synapse/synapse/handlers/federation.py

#
# This file is licensed under the Affero General Public License (AGPL) version 3.
#
# Copyright 2020 Sorunome
# Copyright 2014-2022 The Matrix.org Foundation C.I.C.
# Copyright (C) 2023 New Vector, Ltd
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License as
# published by the Free Software Foundation, either version 3 of the
# License, or (at your option) any later version.
#
# See the GNU Affero General Public License for more details:
# <https://www.gnu.org/licenses/agpl-3.0.html>.
#
# Originally licensed under the Apache License, Version 2.0:
# <http://www.apache.org/licenses/LICENSE-2.0>.
#
# [This file includes modifications made by New Vector Limited]
#
#

"""Contains handlers for federation events."""

import enum
import itertools
import logging
from enum import Enum
from http import HTTPStatus
from typing import (
    TYPE_CHECKING,
    AbstractSet,
    Dict,
    Iterable,
    List,
    Optional,
    Set,
    Tuple,
    Union,
)

import attr
from prometheus_client import Histogram
from signedjson.key import decode_verify_key_bytes
from signedjson.sign import verify_signed_json
from unpaddedbase64 import decode_base64

from synapse import event_auth
from synapse.api.constants import MAX_DEPTH, EventContentFields, EventTypes, Membership
from synapse.api.errors import (
    AuthError,
    CodeMessageException,
    Codes,
    FederationDeniedError,
    FederationError,
    FederationPullAttemptBackoffError,
    HttpResponseException,
    NotFoundError,
    PartialStateConflictError,
    RequestSendFailed,
    SynapseError,
)
from synapse.api.room_versions import KNOWN_ROOM_VERSIONS, RoomVersion
from synapse.crypto.event_signing import compute_event_signature
from synapse.event_auth import validate_event_for_room_version
from synapse.events import EventBase
from synapse.events.snapshot import EventContext, UnpersistedEventContextBase
from synapse.events.validator import EventValidator
from synapse.federation.federation_client import InvalidResponseError
from synapse.handlers.pagination import PURGE_PAGINATION_LOCK_NAME
from synapse.http.servlet import assert_params_in_dict
from synapse.logging.context import nested_logging_context
from synapse.logging.opentracing import SynapseTags, set_tag, tag_args, trace
from synapse.metrics.background_process_metrics import run_as_background_process
from synapse.module_api import NOT_SPAM
from synapse.replication.http.federation import (
    ReplicationCleanRoomRestServlet,
    ReplicationStoreRoomOnOutlierMembershipRestServlet,
)
from synapse.storage.databases.main.events_worker import EventRedactBehaviour
from synapse.types import JsonDict, StrCollection, get_domain_from_id
from synapse.types.state import StateFilter
from synapse.util.async_helpers import Linearizer
from synapse.util.retryutils import NotRetryingDestination
from synapse.visibility import filter_events_for_server

if TYPE_CHECKING:
    from synapse.server import HomeServer

logger = logging.getLogger(__name__)

# Added to debug performance and track progress on optimizations
backfill_processing_before_timer = Histogram(
    "synapse_federation_backfill_processing_before_time_seconds",
    "sec",
    [],
    buckets=(
        0.1,
        0.5,
        1.0,
        2.5,
        5.0,
        7.5,
        10.0,
        15.0,
        20.0,
        30.0,
        40.0,
        60.0,
        80.0,
        "+Inf",
    ),
)


# TODO: We can refactor this away now that there is only one backfill point again
class _BackfillPointType(Enum):
    # a regular backwards extremity (ie, an event which we don't yet have, but which
    # is referred to by other events in the DAG)
    BACKWARDS_EXTREMITY = enum.auto()


@attr.s(slots=True, auto_attribs=True, frozen=True)
class _BackfillPoint:
    """A potential point we might backfill from"""

    event_id: str
    depth: int
    type: _BackfillPointType


class FederationHandler:
    """Handles general incoming federation requests

    Incoming events are *not* handled here, for which see FederationEventHandler.
    """

    def __init__(self, hs: "HomeServer"):
        self.hs = hs

        self.clock = hs.get_clock()
        self.store = hs.get_datastores().main
        self._storage_controllers = hs.get_storage_controllers()
        self._state_storage_controller = self._storage_controllers.state
        self.federation_client = hs.get_federation_client()
        self.state_handler = hs.get_state_handler()
        self.server_name = hs.hostname
        self.keyring = hs.get_keyring()
        self.is_mine_id = hs.is_mine_id
        self.is_mine_server_name = hs.is_mine_server_name
        self._spam_checker_module_callbacks = hs.get_module_api_callbacks().spam_checker
        self.event_creation_handler = hs.get_event_creation_handler()
        self.event_builder_factory = hs.get_event_builder_factory()
        self._event_auth_handler = hs.get_event_auth_handler()
        self._server_notices_mxid = hs.config.servernotices.server_notices_mxid
        self.config = hs.config
        self.http_client = hs.get_proxied_blocklisted_http_client()
        self._replication = hs.get_replication_data_handler()
        self._federation_event_handler = hs.get_federation_event_handler()
        self._device_handler = hs.get_device_handler()
        self._bulk_push_rule_evaluator = hs.get_bulk_push_rule_evaluator()
        self._notifier = hs.get_notifier()
        self._worker_locks = hs.get_worker_locks_handler()

        self._clean_room_for_join_client = ReplicationCleanRoomRestServlet.make_client(
            hs
        )

        if hs.config.worker.worker_app:
            self._maybe_store_room_on_outlier_membership = (
                ReplicationStoreRoomOnOutlierMembershipRestServlet.make_client(hs)
            )
        else:
            self._maybe_store_room_on_outlier_membership = (
                self.store.maybe_store_room_on_outlier_membership
            )

        self._room_backfill = Linearizer("room_backfill")

        self._third_party_event_rules = (
            hs.get_module_api_callbacks().third_party_event_rules
        )

