/
service_entry_point_common.cpp
1379 lines (1174 loc) · 59.2 KB
/
service_entry_point_common.cpp
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
/**
* Copyright (C) 2018-present MongoDB, Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the Server Side Public License, version 1,
* as published by MongoDB, Inc.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* Server Side Public License for more details.
*
* You should have received a copy of the Server Side Public License
* along with this program. If not, see
* <http://www.mongodb.com/licensing/server-side-public-license>.
*
* As a special exception, the copyright holders give permission to link the
* code of portions of this program with the OpenSSL library under certain
* conditions as described in each individual source file and distribute
* linked combinations including the program with the OpenSSL library. You
* must comply with the Server Side Public License in all respects for
* all of the code used other than as permitted herein. If you modify file(s)
* with this exception, you may extend this exception to your version of the
* file(s), but you are not obligated to do so. If you do not wish to do so,
* delete this exception statement from your version. If you delete this
* exception statement from all source files in the program, then also delete
* it in the license file.
*/
#define MONGO_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kCommand
#include "mongo/platform/basic.h"
#include "mongo/db/service_entry_point_common.h"
#include "mongo/base/checked_cast.h"
#include "mongo/bson/mutable/document.h"
#include "mongo/bson/util/bson_extract.h"
#include "mongo/db/audit.h"
#include "mongo/db/auth/authorization_session.h"
#include "mongo/db/auth/impersonation_session.h"
#include "mongo/db/client.h"
#include "mongo/db/command_can_run_here.h"
#include "mongo/db/commands.h"
#include "mongo/db/commands/test_commands_enabled.h"
#include "mongo/db/concurrency/global_lock_acquisition_tracker.h"
#include "mongo/db/curop.h"
#include "mongo/db/curop_metrics.h"
#include "mongo/db/cursor_manager.h"
#include "mongo/db/dbdirectclient.h"
#include "mongo/db/initialize_operation_session_info.h"
#include "mongo/db/introspect.h"
#include "mongo/db/jsobj.h"
#include "mongo/db/lasterror.h"
#include "mongo/db/logical_clock.h"
#include "mongo/db/logical_session_id.h"
#include "mongo/db/logical_session_id_helpers.h"
#include "mongo/db/logical_time_validator.h"
#include "mongo/db/ops/write_ops.h"
#include "mongo/db/ops/write_ops_exec.h"
#include "mongo/db/query/find.h"
#include "mongo/db/read_concern.h"
#include "mongo/db/repl/optime.h"
#include "mongo/db/repl/read_concern_args.h"
#include "mongo/db/repl/repl_client_info.h"
#include "mongo/db/repl/replication_coordinator.h"
#include "mongo/db/s/operation_sharding_state.h"
#include "mongo/db/s/sharded_connection_info.h"
#include "mongo/db/s/sharding_state.h"
#include "mongo/db/service_entry_point_common.h"
#include "mongo/db/session_catalog.h"
#include "mongo/db/stats/counters.h"
#include "mongo/db/stats/top.h"
#include "mongo/rpc/factory.h"
#include "mongo/rpc/get_status_from_command_result.h"
#include "mongo/rpc/message.h"
#include "mongo/rpc/metadata.h"
#include "mongo/rpc/metadata/config_server_metadata.h"
#include "mongo/rpc/metadata/logical_time_metadata.h"
#include "mongo/rpc/metadata/oplog_query_metadata.h"
#include "mongo/rpc/metadata/repl_set_metadata.h"
#include "mongo/rpc/metadata/sharding_metadata.h"
#include "mongo/rpc/metadata/tracking_metadata.h"
#include "mongo/rpc/op_msg.h"
#include "mongo/rpc/reply_builder_interface.h"
#include "mongo/s/grid.h"
#include "mongo/util/fail_point_service.h"
#include "mongo/util/log.h"
#include "mongo/util/scopeguard.h"
namespace mongo {
MONGO_FAIL_POINT_DEFINE(failCommand);
MONGO_FAIL_POINT_DEFINE(rsStopGetMore);
MONGO_FAIL_POINT_DEFINE(respondWithNotPrimaryInCommandDispatch);
MONGO_FAIL_POINT_DEFINE(skipCheckingForNotMasterInCommandDispatch);
namespace {
using logger::LogComponent;
// The command names for which to check out a session. These are commands that support retryable
// writes, readConcern snapshot, or multi-statement transactions. We additionally check out the
// session for commands that can take a lock and then run another whitelisted command in
// DBDirectClient. Otherwise, the nested command would try to check out a session under a lock,
// which is not allowed.
const StringMap<int> sessionCheckoutWhitelist = {{"abortTransaction", 1},
{"aggregate", 1},
{"applyOps", 1},
{"commitTransaction", 1},
{"count", 1},
{"dbHash", 1},
{"delete", 1},
{"distinct", 1},
{"doTxn", 1},
{"eval", 1},
{"$eval", 1},
{"explain", 1},
{"filemd5", 1},
{"find", 1},
{"findandmodify", 1},
{"findAndModify", 1},
{"geoNear", 1},
{"geoSearch", 1},
{"getMore", 1},
{"group", 1},
{"insert", 1},
{"killCursors", 1},
{"mapReduce", 1},
{"parallelCollectionScan", 1},
{"prepareTransaction", 1},
{"refreshLogicalSessionCacheNow", 1},
{"update", 1}};
bool shouldActivateFailCommandFailPoint(const BSONObj& data, StringData cmdName) {
if (cmdName == "configureFailPoint"_sd) // Banned even if in failCommands.
