for missing events. We handle state and auth event seperately,
|
|
# as we want to pull the state from the DB, but we don't for the auth
|
|
# events. (Note: we likely won't use the majority of the auth chain, and
|
|
# it can be *huge* for large rooms, so it's worth ensuring that we don't
|
|
# unnecessarily pull it from the DB).
|
|
missing_state_events = set(state_event_ids) - set(event_map)
|
|
missing_auth_events = set(auth_event_ids) - set(auth_events_in_store)
|
|
if missing_state_events or missing_auth_events:
|
|
await self._get_events_and_persist(
|
|
destination=destination,
|
|
room_id=room_id,
|
|
events=missing_state_events | missing_auth_events,
|
|
)
|
|
|
|
if missing_state_events:
|
|
new_events = await self.store.get_events(
|
|
missing_state_events, allow_rejected=True
|
|
)
|
|
event_map.update(new_events)
|
|
|
|
missing_state_events.difference_update(new_events)
|
|
|
|
if missing_state_events:
|
|
logger.warning(
|
|
"Failed to fetch missing state events for %s %s",
|
|
event_id,
|
|
missing_state_events,
|
|
)
|
|
|
|
if missing_auth_events:
|
|
auth_events_in_store = await self.store.have_seen_events(
|
|
room_id, missing_auth_events
|
|
)
|
|
missing_auth_events.difference_update(auth_events_in_store)
|
|
|
|
if missing_auth_events:
|
|
logger.warning(
|
|
"Failed to fetch missing auth events for %s %s",
|
|
event_id,
|
|
missing_auth_events,
|
|
)
|
|
|
|
remote_state = list(event_map.values())
|
|
|
|
# check for events which were in the wrong room.
|
|
#
|
|
# this can happen if a remote server claims that the state or
|
|
# auth_events at an event in room A are actually events in room B
|
|
|
|
bad_events = [
|
|
(event.event_id, event.room_id)
|
|
for event in remote_state
|
|
if event.room_id != room_id
|
|
]
|
|
|
|
for bad_event_id, bad_room_id in bad_events:
|
|
# This is a bogus situation, but since we may only discover it a long time
|
|
# after it happened, we try our best to carry on, by just omitting the
|
|
# bad events from the returned auth/state set.
|
|
logger.warning(
|
|
"Remote server %s claims event %s in room %s is an auth/state "
|
|
"event in room %s",
|
|
destination,
|
|
bad_event_id,
|
|
bad_room_id,
|
|
room_id,
|
|
)
|
|
|
|
if bad_events:
|
|
remote_state = [e for e in remote_state if e.room_id == room_id]
|
|
|
|
return remote_state
|
|
|
|
async def _get_state_after_missing_prev_event(
|
|
self,
|
|
destination: str,
|
|
room_id: str,
|
|
event_id: str,
|
|
) -> List[EventBase]:
|
|
"""Requests all of the room state at a given event from a remote homeserver.
|
|
|
|
Args:
|
|
destination: The remote homeserver to query for the state.
|
|
room_id: The id of the room we're interested in.
|
|
event_id: The id of the event we want the state at.
|
|
|
|
Returns:
|
|
A list of events in the state, including the event itself
|
|
"""
|
|
# TODO: This function is basically the same as _get_state_for_room. Can
|
|
# we make backfill() use it, rather than having two code paths? I think the
|
|
# only difference is that backfill() persists the prev events separately.
|
|
|
|
(
|
|
state_event_ids,
|
|
auth_event_ids,
|
|
) = await self.federation_client.get_room_state_ids(
|
|
destination, room_id, event_id=event_id
|
|
)
|
|
|
|
logger.debug(
|
|
"state_ids returned %i state events, %i auth events",
|
|
len(state_event_ids),
|
|
len(auth_event_ids),
|
|
)
|
|
|
|
# start by just trying to fetch the events from the store
|
|
desired_events = set(state_event_ids)
|
|
desired_events.add(event_id)
|
|
logger.debug("Fetching %i events from cache/store", len(desired_events))
|
|
fetched_events = await self.store.get_events(
|
|
desired_events, allow_rejected=True
|
|
)
|
|
|
|
missing_desired_events = desired_events - fetched_events.keys()
|
|
logger.debug(
|
|
"We are missing %i events (got %i)",
|
|
len(missing_desired_events),
|
|
len(fetched_events),
|
|
)
|
|
|
|
# We probably won't need most of the auth events, so let's just check which
|
|
# we have for now, rather than thrashing the event cache with them all
|
|
# unnecessarily.
|
|
|
|
# TODO: we probably won't actually need all of the auth events, since we
|
|
# already have a bunch of the state events. It would be nice if the
|
|
# federation api gave us a way of finding out which we actually need.
|
|
|
|
missing_auth_events = set(auth_event_ids) - fetched_events.keys()
|
|
missing_auth_events.difference_update(
|
|
await self.store.have_seen_events(room_id, missing_auth_events)
|
|
)
|
|
logger.debug("We are also missing %i auth events", len(missing_auth_events))
|
|
|
|
missing_events = missing_desired_events | missing_auth_events
|
|
logger.debug("Fetching %i events from remote", len(missing_events))
|
|
await self._get_events_and_persist(
|
|
destination=destination, room_id=room_id, events=missing_events
|
|
)
|
|
|
|
# we need to make sure we re-load from the database to get the rejected
|
|
# state correct.
|
|
fetched_events.update(
|
|
await self.store.get_events(missing_desired_events, allow_rejected=True)
|
|
)
|
|
|
|
# check for events which were in the wrong room.
|
|
#
|
|
# this can happen if a remote server claims that the state or
|
|
# auth_events at an event in room A are actually events in room B
|
|
|
|
bad_events = [
|
|
(event_id, event.room_id)
|
|
for event_id, event in fetched_events.items()
|
|
if event.room_id != room_id
|
|
]
|
|
|
|
for bad_event_id, bad_room_id in bad_events:
|
|
# This is a bogus situation, but since we may only discover it a long time
|
|
# after it happened, we try our best to carry on, by just omitting the
|
|
# bad events from the returned state set.
|
|
logger.warning(
|
|
"Remote server %s claims event %s in room %s is an auth/state "
|
|
"event in room %s",
|
|
destination,
|
|
bad_event_id,
|
|
bad_room_id,
|
|
room_id,
|
|
)
|
|
|
|
del fetched_events[bad_event_id]
|
|
|
|
# if we couldn't get the prev event in question, that's a problem.
|
|
remote_event = fetched_events.get(event_id)
|
|
if not remote_event:
|
|
raise Exception("Unable to get missing prev_event %s" % (event_id,))
|
|
|
|
# missing state at that event is a warning, not a blocker
|
|
# XXX: this doesn't sound right? it means that we'll end up with incomplete
|
|
# state.
|
|
failed_to_fetch = desired_events - fetched_events.keys()
|
|
if failed_to_fetch:
|
|
logger.warning(
|
|
"Failed to fetch missing state events for %s %s",
|
|
event_id,
|
|
failed_to_fetch,
|
|
)
|
|
|
|
remote_state = [
|
|
fetched_events[e_id] for e_id in state_event_ids if e_id in fetched_events
|
|
]
|
|
|
|
if remote_event.is_state() and remote_event.rejected_reason is None:
|
|
remote_state.append(remote_event)
|
|
|
|
return remote_state
|
|
|
|
async def _process_received_pdu(
|
|
self,
|
|
origin: str,
|
|
event: EventBase,
|
|
state: Optional[Iterable[EventBase]],
|
|
) -> None:
|
|
"""Called when we have a new pdu. We need to do auth checks and put it
|
|
through the StateHandler.
|
|
|
|
Args:
|
|
origin: server sending the event
|
|
|
|
event: event to be persisted
|
|
|
|
state: Normally None, but if we are handling a gap in the graph
|
|
(ie, we are missing one or more prev_events), the resolved state at the
|
|
event
|
|
"""
|
|
logger.debug("Processing event: %s", event)
|
|
|
|
try:
|
|
context = await self.state_handler.compute_event_context(
|
|
event, old_state=state
|
|
)
|
|
await self._auth_and_persist_event(origin, event, context, state=state)
|
|
except AuthError as e:
|
|
raise FederationError("ERROR", e.code, e.msg, affected=event.event_id)
|
|
|
|
# For encrypted messages we check that we know about the sending device,
|
|
# if we don't then we mark the device cache for that user as stale.
|
|
if event.type == EventTypes.Encrypted:
|
|
device_id = event.content.get("device_id")
|
|
sender_key = event.content.get("sender_key")
|
|
|
|
cached_devices = await self.store.get_cached_devices_for_user(event.sender)
|
|
|
|
resync = False # Whether we should resync device lists.
|
|
|
|
device = None
|
|
if device_id is not None:
|
|
device = cached_devices.get(device_id)
|
|
if device is None:
|
|
logger.info(
|
|
"Received event from remote device not in our cache: %s %s",
|
|
event.sender,
|
|
device_id,
|
|
)
|
|
resync = True
|
|
|
|
# We also check if the `sender_key` matches what we expect.
|
|
if sender_key is not None:
|
|
# Figure out what sender key we're expecting. If we know the
|
|
# device and recognize the algorithm then we can work out the
|
|
# exact key to expect. Otherwise check it matches any key we
|
|
# have for that device.
|
|
|
|
current_keys: Container[str] = []
|
|
|
|
if device:
|
|
keys = device.get("keys", {}).get("keys", {})
|
|
|
|
if (
|
|
event.content.get("algorithm")
|
|
== RoomEncryptionAlgorithms.MEGOLM_V1_AES_SHA2
|
|
):
|
|
# For this algorithm we expect a curve25519 key.
|
|
key_name = "curve25519:%s" % (device_id,)
|
|
current_keys = [keys.get(key_name)]
|
|
else:
|
|
# We don't know understand the algorithm, so we just
|
|
# check it matches a key for the device.
|
|
current_keys = keys.values()
|
|
elif device_id:
|
|
# We don't have any keys for the device ID.
|
|
pass
|
|
else:
|
|
# The event didn't include a device ID, so we just look for
|
|
# keys across all devices.
|
|
current_keys = [
|
|
key
|
|
for device in cached_devices.values()
|
|
for key in device.get("keys", {}).get("keys", {}).values()
|
|
]
|
|
|
|
# We now check that the sender key matches (one of) the expected
|
|
# keys.
|
|
if sender_key not in current_keys:
|
|
logger.info(
|
|
"Received event from remote device with unexpected sender key: %s %s: %s",
|
|
event.sender,
|
|
device_id or "<no device_id>",
|
|
sender_key,
|
|
)
|
|
resync = True
|
|
|
|
if resync:
|
|
run_as_background_process(
|
|
"resync_device_due_to_pdu", self._resync_device, event.sender
|
|
)
|
|
|
|
await self._handle_marker_event(origin, event)
|
|
|
|
async def _handle_marker_event(self, origin: str, marker_event: EventBase):
|
|
"""Handles backfilling the insertion event when we receive a marker
|
|
event that points to one.
|
|
|
|
Args:
|
|
origin: Origin of the event. Will be called to get the insertion event
|
|
marker_event: The event to process
|
|
"""
|
|
|
|
if marker_event.type != EventTypes.MSC2716_MARKER:
|
|
# Not a marker event
|
|
return
|
|
|
|
if marker_event.rejected_reason is not None:
|
|
# Rejected event
|
|
return
|
|
|
|
# Skip processing a marker event if the room version doesn't
|
|
# support it.
|
|
room_version = await self.store.get_room_version(marker_event.room_id)
|
|
if not room_version.msc2716_historical:
|
|
return
|
|
|
|
logger.debug("_handle_marker_event: received %s", marker_event)
|
|
|
|
insertion_event_id = marker_event.content.get(
|
|
EventContentFields.MSC2716_MARKER_INSERTION
|
|
)
|
|
|
|
if insertion_event_id is None:
|
|
# Nothing to retrieve then (invalid marker)
|
|
return
|
|
|
|
logger.debug(
|
|
"_handle_marker_event: backfilling insertion event %s", insertion_event_id
|
|
)
|
|
|
|
await self._get_events_and_persist(
|
|
origin,
|
|
marker_event.room_id,
|
|
[insertion_event_id],
|
|
)
|
|
|
|
insertion_event = await self.store.get_event(
|
|
insertion_event_id, allow_none=True
|
|
)
|
|
if insertion_event is None:
|
|
logger.warning(
|
|
"_handle_marker_event: server %s didn't return insertion event %s for marker %s",
|
|
origin,
|
|
insertion_event_id,
|
|
marker_event.event_id,
|
|
)
|
|
return
|
|
|
|
logger.debug(
|
|
"_handle_marker_event: succesfully backfilled insertion event %s from marker event %s",
|
|
insertion_event,
|
|
marker_event,
|
|
)
|
|
|
|
await self.store.insert_insertion_extremity(
|
|
insertion_event_id, marker_event.room_id
|
|
)
|
|
|
|
logger.debug(
|
|
"_handle_marker_event: insertion extremity added for %s from marker event %s",
|
|
insertion_event,
|
|
marker_event,
|
|
)
|
|
|
|
async def _resync_device(self, sender: str) -> None:
|
|
"""We have detected that the device list for the given user may be out
|
|
of sync, so we try and resync them.