        # Tracks running partial state syncs by room ID.
        # Partial state syncs currently only run on the main process, so it's okay to
        # track them in-memory for now.
        self._active_partial_state_syncs: Set[str] = set()
        # Tracks partial state syncs we may want to restart.
        # A dictionary mapping room IDs to (initial destination, other destinations)
        # tuples.
        self._partial_state_syncs_maybe_needing_restart: Dict[
            str, Tuple[Optional[str], AbstractSet[str]]
        ] = {}
        # A lock guarding the partial state flag for rooms.
        # When the lock is held for a given room, no other concurrent code may
        # partial state or un-partial state the room.
        self._is_partial_state_room_linearizer = Linearizer(
            name="_is_partial_state_room_linearizer"
        )

        # if this is the main process, fire off a background process to resume
        # any partial-state-resync operations which were in flight when we
        # were shut down.
        if not hs.config.worker.worker_app:
            run_as_background_process(
                "resume_sync_partial_state_room", self._resume_partial_state_room_sync
            )

    @trace
    @tag_args
    async def maybe_backfill(
        self, room_id: str, current_depth: int, limit: int, record_time: bool = True
    ) -> 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.
            record_time: Whether to record the time it takes to backfill.

        Returns:
            True if we actually tried to backfill something, otherwise False.
        """
        # Starting the processing time here so we can include the room backfill
        # linearizer lock queue in the timing
        processing_start_time = self.clock.time_msec() if record_time else 0

        async with self._room_backfill.queue(room_id):
            async with self._worker_locks.acquire_read_write_lock(
                PURGE_PAGINATION_LOCK_NAME, room_id, write=False
            ):
                return await self._maybe_backfill_inner(
                    room_id,
                    current_depth,
                    limit,
                    processing_start_time=processing_start_time,
                )

    @trace
    @tag_args
    async def _maybe_backfill_inner(
        self,
        room_id: str,
        current_depth: int,
        limit: int,
        *,
        processing_start_time: Optional[int],
    ) -> bool:
        """
        Checks whether the `current_depth` is at or approaching any backfill
        points in the room and if so, will backfill. We only care about
        checking backfill points that happened before the `current_depth`
        (meaning less than or equal to the `current_depth`).

        Args:
            room_id: The room to backfill in.
            current_depth: The depth to check at for any upcoming backfill points.
            limit: The max number of events to request from the remote federated server.
            processing_start_time: The time when `maybe_backfill` started processing.
                Only used for timing. If `None`, no timing observation will be made.

        Returns:
            True if we actually tried to backfill something, otherwise False.
        """
        backwards_extremities = [
            _BackfillPoint(event_id, depth, _BackfillPointType.BACKWARDS_EXTREMITY)
            for event_id, depth in await self.store.get_backfill_points_in_room(
                room_id=room_id,
                current_depth=current_depth,
                # We only need to end up with 5 extremities combined with the
                # insertion event extremities to make the `/backfill` request
                # but fetch an order of magnitude more to make sure there is
                # enough even after we filter them by whether visible in the
                # history. This isn't fool-proof as all backfill points within
                # our limit could be filtered out but seems like a good amount
                # to try with at least.
                limit=50,
            )
        ]

        # we now have a list of potential places to backpaginate from. We prefer to
        # start with the most recent (ie, max depth), so let's sort the list.
        sorted_backfill_points: List[_BackfillPoint] = sorted(
            backwards_extremities,
            key=lambda e: -int(e.depth),
        )

        logger.debug(
            "_maybe_backfill_inner: room_id: %s: current_depth: %s, limit: %s, "
            "backfill points (%d): %s",
            room_id,
            current_depth,
            limit,
            len(sorted_backfill_points),
            sorted_backfill_points,
        )
        set_tag(
            SynapseTags.RESULT_PREFIX + "sorted_backfill_points",
            str(sorted_backfill_points),
        )
        set_tag(
            SynapseTags.RESULT_PREFIX + "sorted_backfill_points.length",
            str(len(sorted_backfill_points)),
        )