return false;
for (auto&& failCommand : data.getObjectField("failCommands")) {
if (failCommand.type() == String && failCommand.valueStringData() == cmdName) {
return true;
}
}
return false;
}
void generateLegacyQueryErrorResponse(const AssertionException& exception,
const QueryMessage& queryMessage,
CurOp* curop,
Message* response) {
curop->debug().errInfo = exception.toStatus();
log(LogComponent::kQuery) << "assertion " << exception.toString() << " ns:" << queryMessage.ns
<< " query:" << (queryMessage.query.valid(BSONVersion::kLatest)
? redact(queryMessage.query)
: "query object is corrupt");
if (queryMessage.ntoskip || queryMessage.ntoreturn) {
log(LogComponent::kQuery) << " ntoskip:" << queryMessage.ntoskip
<< " ntoreturn:" << queryMessage.ntoreturn;
}
BSONObjBuilder err;
err.append("$err", exception.reason());
err.append("code", exception.code());
err.append("ok", 0.0);
auto const extraInfo = exception.extraInfo();
if (extraInfo) {
extraInfo->serialize(&err);
}
BSONObj errObj = err.done();
const bool isStaleConfig = exception.code() == ErrorCodes::StaleConfig;
if (isStaleConfig) {
log(LogComponent::kQuery) << "stale version detected during query over " << queryMessage.ns
<< " : " << errObj;
}
BufBuilder bb;
bb.skip(sizeof(QueryResult::Value));
bb.appendBuf((void*)errObj.objdata(), errObj.objsize());
// TODO: call replyToQuery() from here instead of this!!! see dbmessage.h
QueryResult::View msgdata = bb.buf();
QueryResult::View qr = msgdata;
qr.setResultFlags(ResultFlag_ErrSet);
if (isStaleConfig) {
qr.setResultFlags(qr.getResultFlags() | ResultFlag_ShardConfigStale);
}
qr.msgdata().setLen(bb.len());
qr.msgdata().setOperation(opReply);
qr.setCursorId(0);
qr.setStartingFrom(0);
qr.setNReturned(1);
response->setData(bb.release());
}
void registerError(OperationContext* opCtx, const DBException& exception) {
LastError::get(opCtx->getClient()).setLastError(exception.code(), exception.reason());
CurOp::get(opCtx)->debug().errInfo = exception.toStatus();
}
void generateErrorResponse(OperationContext* opCtx,
rpc::ReplyBuilderInterface* replyBuilder,
const DBException& exception,
const BSONObj& replyMetadata,
BSONObj extraFields = {}) {
registerError(opCtx, exception);
// We could have thrown an exception after setting fields in the builder,
// so we need to reset it to a clean state just to be sure.
replyBuilder->reset();
replyBuilder->setCommandReply(exception.toStatus(), extraFields);
replyBuilder->setMetadata(replyMetadata);
}
BSONObj getErrorLabels(const boost::optional<OperationSessionInfoFromClient>& sessionOptions,
const std::string& commandName,
ErrorCodes::Error code) {
// By specifying "autocommit", the user indicates they want to run a transaction.
if (!sessionOptions || !sessionOptions->getAutocommit()) {
return {};
}
bool isRetryable = ErrorCodes::isNotMasterError(code) || ErrorCodes::isShutdownError(code);
bool isTransientTransactionError = code == ErrorCodes::WriteConflict //
|| code == ErrorCodes::SnapshotUnavailable //
|| code == ErrorCodes::NoSuchTransaction //
|| code == ErrorCodes::LockTimeout //
// Clients can retry a single commitTransaction command, but cannot retry the whole
// transaction if commitTransaction fails due to NotMaster.
|| (isRetryable && (commandName != "commitTransaction"));
if (isTransientTransactionError) {
return BSON("errorLabels" << BSON_ARRAY("TransientTransactionError"));
}
return {};
}
/**
* Guard object for making a good-faith effort to enter maintenance mode and leave it when it
* goes out of scope.
*
* Sometimes we cannot set maintenance mode, in which case the call to setMaintenanceMode will
* return a non-OK status. This class does not treat that case as an error which means that
* anybody using it is assuming it is ok to continue execution without maintenance mode.
*
* TODO: This assumption needs to be audited and documented, or this behavior should be moved
* elsewhere.
