summary refs log tree commit diff
path: root/synapse/handlers/federation_event.py
blob: f640b417b39abdf0e8e9b53d45582544cd8cd51b (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
# Copyright 2021 The Matrix.org Foundation C.I.C.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import logging
from http import HTTPStatus
from typing import (
    TYPE_CHECKING,
    Collection,
    Container,
    Dict,
    Iterable,
    List,
    Optional,
    Sequence,
    Set,
    Tuple,
)

from prometheus_client import Counter

from synapse.api.constants import (
    EventContentFields,
    EventTypes,
    GuestAccess,
    Membership,
    RejectedReason,
    RoomEncryptionAlgorithms,
)
from synapse.api.errors import (
    AuthError,
    Codes,
    FederationError,
    HttpResponseException,
    RequestSendFailed,
    SynapseError,
)
from synapse.api.room_versions import KNOWN_ROOM_VERSIONS
from synapse.event_auth import (
    auth_types_for_event,
    check_auth_rules_for_event,
    validate_event_for_room_version,
)
from synapse.events import EventBase
from synapse.events.snapshot import EventContext
from synapse.federation.federation_client import InvalidResponseError
from synapse.logging.context import nested_logging_context, 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 (
    ReplicationFederationSendEventsRestServlet,
)
from synapse.state import StateResolutionStore
from synapse.storage.databases.main.events_worker import EventRedactBehaviour
from synapse.types import (
    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

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",
)


class FederationEventHandler:
    """Handles events that originated from federation.

    Responsible for handing incoming events and passing them on to the rest
    of the homeserver (including auth and state conflict resolutions)
    """

    def __init__(self, hs: "HomeServer"):
        self._store = hs.get_datastore()
        self._storage = hs.get_storage()
        self._state_store = self._storage.state

        self._state_handler = hs.get_state_handler()
        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._action_generator = hs.get_action_generator()
        self._state_resolution_handler = hs.get_state_resolution_handler()
        # avoid a circular dependency by deferring execution here
        self._get_room_member_handler = hs.get_room_member_handler

        self._federation_client = hs.get_federation_client()
        self._third_party_event_rules = hs.get_third_party_event_rules()
        self._notifier = hs.get_notifier()

        self._is_mine_id = hs.is_mine_id
        self._server_name = hs.hostname
        self._instance_name = hs.get_instance_name()

        self._config = hs.config
        self._ephemeral_messages_enabled = hs.config.server.enable_ephemeral_messages

        self._send_events = ReplicationFederationSendEventsRestServlet.make_client(hs)
        if hs.config.worker.worker_app:
            self._user_device_resync = (
                ReplicationUserDevicesResyncRestServlet.make_client(hs)
            )
        else:
            self._device_list_updater = hs.get_device_handler().device_list_updater

        # When joining a room we need to queue any events for that room up.
        # For each room, a list of (pdu, origin) tuples.
        # TODO: replace this with something more elegant, probably based around the
        # federation event staging area.
        self.room_queues: Dict[str, List[Tuple[EventBase, str]]] = {}

        self._room_pdu_linearizer = Linearizer("fed_room_pdu")

    async def on_receive_pdu(self, origin: str, pdu: EventBase) -> None:
        """Process a PDU received via a federation /send/ transaction

        Args:
            origin: server which initiated the /send/ transaction. Will
                be used to fetch missing events or state.
            pdu: received PDU
        """

        # We should never see any outliers here.
        assert not pdu.internal_metadata.outlier

        room_id = pdu.room_id
        event_id = pdu.event_id

        # 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.
        if existing:
            if not existing.internal_metadata.is_outlier():
                logger.info(
                    "Ignoring received event %s which we have already seen", event_id
                )
                return
            if pdu.internal_metadata.is_outlier():
                logger.info(
                    "Ignoring received outlier %s which we already have as an outlier",
                    event_id,
                )
                return
            logger.info("De-outliering event %s", event_id)

        # 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

        # Try to fetch any missing prev events to fill in gaps in the graph
        prevs = set(pdu.prev_event_ids())
        seen = await self._store.have_events_in_timeline(prevs)
        missing_prevs = prevs - seen

        if missing_prevs:
            # 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)

            if min_depth is not None and pdu.depth > min_depth:
                # 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)
                missing_prevs = prevs - seen

                if not missing_prevs:
                    logger.info("Found all missing prev_events")

            if missing_prevs:
                # since 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. (XXX: apart from
                # min_depth?)
                #
                # We therefore reject any such events.
                logger.warning(
                    "Rejecting: failed to fetch %d prev events: %s",
                    len(missing_prevs),
                    shortstr(missing_prevs),
                )
                raise FederationError(
                    "ERROR",
                    403,
                    (
                        "Your server isn't divulging details about prev_events "
                        "referenced in this event."
                    ),
                    affected=pdu.event_id,
                )

        await self._process_received_pdu(origin, pdu, state=None)

    @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.