        # If we have no backfill points lower than the `current_depth` then either we
        # can a) bail or b) still attempt to backfill. We opt to try backfilling anyway
        # just in case we do get relevant events. This is good for eventual consistency
        # sake but we don't need to block the client for something that is just as
        # likely not to return anything relevant so we backfill in the background. The
        # only way, this could return something relevant is if we discover a new branch
        # of history that extends all the way back to where we are currently paginating
        # and it's within the 100 events that are returned from `/backfill`.
        if not sorted_backfill_points and current_depth != MAX_DEPTH:
            # Check that we actually have later backfill points, if not just return.
            have_later_backfill_points = await self.store.get_backfill_points_in_room(
                room_id=room_id,
                current_depth=MAX_DEPTH,
                limit=1,
            )
            if not have_later_backfill_points:
                return False

            logger.debug(
                "_maybe_backfill_inner: all backfill points are *after* current depth. Trying again with later backfill points."
            )
            run_as_background_process(
                "_maybe_backfill_inner_anyway_with_max_depth",
                self.maybe_backfill,
                room_id=room_id,
                # We use `MAX_DEPTH` so that we find all backfill points next
                # time (all events are below the `MAX_DEPTH`)
                current_depth=MAX_DEPTH,
                limit=limit,
                # We don't want to start another timing observation from this
                # nested recursive call. The top-most call can record the time
                # overall otherwise the smaller one will throw off the results.
                record_time=False,
            )
            # We return `False` because we're backfilling in the background and there is
            # no new events immediately for the caller to know about yet.
            return False

        # Even after recursing with `MAX_DEPTH`, we didn't find any
        # backward extremities to backfill from.
        if not sorted_backfill_points:
            logger.debug(
                "_maybe_backfill_inner: Not backfilling as no backward extremeties found."
            )
            return False

        # 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.
        max_depth_of_backfill_points = sorted_backfill_points[0].depth
        if current_depth - 2 * limit > max_depth_of_backfill_points:
            logger.debug(
                "Not backfilling as we don't need to. %d < %d - 2 * %d",
                max_depth_of_backfill_points,
                current_depth,
                limit,
            )
            return False

        # For performance's sake, we only want to paginate from a particular extremity
        # if we can actually see the events we'll get. Otherwise, we'd just spend a lot
        # of resources to get redacted events. We check each extremity in turn and
        # ignore those which users on our server wouldn't be able to see.
        #
        # Additionally, we limit ourselves to backfilling from at most 5 extremities,
        # for two reasons:
        #
        # - The check which determines if we can see an extremity's events can be
        #   expensive (we load the full state for the room at each of the backfill
        #   points, or (worse) their successors)
        # - We want to avoid the server-server API request URI becoming too long.
        #
        # *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) it's
        # relatively rare for there to be any visible history and (b) even when
        # there is it's often sufficiently long ago that clients would stop
        # attempting to paginate before backfill reached the visible history.

        extremities_to_request: List[str] = []
        for bp in sorted_backfill_points:
            if len(extremities_to_request) >= 5:
                break

            # For regular backwards extremities, we don't have the extremity events
            # themselves, so we need to actually check the events that reference them -
            # their "successor" events.
            #
            # TODO: Correctly handle the case where we are allowed to see the
            #   successor 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.
            event_ids_to_check = await self.store.get_successor_events(bp.event_id)

            events_to_check = await self.store.get_events_as_list(
                event_ids_to_check,
                redact_behaviour=EventRedactBehaviour.as_is,
                get_prev_content=False,
            )

            # We unset `filter_out_erased_senders` as we might otherwise get false
            # positives from users having been erased.
            filtered_extremities = await filter_events_for_server(
                self._storage_controllers,
                self.server_name,
                self.server_name,
                events_to_check,
                redact=False,
                filter_out_erased_senders=False,
                filter_out_remote_partial_state_events=False,
            )
            if filtered_extremities:
                extremities_to_request.append(bp.event_id)
            else:
                logger.debug(
                    "_maybe_backfill_inner: skipping extremity %s as it would not be visible",
                    bp,
                )

        if not extremities_to_request:
            logger.debug(
                "_maybe_backfill_inner: found no extremities which would be visible"
            )
            return False

        logger.debug(
            "_maybe_backfill_inner: extremities_to_request %s", extremities_to_request
        )
        set_tag(
            SynapseTags.RESULT_PREFIX + "extremities_to_request",
            str(extremities_to_request),
        )
        set_tag(
            SynapseTags.RESULT_PREFIX + "extremities_to_request.length",
            str(len(extremities_to_request)),
        )

        # Now we need to decide which hosts to hit first.
        # First we try hosts that are already in the room.
        # TODO: HEURISTIC ALERT.
        likely_domains = (
            await self._storage_controllers.state.get_current_hosts_in_room_ordered(
                room_id
            )
        )

        async def try_backfill(domains: StrCollection) -> bool:
            # TODO: Should we try multiple of these at a time?

            # Number of contacted remote homeservers that have denied our backfill
            # request with a 4xx code.
            denied_count = 0

            # Maximum number of contacted remote homeservers that can deny our
            # backfill request with 4xx codes before we give up.
            max_denied_count = 5

            for dom in domains:
                # We don't want to ask our own server for information we don't have
                if self.is_mine_server_name(dom):
                    continue

                try:
                    await self._federation_event_handler.backfill(
                        dom, room_id, limit=100, extremities=extremities_to_request
                    )
                    # If this succeeded then we probably already have the
                    # appropriate stuff.
                    # TODO: We can probably do something more intelligent here.
                    return True
                except NotRetryingDestination as e:
                    logger.info("_maybe_backfill_inner: %s", e)
                    continue
                except FederationDeniedError:
                    logger.info(
                        "_maybe_backfill_inner: Not attempting to backfill from %s because the homeserver is not on our federation whitelist",
                        dom,
                    )
                    continue
                except (SynapseError, InvalidResponseError) as e:
                    logger.info("Failed to backfill from %s because %s", dom, e)
                    continue
                except HttpResponseException as e:
                    if 400 <= e.code < 500:
                        logger.warning(
                            "Backfill denied from %s because %s [%d/%d]",
                            dom,
                            e,
                            denied_count,
                            max_denied_count,
                        )
                        denied_count += 1
                        if denied_count >= max_denied_count:
                            return False
                        continue