*/
class MaintenanceModeSetter {
MONGO_DISALLOW_COPYING(MaintenanceModeSetter);
public:
MaintenanceModeSetter(OperationContext* opCtx)
: _opCtx(opCtx),
_maintenanceModeSet(
repl::ReplicationCoordinator::get(_opCtx)->setMaintenanceMode(true).isOK()) {}
~MaintenanceModeSetter() {
if (_maintenanceModeSet) {
repl::ReplicationCoordinator::get(_opCtx)
->setMaintenanceMode(false)
.transitional_ignore();
}
}
private:
OperationContext* const _opCtx;
const bool _maintenanceModeSet;
};
constexpr auto kLastCommittedOpTimeFieldName = "lastCommittedOpTime"_sd;
// Called from the error contexts where request may not be available.
void appendReplyMetadataOnError(OperationContext* opCtx, BSONObjBuilder* metadataBob) {
const bool isConfig = serverGlobalParams.clusterRole == ClusterRole::ConfigServer;
if (ShardingState::get(opCtx)->enabled() || isConfig) {
auto lastCommittedOpTime =
repl::ReplicationCoordinator::get(opCtx)->getLastCommittedOpTime();
metadataBob->append(kLastCommittedOpTimeFieldName, lastCommittedOpTime.getTimestamp());
}
}
void appendReplyMetadata(OperationContext* opCtx,
const OpMsgRequest& request,
BSONObjBuilder* metadataBob) {
const bool isShardingAware = ShardingState::get(opCtx)->enabled();
const bool isConfig = serverGlobalParams.clusterRole == ClusterRole::ConfigServer;
auto const replCoord = repl::ReplicationCoordinator::get(opCtx);
const bool isReplSet =
replCoord->getReplicationMode() == repl::ReplicationCoordinator::modeReplSet;
if (isReplSet) {
// Attach our own last opTime.
repl::OpTime lastOpTimeFromClient =
repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp();
replCoord->prepareReplMetadata(request.body, lastOpTimeFromClient, metadataBob);
// For commands from mongos, append some info to help getLastError(w) work.
// TODO: refactor out of here as part of SERVER-18236
if (isShardingAware || isConfig) {
rpc::ShardingMetadata(lastOpTimeFromClient, replCoord->getElectionId())
.writeToMetadata(metadataBob)
.transitional_ignore();
}
if (isShardingAware || isConfig) {
auto lastCommittedOpTime = replCoord->getLastCommittedOpTime();
metadataBob->append(kLastCommittedOpTimeFieldName, lastCommittedOpTime.getTimestamp());
}
}
// If we're a shard other than the config shard, attach the last configOpTime we know about.
if (isShardingAware && !isConfig) {
auto opTime = Grid::get(opCtx)->configOpTime();
rpc::ConfigServerMetadata(opTime).writeToMetadata(metadataBob);
}
}
/**
* Given the specified command, returns an effective read concern which should be used or an error
* if the read concern is not valid for the command.
*/
StatusWith<repl::ReadConcernArgs> _extractReadConcern(const CommandInvocation* invocation,
const BSONObj& cmdObj,
bool upconvertToSnapshot) {
repl::ReadConcernArgs readConcernArgs;
auto readConcernParseStatus = readConcernArgs.initialize(cmdObj);
if (!readConcernParseStatus.isOK()) {
return readConcernParseStatus;
}
if (upconvertToSnapshot) {
auto upconvertToSnapshotStatus = readConcernArgs.upconvertReadConcernLevelToSnapshot();
if (!upconvertToSnapshotStatus.isOK()) {
return upconvertToSnapshotStatus;
}
}
if (!invocation->supportsReadConcern(readConcernArgs.getLevel())) {
// We must be in a transaction if the readConcern level was upconverted to snapshot and the
// command must support readConcern level snapshot in order to be supported in transactions.
if (upconvertToSnapshot) {
return {ErrorCodes::OperationNotSupportedInTransaction,
str::stream() << "Command is not supported in a transaction"};
}
return {ErrorCodes::InvalidOptions,
str::stream() << "Command does not support read concern "
<< readConcernArgs.toString()};
}
return readConcernArgs;
}
/**
* For replica set members it returns the last known op time from opCtx. Otherwise will return
* uninitialized cluster time.
*/
LogicalTime getClientOperationTime(OperationContext* opCtx) {
auto const replCoord = repl::ReplicationCoordinator::get(opCtx);
const bool isReplSet =
replCoord->getReplicationMode() == repl::ReplicationCoordinator::modeReplSet;
if (!isReplSet) {
return LogicalTime();
}
return LogicalTime(
repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp().getTimestamp());
}
/**
* Returns the proper operationTime for a command. To construct the operationTime for replica set
* members, it uses the last optime in the oplog for writes, last committed optime for majority
* reads, and the last applied optime for every other read. An uninitialized cluster time is
* returned for non replica set members.
*
* The latest in-memory clusterTime is returned if the start operationTime is uninitialized.