        # First, precalculate the joined hosts so that the federation sender doesn't
        # need to.
        await self._event_creation_handler.cache_joined_hosts_for_event(event, context)

        await self._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,
        )

    @log_function
    async def backfill(
        self, dest: str, room_id: str, limit: int, extremities: List[str]
    ) -> None:
        """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.

        We might also raise an InvalidResponseError if the response from the remote
        server is just bogus.

        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

        # if there are any events in the wrong room, the remote server is buggy and
        # should not be trusted.
        for ev in events:
            if ev.room_id != room_id:
                raise InvalidResponseError(
                    f"Remote server {dest} returned event {ev.event_id} which is in "
                    f"room {ev.room_id}, when we were backfilling in {room_id}"
                )

        await self._process_pulled_events(dest, events, backfilled=True)

    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", len(missing_events))
        await self._process_pulled_events(origin, missing_events, backfilled=False)

    async def _process_pulled_events(
        self, origin: str, events: Iterable[EventBase], backfilled: bool
    ) -> None:
        """Process a batch of events we have pulled from a remote server

        Pulls in any events required to auth the events, persists the received events,
        and notifies clients, if appropriate.

        Assumes the events have already had their signatures and hashes checked.

        Params:
            origin: The server we received these events from
            events: The received events.
            backfilled: True if this is part of a historical batch of events (inhibits
                notification to clients, and validation of device keys.)
        """

        # We want to sort these by depth so we process them and
        # tell clients about them in order.
        sorted_events = sorted(events, key=lambda x: x.depth)

        for ev in sorted_events:
            with nested_logging_context(ev.event_id):
                await self._process_pulled_event(origin, ev, backfilled=backfilled)

    async def _process_pulled_event(
        self, origin: str, event: EventBase, backfilled: bool
    ) -> None:
        """Process a single event that we have pulled from a remote server

        Pulls in any events required to auth the event, persists the received event,
        and notifies clients, if appropriate.

        Assumes the event has already had its signatures and hashes checked.

        This is somewhat equivalent to on_receive_pdu, but applies somewhat different
        logic in the case that we are missing prev_events (in particular, it just
        requests the state at that point, rather than triggering a get_missing_events) -
        so is appropriate when we have pulled the event from a remote server, rather
        than having it pushed to us.

        Params:
            origin: The server we received this event from
            events: The received event
            backfilled: True if this is part of a historical batch of events (inhibits
                notification to clients, and validation of device keys.)
        """
        logger.info("Processing pulled event %s", event)

        # these should not be outliers.
        assert not event.internal_metadata.is_outlier()

        event_id = event.event_id

        existing = await self._store.get_event(
            event_id, allow_none=True, allow_rejected=True
        )
        if existing:
            if not existing.internal_metadata.is_outlier():
                logger.info(
                    "Ignoring received event %s which we have already seen",
                    event_id,
                )
                return
            logger.info("De-outliering event %s", event_id)

        try:
            self._sanity_check_event(event)
        except SynapseError as err:
            logger.warning("Event %s failed sanity check: %s", event_id, err)
            return

        try:
            state = await self._resolve_state_at_missing_prevs(origin, event)
            await self._process_received_pdu(
                origin, event, state=state, backfilled=backfilled
            )
        except FederationError as e:
            if e.code == 403:
                logger.warning("Pulled event %s failed history check.", event_id)
            else:
                raise

    async def _resolve_state_at_missing_prevs(
        self, dest: str, event: EventBase
    ) -> Optional[Iterable[EventBase]]:
        """Calculate the state at an event with missing prev_events.

        This is used when we have pulled a batch of events from a remote server, and
        still don't have all the prev_events.

        If we already have all the prev_events for `event`, this method does nothing.

        Otherwise, the missing prevs become new backwards extremities, and we fall back
        to asking the remote server for the state after each missing `prev_event`,
        and resolving across them.