                    logger.info("Failed to backfill from %s because %s", dom, e)
                    continue
                except CodeMessageException as e:
                    if 400 <= e.code < 500:
                        logger.warning(
                            "Backfill denied from %s because %s [%d/%d]",
                            dom,
                            e,
                            denied_count,
                            max_denied_count,
                        )
                        denied_count += 1
                        if denied_count >= max_denied_count:
                            return False
                        continue

                    logger.info("Failed to backfill from %s because %s", dom, e)
                    continue
                except RequestSendFailed as e:
                    logger.info("Failed to get backfill from %s because %s", dom, e)
                    continue
                except Exception as e:
                    logger.exception("Failed to backfill from %s because %s", dom, e)
                    continue

            return False

        # If we have the `processing_start_time`, then we can make an
        # observation. We wouldn't have the `processing_start_time` in the case
        # where `_maybe_backfill_inner` is recursively called to find any
        # backfill points regardless of `current_depth`.
        if processing_start_time is not None:
            processing_end_time = self.clock.time_msec()
            backfill_processing_before_timer.observe(
                (processing_end_time - processing_start_time) / 1000
            )

        success = await try_backfill(likely_domains)
        if success:
            return True

        # TODO: we could also try servers which were previously in the room, but
        #   are no longer.

        return False

    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.
        # TODO: Before we relax this condition, we need to allow re-syncing of
        # partial room state to happen 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._federation_event_handler.room_queues

        self._federation_event_handler.room_queues[room_id] = []

        is_host_joined = await self.store.is_host_joined(room_id, self.server_name)

        if not is_host_joined:
            # We may have old forward extremities lying around if the homeserver left
            # the room completely in the past. Clear them out.
            #
            # Note that this check-then-clear is subject to races where
            #  * the homeserver is in the room and stops being in the room just after
            #    the check. We won't reset the forward extremities, but that's okay,
            #    since they will be almost up to date.
            #  * the homeserver is not in the room and starts being in the room just
            #    after the check. This can't happen, since `RoomMemberHandler` has a
            #    linearizer lock which prevents concurrent remote joins into the same
            #    room.
            # In short, the races either have an acceptable outcome or should be
            # impossible.
            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

            async with self._is_partial_state_room_linearizer.queue(room_id):
                already_partial_state_room = await self.store.is_partial_state_room(
                    room_id
                )

                ret = await self.federation_client.send_join(
                    host_list,
                    event,
                    room_version_obj,
                    # Perform a full join when we are already in the room and it is a
                    # full state room, since we are not allowed to persist a partial
                    # state join event in a full state room. In the future, we could
                    # optimize this by always performing a partial state join and
                    # computing the state ourselves or retrieving it from the remote
                    # homeserver if necessary.
                    #
                    # There's a race where we leave the room, then perform a full join
                    # anyway. This should end up being fast anyway, since we would
                    # already have the full room state and auth chain persisted.
                    partial_state=not is_host_joined or already_partial_state_room,
                )

                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,
                    state_events=state,
                )

                if ret.partial_state and not already_partial_state_room:
                    # Mark the room as having partial state.
                    # The background process is responsible for unmarking this flag,
                    # even if the join fails.
                    # TODO(faster_joins):
                    #     We may want to reset the partial state info if it's from an
                    #     old, failed partial state join.
                    #     https://github.com/matrix-org/synapse/issues/13000
                    await self.store.store_partial_state_room(
                        room_id=room_id,
                        servers=ret.servers_in_room,
                        device_lists_stream_id=self.store.get_device_stream_token(),
                        joined_via=origin,
                    )

                try:
                    max_stream_id = (
                        await self._federation_event_handler.process_remote_join(
                            origin,
                            room_id,
                            auth_chain,
                            state,
                            event,
                            room_version_obj,
                            partial_state=ret.partial_state,
                        )
                    )
                except PartialStateConflictError:
                    # This should be impossible, since we hold the lock on the room's
                    # partial statedness.
                    logger.error(
                        "Room %s was un-partial stated while processing remote join.",
                        room_id,
                    )
                    raise
                else:
                    # Record the join event id for future use (when we finish the full
                    # join). We have to do this after persisting the event to keep
                    # foreign key constraints intact.
                    if ret.partial_state and not already_partial_state_room:
                        # TODO(faster_joins):
                        #     We may want to reset the partial state info if it's from
                        #     an old, failed partial state join.
                        #     https://github.com/matrix-org/synapse/issues/13000
                        await self.store.write_partial_state_rooms_join_event_id(
                            room_id, event.event_id
                        )
                finally:
                    # Always kick off the background process that asynchronously fetches
                    # state for the room.
                    # If the join failed, the background process is responsible for
                    # cleaning up — including unmarking the room as a partial state
                    # room.
                    if ret.partial_state:
                        # Kick off the process of asynchronously fetching the state for
                        # this room.
                        self._start_partial_state_room_sync(
                            initial_destination=origin,
                            other_destinations=ret.servers_in_room,
                            room_id=room_id,
                        )