|
|
"""
|
|
|
|
try:
|
|
await self.store.mark_remote_user_device_cache_as_stale(sender)
|
|
|
|
# Immediately attempt a resync in the background
|
|
if self.config.worker_app:
|
|
await self._user_device_resync(user_id=sender)
|
|
else:
|
|
await self._device_list_updater.user_device_resync(sender)
|
|
except Exception:
|
|
logger.exception("Failed to resync device for %s", sender)
|
|
|
|
@log_function
|
|
async def backfill(
|
|
self, dest: str, room_id: str, limit: int, extremities: List[str]
|
|
) -> List[EventBase]:
|
|
"""Trigger a backfill request to `dest` for the given `room_id`
|
|
|
|
This will attempt to get more events from the remote. If the other side
|
|
has no new events to offer, this will return an empty list.
|
|
|
|
As the events are received, we check their signatures, and also do some
|
|
sanity-checking on them. If any of the backfilled events are invalid,
|
|
this method throws a SynapseError.
|
|
|
|
TODO: make this more useful to distinguish failures of the remote
|
|
server from invalid events (there is probably no point in trying to
|
|
re-fetch invalid events from every other HS in the room.)
|
|
"""
|
|
if dest == self.server_name:
|
|
raise SynapseError(400, "Can't backfill from self.")
|
|
|
|
events = await self.federation_client.backfill(
|
|
dest, room_id, limit=limit, extremities=extremities
|
|
)
|
|
|
|
if not events:
|
|
return []
|
|
|
|
# ideally we'd sanity check the events here for excess prev_events etc,
|
|
# but it's hard to reject events at this point without completely
|
|
# breaking backfill in the same way that it is currently broken by
|
|
# events whose signature we cannot verify (#3121).
|
|
#
|
|
# So for now we accept the events anyway. #3124 tracks this.
|
|
#
|
|
# for ev in events:
|
|
# self._sanity_check_event(ev)
|
|
|
|
# Don't bother processing events we already have.
|
|
seen_events = await self.store.have_events_in_timeline(
|
|
{e.event_id for e in events}
|
|
)
|
|
|
|
events = [e for e in events if e.event_id not in seen_events]
|
|
|
|
if not events:
|
|
return []
|
|
|
|
event_map = {e.event_id: e for e in events}
|
|
|
|
event_ids = {e.event_id for e in events}
|
|
|
|
# build a list of events whose prev_events weren't in the batch.
|
|
# (XXX: this will include events whose prev_events we already have; that doesn't
|
|
# sound right?)
|
|
edges = [ev.event_id for ev in events if set(ev.prev_event_ids()) - event_ids]
|
|
|
|
logger.info("backfill: Got %d events with %d edges", len(events), len(edges))
|
|
|
|
# For each edge get the current state.
|
|
|
|
state_events = {}
|
|
events_to_state = {}
|
|
for e_id in edges:
|
|
state = await self._get_state_for_room(
|
|
destination=dest,
|
|
room_id=room_id,
|
|
event_id=e_id,
|
|
)
|
|
state_events.update({s.event_id: s for s in state})
|
|
events_to_state[e_id] = state
|
|
|
|
required_auth = {
|
|
a_id
|
|
for event in events + list(state_events.values())
|
|
for a_id in event.auth_event_ids()
|
|
}
|
|
auth_events = await self.store.get_events(required_auth, allow_rejected=True)
|
|
auth_events.update(
|
|
{e_id: event_map[e_id] for e_id in required_auth if e_id in event_map}
|
|
)
|
|
|
|
ev_infos = []
|
|
|
|
# Step 1: persist the events in the chunk we fetched state for (i.e.
|
|
# the backwards extremities), with custom auth events and state
|
|
for e_id in events_to_state:
|
|
# For paranoia we ensure that these events are marked as
|
|
# non-outliers
|
|
ev = event_map[e_id]
|
|
assert not ev.internal_metadata.is_outlier()
|
|
|
|
ev_infos.append(
|
|
_NewEventInfo(
|
|
event=ev,
|
|
state=events_to_state[e_id],
|
|
auth_events={
|
|
(
|
|
auth_events[a_id].type,
|
|
auth_events[a_id].state_key,
|
|
): auth_events[a_id]
|
|
for a_id in ev.auth_event_ids()
|
|
if a_id in auth_events
|
|
},
|
|
)
|
|
)
|
|
|
|
if ev_infos:
|
|
await self._auth_and_persist_events(
|
|
dest, room_id, ev_infos, backfilled=True
|
|
)
|
|
|
|
# Step 2: Persist the rest of the events in the chunk one by one
|
|
events.sort(key=lambda e: e.depth)
|
|
|
|
for event in events:
|
|
if event in events_to_state:
|
|
continue
|
|
|
|
# For paranoia we ensure that these events are marked as
|
|
# non-outliers
|
|
assert not event.internal_metadata.is_outlier()
|
|
|
|
context = await self.state_handler.compute_event_context(event)
|
|
|
|
# We store these one at a time since each event depends on the
|
|
# previous to work out the state.
|
|
# TODO: We can probably do something more clever here.
|
|
await self._auth_and_persist_event(dest, event, context, backfilled=True)
|
|
|
|
return events
|
|
|
|
async def maybe_backfill(
|
|
self, room_id: str, current_depth: int, limit: int
|
|
) -> bool:
|
|
"""Checks the database to see if we should backfill before paginating,
|
|
and if so do.
|
|
|
|
Args:
|
|
room_id
|
|
current_depth: The depth from which we're paginating from. This is
|
|
used to decide if we should backfill and what extremities to
|
|
use.
|
|
limit: The number of events that the pagination request will
|
|
return. This is used as part of the heuristic to decide if we
|
|
should back paginate.
|
|
"""
|
|
with (await self._room_backfill.queue(room_id)):
|
|
return await self._maybe_backfill_inner(room_id, current_depth, limit)
|
|
|
|
async def _maybe_backfill_inner(
|
|
self, room_id: str, current_depth: int, limit: int
|
|
) -> bool:
|
|
oldest_events_with_depth = (
|
|
await self.store.get_oldest_event_ids_with_depth_in_room(room_id)
|
|
)
|
|
insertion_events_to_be_backfilled = (
|
|
await self.store.get_insertion_event_backwards_extremities_in_room(room_id)
|
|
)
|
|
logger.debug(
|
|
"_maybe_backfill_inner: extremities oldest_events_with_depth=%s insertion_events_to_be_backfilled=%s",
|
|
oldest_events_with_depth,
|
|
insertion_events_to_be_backfilled,
|
|
)
|
|
|
|
if not oldest_events_with_depth and not insertion_events_to_be_backfilled:
|
|
logger.debug("Not backfilling as no extremeties found.")
|
|
return False
|
|
|
|
# We only want to paginate if we can actually see the events we'll get,
|
|
# as otherwise we'll just spend a lot of resources to get redacted
|
|
# events.
|
|
#
|
|
# We do this by filtering all the backwards extremities and seeing if
|
|
# any remain. Given we don't have the extremity events themselves, we
|
|
# need to actually check the events that reference them.
|
|
#
|
|
# *Note*: the spec wants us to keep backfilling until we reach the start
|
|
# of the room in case we are allowed to see some of the history. However
|
|
# in practice that causes more issues than its worth, as a) its
|
|
# relatively rare for there to be any visible history and b) even when
|
|
# there is its often sufficiently long ago that clients would stop
|
|
# attempting to paginate before backfill reached the visible history.
|
|
#
|
|
# TODO: If we do do a backfill then we should filter the backwards
|
|
# extremities to only include those that point to visible portions of
|
|
# history.
|
|
#
|
|
# TODO: Correctly handle the case where we are allowed to see the
|
|
# forward event but not the backward extremity, e.g. in the case of
|
|
# initial join of the server where we are allowed to see the join
|
|
# event but not anything before it. This would require looking at the
|
|
# state *before* the event, ignoring the special casing certain event
|
|
# types have.
|
|
|
|
forward_event_ids = await self.store.get_successor_events(
|
|
list(oldest_events_with_depth)
|
|
)
|
|
|
|
extremities_events = await self.store.get_events(
|
|
forward_event_ids,
|
|
redact_behaviour=EventRedactBehaviour.AS_IS,
|
|
get_prev_content=False,
|
|
)
|
|
|
|
# We set `check_history_visibility_only` as we might otherwise get false
|
|
# positives from users having been erased.
|
|
filtered_extremities = await filter_events_for_server(
|
|
self.storage,
|
|
self.server_name,
|
|
list(extremities_events.values()),
|
|
redact=False,
|
|
check_history_visibility_only=True,
|
|
)
|
|
logger.debug(
|
|
"_maybe_backfill_inner: filtered_extremities %s", filtered_extremities
|
|
)
|
|
|
|
if not filtered_extremities and not insertion_events_to_be_backfilled:
|
|
return False
|
|
|
|
extremities = {
|
|
**oldest_events_with_depth,
|
|
# TODO: insertion_events_to_be_backfilled is currently skipping the filtered_extremities checks
|
|
**insertion_events_to_be_backfilled,
|
|
}
|
|
|
|
# Check if we reached a point where we should start backfilling.
|
|
sorted_extremeties_tuple = sorted(extremities.items(), key=lambda e: -int(e[1]))
|
|
max_depth = sorted_extremeties_tuple[0][1]
|
|
|
|
# If we're approaching an extremity we trigger a backfill, otherwise we
|
|
# no-op.
|
|
#
|
|
# We chose twice the limit here as then clients paginating backwards
|
|
# will send pagination requests that trigger backfill at least twice
|
|
# using the most recent extremity before it gets removed (see below). We
|
|
# chose more than one times the limit in case of failure, but choosing a
|
|
# much larger factor will result in triggering a backfill request much
|
|
# earlier than necessary.
|
|
if current_depth - 2 * limit > max_depth:
|
|
logger.debug(
|
|
"Not backfilling as we don't need to. %d < %d - 2 * %d",
|
|
max_depth,
|
|
current_depth,
|
|
limit,
|
|
)
|
|
return False
|
|
|
|
logger.debug(
|
|
"room_id: %s, backfill: current_depth: %s, max_depth: %s, extrems: %s",
|
|
room_id,
|
|
current_depth,
|
|
max_depth,
|
|
sorted_extremeties_tuple,
|
|
)
|
|
|
|
# We ignore extremities that have a greater depth than our current depth
|
|
# as:
|
|
# 1. we don't really care about getting events that have happened
|
|
# before our current position; and
|
|
# 2. we have likely previously tried and failed to backfill from that
|
|
# extremity, so to avoid getting "stuck" requesting the same
|
|
# backfill repeatedly we drop those extremities.
|
|
filtered_sorted_extremeties_tuple = [
|
|
t for t in sorted_extremeties_tuple if int(t[1]) <= current_depth
|
|
]
|
|
|
|
# However, we need to check that the filtered extremities are non-empty.
|
|
# If they are empty then either we can a) bail or b) still attempt to
|
|
# backill. We opt to try backfilling anyway just in case we do get
|
|
# relevant events.
|
|
if filtered_sorted_extremeties_tuple:
|
|
sorted_extremeties_tuple = filtered_sorted_extremeties_tuple
|
|
|
|
# We don't want to specify too many extremities as it causes the backfill
|
|
# request URI to be too long.
|
|
extremities = dict(sorted_extremeties_tuple[:5])
|
|
|
|
# Now we need to decide which hosts to hit first.
|
|
|
|
# First we try hosts that are already in the room
|
|
# TODO: HEURISTIC ALERT.
|
|
|
|
curr_state = await self.state_handler.get_current_state(room_id)
|
|
|
|
def get_domains_from_state(state: StateMap[EventBase]) -> List[Tuple[str, int]]:
|
|
"""Get joined domains from state
|
|
|
|
Args:
|
|
state: State map from type/state key to event.
|
|
|
|
Returns:
|
|
Returns a list of servers with the lowest depth of their joins.
|
|
Sorted by lowest depth first.