        That's ok provided we then resolve the state against other bits of the DAG
        before using it - in other words, that the received event `event` is not going
        to become the only forwards_extremity in the room (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).

        In other words: we should only call this method if `event` has been *pulled*
        as part of a batch of missing prev events, or similar.

        Params:
            dest: the remote server to ask for state at the missing prevs. Typically,
                this will be the server we got `event` from.
            event: an event to check for missing prevs.

        Returns:
            if we already had all the prev events, `None`. Otherwise, returns a list of
            the events in the state at `event`.
        """
        room_id = event.room_id
        event_id = event.event_id

        prevs = set(event.prev_event_ids())
        seen = await self._store.have_events_in_timeline(prevs)
        missing_prevs = prevs - seen

        if not missing_prevs:
            return None

        logger.info(
            "Event %s is missing prev_events %s: calculating state for a "
            "backwards extremity",
            event_id,
            shortstr(missing_prevs),
        )
        # 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: event}
        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 missing_prevs:
                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(
                        dest, 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,
            )
        return 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
        """
        (
            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, event_ids=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]],
        backfilled: bool = False,
    ) -> None:
        """Called when we have a new non-outlier event.

        This is called when we have a new event to add to the room DAG - either directly
        via a /send request, retrieved via get_missing_events after a /send request, or
        backfilled after a client request.

        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

            backfilled: True if this is part of a historical batch of events (inhibits
                notification to clients, and validation of device keys.)
        """
        logger.debug("Processing event: %s", event)
        assert not event.internal_metadata.outlier

        try:
            context = await self._state_handler.compute_event_context(
                event, old_state=state
            )
            context = await self._check_event_auth(
                origin,
                event,
                context,
                state=state,
                backfilled=backfilled,
            )
        except AuthError as e:
            # FIXME richvdh 2021/10/07 I don't think this is reachable. Let's log it
            #   for now
            logger.exception("Unexpected AuthError from _check_event_auth")
            raise FederationError("ERROR", e.code, e.msg, affected=event.event_id)

        await self._run_push_actions_and_persist_event(event, context, backfilled)

        if backfilled:
            return

        # 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 _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.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)

    async def _handle_marker_event(self, origin: str, marker_event: EventBase) -> None:
        """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 or the event is not from the room creator.
        room_version = await self._store.get_room_version(marker_event.room_id)
        create_event = await self._store.get_create_event_for_room(marker_event.room_id)
        room_creator = create_event.content.get(EventContentFields.ROOM_CREATOR)
        if not room_version.msc2716_historical and (
            not self._config.experimental.msc2716_enabled
            or marker_event.sender != room_creator
        ):
            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 _get_events_and_persist(
        self, destination: str, room_id: str, event_ids: Collection[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 `event_ids` 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)

        events: List[EventBase] = []

        async def get_event(event_id: str) -> None:
            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
                    events.append(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, event_ids, 5)
        logger.info("Fetched %i events of %i requested", len(events), len(event_ids))
        await self._auth_and_persist_fetched_events(destination, room_id, events)

    async def _auth_and_persist_fetched_events(
        self, origin: str, room_id: str, events: Iterable[EventBase]
    ) -> None:
        """Persist the events fetched by _get_events_and_persist or _get_remote_auth_chain_for_event

        The events to be persisted must be outliers.

        We first sort the events to make sure that we process each event's auth_events
        before the event itself, and then auth and persist them.

        Notifies about the events where appropriate.

        Params:
            origin: where the events came from
            room_id: the room that the events are meant to be in (though this has
               not yet been checked)
            events: the events that have been fetched
        """
        event_map = {event.event_id: event for event in events}

        # XXX: it might be possible to kick this process off in parallel with fetching
        # the events.
        while event_map:
            # build a list of events whose auth events are not in the queue.
            roots = tuple(
                ev
                for ev in event_map.values()
                if not any(aid in event_map for aid in ev.auth_event_ids())
            )

            if not roots:
                # if *none* of the remaining events are ready, that means
                # we have a loop. This either means a bug in our logic, or that
                # somebody has managed to create a loop (which requires finding a
                # hash collision in room v2 and later).
                logger.warning(
                    "Loop found in auth events while fetching missing state/auth "
                    "events: %s",
                    shortstr(event_map.keys()),
                )
                return

            logger.info(
                "Persisting %i of %i remaining outliers: %s",
                len(roots),
                len(event_map),
                shortstr(e.event_id for e in roots),
            )

            await self._auth_and_persist_fetched_events_inner(origin, room_id, roots)

            for ev in roots:
                del event_map[ev.event_id]

    async def _auth_and_persist_fetched_events_inner(
        self, origin: str, room_id: str, fetched_events: Collection[EventBase]
    ) -> None:
        """Helper for _auth_and_persist_fetched_events

        Persists a batch of events where we have (theoretically) already persisted all
        of their auth events.