            # 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._federation_event_handler.room_queues[room_id]
            del self._federation_event_handler.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_as_background_process(
                "handle_queued_pdus", self._handle_queued_pdus, room_queue
            )

    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
        )

        # Mark the knock as an outlier as we don't yet have the state at this point in
        # the DAG.
        event.internal_metadata.outlier = True

        # ... but tell /sync to send it to clients anyway.
        event.internal_metadata.out_of_band_membership = True

        # 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
        knock_response = await self.federation_client.send_knock(target_hosts, event)

        # Store any stripped room state events in the "unsigned" key of the event.
        # This is a bit of a hack and is cribbing off of invites. Basically we
        # store the room state here and retrieve it again when this event appears
        # in the invitee's sync stream. It is stripped out for all other local users.
        stripped_room_state = knock_response.get("knock_room_state")

        if stripped_room_state is None:
            raise KeyError("Missing 'knock_room_state' field in send_knock response")

        event.unsigned["knock_room_state"] = stripped_room_state

        context = EventContext.for_outlier(self._storage_controllers)
        stream_id = await self._federation_event_handler.persist_events_and_notify(
            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",
                    p,
                )
                with nested_logging_context(p.event_id):
                    await self._federation_event_handler.on_receive_pdu(origin, p)
            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)

        if await self.store.is_partial_state_room(room_id):
            # If our server is still only partially joined, we can't give a complete
            # response to /make_join, so return a 404 as we would if we weren't in the
            # room at all.
            # The main reason we can't respond properly is that we need to know about
            # the auth events for the join event that we would return.
            # We also should not bother entertaining the /make_join since we cannot
            # handle the /send_join.
            logger.info(
                "Rejecting /make_join to %s because it's a partial state room", room_id
            )
            raise SynapseError(
                404,
                "Unable to handle /make_join right now; this server is not fully joined.",
                errcode=Codes.NOT_FOUND,
            )

        # now check that we are *still* in the room
        is_in_room = await self._event_auth_handler.is_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.
        prev_event_ids = None
        if room_version.restricted_join_rule:
            # Note that the room's state can change out from under us and render our
            # nice join rules-conformant event non-conformant by the time we build the
            # event. When this happens, our validation at the end fails and we respond
            # to the requesting server with a 403, which is misleading — it indicates
            # that the user is not allowed to join the room and the joining server
            # should not bother retrying via this homeserver or any others, when
            # in fact we've just messed up with building the event.
            #
            # To reduce the likelihood of this race, we capture the forward extremities
            # of the room (prev_event_ids) just before fetching the current state, and
            # hope that the state we fetch corresponds to the prev events we chose.
            prev_event_ids = await self.store.get_prev_events_for_room(room_id)
            state_ids = await self._state_storage_controller.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[EventContentFields.AUTHORISING_USER] = (
                        await self._event_auth_handler.get_user_which_could_invite(
                            room_id,
                            state_ids,
                        )
                    )

        builder = self.event_builder_factory.for_room_version(
            room_version,
            {
                "type": EventTypes.Member,
                "content": event_content,
                "room_id": room_id,
                "sender": user_id,
                "state_key": user_id,
            },
        )

        try:
            (
                event,
                unpersisted_context,
            ) = await self.event_creation_handler.create_new_client_event(
                builder=builder,
                prev_event_ids=prev_event_ids,
            )
        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._federation_event_handler.check_join_restrictions(
            unpersisted_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_auth_rules_from_context(event)
        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.server.block_non_admin_invites:
            raise SynapseError(403, "This server does not accept room invites")

        spam_check = await self._spam_checker_module_callbacks.user_may_invite(
            event.sender, event.state_key, event.room_id
        )
        if spam_check != NOT_SPAM:
            raise SynapseError(
                403,
                "This user is not permitted to send invites to this server/user",
                errcode=spam_check[0],
                additional_fields=spam_check[1],
            )

        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 = EventContext.for_outlier(self._storage_controllers)

        await self._bulk_push_rule_evaluator.action_for_events_by_user(
            [(event, context)]
        )
        try:
            await self._federation_event_handler.persist_events_and_notify(
                event.room_id, [(event, context)]
            )
        except Exception:
            await self.store.remove_push_actions_from_staging(event.event_id)
            raise

        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 = EventContext.for_outlier(self._storage_controllers)
        stream_id = await self._federation_event_handler.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_obj = await self.store.get_room_version(room_id)
        builder = self.event_builder_factory.for_room_version(
            room_version_obj,
            {
                "type": EventTypes.Member,
                "content": {"membership": Membership.LEAVE},
                "room_id": room_id,
                "sender": user_id,
                "state_key": user_id,
            },
        )

        event, _ = 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_auth_rules_from_context(event)
        except AuthError as e:
            logger.warning("Failed to create new leave %r because %s", event, e)
            raise e

        return event

    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_obj = await self.store.get_room_version(room_id)

        builder = self.event_builder_factory.for_room_version(
            room_version_obj,
            {
                "type": EventTypes.Member,
                "content": {"membership": Membership.KNOCK},
                "room_id": room_id,
                "sender": user_id,
                "state_key": user_id,
            },
        )