|
|
"""
|
|
joined_users = [
|
|
(state_key, int(event.depth))
|
|
for (e_type, state_key), event in state.items()
|
|
if e_type == EventTypes.Member and event.membership == Membership.JOIN
|
|
]
|
|
|
|
joined_domains: Dict[str, int] = {}
|
|
for u, d in joined_users:
|
|
try:
|
|
dom = get_domain_from_id(u)
|
|
old_d = joined_domains.get(dom)
|
|
if old_d:
|
|
joined_domains[dom] = min(d, old_d)
|
|
else:
|
|
joined_domains[dom] = d
|
|
except Exception:
|
|
pass
|
|
|
|
return sorted(joined_domains.items(), key=lambda d: d[1])
|
|
|
|
curr_domains = get_domains_from_state(curr_state)
|
|
|
|
likely_domains = [
|
|
domain for domain, depth in curr_domains if domain != self.server_name
|
|
]
|
|
|
|
async def try_backfill(domains: List[str]) -> bool:
|
|
# TODO: Should we try multiple of these at a time?
|
|
for dom in domains:
|
|
try:
|
|
await self.backfill(
|
|
dom, room_id, limit=100, extremities=extremities
|
|
)
|
|
# If this succeeded then we probably already have the
|
|
# appropriate stuff.
|
|
# TODO: We can probably do something more intelligent here.
|
|
return True
|
|
except SynapseError as e:
|
|
logger.info("Failed to backfill from %s because %s", dom, e)
|
|
continue
|
|
except HttpResponseException as e:
|
|
if 400 <= e.code < 500:
|
|
raise e.to_synapse_error()
|
|
|
|
logger.info("Failed to backfill from %s because %s", dom, e)
|
|
continue
|
|
except CodeMessageException as e:
|
|
if 400 <= e.code < 500:
|
|
raise
|
|
|
|
logger.info("Failed to backfill from %s because %s", dom, e)
|
|
continue
|
|
except NotRetryingDestination as e:
|
|
logger.info(str(e))
|
|
continue
|
|
except RequestSendFailed as e:
|
|
logger.info("Failed to get backfill from %s because %s", dom, e)
|
|
continue
|
|
except FederationDeniedError as e:
|
|
logger.info(e)
|
|
continue
|
|
except Exception as e:
|
|
logger.exception("Failed to backfill from %s because %s", dom, e)
|
|
continue
|
|
|
|
return False
|
|
|
|
success = await try_backfill(likely_domains)
|
|
if success:
|
|
return True
|
|
|
|
# Huh, well *those* domains didn't work out. Lets try some domains
|
|
# from the time.
|
|
|
|
tried_domains = set(likely_domains)
|
|
tried_domains.add(self.server_name)
|
|
|
|
event_ids = list(extremities.keys())
|
|
|
|
logger.debug("calling resolve_state_groups in _maybe_backfill")
|
|
resolve = preserve_fn(self.state_handler.resolve_state_groups_for_events)
|
|
states = await make_deferred_yieldable(
|
|
defer.gatherResults(
|
|
[resolve(room_id, [e]) for e in event_ids], consumeErrors=True
|
|
)
|
|
)
|
|
|
|
# dict[str, dict[tuple, str]], a map from event_id to state map of
|
|
# event_ids.
|
|
states = dict(zip(event_ids, [s.state for s in states]))
|
|
|
|
state_map = await self.store.get_events(
|
|
[e_id for ids in states.values() for e_id in ids.values()],
|
|
get_prev_content=False,
|
|
)
|
|
states = {
|
|
key: {
|
|
k: state_map[e_id]
|
|
for k, e_id in state_dict.items()
|
|
if e_id in state_map
|
|
}
|
|
for key, state_dict in states.items()
|
|
}
|
|
|
|
for e_id, _ in sorted_extremeties_tuple:
|
|
likely_extremeties_domains = get_domains_from_state(states[e_id])
|
|
|
|
success = await try_backfill(
|
|
[
|
|
dom
|
|
for dom, _ in likely_extremeties_domains
|
|
if dom not in tried_domains
|
|
]
|
|
)
|
|
if success:
|
|
return True
|
|
|
|
tried_domains.update(dom for dom, _ in likely_extremeties_domains)
|
|
|
|
return False
|
|
|
|
async def _get_events_and_persist(
|
|
self, destination: str, room_id: str, events: Iterable[str]
|
|
) -> None:
|
|
"""Fetch the given events from a server, and persist them as outliers.
|
|
|
|
This function *does not* recursively get missing auth events of the
|
|
newly fetched events. Callers must include in the `events` argument
|
|
any missing events from the auth chain.
|
|
|
|
Logs a warning if we can't find the given event.
|
|
"""
|
|
|
|
room_version = await self.store.get_room_version(room_id)
|
|
|
|
event_map: Dict[str, EventBase] = {}
|
|
|
|
async def get_event(event_id: str):
|
|
with nested_logging_context(event_id):
|
|
try:
|
|
event = await self.federation_client.get_pdu(
|
|
[destination],
|
|
event_id,
|
|
room_version,
|
|
outlier=True,
|
|
)
|
|
if event is None:
|
|
logger.warning(
|
|
"Server %s didn't return event %s",
|
|
destination,
|
|
event_id,
|
|
)
|
|
return
|
|
|
|
event_map[event.event_id] = event
|
|
|
|
except Exception as e:
|
|
logger.warning(
|
|
"Error fetching missing state/auth event %s: %s %s",
|
|
event_id,
|
|
type(e),
|
|
e,
|
|
)
|
|
|
|
await concurrently_execute(get_event, events, 5)
|
|
|
|
# Make a map of auth events for each event. We do this after fetching
|
|
# all the events as some of the events' auth events will be in the list
|
|
# of requested events.
|
|
|
|
auth_events = [
|
|
aid
|
|
for event in event_map.values()
|
|
for aid in event.auth_event_ids()
|
|
if aid not in event_map
|
|
]
|
|
persisted_events = await self.store.get_events(
|
|
auth_events,
|
|
allow_rejected=True,
|
|
)
|
|
|
|
event_infos = []
|
|
for event in event_map.values():
|
|
auth = {}
|
|
for auth_event_id in event.auth_event_ids():
|
|
ae = persisted_events.get(auth_event_id) or event_map.get(auth_event_id)
|
|
if ae:
|
|
auth[(ae.type, ae.state_key)] = ae
|
|
else:
|
|
logger.info("Missing auth event %s", auth_event_id)
|
|
|
|
event_infos.append(_NewEventInfo(event, None, auth))
|
|
|
|
if event_infos:
|
|
await self._auth_and_persist_events(
|
|
destination,
|
|
room_id,
|
|
event_infos,
|
|
)
|
|
|
|
def _sanity_check_event(self, ev: EventBase) -> None:
|
|
"""
|
|
Do some early sanity checks of a received event
|
|
|
|
In particular, checks it doesn't have an excessive number of
|
|
prev_events or auth_events, which could cause a huge state resolution
|
|
or cascade of event fetches.
|
|
|
|
Args:
|
|
ev: event to be checked
|
|
|
|
Raises:
|
|
SynapseError if the event does not pass muster
|
|
"""
|
|
if len(ev.prev_event_ids()) > 20:
|
|
logger.warning(
|
|
"Rejecting event %s which has %i prev_events",
|
|
ev.event_id,
|
|
len(ev.prev_event_ids()),
|
|
)
|
|
raise SynapseError(HTTPStatus.BAD_REQUEST, "Too many prev_events")
|
|
|
|
if len(ev.auth_event_ids()) > 10:
|
|
logger.warning(
|
|
"Rejecting event %s which has %i auth_events",
|
|
ev.event_id,
|
|
len(ev.auth_event_ids()),
|
|
)
|
|
raise SynapseError(HTTPStatus.BAD_REQUEST, "Too many auth_events")
|
|
|
|
async def send_invite(self, target_host: str, event: EventBase) -> EventBase:
|
|
"""Sends the invite to the remote server for signing.
|
|
|
|
Invites must be signed by the invitee's server before distribution.
|
|
"""
|
|
try:
|
|
pdu = await self.federation_client.send_invite(
|
|
destination=target_host,
|
|
room_id=event.room_id,
|
|
event_id=event.event_id,
|
|
pdu=event,
|
|
)
|
|
except RequestSendFailed:
|
|
raise SynapseError(502, f"Can't connect to server {target_host}")
|
|
|
|
return pdu
|
|
|
|
async def on_event_auth(self, event_id: str) -> List[EventBase]:
|
|
event = await self.store.get_event(event_id)
|
|
auth = await self.store.get_auth_chain(
|
|
event.room_id, list(event.auth_event_ids()), include_given=True
|
|
)
|
|
return list(auth)
|
|
|
|
async def do_invite_join(
|
|
self, target_hosts: Iterable[str], room_id: str, joinee: str, content: JsonDict
|
|
) -> Tuple[str, int]:
|
|
"""Attempts to join the `joinee` to the room `room_id` via the
|
|
servers contained in `target_hosts`.
|
|
|
|
This first triggers a /make_join/ request that returns a partial
|
|
event that we can fill out and sign. This is then sent to the
|
|
remote server via /send_join/ which responds with the state at that
|
|
event and the auth_chains.
|
|
|
|
We suspend processing of any received events from this room until we
|
|
have finished processing the join.
|
|
|
|
Args:
|
|
target_hosts: List of servers to attempt to join the room with.
|
|
|
|
room_id: The ID of the room to join.
|
|
|
|
joinee: The User ID of the joining user.
|
|
|
|
content: The event content to use for the join event.
|
|
"""
|
|
# TODO: We should be able to call this on workers, but the upgrading of
|
|
# room stuff after join currently doesn't work on workers.
|
|
assert self.config.worker.worker_app is None
|
|
|
|
logger.debug("Joining %s to %s", joinee, room_id)
|
|
|
|
origin, event, room_version_obj = await self._make_and_verify_event(
|
|
target_hosts,
|
|
room_id,
|
|
joinee,
|
|
"join",
|
|
content,
|
|
params={"ver": KNOWN_ROOM_VERSIONS},
|
|
)
|
|
|
|
# This shouldn't happen, because the RoomMemberHandler has a
|
|
# linearizer lock which only allows one operation per user per room
|
|
# at a time - so this is just paranoia.
|
|
assert room_id not in self.room_queues
|
|
|
|
self.room_queues[room_id] = []
|
|
|
|
await self._clean_room_for_join(room_id)
|
|
|
|
try:
|
|
# Try the host we successfully got a response to /make_join/
|
|
# request first.
|
|
host_list = list(target_hosts)
|
|
try:
|
|
host_list.remove(origin)
|
|
host_list.insert(0, origin)
|
|
except ValueError:
|
|
pass
|
|
|
|
ret = await self.federation_client.send_join(
|
|
host_list, event, room_version_obj
|
|
)
|
|
|
|
event = ret.event
|
|
origin = ret.origin
|
|
state = ret.state
|
|
auth_chain = ret.auth_chain
|
|
auth_chain.sort(key=lambda e: e.depth)
|
|
|
|
logger.debug("do_invite_join auth_chain: %s", auth_chain)
|
|
logger.debug("do_invite_join state: %s", state)
|
|
|
|
logger.debug("do_invite_join event: %s", event)
|
|
|
|
# if this is the first time we've joined this room, it's time to add
|
|
# a row to `rooms` with the correct room version. If there's already a
|
|
# row there, we should override it, since it may have been populated
|
|
# based on an invite request which lied about the room version.
|
|
#
|
|
# federation_client.send_join has already checked that the room
|
|
# version in the received create event is the same as room_version_obj,
|
|
# so we can rely on it now.
|
|
#
|
|
await self.store.upsert_room_on_join(
|
|
room_id=room_id,
|
|
room_version=room_version_obj,
|
|
)
|
|
|
|
max_stream_id = await self._persist_auth_tree(
|
|
origin, room_id, auth_chain, state, event, room_version_obj
|
|
)
|
|
|
|
# We wait here until this instance has seen the events come down
|
|
# replication (if we're using replication) as the below uses caches.
|
|
await self._replication.wait_for_stream_position(
|
|
self.config.worker.events_shard_config.get_instance(room_id),
|
|
"events",
|
|
max_stream_id,
|
|
)
|
|
|
|
# Check whether this room is the result of an upgrade of a room we already know
|
|
# about. If so, migrate over user information
|
|
predecessor = await self.store.get_room_predecessor(room_id)
|
|
if not predecessor or not isinstance(predecessor.get("room_id"), str):
|
|
return event.event_id, max_stream_id
|
|
old_room_id = predecessor["room_id"]
|
|
logger.debug(
|
|
"Found predecessor for %s during remote join: %s", room_id, old_room_id
|
|
)
|
|
|
|
# We retrieve the room member handler here as to not cause a cyclic dependency
|
|
member_handler = self.hs.get_room_member_handler()
|
|
await member_handler.transfer_room_state_on_room_upgrade(
|
|
old_room_id, room_id
|
|
)
|
|
|
|
logger.debug("Finished joining %s to %s", joinee, room_id)
|
|
return event.event_id, max_stream_id
|
|
finally:
|
|
room_queue = self.room_queues[room_id]
|
|
del self.room_queues[room_id]
|
|
|
|
# we don't need to wait for the queued events to be processed -
|
|
# it's just a best-effort thing at this point. We do want to do
|
|
# them roughly in order, though, otherwise we'll end up making
|
|
# lots of requests for missing prev_events which we do actually
|
|
# have. Hence we fire off the background task, but don't wait for it.
|
|
|
|
run_in_background(self._handle_queued_pdus, room_queue)
|
|
|
|
@log_function
|
|
async def do_knock(
|
|
self,
|
|
target_hosts: List[str],
|
|
room_id: str,
|
|
knockee: str,
|
|
content: JsonDict,
|
|
) -> Tuple[str, int]:
|
|
"""Sends the knock to the remote server.