*/
LogicalTime computeOperationTime(OperationContext* opCtx, LogicalTime startOperationTime) {
auto const replCoord = repl::ReplicationCoordinator::get(opCtx);
const bool isReplSet =
replCoord->getReplicationMode() == repl::ReplicationCoordinator::modeReplSet;
invariant(isReplSet);
if (startOperationTime == LogicalTime::kUninitialized) {
return LogicalTime(replCoord->getMyLastAppliedOpTime().getTimestamp());
}
auto operationTime = getClientOperationTime(opCtx);
invariant(operationTime >= startOperationTime);
// If the last operationTime has not changed, consider this command a read, and, for replica set
// members, construct the operationTime with the proper optime for its read concern level.
if (operationTime == startOperationTime) {
auto& readConcernArgs = repl::ReadConcernArgs::get(opCtx);
// Note: ReadConcernArgs::getLevel returns kLocal if none was set.
if (readConcernArgs.getLevel() == repl::ReadConcernLevel::kMajorityReadConcern) {
operationTime = LogicalTime(replCoord->getLastCommittedOpTime().getTimestamp());
} else {
operationTime = LogicalTime(replCoord->getMyLastAppliedOpTime().getTimestamp());
}
}
return operationTime;
}
/**
* Computes the proper $clusterTime and operationTime values to include in the command response and
* appends them to it. $clusterTime is added as metadata and operationTime as a command body field.
*
* The command body BSONObjBuilder is either the builder for the command body itself, or a builder
* for extra fields to be added to the reply when generating an error response.
*/
void appendClusterAndOperationTime(OperationContext* opCtx,
BSONObjBuilder* commandBodyFieldsBob,
BSONObjBuilder* metadataBob,
LogicalTime startTime) {
if (repl::ReplicationCoordinator::get(opCtx)->getReplicationMode() !=
repl::ReplicationCoordinator::modeReplSet ||
!LogicalClock::get(opCtx)->isEnabled()) {
return;
}
// Authorized clients always receive operationTime and dummy signed $clusterTime.
if (LogicalTimeValidator::isAuthorizedToAdvanceClock(opCtx)) {
auto operationTime = computeOperationTime(opCtx, startTime);
auto signedTime = SignedLogicalTime(
LogicalClock::get(opCtx)->getClusterTime(), TimeProofService::TimeProof(), 0);
// TODO SERVER-35663: invariant that signedTime.getTime() >= operationTime.
rpc::LogicalTimeMetadata(signedTime).writeToMetadata(metadataBob);
operationTime.appendAsOperationTime(commandBodyFieldsBob);
return;
}
// Servers without validators (e.g. a shard server not yet added to a cluster) do not return
// logical times to unauthorized clients.
auto validator = LogicalTimeValidator::get(opCtx);
if (!validator) {
return;
}
auto operationTime = computeOperationTime(opCtx, startTime);
auto signedTime = validator->trySignLogicalTime(LogicalClock::get(opCtx)->getClusterTime());
// If there were no keys, do not return $clusterTime or operationTime to unauthorized clients.
if (signedTime.getKeyId() == 0) {
return;
}
// TODO SERVER-35663: invariant that signedTime.getTime() >= operationTime.
rpc::LogicalTimeMetadata(signedTime).writeToMetadata(metadataBob);
operationTime.appendAsOperationTime(commandBodyFieldsBob);
}
void invokeInTransaction(OperationContext* opCtx,
CommandInvocation* invocation,
CommandReplyBuilder* replyBuilder) {
auto session = OperationContextSession::get(opCtx);
invariant(session);
invariant(opCtx->getTxnNumber() || opCtx->getClient()->isInDirectClient());
session->unstashTransactionResources(opCtx, invocation->definition()->getName());
ScopeGuard guard = MakeGuard([session, opCtx]() { session->abortActiveTransaction(opCtx); });
invocation->run(opCtx, replyBuilder);
if (auto okField = replyBuilder->getBodyBuilder().asTempObj()["ok"]) {
// If ok is present, use its truthiness.
if (!okField.trueValue()) {
return;
}
}
// Stash or commit the transaction when the command succeeds.
session->stashTransactionResources(opCtx);
guard.Dismiss();
}
bool runCommandImpl(OperationContext* opCtx,
CommandInvocation* invocation,
const OpMsgRequest& request,
rpc::ReplyBuilderInterface* replyBuilder,
LogicalTime startOperationTime,
const ServiceEntryPointCommon::Hooks& behaviors,
BSONObjBuilder* extraFieldsBuilder,
const boost::optional<OperationSessionInfoFromClient>& sessionOptions) {
const Command* command = invocation->definition();
auto bytesToReserve = command->reserveBytesForReply();
// SERVER-22100: In Windows DEBUG builds, the CRT heap debugging overhead, in conjunction with the
// additional memory pressure introduced by reply buffer pre-allocation, causes the concurrency
// suite to run extremely slowly. As a workaround we do not pre-allocate in Windows DEBUG builds.