        (
            event,
            unpersisted_context,
        ) = await self.event_creation_handler.create_new_client_event(builder=builder)

        event_allowed, _ = await self._third_party_event_rules.check_event_allowed(
            event, unpersisted_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_auth_rules_from_context(event)
        except AuthError as e:
            logger.warning("Failed to create new knock %r because %s", event, e)
            raise e

        return event

    @trace
    @tag_args
    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)
        if event.internal_metadata.outlier:
            raise NotFoundError("State not known at event %s" % (event_id,))

        state_groups = await self._state_storage_controller.get_state_groups_ids(
            room_id, [event_id]
        )

        # get_state_groups_ids should return exactly one result
        assert len(state_groups) == 1

        state_map = next(iter(state_groups.values()))

        state_key = event.get_state_key()
        if state_key is not None:
            # the event was not rejected (get_event raises a NotFoundError for rejected
            # events) so the state at the event should include the event itself.
            assert (
                state_map.get((event.type, state_key)) == event.event_id
            ), "State at event did not include event itself"

            # ... but we need the state *before* that event
            if "replaces_state" in event.unsigned:
                prev_id = event.unsigned["replaces_state"]
                state_map[(event.type, state_key)] = prev_id
            else:
                del state_map[(event.type, state_key)]

        return list(state_map.values())

    async def on_backfill_request(
        self, origin: str, room_id: str, pdu_list: List[str], limit: int
    ) -> List[EventBase]:
        # We allow partially joined rooms since in this case we are filtering out
        # non-local events in `filter_events_for_server`.
        await self._event_auth_handler.assert_host_in_room(room_id, origin, True)

        # Synapse asks for 100 events per backfill request. Do not allow more.
        limit = min(limit, 100)

        events = await self.store.get_backfill_events(room_id, pdu_list, limit)
        logger.debug(
            "on_backfill_request: backfill events=%s",
            [
                "event_id=%s,depth=%d,body=%s,prevs=%s\n"
                % (
                    event.event_id,
                    event.depth,
                    event.content.get("body", event.type),
                    event.prev_event_ids(),
                )
                for event in events
            ],
        )

        events = await filter_events_for_server(
            self._storage_controllers,
            origin,
            self.server_name,
            events,
            redact=True,
            filter_out_erased_senders=True,
            filter_out_remote_partial_state_events=True,
        )

        return events

    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 not event:
            return None

        await self._event_auth_handler.assert_host_in_room(event.room_id, origin)

        events = await filter_events_for_server(
            self._storage_controllers,
            origin,
            self.server_name,
            [event],
            redact=True,
            filter_out_erased_senders=True,
            filter_out_remote_partial_state_events=True,
        )
        event = events[0]
        return event

    async def on_get_missing_events(
        self,
        origin: str,
        room_id: str,
        earliest_events: List[str],
        latest_events: List[str],
        limit: int,
    ) -> List[EventBase]:
        # We allow partially joined rooms since in this case we are filtering out
        # non-local events in `filter_events_for_server`.
        await self._event_auth_handler.assert_host_in_room(room_id, origin, True)

        # Only allow up to 20 events to be retrieved per request.
        limit = min(limit, 20)

        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_controllers,
            origin,
            self.server_name,
            missing_events,
            redact=True,
            filter_out_erased_senders=True,
            filter_out_remote_partial_state_events=True,
        )

        return missing_events

    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.is_host_in_room(room_id, self.hs.hostname):
            room_version_obj = await self.store.get_room_version(room_id)
            builder = self.event_builder_factory.for_room_version(
                room_version_obj, event_dict
            )

            EventValidator().validate_builder(builder, self.hs.config)

            # Try several times, it could fail with PartialStateConflictError
            # in send_membership_event, cf comment in except block.
            max_retries = 5
            for i in range(max_retries):
                try:
                    (
                        event,
                        unpersisted_context,
                    ) = await self.event_creation_handler.create_new_client_event(
                        builder=builder
                    )

                    (
                        event,
                        unpersisted_context,
                    ) = await self.add_display_name_to_third_party_invite(
                        room_version_obj, event_dict, event, unpersisted_context
                    )

                    context = await unpersisted_context.persist(event)

                    EventValidator().validate_new(event, self.config)

                    # We need to tell the transaction queue to send this out, even
                    # though the sender isn't a local user.
                    event.internal_metadata.send_on_behalf_of = self.hs.hostname

                    try:
                        validate_event_for_room_version(event)
                        await self._event_auth_handler.check_auth_rules_from_context(
                            event
                        )
                    except AuthError as e:
                        logger.warning(
                            "Denying new third party invite %r because %s", event, e
                        )
                        raise e

                    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)

                    break
                except PartialStateConflictError as e:
                    # Persisting couldn't happen because the room got un-partial stated
                    # in the meantime and context needs to be recomputed, so let's do so.
                    if i == max_retries - 1:
                        raise e
        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_obj = await self.store.get_room_version(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.for_room_version(
            room_version_obj, event_dict
        )