        Notifies about the events where appropriate.

        Params:
            origin: where the events came from
            room_id: the room that the events are meant to be in (though this has
               not yet been checked)
            fetched_events: the events to persist
        """
        # get all the auth events for all the events in this batch. By now, they should
        # have been persisted.
        auth_events = {
            aid for event in fetched_events for aid in event.auth_event_ids()
        }
        persisted_events = await self._store.get_events(
            auth_events,
            allow_rejected=True,
        )

        room_version = await self._store.get_room_version_id(room_id)
        room_version_obj = KNOWN_ROOM_VERSIONS[room_version]

        def prep(event: EventBase) -> Optional[Tuple[EventBase, EventContext]]:
            with nested_logging_context(suffix=event.event_id):
                auth = {}
                for auth_event_id in event.auth_event_ids():
                    ae = persisted_events.get(auth_event_id)
                    if not ae:
                        logger.warning(
                            "Event %s relies on auth_event %s, which could not be found.",
                            event,
                            auth_event_id,
                        )
                        # the fact we can't find the auth event doesn't mean it doesn't
                        # exist, which means it is premature to reject `event`. Instead we
                        # just ignore it for now.
                        return None
                    auth[(ae.type, ae.state_key)] = ae

                context = EventContext.for_outlier()
                try:
                    validate_event_for_room_version(room_version_obj, event)
                    check_auth_rules_for_event(room_version_obj, event, auth)
                except AuthError as e:
                    logger.warning("Rejecting %r because %s", event, e)
                    context.rejected = RejectedReason.AUTH_ERROR

            return event, context

        events_to_persist = (x for x in (prep(event) for event in fetched_events) if x)
        await self.persist_events_and_notify(room_id, tuple(events_to_persist))

    async def _check_event_auth(
        self,
        origin: str,
        event: EventBase,
        context: EventContext,
        state: Optional[Iterable[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.

            state:
                The state events used to check the event for soft-fail. If this is
                not provided the current state events will be used.

            backfilled: True if the event was backfilled.

        Returns:
            The updated context object.
        """
        # This method should only be used for non-outliers
        assert not event.internal_metadata.outlier

        # first of all, check that the event itself is valid.
        room_version = await self._store.get_room_version_id(event.room_id)
        room_version_obj = KNOWN_ROOM_VERSIONS[room_version]

        try:
            validate_event_for_room_version(room_version_obj, event)
        except AuthError as e:
            logger.warning("While validating received event %r: %s", event, e)
            # TODO: use a different rejected reason here?
            context.rejected = RejectedReason.AUTH_ERROR
            return context

        # calculate what the auth events *should* be, to use as a basis for auth.
        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)
        calculated_auth_event_map = {
            (e.type, e.state_key): e for e in auth_events_x.values()
        }

        try:
            (
                context,
                auth_events_for_auth,
            ) = await self._update_auth_events_and_context_for_auth(
                origin,
                event,
                context,
                calculated_auth_event_map=calculated_auth_event_map,
            )
        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,
            )
            auth_events_for_auth = calculated_auth_event_map

        try:
            check_auth_rules_for_event(room_version_obj, event, auth_events_for_auth)
        except AuthError as e:
            logger.warning("Failed auth resolution for %r because %s", event, e)
            context.rejected = RejectedReason.AUTH_ERROR
            return context

        await self._check_for_soft_fail(event, state, backfilled, origin=origin)
        await self._maybe_kick_guest_users(event)

        return context

    async def _maybe_kick_guest_users(self, event: EventBase) -> None:
        if event.type != EventTypes.GuestAccess:
            return

        guest_access = event.content.get(EventContentFields.GUEST_ACCESS)
        if guest_access == GuestAccess.CAN_JOIN:
            return

        current_state_map = await self._state_handler.get_current_state(event.room_id)
        current_state = list(current_state_map.values())
        await self._get_room_member_handler().kick_guest_users(current_state)

    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:
            check_auth_rules_for_event(room_version_obj, event, 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 _update_auth_events_and_context_for_auth(
        self,
        origin: str,
        event: EventBase,
        context: EventContext,
        calculated_auth_event_map: StateMap[EventBase],
    ) -> Tuple[EventContext, StateMap[EventBase]]:
        """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:

            calculated_auth_event_map:
                Our calculated auth_events based on the state of the room
                at the event's position in the DAG.