#ifdef _WIN32
if (kDebugBuild)
bytesToReserve = 0;
#endif
CommandReplyBuilder crb(replyBuilder->getInPlaceReplyBuilder(bytesToReserve));
auto session = OperationContextSession::get(opCtx);
if (!invocation->supportsWriteConcern()) {
behaviors.uassertCommandDoesNotSpecifyWriteConcern(request.body);
if (session) {
invokeInTransaction(opCtx, invocation, &crb);
} else {
invocation->run(opCtx, &crb);
}
} else {
auto wcResult = uassertStatusOK(extractWriteConcern(opCtx, request.body));
uassert(ErrorCodes::InvalidOptions,
"writeConcern is not allowed within a multi-statement transaction",
wcResult.usedDefault || !session ||
!session->inActiveOrKilledMultiDocumentTransaction() ||
invocation->definition()->getName() == "commitTransaction" ||
invocation->definition()->getName() == "abortTransaction" ||
invocation->definition()->getName() == "doTxn");
auto lastOpBeforeRun = repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp();
// Change the write concern while running the command.
const auto oldWC = opCtx->getWriteConcern();
ON_BLOCK_EXIT([&] { opCtx->setWriteConcern(oldWC); });
opCtx->setWriteConcern(wcResult);
auto waitForWriteConcern = [&](auto&& bb) {
MONGO_FAIL_POINT_BLOCK_IF(failCommand, data, [&](const BSONObj& data) {
return shouldActivateFailCommandFailPoint(data, request.getCommandName()) &&
data.hasField("writeConcernError");
}) {
bb.append(data.getData()["writeConcernError"]);
return; // Don't do normal waiting.
}
behaviors.waitForWriteConcern(opCtx, invocation, lastOpBeforeRun, bb);
};
try {
if (session) {
invokeInTransaction(opCtx, invocation, &crb);
} else {
invocation->run(opCtx, &crb);
}
} catch (const DBException&) {
waitForWriteConcern(*extraFieldsBuilder);
throw;
}
waitForWriteConcern(crb.getBodyBuilder());
// Nothing in run() should change the writeConcern.
dassert(SimpleBSONObjComparator::kInstance.evaluate(opCtx->getWriteConcern().toBSON() ==
wcResult.toBSON()));
}
behaviors.waitForLinearizableReadConcern(opCtx);
const bool ok = [&] {
auto body = crb.getBodyBuilder();
return CommandHelpers::extractOrAppendOk(body);
}();
behaviors.attachCurOpErrInfo(opCtx, crb.getBodyBuilder().asTempObj());
if (!ok) {
auto response = crb.getBodyBuilder().asTempObj();
auto codeField = response["code"];
if (codeField.isNumber()) {
auto code = ErrorCodes::Error(codeField.numberInt());
// Append the error labels for transient transaction errors.
auto errorLabels = getErrorLabels(sessionOptions, command->getName(), code);
crb.getBodyBuilder().appendElements(errorLabels);
}
}
BSONObjBuilder metadataBob;
appendReplyMetadata(opCtx, request, &metadataBob);
{
auto commandBodyBob = crb.getBodyBuilder();
appendClusterAndOperationTime(opCtx, &commandBodyBob, &metadataBob, startOperationTime);
}
replyBuilder->setMetadata(metadataBob.obj());
return ok;
}
/**
* Maybe uassert according to the 'failCommand' fail point.
*/
void evaluateFailCommandFailPoint(OperationContext* opCtx, StringData commandName) {
MONGO_FAIL_POINT_BLOCK_IF(failCommand, data, [&](const BSONObj& data) {
return shouldActivateFailCommandFailPoint(data, commandName) &&
(data.hasField("closeConnection") || data.hasField("errorCode"));
}) {
bool closeConnection;
if (bsonExtractBooleanField(data.getData(), "closeConnection", &closeConnection).isOK() &&
closeConnection) {
opCtx->getClient()->session()->end();
log() << "Failing command '" << commandName
<< "' via 'failCommand' failpoint. Action: closing connection.";
uasserted(50838, "Failing command due to 'failCommand' failpoint");
}
long long errorCode;
if (bsonExtractIntegerField(data.getData(), "errorCode", &errorCode).isOK()) {
log() << "Failing command '" << commandName
<< "' via 'failCommand' failpoint. Action: returning error code " << errorCode
<< ".";
uasserted(ErrorCodes::Error(errorCode),
"Failing command due to 'failCommand' failpoint");
}
}
}
/**
* Executes a command after stripping metadata, performing authorization checks,
* handling audit impersonation, and (potentially) setting maintenance mode. This method
* also checks that the command is permissible to run on the node given its current
* replication state. All the logic here is independent of any particular command; any
* functionality relevant to a specific command should be confined to its run() method.
*/
void execCommandDatabase(OperationContext* opCtx,
Command* command,
const OpMsgRequest& request,
rpc::ReplyBuilderInterface* replyBuilder,
const ServiceEntryPointCommon::Hooks& behaviors) {
CommandHelpers::uassertShouldAttemptParse(opCtx, command, request);
BSONObjBuilder extraFieldsBuilder;
auto startOperationTime = getClientOperationTime(opCtx);
auto invocation = command->parse(opCtx, request);
boost::optional<OperationSessionInfoFromClient> sessionOptions = boost::none;
try {
{
stdx::lock_guard<Client> lk(*opCtx->getClient());
CurOp::get(opCtx)->setCommand_inlock(command);
}
// TODO: move this back to runCommands when mongos supports OperationContext
// see SERVER-18515 for details.