|
|
|
|
This first triggers a make_knock request that returns a partial
|
|
event that we can fill out and sign. This is then sent to the
|
|
remote server via send_knock.
|
|
|
|
Knock events must be signed by the knockee's server before distributing.
|
|
|
|
Args:
|
|
target_hosts: A list of hosts that we want to try knocking through.
|
|
room_id: The ID of the room to knock on.
|
|
knockee: The ID of the user who is knocking.
|
|
content: The content of the knock event.
|
|
|
|
Returns:
|
|
A tuple of (event ID, stream ID).
|
|
|
|
Raises:
|
|
SynapseError: If the chosen remote server returns a 3xx/4xx code.
|
|
RuntimeError: If no servers were reachable.
|
|
"""
|
|
logger.debug("Knocking on room %s on behalf of user %s", room_id, knockee)
|
|
|
|
# Inform the remote server of the room versions we support
|
|
supported_room_versions = list(KNOWN_ROOM_VERSIONS.keys())
|
|
|
|
# Ask the remote server to create a valid knock event for us. Once received,
|
|
# we sign the event
|
|
params: Dict[str, Iterable[str]] = {"ver": supported_room_versions}
|
|
origin, event, event_format_version = await self._make_and_verify_event(
|
|
target_hosts, room_id, knockee, Membership.KNOCK, content, params=params
|
|
)
|
|
|
|
# Record the room ID and its version so that we have a record of the room
|
|
await self._maybe_store_room_on_outlier_membership(
|
|
room_id=event.room_id, room_version=event_format_version
|
|
)
|
|
|
|
# Initially try the host that we successfully called /make_knock on
|
|
try:
|
|
target_hosts.remove(origin)
|
|
target_hosts.insert(0, origin)
|
|
except ValueError:
|
|
pass
|
|
|
|
# Send the signed event back to the room, and potentially receive some
|
|
# further information about the room in the form of partial state events
|
|
stripped_room_state = await self.federation_client.send_knock(
|
|
target_hosts, event
|
|
)
|
|
|
|
# Store any stripped room state events in the "unsigned" key of the event.
|
|
# This is a bit of a hack and is cribbing off of invites. Basically we
|
|
# store the room state here and retrieve it again when this event appears
|
|
# in the invitee's sync stream. It is stripped out for all other local users.
|
|
event.unsigned["knock_room_state"] = stripped_room_state["knock_state_events"]
|
|
|
|
context = await self.state_handler.compute_event_context(event)
|
|
stream_id = await self.persist_events_and_notify(
|
|
event.room_id, [(event, context)]
|
|
)
|
|
return event.event_id, stream_id
|
|
|
|
async def _handle_queued_pdus(
|
|
self, room_queue: List[Tuple[EventBase, str]]
|
|
) -> None:
|
|
"""Process PDUs which got queued up while we were busy send_joining.
|
|
|
|
Args:
|
|
room_queue: list of PDUs to be processed and the servers that sent them
|
|
"""
|
|
for p, origin in room_queue:
|
|
try:
|
|
logger.info(
|
|
"Processing queued PDU %s which was received "
|
|
"while we were joining %s",
|
|
p.event_id,
|
|
p.room_id,
|
|
)
|
|
with nested_logging_context(p.event_id):
|
|
await self.on_receive_pdu(origin, p, sent_to_us_directly=True)
|
|
except Exception as e:
|
|
logger.warning(
|
|
"Error handling queued PDU %s from %s: %s", p.event_id, origin, e
|
|
)
|
|
|
|
async def on_make_join_request(
|
|
self, origin: str, room_id: str, user_id: str
|
|
) -> EventBase:
|
|
"""We've received a /make_join/ request, so we create a partial
|
|
join event for the room and return that. We do *not* persist or
|
|
process it until the other server has signed it and sent it back.
|
|
|
|
Args:
|
|
origin: The (verified) server name of the requesting server.
|
|
room_id: Room to create join event in
|
|
user_id: The user to create the join for
|
|
"""
|
|
if get_domain_from_id(user_id) != origin:
|
|
logger.info(
|
|
"Got /make_join request for user %r from different origin %s, ignoring",
|
|
user_id,
|
|
origin,
|
|
)
|
|
raise SynapseError(403, "User not from origin", Codes.FORBIDDEN)
|
|
|
|
# checking the room version will check that we've actually heard of the room
|
|
# (and return a 404 otherwise)
|
|
room_version = await self.store.get_room_version(room_id)
|
|
|
|
# now check that we are *still* in the room
|
|
is_in_room = await self._event_auth_handler.check_host_in_room(
|
|
room_id, self.server_name
|
|
)
|
|
if not is_in_room:
|
|
logger.info(
|
|
"Got /make_join request for room %s we are no longer in",
|
|
room_id,
|
|
)
|
|
raise NotFoundError("Not an active room on this server")
|
|
|
|
event_content = {"membership": Membership.JOIN}
|
|
|
|
# If the current room is using restricted join rules, additional information
|
|
# may need to be included in the event content in order to efficiently
|
|
# validate the event.
|
|
#
|
|
# Note that this requires the /send_join request to come back to the
|
|
# same server.
|
|
if room_version.msc3083_join_rules:
|
|
state_ids = await self.store.get_current_state_ids(room_id)
|
|
if await self._event_auth_handler.has_restricted_join_rules(
|
|
state_ids, room_version
|
|
):
|
|
prev_member_event_id = state_ids.get((EventTypes.Member, user_id), None)
|
|
# If the user is invited or joined to the room already, then
|
|
# no additional info is needed.
|
|
include_auth_user_id = True
|
|
if prev_member_event_id:
|
|
prev_member_event = await self.store.get_event(prev_member_event_id)
|
|
include_auth_user_id = prev_member_event.membership not in (
|
|
Membership.JOIN,
|
|
Membership.INVITE,
|
|
)
|
|
|
|
if include_auth_user_id:
|
|
event_content[
|
|
"join_authorised_via_users_server"
|
|
] = await self._event_auth_handler.get_user_which_could_invite(
|
|
room_id,
|
|
state_ids,
|
|
)
|
|
|
|
builder = self.event_builder_factory.new(
|
|
room_version.identifier,
|
|
{
|
|
"type": EventTypes.Member,
|
|
"content": event_content,
|
|
"room_id": room_id,
|
|
"sender": user_id,
|
|
"state_key": user_id,
|
|
},
|
|
)
|
|
|
|
try:
|
|
event, context = await self.event_creation_handler.create_new_client_event(
|
|
builder=builder
|
|
)
|
|
except SynapseError as e:
|
|
logger.warning("Failed to create join to %s because %s", room_id, e)
|
|
raise
|
|
|
|
# Ensure the user can even join the room.
|
|
await self._check_join_restrictions(context, event)
|
|
|
|
# The remote hasn't signed it yet, obviously. We'll do the full checks
|
|
# when we get the event back in `on_send_join_request`
|
|
await self._event_auth_handler.check_from_context(
|
|
room_version.identifier, event, context, do_sig_check=False
|
|
)
|
|
|
|
return event
|
|
|
|
async def on_invite_request(
|
|
self, origin: str, event: EventBase, room_version: RoomVersion
|
|
) -> EventBase:
|
|
"""We've got an invite event. Process and persist it. Sign it.
|
|
|
|
Respond with the now signed event.
|
|
"""
|
|
if event.state_key is None:
|
|
raise SynapseError(400, "The invite event did not have a state key")
|
|
|
|
is_blocked = await self.store.is_room_blocked(event.room_id)
|
|
if is_blocked:
|
|
raise SynapseError(403, "This room has been blocked on this server")
|
|
|
|
if self.hs.config.block_non_admin_invites:
|
|
raise SynapseError(403, "This server does not accept room invites")
|
|
|
|
if not await self.spam_checker.user_may_invite(
|
|
event.sender, event.state_key, event.room_id
|
|
):
|
|
raise SynapseError(
|
|
403, "This user is not permitted to send invites to this server/user"
|
|
)
|
|
|
|
membership = event.content.get("membership")
|
|
if event.type != EventTypes.Member or membership != Membership.INVITE:
|
|
raise SynapseError(400, "The event was not an m.room.member invite event")
|
|
|
|
sender_domain = get_domain_from_id(event.sender)
|
|
if sender_domain != origin:
|
|
raise SynapseError(
|
|
400, "The invite event was not from the server sending it"
|
|
)
|
|
|
|
if not self.is_mine_id(event.state_key):
|
|
raise SynapseError(400, "The invite event must be for this server")
|
|
|
|
# block any attempts to invite the server notices mxid
|
|
if event.state_key == self._server_notices_mxid:
|
|
raise SynapseError(HTTPStatus.FORBIDDEN, "Cannot invite this user")
|
|
|
|
# We retrieve the room member handler here as to not cause a cyclic dependency
|
|
member_handler = self.hs.get_room_member_handler()
|
|
# We don't rate limit based on room ID, as that should be done by
|
|
# sending server.
|
|
await member_handler.ratelimit_invite(None, None, event.state_key)
|
|
|
|
# keep a record of the room version, if we don't yet know it.
|
|
# (this may get overwritten if we later get a different room version in a
|
|
# join dance).
|
|
await self._maybe_store_room_on_outlier_membership(
|
|
room_id=event.room_id, room_version=room_version
|
|
)
|
|
|
|
event.internal_metadata.outlier = True
|
|
event.internal_metadata.out_of_band_membership = True
|
|
|
|
event.signatures.update(
|
|
compute_event_signature(
|
|
room_version,
|
|
event.get_pdu_json(),
|
|
self.hs.hostname,
|
|
self.hs.signing_key,
|
|
)
|
|
)
|
|
|
|
context = await self.state_handler.compute_event_context(event)
|
|
await self.persist_events_and_notify(event.room_id, [(event, context)])
|
|
|
|
return event
|
|
|
|
async def do_remotely_reject_invite(
|
|
self, target_hosts: Iterable[str], room_id: str, user_id: str, content: JsonDict
|
|
) -> Tuple[EventBase, int]:
|
|
origin, event, room_version = await self._make_and_verify_event(
|
|
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.
|
|
event.internal_metadata.outlier = True
|
|
event.internal_metadata.out_of_band_membership = True
|
|
|
|
# Try the host that we successfully called /make_leave/ on first for
|
|
# the /send_leave/ request.
|
|
host_list = list(target_hosts)
|
|
try:
|
|
host_list.remove(origin)
|
|
host_list.insert(0, origin)
|
|
except ValueError:
|
|
pass
|
|
|
|
await self.federation_client.send_leave(host_list, event)
|
|
|
|
context = await self.state_handler.compute_event_context(event)
|
|
stream_id = await self.persist_events_and_notify(
|
|
event.room_id, [(event, context)]
|
|
)
|
|
|
|
return event, stream_id
|
|
|
|
async def _make_and_verify_event(
|
|
self,
|
|
target_hosts: Iterable[str],
|
|
room_id: str,
|
|
user_id: str,
|
|
membership: str,
|
|
content: JsonDict,
|
|
params: Optional[Dict[str, Union[str, Iterable[str]]]] = None,
|
|
) -> Tuple[str, EventBase, RoomVersion]:
|
|
(
|
|
origin,
|
|
event,
|
|
room_version,
|
|
) = await self.federation_client.make_membership_event(
|
|
target_hosts, room_id, user_id, membership, content, params=params
|
|
)
|
|
|
|
logger.debug("Got response to make_%s: %s", membership, event)
|
|
|
|
# We should assert some things.
|
|
# FIXME: Do this in a nicer way
|
|
assert event.type == EventTypes.Member
|
|
assert event.user_id == user_id
|
|
assert event.state_key == user_id
|
|
assert event.room_id == room_id
|
|
return origin, event, room_version
|
|
|
|
async def on_make_leave_request(
|
|
self, origin: str, room_id: str, user_id: str
|
|
) -> EventBase:
|
|
"""We've received a /make_leave/ request, so we create a partial
|
|
leave event for the room and return that. We do *not* persist or
|
|
process it until the other server has signed it and sent it back.
|
|
|
|
Args:
|
|
origin: The (verified) server name of the requesting server.
|
|
room_id: Room to create leave event in
|
|
user_id: The user to create the leave for
|
|
"""
|
|
if get_domain_from_id(user_id) != origin:
|
|
logger.info(
|
|
"Got /make_leave request for user %r from different origin %s, ignoring",
|
|
user_id,
|
|
origin,
|
|
)
|
|
raise SynapseError(403, "User not from origin", Codes.FORBIDDEN)
|
|
|
|
room_version = await self.store.get_room_version_id(room_id)
|
|
builder = self.event_builder_factory.new(
|
|
room_version,
|
|
{
|
|
"type": EventTypes.Member,
|
|
"content": {"membership": Membership.LEAVE},
|
|
"room_id": room_id,
|
|
"sender": user_id,
|
|
"state_key": user_id,
|
|
},
|
|
)
|
|
|
|
event, context = await self.event_creation_handler.create_new_client_event(
|
|
builder=builder
|
|
)
|
|
|
|
try:
|
|
# The remote hasn't signed it yet, obviously. We'll do the full checks
|
|
# when we get the event back in `on_send_leave_request`
|
|
await self._event_auth_handler.check_from_context(
|
|
room_version, event, context, do_sig_check=False
|
|
)
|
|
except AuthError as e:
|
|
logger.warning("Failed to create new leave %r because %s", event, e)
|
|
raise e
|
|
|
|
return event
|
|
|
|
@log_function
|
|
async def on_make_knock_request(
|
|
self, origin: str, room_id: str, user_id: str
|
|
) -> EventBase:
|
|
"""We've received a make_knock request, so we create a partial
|
|
knock event for the room and return that. We do *not* persist or
|
|
process it until the other server has signed it and sent it back.