        # Try several times, it could fail with PartialStateConflictError
        # in send_membership_event, cf comment in except block.
        max_retries = 5
        for i in range(max_retries):
            try:
                (
                    event,
                    unpersisted_context,
                ) = await self.event_creation_handler.create_new_client_event(
                    builder=builder
                )
                (
                    event,
                    unpersisted_context,
                ) = await self.add_display_name_to_third_party_invite(
                    room_version_obj, event_dict, event, unpersisted_context
                )

                context = await unpersisted_context.persist(event)

                try:
                    validate_event_for_room_version(event)
                    await self._event_auth_handler.check_auth_rules_from_context(event)
                except AuthError as e:
                    logger.warning("Denying third party invite %r because %s", event, e)
                    raise e
                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)

                break
            except PartialStateConflictError as e:
                # Persisting couldn't happen because the room got un-partial stated
                # in the meantime and context needs to be recomputed, so let's do so.
                if i == max_retries - 1:
                    raise e

    async def add_display_name_to_third_party_invite(
        self,
        room_version_obj: RoomVersion,
        event_dict: JsonDict,
        event: EventBase,
        context: UnpersistedEventContextBase,
    ) -> Tuple[EventBase, UnpersistedEventContextBase]:
        key = (
            EventTypes.ThirdPartyInvite,
            event.content["third_party_invite"]["signed"]["token"],
        )
        original_invite = None
        prev_state_ids = await context.get_prev_state_ids(StateFilter.from_types([key]))
        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.for_room_version(
            room_version_obj, event_dict
        )
        EventValidator().validate_builder(builder, self.hs.config)

        (
            event,
            unpersisted_context,
        ) = await self.event_creation_handler.create_new_client_event(builder=builder)

        EventValidator().validate_new(event, self.config)
        return event, unpersisted_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(
            StateFilter.from_types([(EventTypes.ThirdPartyInvite, token)])
        )
        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 _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.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: The remote servers to ask.
            room_id: 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

    async def _resume_partial_state_room_sync(self) -> None:
        """Resumes resyncing of all partial-state rooms after a restart."""
        assert not self.config.worker.worker_app

        partial_state_rooms = await self.store.get_partial_state_room_resync_info()
        for room_id, resync_info in partial_state_rooms.items():
            self._start_partial_state_room_sync(
                initial_destination=resync_info.joined_via,
                other_destinations=resync_info.servers_in_room,
                room_id=room_id,
            )

    def _start_partial_state_room_sync(
        self,
        initial_destination: Optional[str],
        other_destinations: AbstractSet[str],
        room_id: str,
    ) -> None:
        """Starts the background process to resync the state of a partial state room,
        if it is not already running.

        Args:
            initial_destination: the initial homeserver to pull the state from
            other_destinations: other homeservers to try to pull the state from, if
                `initial_destination` is unavailable
            room_id: room to be resynced
        """

        async def _sync_partial_state_room_wrapper() -> None:
            if room_id in self._active_partial_state_syncs:
                # Another local user has joined the room while there is already a
                # partial state sync running. This implies that there is a new join
                # event to un-partial state. We might find ourselves in one of a few
                # scenarios:
                #  1. There is an existing partial state sync. The partial state sync
                #     un-partial states the new join event before completing and all is
                #     well.
                #  2. Before the latest join, the homeserver was no longer in the room
                #     and there is an existing partial state sync from our previous
                #     membership of the room. The partial state sync may have:
                #      a) succeeded, but not yet terminated. The room will not be
                #         un-partial stated again unless we restart the partial state
                #         sync.
                #      b) failed, because we were no longer in the room and remote
                #         homeservers were refusing our requests, but not yet
                #         terminated. After the latest join, remote homeservers may
                #         start answering our requests again, so we should restart the
                #         partial state sync.
                # In the cases where we would want to restart the partial state sync,
                # the room would have the partial state flag when the partial state sync
                # terminates.
                self._partial_state_syncs_maybe_needing_restart[room_id] = (
                    initial_destination,
                    other_destinations,
                )
                return

            self._active_partial_state_syncs.add(room_id)

            try:
                await self._sync_partial_state_room(
                    initial_destination=initial_destination,
                    other_destinations=other_destinations,
                    room_id=room_id,
                )
            finally:
                # Read the room's partial state flag while we still hold the claim to
                # being the active partial state sync (so that another partial state
                # sync can't come along and mess with it under us).
                # Normally, the partial state flag will be gone. If it isn't, then we
                # may find ourselves in scenario 2a or 2b as described in the comment
                # above, where we want to restart the partial state sync.
                is_still_partial_state_room = await self.store.is_partial_state_room(
                    room_id
                )
                self._active_partial_state_syncs.remove(room_id)

                if room_id in self._partial_state_syncs_maybe_needing_restart:
                    (
                        restart_initial_destination,
                        restart_other_destinations,
                    ) = self._partial_state_syncs_maybe_needing_restart.pop(room_id)

                    if is_still_partial_state_room:
                        self._start_partial_state_room_sync(
                            initial_destination=restart_initial_destination,
                            other_destinations=restart_other_destinations,
                            room_id=room_id,
                        )

        run_as_background_process(
            desc="sync_partial_state_room", func=_sync_partial_state_room_wrapper
        )

    async def _sync_partial_state_room(
        self,
        initial_destination: Optional[str],
        other_destinations: AbstractSet[str],
        room_id: str,
    ) -> None:
        """Background process to resync the state of a partial-state room