rpc::readRequestMetadata(opCtx, request.body, command->requiresAuth());
rpc::TrackingMetadata::get(opCtx).initWithOperName(command->getName());
auto const replCoord = repl::ReplicationCoordinator::get(opCtx);
sessionOptions = initializeOperationSessionInfo(
opCtx,
request.body,
command->requiresAuth(),
replCoord->getReplicationMode() == repl::ReplicationCoordinator::modeReplSet,
opCtx->getServiceContext()->getStorageEngine()->supportsDocLocking());
evaluateFailCommandFailPoint(opCtx, command->getName());
const auto dbname = request.getDatabase().toString();
uassert(
ErrorCodes::InvalidNamespace,
str::stream() << "Invalid database name: '" << dbname << "'",
NamespaceString::validDBName(dbname, NamespaceString::DollarInDbNameBehavior::Allow));
// Session ids are forwarded in requests, so commands that require roundtrips between
// servers may result in a deadlock when a server tries to check out a session it is already
// using to service an earlier operation in the command's chain. To avoid this, only check
// out sessions for commands that require them.
const bool shouldCheckoutSession = static_cast<bool>(opCtx->getTxnNumber()) &&
sessionCheckoutWhitelist.find(command->getName()) != sessionCheckoutWhitelist.cend();
// Parse the arguments specific to multi-statement transactions.
boost::optional<bool> startMultiDocTxn = boost::none;
boost::optional<bool> autocommitVal = boost::none;
if (sessionOptions) {
startMultiDocTxn = sessionOptions->getStartTransaction();
autocommitVal = sessionOptions->getAutocommit();
if (command->getName() == "doTxn") {
// Autocommit and 'startMultiDocTxn' are overridden for 'doTxn' to get the oplog
// entry generation behavior used for multi-document transactions. The 'doTxn'
// command still logically behaves as a commit.
autocommitVal = false;
startMultiDocTxn = true;
}
}
// Reject commands with 'txnNumber' that do not check out the Session, since no retryable
// writes or transaction machinery will be used to execute commands that do not check out
// the Session. Do not check this if we are in DBDirectClient because the outer command is
// responsible for checking out the Session.
if (!opCtx->getClient()->isInDirectClient()) {
uassert(ErrorCodes::OperationNotSupportedInTransaction,
str::stream() << "It is illegal to run command " << command->getName()
<< " in a multi-document transaction.",
shouldCheckoutSession || !autocommitVal || command->getName() == "doTxn");
uassert(50768,
str::stream() << "It is illegal to provide a txnNumber for command "
<< command->getName(),
shouldCheckoutSession || !opCtx->getTxnNumber());
}
// This constructor will check out the session and start a transaction, if necessary. It
// handles the appropriate state management for both multi-statement transactions and
// retryable writes.
OperationContextSession sessionTxnState(opCtx,
shouldCheckoutSession,
autocommitVal,
startMultiDocTxn,
dbname,
command->getName());
std::unique_ptr<MaintenanceModeSetter> mmSetter;
BSONElement cmdOptionMaxTimeMSField;
BSONElement allowImplicitCollectionCreationField;
BSONElement helpField;
StringMap<int> topLevelFields;
for (auto&& element : request.body) {
StringData fieldName = element.fieldNameStringData();
if (fieldName == QueryRequest::cmdOptionMaxTimeMS) {
cmdOptionMaxTimeMSField = element;
} else if (fieldName == "allowImplicitCollectionCreation") {
allowImplicitCollectionCreationField = element;
} else if (fieldName == CommandHelpers::kHelpFieldName) {
helpField = element;
} else if (fieldName == QueryRequest::queryOptionMaxTimeMS) {
uasserted(ErrorCodes::InvalidOptions,
"no such command option $maxTimeMs; use maxTimeMS instead");
}
uassert(ErrorCodes::FailedToParse,
str::stream() << "Parsed command object contains duplicate top level key: "
<< fieldName,
topLevelFields[fieldName]++ == 0);
}
if (CommandHelpers::isHelpRequest(helpField)) {
CurOp::get(opCtx)->ensureStarted();
// We disable last-error for help requests due to SERVER-11492, because config servers
// use help requests to determine which commands are database writes, and so must be
// forwarded to all config servers.