|
|
|
|
Args:
|
|
origin: The (verified) server name of the requesting server.
|
|
room_id: The room to create the knock event in.
|
|
user_id: The user to create the knock for.
|
|
|
|
Returns:
|
|
The partial knock event.
|
|
"""
|
|
if get_domain_from_id(user_id) != origin:
|
|
logger.info(
|
|
"Get /make_knock request for user %r from different origin %s, ignoring",
|
|
user_id,
|
|
origin,
|
|
)
|
|
raise SynapseError(403, "User not from origin", Codes.FORBIDDEN)
|
|
|
|
room_version = await self.store.get_room_version_id(room_id)
|
|
|
|
builder = self.event_builder_factory.new(
|
|
room_version,
|
|
{
|
|
"type": EventTypes.Member,
|
|
"content": {"membership": Membership.KNOCK},
|
|
"room_id": room_id,
|
|
"sender": user_id,
|
|
"state_key": user_id,
|
|
},
|
|
)
|
|
|
|
event, context = await self.event_creation_handler.create_new_client_event(
|
|
builder=builder
|
|
)
|
|
|
|
event_allowed, _ = await self.third_party_event_rules.check_event_allowed(
|
|
event, context
|
|
)
|
|
if not event_allowed:
|
|
logger.warning("Creation of knock %s forbidden by third-party rules", event)
|
|
raise SynapseError(
|
|
403, "This event is not allowed in this context", Codes.FORBIDDEN
|
|
)
|
|
|
|
try:
|
|
# The remote hasn't signed it yet, obviously. We'll do the full checks
|
|
# when we get the event back in `on_send_knock_request`
|
|
await self._event_auth_handler.check_from_context(
|
|
room_version, event, context, do_sig_check=False
|
|
)
|
|
except AuthError as e:
|
|
logger.warning("Failed to create new knock %r because %s", event, e)
|
|
raise e
|
|
|
|
return event
|
|
|
|
@log_function
|
|
async def on_send_membership_event(
|
|
self, origin: str, event: EventBase
|
|
) -> Tuple[EventBase, EventContext]:
|
|
"""
|
|
We have received a join/leave/knock event for a room via send_join/leave/knock.
|
|
|
|
Verify that event and send it into the room on the remote homeserver's behalf.
|
|
|
|
This is quite similar to on_receive_pdu, with the following principal
|
|
differences:
|
|
* only membership events are permitted (and only events with
|
|
sender==state_key -- ie, no kicks or bans)
|
|
* *We* send out the event on behalf of the remote server.
|
|
* We enforce the membership restrictions of restricted rooms.
|
|
* Rejected events result in an exception rather than being stored.
|
|
|
|
There are also other differences, however it is not clear if these are by
|
|
design or omission. In particular, we do not attempt to backfill any missing
|
|
prev_events.
|
|
|
|
Args:
|
|
origin: The homeserver of the remote (joining/invited/knocking) user.
|
|
event: The member event that has been signed by the remote homeserver.
|
|
|
|
Returns:
|
|
The event and context of the event after inserting it into the room graph.
|
|
|
|
Raises:
|
|
SynapseError if the event is not accepted into the room
|
|
"""
|
|
logger.debug(
|
|
"on_send_membership_event: Got event: %s, signatures: %s",
|
|
event.event_id,
|
|
event.signatures,
|
|
)
|
|
|
|
if get_domain_from_id(event.sender) != origin:
|
|
logger.info(
|
|
"Got send_membership request for user %r from different origin %s",
|
|
event.sender,
|
|
origin,
|
|
)
|
|
raise SynapseError(403, "User not from origin", Codes.FORBIDDEN)
|
|
|
|
if event.sender != event.state_key:
|
|
raise SynapseError(400, "state_key and sender must match", Codes.BAD_JSON)
|
|
|
|
assert not event.internal_metadata.outlier
|
|
|
|
# Send this event on behalf of the other server.
|
|
#
|
|
# The remote server isn't a full participant in the room at this point, so
|
|
# may not have an up-to-date list of the other homeservers participating in
|
|
# the room, so we send it on their behalf.
|
|
event.internal_metadata.send_on_behalf_of = origin
|
|
|
|
context = await self.state_handler.compute_event_context(event)
|
|
context = await self._check_event_auth(origin, event, context)
|
|
if context.rejected:
|
|
raise SynapseError(
|
|
403, f"{event.membership} event was rejected", Codes.FORBIDDEN
|
|
)
|
|
|
|
# for joins, we need to check the restrictions of restricted rooms
|
|
if event.membership == Membership.JOIN:
|
|
await self._check_join_restrictions(context, event)
|
|
|
|
# for knock events, we run the third-party event rules. It's not entirely clear
|
|
# why we don't do this for other sorts of membership events.
|
|
if event.membership == Membership.KNOCK:
|
|
event_allowed, _ = await self.third_party_event_rules.check_event_allowed(
|
|
event, context
|
|
)
|
|
if not event_allowed:
|
|
logger.info("Sending of knock %s forbidden by third-party rules", event)
|
|
raise SynapseError(
|
|
403, "This event is not allowed in this context", Codes.FORBIDDEN
|
|
)
|
|
|
|
# all looks good, we can persist the event.
|
|
await self._run_push_actions_and_persist_event(event, context)
|
|
return event, context
|
|
|
|
async def _check_join_restrictions(
|
|
self, context: EventContext, event: EventBase
|
|
) -> None:
|
|
"""Check that restrictions in restricted join rules are matched
|
|
|
|
Called when we receive a join event via send_join.
|
|
|
|
Raises an auth error if the restrictions are not matched.
|
|
"""
|
|
prev_state_ids = await context.get_prev_state_ids()
|
|
|
|
# Check if the user is already in the room or invited to the room.
|
|
user_id = event.state_key
|
|
prev_member_event_id = prev_state_ids.get((EventTypes.Member, user_id), None)
|
|
prev_member_event = None
|
|
if prev_member_event_id:
|
|
prev_member_event = await self.store.get_event(prev_member_event_id)
|
|
|
|
# Check if the member should be allowed access via membership in a space.
|
|
await self._event_auth_handler.check_restricted_join_rules(
|
|
prev_state_ids,
|
|
event.room_version,
|
|
user_id,
|
|
prev_member_event,
|
|
)
|
|
|
|
async def get_state_for_pdu(self, room_id: str, event_id: str) -> List[EventBase]:
|
|
"""Returns the state at the event. i.e. not including said event."""
|
|
|
|
event = await self.store.get_event(event_id, check_room_id=room_id)
|
|
|
|
state_groups = await self.state_store.get_state_groups(room_id, [event_id])
|
|
|
|
if state_groups:
|
|
_, state = list(state_groups.items()).pop()
|
|
results = {(e.type, e.state_key): e for e in state}
|
|
|
|
if event.is_state():
|
|
# Get previous state
|
|
if "replaces_state" in event.unsigned:
|
|
prev_id = event.unsigned["replaces_state"]
|
|
if prev_id != event.event_id:
|
|
prev_event = await self.store.get_event(prev_id)
|
|
results[(event.type, event.state_key)] = prev_event
|
|
else:
|
|
del results[(event.type, event.state_key)]
|
|
|
|
res = list(results.values())
|
|
return res
|
|
else:
|
|
return []
|
|
|
|
async def get_state_ids_for_pdu(self, room_id: str, event_id: str) -> List[str]:
|
|
"""Returns the state at the event. i.e. not including said event."""
|
|
event = await self.store.get_event(event_id, check_room_id=room_id)
|
|
|
|
state_groups = await self.state_store.get_state_groups_ids(room_id, [event_id])
|
|
|
|
if state_groups:
|
|
_, state = list(state_groups.items()).pop()
|
|
results = state
|
|
|
|
if event.is_state():
|
|
# Get previous state
|
|
if "replaces_state" in event.unsigned:
|
|
prev_id = event.unsigned["replaces_state"]
|
|
if prev_id != event.event_id:
|
|
results[(event.type, event.state_key)] = prev_id
|
|
else:
|
|
results.pop((event.type, event.state_key), None)
|
|
|
|
return list(results.values())
|
|
else:
|
|
return []
|
|
|
|
@log_function
|
|
async def on_backfill_request(
|
|
self, origin: str, room_id: str, pdu_list: List[str], limit: int
|
|
) -> List[EventBase]:
|
|
in_room = await self._event_auth_handler.check_host_in_room(room_id, origin)
|
|
if not in_room:
|
|
raise AuthError(403, "Host not in room.")
|
|
|
|
# Synapse asks for 100 events per backfill request. Do not allow more.
|
|
limit = min(limit, 100)
|
|
|
|
events = await self.store.get_backfill_events(room_id, pdu_list, limit)
|
|
|
|
events = await filter_events_for_server(self.storage, origin, events)
|
|
|
|
return events
|
|
|
|
@log_function
|
|
async def get_persisted_pdu(
|
|
self, origin: str, event_id: str
|
|
) -> Optional[EventBase]:
|
|
"""Get an event from the database for the given server.
|
|
|
|
Args:
|
|
origin: hostname of server which is requesting the event; we
|
|
will check that the server is allowed to see it.
|
|
event_id: id of the event being requested
|
|
|
|
Returns:
|
|
None if we know nothing about the event; otherwise the (possibly-redacted) event.
|
|
|
|
Raises:
|
|
AuthError if the server is not currently in the room
|
|
"""
|
|
event = await self.store.get_event(
|
|
event_id, allow_none=True, allow_rejected=True
|
|
)
|
|
|
|
if event:
|
|
in_room = await self._event_auth_handler.check_host_in_room(
|
|
event.room_id, origin
|
|
)
|
|
if not in_room:
|
|
raise AuthError(403, "Host not in room.")
|
|
|
|
events = await filter_events_for_server(self.storage, origin, [event])
|
|
event = events[0]
|
|
return event
|
|
else:
|
|
return None
|
|
|
|
async def get_min_depth_for_context(self, context: str) -> int:
|
|
return await self.store.get_min_depth(context)
|
|
|
|
async def _auth_and_persist_event(
|
|
self,
|
|
origin: str,
|
|
event: EventBase,
|
|
context: EventContext,
|
|
state: Optional[Iterable[EventBase]] = None,
|
|
auth_events: Optional[MutableStateMap[EventBase]] = None,
|
|
backfilled: bool = False,
|
|
) -> None:
|
|
"""
|
|
Process an event by performing auth checks and then persisting to the database.
|
|
|
|
Args:
|
|
origin: The host the event originates from.
|
|
event: The event itself.
|
|
context:
|
|
The event context.
|
|
|
|
NB that this function potentially modifies it.
|
|
state:
|
|
The state events used to check the event for soft-fail. If this is
|
|
not provided the current state events will be used.
|
|
auth_events:
|
|
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.
|
|
backfilled: True if the event was backfilled.
|
|
"""
|
|
context = await self._check_event_auth(
|
|
origin,
|
|
event,
|
|
context,
|
|
state=state,
|
|
auth_events=auth_events,
|
|
backfilled=backfilled,
|
|
)
|
|
|
|
await self._run_push_actions_and_persist_event(event, context, backfilled)
|
|
|
|
async def _run_push_actions_and_persist_event(
|
|
self, event: EventBase, context: EventContext, backfilled: bool = False
|
|
):
|
|
"""Run the push actions for a received event, and persist it.
|
|
|
|
Args:
|
|
event: The event itself.
|
|
context: The event context.
|
|
backfilled: True if the event was backfilled.
|
|
"""
|
|
try:
|
|
if (
|
|
not event.internal_metadata.is_outlier()
|
|
and not backfilled
|
|
and not context.rejected
|
|
):
|
|
await self.action_generator.handle_push_actions_for_event(
|
|
event, context
|
|
)
|
|
|
|
await self.persist_events_and_notify(
|
|
event.room_id, [(event, context)], backfilled=backfilled
|
|
)
|
|
except Exception:
|
|
run_in_background(
|
|
self.store.remove_push_actions_from_staging, event.event_id
|
|
)
|
|
raise
|
|
|
|
async def _auth_and_persist_events(
|
|
self,
|
|
origin: str,
|
|
room_id: str,
|
|
event_infos: Collection[_NewEventInfo],
|
|
backfilled: bool = False,
|
|
) -> None:
|
|
"""Creates the appropriate contexts and persists events. The events
|
|
should not depend on one another, e.g. this should be used to persist
|
|
a bunch of outliers, but not a chunk of individual events that depend
|
|
on each other for state calculations.