        Args:
            initial_destination: the initial homeserver to pull the state from
            other_destinations: other homeservers to try to pull the state from, if
                `initial_destination` is unavailable
            room_id: room to be resynced
        """
        # Assume that we run on the main process for now.
        # TODO(faster_joins,multiple workers)
        # When moving the sync to workers, we need to ensure that
        #  * `_start_partial_state_room_sync` still prevents duplicate resyncs
        #  * `_is_partial_state_room_linearizer` correctly guards partial state flags
        #    for rooms between the workers doing remote joins and resync.
        assert not self.config.worker.worker_app

        # TODO(faster_joins): do we need to lock to avoid races? What happens if other
        #   worker processes kick off a resync in parallel? Perhaps we should just elect
        #   a single worker to do the resync.
        #   https://github.com/matrix-org/synapse/issues/12994
        #
        # TODO(faster_joins): what happens if we leave the room during a resync? if we
        #   really leave, that might mean we have difficulty getting the room state over
        #   federation.
        #   https://github.com/matrix-org/synapse/issues/12802

        # Make an infinite iterator of destinations to try. Once we find a working
        # destination, we'll stick with it until it flakes.
        destinations = _prioritise_destinations_for_partial_state_resync(
            initial_destination, other_destinations, room_id
        )
        destination_iter = itertools.cycle(destinations)

        # `destination` is the current remote homeserver we're pulling from.
        destination = next(destination_iter)
        logger.info("Syncing state for room %s via %s", room_id, destination)

        # we work through the queue in order of increasing stream ordering.
        while True:
            batch = await self.store.get_partial_state_events_batch(room_id)
            if not batch:
                # all the events are updated, so we can update current state and
                # clear the lazy-loading flag.
                logger.info("Updating current state for %s", room_id)
                # TODO(faster_joins): notify workers in notify_room_un_partial_stated
                #   https://github.com/matrix-org/synapse/issues/12994
                #
                # NB: there's a potential race here. If room is purged just before we
                # call this, we _might_ end up inserting rows into current_state_events.
                # (The logic is hard to chase through.) We think this is fine, but if
                # not the HS admin should purge the room again.
                await self.state_handler.update_current_state(room_id)

                logger.info("Handling any pending device list updates")
                await self._device_handler.handle_room_un_partial_stated(room_id)

                async with self._is_partial_state_room_linearizer.queue(room_id):
                    logger.info("Clearing partial-state flag for %s", room_id)
                    new_stream_id = await self.store.clear_partial_state_room(room_id)

                if new_stream_id is not None:
                    logger.info("State resync complete for %s", room_id)
                    self._storage_controllers.state.notify_room_un_partial_stated(
                        room_id
                    )

                    await self._notifier.on_un_partial_stated_room(
                        room_id, new_stream_id
                    )
                    return

                # we raced against more events arriving with partial state. Go round
                # the loop again. We've already logged a warning, so no need for more.
                continue

            events = await self.store.get_events_as_list(
                batch,
                redact_behaviour=EventRedactBehaviour.as_is,
                allow_rejected=True,
            )
            for event in events:
                for attempt in itertools.count():
                    # We try a new destination on every iteration.
                    try:
                        while True:
                            try:
                                await self._federation_event_handler.update_state_for_partial_state_event(
                                    destination, event
                                )
                                break
                            except FederationPullAttemptBackoffError as e:
                                # We are in the backoff period for one of the event's
                                # prev_events. Wait it out and try again after.
                                logger.warning(
                                    "%s; waiting for %d ms...", e, e.retry_after_ms
                                )
                                await self.clock.sleep(e.retry_after_ms / 1000)

                        # Success, no need to try the rest of the destinations.
                        break
                    except FederationError as e:
                        if attempt == len(destinations) - 1:
                            # We have tried every remote server for this event. Give up.
                            # TODO(faster_joins) giving up isn't the right thing to do
                            #   if there's a temporary network outage. retrying
                            #   indefinitely is also not the right thing to do if we can
                            #   reach all homeservers and they all claim they don't have
                            #   the state we want.
                            #   https://github.com/matrix-org/synapse/issues/13000
                            logger.error(
                                "Failed to get state for %s at %s from %s because %s, "
                                "giving up!",
                                room_id,
                                event,
                                destination,
                                e,
                            )
                            # TODO: We should `record_event_failed_pull_attempt` here,
                            #   see https://github.com/matrix-org/synapse/issues/13700
                            raise

                        # Try the next remote server.
                        logger.info(
                            "Failed to get state for %s at %s from %s because %s",
                            room_id,
                            event,
                            destination,
                            e,
                        )
                        destination = next(destination_iter)
                        logger.info(
                            "Syncing state for room %s via %s instead",
                            room_id,
                            destination,
                        )


def _prioritise_destinations_for_partial_state_resync(
    initial_destination: Optional[str],
    other_destinations: AbstractSet[str],
    room_id: str,
) -> StrCollection:
    """Work out the order in which we should ask servers to resync events.

    If an `initial_destination` is given, it takes top priority. Otherwise
    all servers are treated equally.

    :raises ValueError: if no destination is provided at all.
    """
    if initial_destination is None and len(other_destinations) == 0:
        raise ValueError(f"Cannot resync state of {room_id}: no destinations provided")

    if initial_destination is None:
        return other_destinations

    # Move `initial_destination` to the front of the list.
    destinations = list(other_destinations)
    if initial_destination in destinations:
        destinations.remove(initial_destination)
    destinations = [initial_destination] + destinations
    return destinations