LastError::get(opCtx->getClient()).disable();
Command::generateHelpResponse(opCtx, replyBuilder, *command);
return;
}
ImpersonationSessionGuard guard(opCtx);
invocation->checkAuthorization(opCtx, request);
const bool iAmPrimary = replCoord->canAcceptWritesForDatabase_UNSAFE(opCtx, dbname);
if (!opCtx->getClient()->isInDirectClient() &&
!MONGO_FAIL_POINT(skipCheckingForNotMasterInCommandDispatch)) {
auto allowed = command->secondaryAllowed(opCtx->getServiceContext());
bool alwaysAllowed = allowed == Command::AllowedOnSecondary::kAlways;
bool couldHaveOptedIn = allowed == Command::AllowedOnSecondary::kOptIn;
bool optedIn =
couldHaveOptedIn && ReadPreferenceSetting::get(opCtx).canRunOnSecondary();
bool canRunHere = commandCanRunHere(opCtx, dbname, command);
if (!canRunHere && couldHaveOptedIn) {
uasserted(ErrorCodes::NotMasterNoSlaveOk, "not master and slaveOk=false");
}
if (MONGO_FAIL_POINT(respondWithNotPrimaryInCommandDispatch)) {
uassert(ErrorCodes::NotMaster, "not primary", canRunHere);
} else {
uassert(ErrorCodes::NotMaster, "not master", canRunHere);
}
if (!command->maintenanceOk() &&
replCoord->getReplicationMode() == repl::ReplicationCoordinator::modeReplSet &&
!replCoord->canAcceptWritesForDatabase_UNSAFE(opCtx, dbname) &&
!replCoord->getMemberState().secondary()) {
uassert(ErrorCodes::NotMasterOrSecondary,
"node is recovering",
!replCoord->getMemberState().recovering());
uassert(ErrorCodes::NotMasterOrSecondary,
"node is not in primary or recovering state",
replCoord->getMemberState().primary());
// Check ticket SERVER-21432, slaveOk commands are allowed in drain mode
uassert(ErrorCodes::NotMasterOrSecondary,
"node is in drain mode",
optedIn || alwaysAllowed);
}
}
if (command->adminOnly()) {
LOG(2) << "command: " << request.getCommandName();
}
if (command->maintenanceMode()) {
mmSetter.reset(new MaintenanceModeSetter(opCtx));
}
if (command->shouldAffectCommandCounter()) {
OpCounters* opCounters = &globalOpCounters;
opCounters->gotCommand();
}
// Parse the 'maxTimeMS' command option, and use it to set a deadline for the operation on
// the OperationContext. The 'maxTimeMS' option unfortunately has a different meaning for a
// getMore command, where it is used to communicate the maximum time to wait for new inserts
// on tailable cursors, not as a deadline for the operation.
// TODO SERVER-34277 Remove the special handling for maxTimeMS for getMores. This will
// require introducing a new 'max await time' parameter for getMore, and eventually banning
// maxTimeMS altogether on a getMore command.
const int maxTimeMS =
uassertStatusOK(QueryRequest::parseMaxTimeMS(cmdOptionMaxTimeMSField));
if (maxTimeMS > 0 && command->getLogicalOp() != LogicalOp::opGetMore) {
uassert(40119,
"Illegal attempt to set operation deadline within DBDirectClient",
!opCtx->getClient()->isInDirectClient());
opCtx->setDeadlineAfterNowBy(Milliseconds{maxTimeMS}, ErrorCodes::MaxTimeMSExpired);
}
auto& readConcernArgs = repl::ReadConcernArgs::get(opCtx);
auto session = OperationContextSession::get(opCtx);
if (!opCtx->getClient()->isInDirectClient() || !session ||
!session->inActiveOrKilledMultiDocumentTransaction()) {
const bool upconvertToSnapshot = session &&
session->inActiveOrKilledMultiDocumentTransaction() && sessionOptions &&
(sessionOptions->getStartTransaction() == boost::optional<bool>(true));
readConcernArgs = uassertStatusOK(
_extractReadConcern(invocation.get(), request.body, upconvertToSnapshot));
}
if (readConcernArgs.getArgsAtClusterTime()) {
uassert(ErrorCodes::InvalidOptions,
"atClusterTime is only used for testing",
getTestCommandsEnabled());
}
if (readConcernArgs.getLevel() == repl::ReadConcernLevel::kSnapshotReadConcern) {
auto session = OperationContextSession::get(opCtx);
uassert(ErrorCodes::InvalidOptions,
"readConcern level snapshot is only valid in multi-statement transactions",
session && session->inActiveOrKilledMultiDocumentTransaction());
uassert(ErrorCodes::InvalidOptions,
"readConcern level snapshot requires a session ID",
opCtx->getLogicalSessionId());
uassert(ErrorCodes::InvalidOptions,
"readConcern level snapshot requires a txnNumber",
opCtx->getTxnNumber());
opCtx->lockState()->setSharedLocksShouldTwoPhaseLock(true);
}
auto& oss = OperationShardingState::get(opCtx);
if (!opCtx->getClient()->isInDirectClient() &&
readConcernArgs.getLevel() != repl::ReadConcernLevel::kAvailableReadConcern &&
(iAmPrimary ||
(readConcernArgs.hasLevel() || readConcernArgs.getArgsAfterClusterTime()))) {
oss.initializeClientRoutingVersions(invocation->ns(), request.body);
auto const shardingState = ShardingState::get(opCtx);
if (oss.hasShardVersion() || oss.hasDbVersion()) {
uassertStatusOK(shardingState->canAcceptShardedCommands());
}
behaviors.advanceConfigOptimeFromRequestMetadata(opCtx);
}
oss.setAllowImplicitCollectionCreation(allowImplicitCollectionCreationField);
auto scoped = behaviors.scopedOperationCompletionShardingActions(opCtx);
// This may trigger the maxTimeAlwaysTimeOut failpoint.
auto status = opCtx->checkForInterruptNoAssert();
// We still proceed if the primary stepped down, but accept other kinds of interruptions.