|
|
|
|
Notifies about the events where appropriate.
|
|
"""
|
|
|
|
if not event_infos:
|
|
return
|
|
|
|
async def prep(ev_info: _NewEventInfo):
|
|
event = ev_info.event
|
|
with nested_logging_context(suffix=event.event_id):
|
|
res = await self.state_handler.compute_event_context(
|
|
event, old_state=ev_info.state
|
|
)
|
|
res = await self._check_event_auth(
|
|
origin,
|
|
event,
|
|
res,
|
|
state=ev_info.state,
|
|
auth_events=ev_info.auth_events,
|
|
backfilled=backfilled,
|
|
)
|
|
return res
|
|
|
|
contexts = await make_deferred_yieldable(
|
|
defer.gatherResults(
|
|
[run_in_background(prep, ev_info) for ev_info in event_infos],
|
|
consumeErrors=True,
|
|
)
|
|
)
|
|
|
|
await self.persist_events_and_notify(
|
|
room_id,
|
|
[
|
|
(ev_info.event, context)
|
|
for ev_info, context in zip(event_infos, contexts)
|
|
],
|
|
backfilled=backfilled,
|
|
)
|
|
|
|
async def _persist_auth_tree(
|
|
self,
|
|
origin: str,
|
|
room_id: str,
|
|
auth_events: List[EventBase],
|
|
state: List[EventBase],
|
|
event: EventBase,
|
|
room_version: RoomVersion,
|
|
) -> int:
|
|
"""Checks the auth chain is valid (and passes auth checks) for the
|
|
state and event. Then persists the auth chain and state atomically.
|
|
Persists the event separately. Notifies about the persisted events
|
|
where appropriate.
|
|
|
|
Will attempt to fetch missing auth events.
|
|
|
|
Args:
|
|
origin: Where the events came from
|
|
room_id,
|
|
auth_events
|
|
state
|
|
event
|
|
room_version: The room version we expect this room to have, and
|
|
will raise if it doesn't match the version in the create event.
|
|
"""
|
|
events_to_context = {}
|
|
for e in itertools.chain(auth_events, state):
|
|
e.internal_metadata.outlier = True
|
|
ctx = await self.state_handler.compute_event_context(e)
|
|
events_to_context[e.event_id] = ctx
|
|
|
|
event_map = {
|
|
e.event_id: e for e in itertools.chain(auth_events, state, [event])
|
|
}
|
|
|
|
create_event = None
|
|
for e in auth_events:
|
|
if (e.type, e.state_key) == (EventTypes.Create, ""):
|
|
create_event = e
|
|
break
|
|
|
|
if create_event is None:
|
|
# If the state doesn't have a create event then the room is
|
|
# invalid, and it would fail auth checks anyway.
|
|
raise SynapseError(400, "No create event in state")
|
|
|
|
room_version_id = create_event.content.get(
|
|
"room_version", RoomVersions.V1.identifier
|
|
)
|
|
|
|
if room_version.identifier != room_version_id:
|
|
raise SynapseError(400, "Room version mismatch")
|
|
|
|
missing_auth_events = set()
|
|
for e in itertools.chain(auth_events, state, [event]):
|
|
for e_id in e.auth_event_ids():
|
|
if e_id not in event_map:
|
|
missing_auth_events.add(e_id)
|
|
|
|
for e_id in missing_auth_events:
|
|
m_ev = await self.federation_client.get_pdu(
|
|
[origin],
|
|
e_id,
|
|
room_version=room_version,
|
|
outlier=True,
|
|
timeout=10000,
|
|
)
|
|
if m_ev and m_ev.event_id == e_id:
|
|
event_map[e_id] = m_ev
|
|
else:
|
|
logger.info("Failed to find auth event %r", e_id)
|
|
|
|
for e in itertools.chain(auth_events, state, [event]):
|
|
auth_for_e = {
|
|
(event_map[e_id].type, event_map[e_id].state_key): event_map[e_id]
|
|
for e_id in e.auth_event_ids()
|
|
if e_id in event_map
|
|
}
|
|
if create_event:
|
|
auth_for_e[(EventTypes.Create, "")] = create_event
|
|
|
|
try:
|
|
event_auth.check(room_version, e, auth_events=auth_for_e)
|
|
except SynapseError as err:
|
|
# we may get SynapseErrors here as well as AuthErrors. For
|
|
# instance, there are a couple of (ancient) events in some
|
|
# rooms whose senders do not have the correct sigil; these
|
|
# cause SynapseErrors in auth.check. We don't want to give up
|
|
# the attempt to federate altogether in such cases.
|
|
|
|
logger.warning("Rejecting %s because %s", e.event_id, err.msg)
|
|
|
|
if e == event:
|
|
raise
|
|
events_to_context[e.event_id].rejected = RejectedReason.AUTH_ERROR
|
|
|
|
if auth_events or state:
|
|
await self.persist_events_and_notify(
|
|
room_id,
|
|
[
|
|
(e, events_to_context[e.event_id])
|
|
for e in itertools.chain(auth_events, state)
|
|
],
|
|
)
|
|
|
|
new_event_context = await self.state_handler.compute_event_context(
|
|
event, old_state=state
|
|
)
|
|
|
|
return await self.persist_events_and_notify(
|
|
room_id, [(event, new_event_context)]
|
|
)
|
|
|
|
async def _check_for_soft_fail(
|
|
self,
|
|
event: EventBase,
|
|
state: Optional[Iterable[EventBase]],
|
|
backfilled: bool,
|
|
origin: str,
|
|
) -> None:
|
|
"""Checks if we should soft fail the event; if so, marks the event as
|
|
such.
|
|
|
|
Args:
|
|
event
|
|
state: The state at the event if we don't have all the event's prev events
|
|
backfilled: Whether the event is from backfill
|
|
origin: The host the event originates from.
|
|
"""
|
|
# For new (non-backfilled and non-outlier) events we check if the event
|
|
# passes auth based on the current state. If it doesn't then we
|
|
# "soft-fail" the event.
|
|
if backfilled or event.internal_metadata.is_outlier():
|
|
return
|
|
|
|
extrem_ids_list = await self.store.get_latest_event_ids_in_room(event.room_id)
|
|
extrem_ids = set(extrem_ids_list)
|
|
prev_event_ids = set(event.prev_event_ids())
|
|
|
|
if extrem_ids == prev_event_ids:
|
|
# If they're the same then the current state is the same as the
|
|
# state at the event, so no point rechecking auth for soft fail.
|
|
return
|
|
|
|
room_version = await self.store.get_room_version_id(event.room_id)
|
|
room_version_obj = KNOWN_ROOM_VERSIONS[room_version]
|
|
|
|
# Calculate the "current state".
|
|
if state is not None:
|
|
# If we're explicitly given the state then we won't have all the
|
|
# prev events, and so we have a gap in the graph. In this case
|
|
# we want to be a little careful as we might have been down for
|
|
# a while and have an incorrect view of the current state,
|
|
# however we still want to do checks as gaps are easy to
|
|
# maliciously manufacture.
|
|
#
|
|
# So we use a "current state" that is actually a state
|
|
# resolution across the current forward extremities and the
|
|
# given state at the event. This should correctly handle cases
|
|
# like bans, especially with state res v2.
|
|
|
|
state_sets_d = await self.state_store.get_state_groups(
|
|
event.room_id, extrem_ids
|
|
)
|
|
state_sets: List[Iterable[EventBase]] = list(state_sets_d.values())
|
|
state_sets.append(state)
|
|
current_states = await self.state_handler.resolve_events(
|
|
room_version, state_sets, event
|
|
)
|
|
current_state_ids: StateMap[str] = {
|
|
k: e.event_id for k, e in current_states.items()
|
|
}
|
|
else:
|
|
current_state_ids = await self.state_handler.get_current_state_ids(
|
|
event.room_id, latest_event_ids=extrem_ids
|
|
)
|
|
|
|
logger.debug(
|
|
"Doing soft-fail check for %s: state %s",
|
|
event.event_id,
|
|
current_state_ids,
|
|
)
|
|
|
|
# Now check if event pass auth against said current state
|
|
auth_types = auth_types_for_event(room_version_obj, event)
|
|
current_state_ids_list = [
|
|
e for k, e in current_state_ids.items() if k in auth_types
|
|
]
|
|
|
|
auth_events_map = await self.store.get_events(current_state_ids_list)
|
|
current_auth_events = {
|
|
(e.type, e.state_key): e for e in auth_events_map.values()
|
|
}
|
|
|
|
try:
|
|
event_auth.check(room_version_obj, event, auth_events=current_auth_events)
|
|
except AuthError as e:
|
|
logger.warning(
|
|
"Soft-failing %r (from %s) because %s",
|
|
event,
|
|
e,
|
|
origin,
|
|
extra={
|
|
"room_id": event.room_id,
|
|
"mxid": event.sender,
|
|
"hs": origin,
|
|
},
|
|
)
|
|
soft_failed_event_counter.inc()
|
|
event.internal_metadata.soft_failed = True
|
|
|
|
async def on_get_missing_events(
|
|
self,
|
|
origin: str,
|
|
room_id: str,
|
|
earliest_events: List[str],
|
|
latest_events: List[str],
|
|
limit: int,
|
|
) -> List[EventBase]:
|
|
in_room = await self._event_auth_handler.check_host_in_room(room_id, origin)
|
|
if not in_room:
|
|
raise AuthError(403, "Host not in room.")
|
|
|
|
# Only allow up to 20 events to be retrieved per request.
|
|
limit = min(limit, 20)
|
|
|
|
missing_events = await self.store.get_missing_events(
|
|
room_id=room_id,
|
|
earliest_events=earliest_events,
|
|
latest_events=latest_events,
|
|
limit=limit,
|
|
)
|
|
|
|
missing_events = await filter_events_for_server(
|
|
self.storage, origin, missing_events
|
|
)
|
|
|
|
return missing_events
|
|
|
|
async def _check_event_auth(
|
|
self,
|
|
origin: str,
|
|
event: EventBase,
|
|
context: EventContext,
|
|
state: Optional[Iterable[EventBase]] = None,
|
|
auth_events: Optional[MutableStateMap[EventBase]] = None,
|
|
backfilled: bool = False,
|
|
) -> EventContext:
|
|
"""
|
|
Checks whether an event should be rejected (for failing auth checks).
|
|
|
|
Args:
|
|
origin: The host the event originates from.
|
|
event: The event itself.
|
|
context:
|
|
The event context.
|
|
|
|
NB that this function potentially modifies it.
|
|
state:
|
|
The state events used to check the event for soft-fail. If this is
|
|
not provided the current state events will be used.
|
|
auth_events:
|
|
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.
|
|
|
|
If this is not provided, it is calculated from the previous state IDs.
|
|
backfilled: True if the event was backfilled.
|
|
|
|
Returns:
|
|
The updated context object.
|
|
"""
|
|
room_version = await self.store.get_room_version_id(event.room_id)
|
|
room_version_obj = KNOWN_ROOM_VERSIONS[room_version]
|
|
|
|
if not auth_events:
|
|
prev_state_ids = await context.get_prev_state_ids()
|
|
auth_events_ids = self._event_auth_handler.compute_auth_events(
|
|
event, prev_state_ids, for_verification=True
|
|
)
|
|
auth_events_x = await self.store.get_events(auth_events_ids)
|
|
auth_events = {(e.type, e.state_key): e for e in auth_events_x.values()}
|
|
|
|
# This is a hack to fix some old rooms where the initial join event
|
|
# didn't reference the create event in its auth events.
|
|
if event.type == EventTypes.Member and not event.auth_event_ids():
|
|
if len(event.prev_event_ids()) == 1 and event.depth < 5:
|
|
c = await self.store.get_event(
|
|
event.prev_event_ids()[0], allow_none=True
|
|
)
|
|
if c and c.type == EventTypes.Create:
|
|
auth_events[(c.type, c.state_key)] = c
|
|
|
|
try:
|
|
context = await self._update_auth_events_and_context_for_auth(
|
|
origin, event, context, auth_events
|
|
)
|
|
except Exception:
|
|
# We don't really mind if the above fails, so lets not fail
|
|
# processing if it does. However, it really shouldn't fail so
|
|
# let's still log as an exception since we'll still want to fix
|
|
# any bugs.
|
|
logger.exception(
|
|
"Failed to double check auth events for %s with remote. "
|
|
"Ignoring failure and continuing processing of event.",
|
|
event.event_id,
|
|
)
|
|
|
|
try:
|
|
event_auth.check(room_version_obj, event, auth_events=auth_events)
|
|
except AuthError as e:
|
|
logger.warning("Failed auth resolution for %r because %s", event, e)
|
|
context.rejected = RejectedReason.AUTH_ERROR
|
|
|
|
if not context.rejected:
|
|
await self._check_for_soft_fail(event, state, backfilled, origin=origin)
|
|
|
|
if event.type == EventTypes.GuestAccess and not context.rejected:
|
|
await self.maybe_kick_guest_users(event)
|
|
|
|
# If we are going to send this event over federation we precaclculate
|
|
# the joined hosts.
|
|
if event.internal_metadata.get_send_on_behalf_of():
|
|
await self.event_creation_handler.cache_joined_hosts_for_event(
|
|
event, context
|
|
)
|
|
|
|
return context
|
|
|
|
async def _update_auth_events_and_context_for_auth(
|
|
self,
|
|
origin: str,
|
|
event: EventBase,
|
|
context: EventContext,
|
|
auth_events: MutableStateMap[EventBase],
|
|
) -> EventContext:
|
|
"""Helper for _check_event_auth. See there for docs.