        Returns:
            updated context, updated auth event map
        """
        assert not event.internal_metadata.outlier

        # take a copy of calculated_auth_event_map before we modify it.
        auth_events: MutableStateMap[EventBase] = dict(calculated_auth_event_map)

        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)

        # missing_auth is now the set of event_ids which:
        #  a. are listed in event.auth_events, *and*
        #  b. are *not* part of our calculated auth events based on room state, *and*
        #  c. are *not* yet in our database.

        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:
                await self._get_remote_auth_chain_for_event(
                    origin, event.room_id, event.event_id
                )
            except Exception:
                logger.exception("Failed to get auth chain")
            else:
                # load any auth events we might have persisted from the database. This
                # has the side-effect of correctly setting the rejected_reason on them.
                auth_events.update(
                    {
                        (ae.type, ae.state_key): ae
                        for ae in await self._store.get_events_as_list(
                            missing_auth, allow_rejected=True
                        )
                    }
                )

        # auth_events now contains
        #   1. our *calculated* auth events based on the room state, plus:
        #   2. any events which:
        #       a. are listed in `event.auth_events`, *and*
        #       b. are not part of our calculated auth events, *and*
        #       c. were not in our database before the call to /event_auth
        #       d. have since been added to our database (most likely by /event_auth).

        different_auth = event_auth_events.difference(
            e.event_id for e in auth_events.values()
        )

        # different_auth is the set of events which *are* in `event.auth_events`, but
        # which are *not* in `auth_events`. Comparing with (2.) above, this means
        # exclusively the set of `event.auth_events` which we already had in our
        # database before any call to /event_auth.
        #
        # I'm reasonably sure that the fact that events returned by /event_auth are
        # blindly added to auth_events (and hence excluded from different_auth) is a bug
        # - though it's a very long-standing one (see
        # https://github.com/matrix-org/synapse/commit/78015948a7febb18e000651f72f8f58830a55b93#diff-0bc92da3d703202f5b9be2d3f845e375f5b1a6bc6ba61705a8af9be1121f5e42R786
        # from Jan 2015 which seems to add it, though it actually just moves it from
        # elsewhere (before that, it gets lost in a mess of huge "various bug fixes"
        # PRs).

        if not different_auth:
            return context, auth_events

        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, auth_events

        # 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, auth_events

    async def _get_remote_auth_chain_for_event(
        self, destination: str, room_id: str, event_id: str
    ) -> None:
        """If we are missing some of an event's auth events, attempt to request them

        Args:
            destination: where to fetch the auth tree from
            room_id: the room in which we are lacking auth events
            event_id: the event for which we are lacking auth events
        """
        try:
            remote_event_map = {
                e.event_id: e
                for e in await self._federation_client.get_event_auth(
                    destination, room_id, 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

        logger.info("/event_auth returned %i events", len(remote_event_map))

        # `event` may be returned, but we should not yet process it.
        remote_event_map.pop(event_id, None)

        # nor should we reprocess any events we have already seen.
        seen_remotes = await self._store.have_seen_events(
            room_id, remote_event_map.keys()
        )
        for s in seen_remotes:
            remote_event_map.pop(s, None)

        await self._auth_and_persist_fetched_events(
            destination, room_id, remote_event_map.values()
        )

    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 _run_push_actions_and_persist_event(
        self, event: EventBase, context: EventContext, backfilled: bool = False
    ) -> None:
        """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.
        """
        # this method should not be called on outliers (those code paths call
        # persist_events_and_notify directly.)
        assert not event.internal_metadata.outlier

        try:
            if (
                not backfilled
                and not context.rejected
                and (await self._store.get_min_depth(event.room_id)) <= event.depth
            ):
                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 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_token: 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
        )

    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 get_min_depth_for_context(self, context: str) -> int:
        return await self._store.get_min_depth(context)