// We defer to individual commands to allow themselves to be interruptible by stepdowns,
// since commands like 'voteRequest' should conversely continue executing.
if (status != ErrorCodes::PrimarySteppedDown &&
status != ErrorCodes::InterruptedDueToReplStateChange) {
uassertStatusOK(status);
}
CurOp::get(opCtx)->ensureStarted();
command->incrementCommandsExecuted();
if (logger::globalLogDomain()->shouldLog(logger::LogComponent::kTracking,
logger::LogSeverity::Debug(1)) &&
rpc::TrackingMetadata::get(opCtx).getParentOperId()) {
MONGO_LOG_COMPONENT(1, logger::LogComponent::kTracking)
<< rpc::TrackingMetadata::get(opCtx).toString();
rpc::TrackingMetadata::get(opCtx).setIsLogged(true);
}
behaviors.waitForReadConcern(opCtx, invocation.get(), request);
try {
if (!runCommandImpl(opCtx,
invocation.get(),
request,
replyBuilder,
startOperationTime,
behaviors,
&extraFieldsBuilder,
sessionOptions)) {
command->incrementCommandsFailed();
}
} catch (const DBException&) {
command->incrementCommandsFailed();
throw;
}
} catch (const DBException& e) {
behaviors.handleException(e, opCtx);
// Append the error labels for transient transaction errors.
auto errorLabels = getErrorLabels(sessionOptions, command->getName(), e.code());
extraFieldsBuilder.appendElements(errorLabels);
BSONObjBuilder metadataBob;
appendReplyMetadata(opCtx, request, &metadataBob);
// The read concern may not have yet been placed on the operation context, so attempt to
// parse it here, so if it is valid it can be used to compute the proper operationTime.
auto& readConcernArgs = repl::ReadConcernArgs::get(opCtx);
if (readConcernArgs.isEmpty()) {
auto readConcernArgsStatus = _extractReadConcern(invocation.get(), request.body, false);
if (readConcernArgsStatus.isOK()) {
readConcernArgs = readConcernArgsStatus.getValue();
}
}
appendClusterAndOperationTime(opCtx, &extraFieldsBuilder, &metadataBob, startOperationTime);
LOG(1) << "assertion while executing command '" << request.getCommandName() << "' "
<< "on database '" << request.getDatabase() << "' "
<< "with arguments '"
<< redact(ServiceEntryPointCommon::getRedactedCopyForLogging(command, request.body))
<< "': " << redact(e.toString());
generateErrorResponse(opCtx, replyBuilder, e, metadataBob.obj(), extraFieldsBuilder.obj());
}
}
/**
* Fills out CurOp / OpDebug with basic command info.
*/
void curOpCommandSetup(OperationContext* opCtx, const OpMsgRequest& request) {
auto curop = CurOp::get(opCtx);
curop->debug().iscommand = true;
// We construct a legacy $cmd namespace so we can fill in curOp using
// the existing logic that existed for OP_QUERY commands
NamespaceString nss(request.getDatabase(), "$cmd");
stdx::lock_guard<Client> lk(*opCtx->getClient());
curop->setOpDescription_inlock(request.body);
curop->markCommand_inlock();
curop->setNS_inlock(nss.ns());
}
DbResponse receivedCommands(OperationContext* opCtx,
const Message& message,
const ServiceEntryPointCommon::Hooks& behaviors) {
auto replyBuilder = rpc::makeReplyBuilder(rpc::protocolForMessage(message));
[&] {
OpMsgRequest request;
try { // Parse.
request = rpc::opMsgRequestFromAnyProtocol(message);
} catch (const DBException& ex) {
// If this error needs to fail the connection, propagate it out.
if (ErrorCodes::isConnectionFatalMessageParseError(ex.code()))
throw;
BSONObjBuilder metadataBob;
appendReplyMetadataOnError(opCtx, &metadataBob);
BSONObjBuilder extraFieldsBuilder;
appendClusterAndOperationTime(
opCtx, &extraFieldsBuilder, &metadataBob, LogicalTime::kUninitialized);
// Otherwise, reply with the parse error. This is useful for cases where parsing fails
// due to user-supplied input, such as the document too deep error. Since we failed
// during parsing, we can't log anything about the command.
LOG(1) << "assertion while parsing command: " << ex.toString();
generateErrorResponse(
opCtx, replyBuilder.get(), ex, metadataBob.obj(), extraFieldsBuilder.obj());
return; // From lambda. Don't try executing if parsing failed.
}
try { // Execute.
curOpCommandSetup(opCtx, request);
Command* c = nullptr;
// In the absence of a Command object, no redaction is possible. Therefore
// to avoid displaying potentially sensitive information in the logs,
// we restrict the log message to the name of the unrecognized command.
// However, the complete command object will still be echoed to the client.
if (!(c = CommandHelpers::findCommand(request.getCommandName()))) {
globalCommandRegistry()->incrementUnknownCommands();
std::string msg = str::stream() << "no such command: '" << request.getCommandName()
<< "'";
LOG(2) << msg;