|
|
|
|
Checks whether a given event has the expected auth events. If it
|
|
doesn't then we talk to the remote server to compare state to see if
|
|
we can come to a consensus (e.g. if one server missed some valid
|
|
state).
|
|
|
|
This attempts to resolve any potential divergence of state between
|
|
servers, but is not essential and so failures should not block further
|
|
processing of the event.
|
|
|
|
Args:
|
|
origin:
|
|
event:
|
|
context:
|
|
|
|
auth_events:
|
|
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:
|
|
updated context
|
|
"""
|
|
event_auth_events = set(event.auth_event_ids())
|
|
|
|
# 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:
|
|
have_events = await self.store.have_seen_events(event.room_id, missing_auth)
|
|
logger.debug("Events %s are in the store", have_events)
|
|
missing_auth.difference_update(have_events)
|
|
|
|
if missing_auth:
|
|
# If we don't have all the auth events, we need to get them.
|
|
logger.info("auth_events contains unknown events: %s", missing_auth)
|
|
try:
|
|
try:
|
|
remote_auth_chain = await self.federation_client.get_event_auth(
|
|
origin, event.room_id, event.event_id
|
|
)
|
|
except RequestSendFailed as e1:
|
|
# The other side isn't around or doesn't implement the
|
|
# endpoint, so lets just bail out.
|
|
logger.info("Failed to get event auth from remote: %s", e1)
|
|
return context
|
|
|
|
seen_remotes = await self.store.have_seen_events(
|
|
event.room_id, [e.event_id for e in remote_auth_chain]
|
|
)
|
|
|
|
for e in remote_auth_chain:
|
|
if e.event_id in seen_remotes:
|
|
continue
|
|
|
|
if e.event_id == event.event_id:
|
|
continue
|
|
|
|
try:
|
|
auth_ids = e.auth_event_ids()
|
|
auth = {
|
|
(e.type, e.state_key): e
|
|
for e in remote_auth_chain
|
|
if e.event_id in auth_ids or e.type == EventTypes.Create
|
|
}
|
|
e.internal_metadata.outlier = True
|
|
|
|
logger.debug(
|
|
"_check_event_auth %s missing_auth: %s",
|
|
event.event_id,
|
|
e.event_id,
|
|
)
|
|
missing_auth_event_context = (
|
|
await self.state_handler.compute_event_context(e)
|
|
)
|
|
await self._auth_and_persist_event(
|
|
origin, e, missing_auth_event_context, auth_events=auth
|
|
)
|
|
|
|
if e.event_id in event_auth_events:
|
|
auth_events[(e.type, e.state_key)] = e
|
|
except AuthError:
|
|
pass
|
|
|
|
except Exception:
|
|
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
|
|
|
|
different_auth = event_auth_events.difference(
|
|
e.event_id for e in auth_events.values()
|
|
)
|
|
|
|
if not different_auth:
|
|
return context
|
|
|
|
logger.info(
|
|
"auth_events refers to events which are not in our calculated auth "
|
|
"chain: %s",
|
|
different_auth,
|
|
)
|
|
|
|
# XXX: currently this checks for redactions but I'm not convinced that is
|
|
# necessary?
|
|
different_events = await self.store.get_events_as_list(different_auth)
|
|
|
|
for d in different_events:
|
|
if d.room_id != event.room_id:
|
|
logger.warning(
|
|
"Event %s refers to auth_event %s which is in a different room",
|
|
event.event_id,
|
|
d.event_id,
|
|
)
|
|
|
|
# don't attempt to resolve the claimed auth events against our own
|
|
# in this case: just use our own auth events.
|
|
#
|
|
# XXX: should we reject the event in this case? It feels like we should,
|
|
# but then shouldn't we also do so if we've failed to fetch any of the
|
|
# auth events?
|
|
return context
|
|
|
|
# now we state-resolve between our own idea of the auth events, and the remote's
|
|
# idea of them.
|
|
|
|
local_state = auth_events.values()
|
|
remote_auth_events = dict(auth_events)
|
|
remote_auth_events.update({(d.type, d.state_key): d for d in different_events})
|
|
remote_state = remote_auth_events.values()
|
|
|
|
room_version = await self.store.get_room_version_id(event.room_id)
|
|
new_state = await self.state_handler.resolve_events(
|
|
room_version, (local_state, remote_state), event
|
|
)
|
|
|
|
logger.info(
|
|
"After state res: updating auth_events with new state %s",
|
|
{
|
|
(d.type, d.state_key): d.event_id
|
|
for d in new_state.values()
|
|
if auth_events.get((d.type, d.state_key)) != d
|
|
},
|
|
)
|
|
|
|
auth_events.update(new_state)
|
|
|
|
context = await self._update_context_for_auth_events(
|
|
event, context, auth_events
|
|
)
|
|
|
|
return context
|
|
|
|
async def _update_context_for_auth_events(
|
|
self, event: EventBase, context: EventContext, auth_events: StateMap[EventBase]
|
|
) -> EventContext:
|
|
"""Update the state_ids in an event context after auth event resolution,
|
|
storing the changes as a new state group.
|
|
|
|
Args:
|
|
event: The event we're handling the context for
|
|
|
|
context: initial event context
|
|
|
|
auth_events: Events to update in the event context.
|
|
|
|
Returns:
|
|
new event context
|
|
"""
|
|
# exclude the state key of the new event from the current_state in the context.
|
|
if event.is_state():
|
|
event_key: Optional[Tuple[str, str]] = (event.type, event.state_key)
|
|
else:
|
|
event_key = None
|
|
state_updates = {
|
|
k: a.event_id for k, a in auth_events.items() if k != event_key
|
|
}
|
|
|
|
current_state_ids = await context.get_current_state_ids()
|
|
current_state_ids = dict(current_state_ids) # type: ignore
|
|
|
|
current_state_ids.update(state_updates)
|
|
|
|
prev_state_ids = await context.get_prev_state_ids()
|
|
prev_state_ids = dict(prev_state_ids)
|
|
|
|
prev_state_ids.update({k: a.event_id for k, a in auth_events.items()})
|
|
|
|
# create a new state group as a delta from the existing one.
|
|
prev_group = context.state_group
|
|
state_group = await self.state_store.store_state_group(
|
|
event.event_id,
|
|
event.room_id,
|
|
prev_group=prev_group,
|
|
delta_ids=state_updates,
|
|
current_state_ids=current_state_ids,
|
|
)
|
|
|
|
return EventContext.with_state(
|
|
state_group=state_group,
|
|
state_group_before_event=context.state_group_before_event,
|
|
current_state_ids=current_state_ids,
|
|
prev_state_ids=prev_state_ids,
|
|
prev_group=prev_group,
|
|
delta_ids=state_updates,
|
|
)
|
|
|
|
async def construct_auth_difference(
|
|
self, local_auth: Iterable[EventBase], remote_auth: Iterable[EventBase]
|
|
) -> Dict:
|
|
"""Given a local and remote auth chain, find the differences. This
|
|
assumes that we have already processed all events in remote_auth
|
|
|
|
Params:
|
|
local_auth
|
|
remote_auth
|
|
|
|
Returns:
|
|
dict
|
|
"""
|
|
|
|
logger.debug("construct_auth_difference Start!")
|
|
|
|
# TODO: Make sure we are OK with local_auth or remote_auth having more
|
|
# auth events in them than strictly necessary.
|
|
|
|
def sort_fun(ev):
|
|
return ev.depth, ev.event_id
|
|
|
|
logger.debug("construct_auth_difference after sort_fun!")
|
|
|
|
# We find the differences by starting at the "bottom" of each list
|
|
# and iterating up on both lists. The lists are ordered by depth and
|
|
# then event_id, we iterate up both lists until we find the event ids
|
|
# don't match. Then we look at depth/event_id to see which side is
|
|
# missing that event, and iterate only up that list. Repeat.
|
|
|
|
remote_list = list(remote_auth)
|
|
remote_list.sort(key=sort_fun)
|
|
|
|
local_list = list(local_auth)
|
|
local_list.sort(key=sort_fun)
|
|
|
|
local_iter = iter(local_list)
|
|
remote_iter = iter(remote_list)
|
|
|
|
logger.debug("construct_auth_difference before get_next!")
|
|
|
|
def get_next(it, opt=None):
|
|
try:
|
|
return next(it)
|
|
except Exception:
|
|
return opt
|
|
|
|
current_local = get_next(local_iter)
|
|
current_remote = get_next(remote_iter)
|
|
|
|
logger.debug("construct_auth_difference before while")
|
|
|
|
missing_remotes = []
|
|
missing_locals = []
|
|
while current_local or current_remote:
|
|
if current_remote is None:
|
|
missing_locals.append(current_local)
|
|
current_local = get_next(local_iter)
|
|
continue
|
|
|
|
if current_local is None:
|
|
missing_remotes.append(current_remote)
|
|
current_remote = get_next(remote_iter)
|
|
continue
|
|
|
|
if current_local.event_id == current_remote.event_id:
|
|
current_local = get_next(local_iter)
|
|
current_remote = get_next(remote_iter)
|
|
continue
|
|
|
|
if current_local.depth < current_remote.depth:
|
|
missing_locals.append(current_local)
|
|
current_local = get_next(local_iter)
|
|
continue
|
|
|
|
if current_local.depth > current_remote.depth:
|
|
missing_remotes.append(current_remote)
|
|
current_remote = get_next(remote_iter)
|
|
continue
|
|
|
|
# They have the same depth, so we fall back to the event_id order
|
|
if current_local.event_id < current_remote.event_id:
|
|
missing_locals.append(current_local)
|
|
current_local = get_next(local_iter)
|
|
|
|
if current_local.event_id > current_remote.event_id:
|
|
missing_remotes.append(current_remote)
|
|
current_remote = get_next(remote_iter)
|
|
continue
|
|
|
|
logger.debug("construct_auth_difference after while")
|
|
|
|
# missing locals should be sent to the server
|
|
# We should find why we are missing remotes, as they will have been
|
|
# rejected.
|
|
|
|
# Remove events from missing_remotes if they are referencing a missing
|
|
# remote. We only care about the "root" rejected ones.
|
|
missing_remote_ids = [e.event_id for e in missing_remotes]
|
|
base_remote_rejected = list(missing_remotes)
|
|
for e in missing_remotes:
|
|
for e_id in e.auth_event_ids():
|
|
if e_id in missing_remote_ids:
|
|
try:
|
|
base_remote_rejected.remove(e)
|
|
except ValueError:
|
|
pass
|
|
|
|
reason_map = {}
|
|
|
|
for e in base_remote_rejected:
|
|
reason = await self.store.get_rejection_reason(e.event_id)
|
|
if reason is None:
|
|
# TODO: e is not in the current state, so we should
|
|
# construct some proof of that.
|
|
continue
|
|
|
|
reason_map[e.event_id] = reason
|
|
|
|
logger.debug("construct_auth_difference returning")
|
|
|
|
return {
|
|
"auth_chain": local_auth,
|
|
"rejects": {
|
|
e.event_id: {"reason": reason_map[e.event_id], "proof": None}
|
|
for e in base_remote_rejected
|
|
},
|
|
"missing": [e.event_id for e in missing_locals],
|
|
}
|
|
|
|
@log_function
|
|
async def exchange_third_party_invite(
|
|
self, sender_user_id: str, target_user_id: str, room_id: str, signed: JsonDict
|
|
) -> None:
|
|
third_party_invite = {"signed": signed}
|
|
|
|
event_dict = {
|
|
"type": EventTypes.Member,
|
|
"content": {
|
|
"membership": Membership.INVITE,
|
|
"third_party_invite": third_party_invite,
|
|
},
|
|
"room_id": room_id,
|
|
"sender": sender_user_id,
|
|
"state_key": target_user_id,
|
|
}
|
|
|
|
if await self._event_auth_handler.check_host_in_room(room_id, self.hs.hostname):
|
|
room_version = await self.store.get_room_version_id(room_id)
|
|
builder = self.event_builder_factory.new(room_version, event_dict)
|
|
|
|
EventValidator().validate_builder(builder)
|
|
event, context = await self.event_creation_handler.create_new_client_event(
|
|
builder=builder
|
|
)
|
|
|
|
event, context = await self.add_display_name_to_third_party_invite(
|
|
room_version, event_dict, event, context
|
|
)
|
|
|
|
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.
|
|
event.internal_metadata.send_on_behalf_of = self.hs.hostname
|
|
|
|
try:
|
|
await self._event_auth_handler.check_from_context(
|
|
room_version, event, context
|
|
)
|
|
except AuthError as e:
|
|
logger.warning("Denying new third party invite %r because %s", event, e)
|
|
raise e
|
|
|
|
await self._check_signature(event, context)
|
|
|
|
# We retrieve the room member handler here as to not cause a cyclic dependency
|
|
member_handler = self.hs.get_room_member_handler()
|
|
await member_handler.send_membership_event(None, event, context)
|
|
else:
|
|
destinations = {x.split(":", 1)[-1] for x in (sender_user_id, room_id)}
|
|
|
|
try:
|
|
await self.federation_client.forward_third_party_invite(
|
|
destinations, room_id, event_dict
|
|
)
|
|
except (RequestSendFailed, HttpResponseException):
|
|
raise SynapseError(502, "Failed to forward third party invite")
|
|
|
|
async def on_exchange_third_party_invite_request(
|
|
self, event_dict: JsonDict
|
|
) -> None:
|
|
"""Handle an exchange_third_party_invite request from a remote server
|
|
|
|
The remote server will call this when it wants to turn a 3pid invite
|
|
into a normal m.room.member invite.
|
|
|
|
Args:
|
|
event_dict: Dictionary containing the event body.
|
|
|
|
"""
|
|
assert_params_in_dict(event_dict, ["room_id"])
|
|
room_version = await self.store.get_room_version_id(event_dict["room_id"])
|
|
|
|
# NB: event_dict has a particular specced format we might need to fudge
|
|
# if we change event formats too much.
|
|
builder = self.event_builder_factory.new(room_version, event_dict)
|
|
|
|
event, context = await self.event_creation_handler.create_new_client_event(
|
|
builder=builder
|
|
)
|
|
event, context = await self.add_display_name_to_third_party_invite(
|
|
room_version, event_dict, event, context
|
|
)
|
|
|
|
try:
|
|
await self._event_auth_handler.check_from_context(
|
|
room_version, event, context
|
|
)
|
|
except AuthError as e:
|
|
logger.warning("Denying third party invite %r because %s", event, e)
|
|
raise e
|
|
await self._check_signature(event, context)
|
|
|
|
# We need to tell the transaction queue to send this out, even
|
|
# though the sender isn't a local user.
|
|
event.internal_metadata.send_on_behalf_of = get_domain_from_id(event.sender)
|
|
|
|
# We retrieve the room member handler here as to not cause a cyclic dependency
|
|
member_handler = self.hs.get_room_member_handler()
|
|
await member_handler.send_membership_event(None, event, context)
|
|
|
|
async def add_display_name_to_third_party_invite(
|
|
self,
|
|
room_version: str,
|
|
event_dict: JsonDict,
|
|
event: EventBase,
|
|
context: EventContext,
|
|
) -> Tuple[EventBase, EventContext]:
|
|
key = (
|
|
EventTypes.ThirdPartyInvite,
|
|
event.content["third_party_invite"]["signed"]["token"],
|
|
)
|
|
original_invite = None
|
|
prev_state_ids = await context.get_prev_state_ids()
|
|
original_invite_id = prev_state_ids.get(key)
|
|
if original_invite_id:
|
|
original_invite = await self.store.get_event(
|
|
original_invite_id, allow_none=True
|
|
)
|
|
if original_invite:
|
|
# If the m.room.third_party_invite event's content is empty, it means the
|
|
# invite has been revoked. In this case, we don't have to raise an error here
|
|
# because the auth check will fail on the invite (because it's not able to
|
|
# fetch public keys from the m.room.third_party_invite event's content, which
|
|
# is empty).
|
|
display_name = original_invite.content.get("display_name")
|
|
event_dict["content"]["third_party_invite"]["display_name"] = display_name
|
|
else:
|
|
logger.info(
|
|
"Could not find invite event for third_party_invite: %r", event_dict
|
|
)
|
|
# We don't discard here as this is not the appropriate place to do
|
|
# auth checks. If we need the invite and don't have it then the
|
|
# auth check code will explode appropriately.
|
|
|
|
builder = self.event_builder_factory.new(room_version, event_dict)
|
|
EventValidator().validate_builder(builder)
|
|
event, context = await self.event_creation_handler.create_new_client_event(
|
|
builder=builder
|
|
)
|
|
EventValidator().validate_new(event, self.config)
|
|
return (event, context)
|
|
|
|
async def _check_signature(self, event: EventBase, context: EventContext) -> None:
|
|
"""
|
|
Checks that the signature in the event is consistent with its invite.
|
|
|
|
Args:
|
|
event: The m.room.member event to check
|
|
context:
|
|
|
|
Raises:
|
|
AuthError: if signature didn't match any keys, or key has been
|
|
revoked,
|
|
SynapseError: if a transient error meant a key couldn't be checked
|
|
for revocation.
|
|
"""
|
|
signed = event.content["third_party_invite"]["signed"]
|
|
token = signed["token"]
|
|
|
|
prev_state_ids = await context.get_prev_state_ids()
|
|
invite_event_id = prev_state_ids.get((EventTypes.ThirdPartyInvite, token))
|
|
|
|
invite_event = None
|
|
if invite_event_id:
|
|
invite_event = await self.store.get_event(invite_event_id, allow_none=True)
|
|
|
|
if not invite_event:
|
|
raise AuthError(403, "Could not find invite")
|
|
|
|
logger.debug("Checking auth on event %r", event.content)
|
|
|
|
last_exception: Optional[Exception] = None
|
|
|
|
# for each public key in the 3pid invite event
|
|
for public_key_object in event_auth.get_public_keys(invite_event):
|
|
try:
|
|
# for each sig on the third_party_invite block of the actual invite
|
|
for server, signature_block in signed["signatures"].items():
|
|
for key_name in signature_block.keys():
|
|
if not key_name.startswith("ed25519:"):
|
|
continue
|
|
|
|
logger.debug(
|
|
"Attempting to verify sig with key %s from %r "
|
|
"against pubkey %r",
|
|
key_name,
|
|
server,
|
|
public_key_object,
|
|
)
|
|
|
|
try:
|
|
public_key = public_key_object["public_key"]
|
|
verify_key = decode_verify_key_bytes(
|
|
key_name, decode_base64(public_key)
|
|
)
|
|
verify_signed_json(signed, server, verify_key)
|
|
logger.debug(
|
|
"Successfully verified sig with key %s from %r "
|
|
"against pubkey %r",
|
|
key_name,
|
|
server,
|
|
public_key_object,
|
|
)
|
|
except Exception:
|
|
logger.info(
|
|
"Failed to verify sig with key %s from %r "
|
|
"against pubkey %r",
|
|
key_name,
|
|
server,
|
|
public_key_object,
|
|
)
|
|
raise
|
|
try:
|
|
if "key_validity_url" in public_key_object:
|
|
await self._check_key_revocation(
|
|
public_key, public_key_object["key_validity_url"]
|
|
)
|
|
except Exception:
|
|
logger.info(
|
|
"Failed to query key_validity_url %s",
|
|
public_key_object["key_validity_url"],
|
|
)
|
|
raise
|
|
return
|
|
except Exception as e:
|
|
last_exception = e
|
|
|
|
if last_exception is None:
|
|
# we can only get here if get_public_keys() returned an empty list
|
|
# TODO: make this better
|
|
raise RuntimeError("no public key in invite event")
|
|
|
|
raise last_exception
|
|
|
|
async def _check_key_revocation(self, public_key: str, url: str) -> None:
|
|
"""
|
|
Checks whether public_key has been revoked.
|
|
|
|
Args:
|
|
public_key: base-64 encoded public key.
|
|
url: Key revocation URL.
|
|
|
|
Raises:
|
|
AuthError: if they key has been revoked.
|
|
SynapseError: if a transient error meant a key couldn't be checked
|
|
for revocation.
|
|
"""
|
|
try:
|
|
response = await self.http_client.get_json(url, {"public_key": public_key})
|
|
except Exception:
|
|
raise SynapseError(502, "Third party certificate could not be checked")
|
|
if "valid" not in response or not response["valid"]:
|
|
raise AuthError(403, "Third party certificate was invalid")
|
|
|
|
async def persist_events_and_notify(
|
|
self,
|
|
room_id: str,
|
|
event_and_contexts: Sequence[Tuple[EventBase, EventContext]],
|
|
backfilled: bool = False,
|
|
) -> int:
|
|
"""Persists events and tells the notifier/pushers about them, if
|
|
necessary.
|
|
|
|
Args:
|
|
room_id: The room ID of events being persisted.
|
|
event_and_contexts: Sequence of events with their associated
|
|
context that should be persisted. All events must belong to
|
|
the same room.
|
|
backfilled: Whether these events are a result of
|
|
backfilling or not
|
|
|
|
Returns:
|
|
The stream ID after which all events have been persisted.
|
|
"""
|
|
if not event_and_contexts:
|
|
return self.store.get_current_events_token()
|
|
|
|
instance = self.config.worker.events_shard_config.get_instance(room_id)
|
|
if instance != self._instance_name:
|
|
# Limit the number of events sent over replication. We choose 200
|
|
# here as that is what we default to in `max_request_body_size(..)`
|
|
for batch in batch_iter(event_and_contexts, 200):
|
|
result = await self._send_events(
|
|
instance_name=instance,
|
|
store=self.store,
|
|
room_id=room_id,
|
|
event_and_contexts=batch,
|
|
backfilled=backfilled,
|
|
)
|
|
return result["max_stream_id"]
|
|
else:
|
|
assert self.storage.persistence
|
|
|
|
# Note that this returns the events that were persisted, which may not be
|
|
# the same as were passed in if some were deduplicated due to transaction IDs.
|
|
events, max_stream_token = await self.storage.persistence.persist_events(
|
|
event_and_contexts, backfilled=backfilled
|
|
)
|
|
|
|
if self._ephemeral_messages_enabled:
|
|
for event in events:
|
|
# If there's an expiry timestamp on the event, schedule its expiry.
|
|
self._message_handler.maybe_schedule_expiry(event)
|
|
|
|
if not backfilled: # Never notify for backfilled events
|
|
for event in events:
|
|
await self._notify_persisted_event(event, max_stream_token)
|
|
|
|
return max_stream_token.stream
|
|
|
|
async def _notify_persisted_event(
|
|
self, event: EventBase, max_stream_token: RoomStreamToken
|
|
) -> None:
|
|
"""Checks to see if notifier/pushers should be notified about the
|
|
event or not.
|
|
|
|
Args:
|
|
event:
|
|
max_stream_id: The max_stream_id returned by persist_events
|
|
"""
|
|
|
|
extra_users = []
|
|
if event.type == EventTypes.Member:
|
|
target_user_id = event.state_key
|
|
|
|
# We notify for memberships if its an invite for one of our
|
|
# users
|
|
if event.internal_metadata.is_outlier():
|
|
if event.membership != Membership.INVITE:
|
|
if not self.is_mine_id(target_user_id):
|
|
return
|
|
|
|
target_user = UserID.from_string(target_user_id)
|
|
extra_users.append(target_user)
|
|
elif event.internal_metadata.is_outlier():
|
|
return
|
|
|
|
# the event has been persisted so it should have a stream ordering.
|
|
assert event.internal_metadata.stream_ordering
|
|
|
|
event_pos = PersistedEventPosition(
|
|
self._instance_name, event.internal_metadata.stream_ordering
|
|
)
|
|
self.notifier.on_new_room_event(
|
|
event, event_pos, max_stream_token, extra_users=extra_users
|
|
)
|
|
|
|
async def _clean_room_for_join(self, room_id: str) -> None:
|
|
"""Called to clean up any data in DB for a given room, ready for the
|
|
server to join the room.
|
|
|
|
Args:
|
|
room_id
|
|
"""
|
|
if self.config.worker_app:
|
|
await self._clean_room_for_join_client(room_id)
|
|
else:
|
|
await self.store.clean_room_for_join(room_id)
|
|
|
|
async def get_room_complexity(
|
|
self, remote_room_hosts: List[str], room_id: str
|
|
) -> Optional[dict]:
|
|
"""
|
|
Fetch the complexity of a remote room over federation.
|
|
|
|
Args:
|
|
remote_room_hosts (list[str]): The remote servers to ask.
|
|
room_id (str): The room ID to ask about.
|
|
|
|
Returns:
|
|
Dict contains the complexity
|
|
metric versions, while None means we could not fetch the complexity.
|
|
"""
|
|
|
|
for host in remote_room_hosts:
|
|
res = await self.federation_client.get_room_complexity(host, room_id)
|
|
|
|
# We got a result, return it.
|
|
if res:
|
|
return res
|
|
|
|
# We fell off the bottom, couldn't get the complexity from anyone. Oh
|
|
# well.
|
|
return None
|