forked from mongodb/mongo-go-driver
-
Notifications
You must be signed in to change notification settings - Fork 0
/
operation.go
1315 lines (1166 loc) · 43.9 KB
/
operation.go
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
package driver
import (
"bytes"
"context"
"errors"
"fmt"
"strconv"
"time"
"go.mongodb.org/mongo-driver/bson"
"go.mongodb.org/mongo-driver/bson/bsontype"
"go.mongodb.org/mongo-driver/bson/primitive"
"go.mongodb.org/mongo-driver/event"
"go.mongodb.org/mongo-driver/mongo/readconcern"
"go.mongodb.org/mongo-driver/mongo/readpref"
"go.mongodb.org/mongo-driver/mongo/writeconcern"
"go.mongodb.org/mongo-driver/x/bsonx/bsoncore"
"go.mongodb.org/mongo-driver/x/mongo/driver/description"
"go.mongodb.org/mongo-driver/x/mongo/driver/session"
"go.mongodb.org/mongo-driver/x/mongo/driver/wiremessage"
)
const defaultLocalThreshold = 15 * time.Millisecond
var dollarCmd = [...]byte{'.', '$', 'c', 'm', 'd'}
var (
// ErrNoDocCommandResponse occurs when the server indicated a response existed, but none was found.
ErrNoDocCommandResponse = errors.New("command returned no documents")
// ErrMultiDocCommandResponse occurs when the server sent multiple documents in response to a command.
ErrMultiDocCommandResponse = errors.New("command returned multiple documents")
// ErrReplyDocumentMismatch occurs when the number of documents returned in an OP_QUERY does not match the numberReturned field.
ErrReplyDocumentMismatch = errors.New("number of documents returned does not match numberReturned field")
// ErrNonPrimaryReadPref is returned when a read is attempted in a transaction with a non-primary read preference.
ErrNonPrimaryReadPref = errors.New("read preference in a transaction must be primary")
)
const (
// maximum BSON object size when client side encryption is enabled
cryptMaxBsonObjectSize uint32 = 2097152
// minimum wire version necessary to use automatic encryption
cryptMinWireVersion int32 = 8
)
// InvalidOperationError is returned from Validate and indicates that a required field is missing
// from an instance of Operation.
type InvalidOperationError struct{ MissingField string }
func (err InvalidOperationError) Error() string {
return "the " + err.MissingField + " field must be set on Operation"
}
// opReply stores information returned in an OP_REPLY response from the server.
// The err field stores any error that occurred when decoding or validating the OP_REPLY response.
type opReply struct {
responseFlags wiremessage.ReplyFlag
cursorID int64
startingFrom int32
numReturned int32
documents []bsoncore.Document
err error
}
// startedInformation keeps track of all of the information necessary for monitoring started events.
type startedInformation struct {
cmd bsoncore.Document
requestID int32
cmdName string
documentSequenceIncluded bool
connID string
}
// finishedInformation keeps track of all of the information necessary for monitoring success and failure events.
type finishedInformation struct {
cmdName string
requestID int32
response bsoncore.Document
cmdErr error
connID string
startTime time.Time
}
// Operation is used to execute an operation. It contains all of the common code required to
// select a server, transform an operation into a command, write the command to a connection from
// the selected server, read a response from that connection, process the response, and potentially
// retry.
//
// The required fields are Database, CommandFn, and Deployment. All other fields are optional.
//
// While an Operation can be constructed manually, drivergen should be used to generate an
// implementation of an operation instead. This will ensure that there are helpers for constructing
// the operation and that this type isn't configured incorrectly.
type Operation struct {
// CommandFn is used to create the command that will be wrapped in a wire message and sent to
// the server. This function should only add the elements of the command and not start or end
// the enclosing BSON document. Per the command API, the first element must be the name of the
// command to run. This field is required.
CommandFn func(dst []byte, desc description.SelectedServer) ([]byte, error)
// Database is the database that the command will be run against. This field is required.
Database string
// Deployment is the MongoDB Deployment to use. While most of the time this will be multiple
// servers, commands that need to run against a single, preselected server can use the
// SingleServerDeployment type. Commands that need to run on a preselected connection can use
// the SingleConnectionDeployment type.
Deployment Deployment
// ProcessResponseFn is called after a response to the command is returned. The server is
// provided for types like Cursor that are required to run subsequent commands using the same
// server.
ProcessResponseFn func(response bsoncore.Document, srvr Server, desc description.Server) error
// Selector is the server selector that's used during both initial server selection and
// subsequent selection for retries. Depending on the Deployment implementation, the
// SelectServer method may not actually be called.
Selector description.ServerSelector
// ReadPreference is the read preference that will be attached to the command. If this field is
// not specified a default read preference of primary will be used.
ReadPreference *readpref.ReadPref
// ReadConcern is the read concern used when running read commands. This field should not be set
// for write operations. If this field is set, it will be encoded onto the commands sent to the
// server.
ReadConcern *readconcern.ReadConcern
// MinimumReadConcernWireVersion specifies the minimum wire version to add the read concern to
// the command being executed.
MinimumReadConcernWireVersion int32
// WriteConcern is the write concern used when running write commands. This field should not be
// set for read operations. If this field is set, it will be encoded onto the commands sent to
// the server.
WriteConcern *writeconcern.WriteConcern
// MinimumWriteConcernWireVersion specifies the minimum wire version to add the write concern to
// the command being executed.
MinimumWriteConcernWireVersion int32
// Client is the session used with this operation. This can be either an implicit or explicit
// session. If the server selected does not support sessions and Client is specified the
// behavior depends on the session type. If the session is implicit, the session fields will not
// be encoded onto the command. If the session is explicit, an error will be returned. The
// caller is responsible for ensuring that this field is nil if the Deployment does not support
// sessions.
Client *session.Client
// Clock is a cluster clock, different from the one contained within a session.Client. This
// allows updating cluster times for a global cluster clock while allowing individual session's
// cluster clocks to be only updated as far as the last command that's been run.
Clock *session.ClusterClock
// RetryMode specifies how to retry. There are three modes that enable retry: RetryOnce,
// RetryOncePerCommand, and RetryContext. For more information about what these modes do, please
// refer to their definitions. Both RetryMode and Type must be set for retryability to be enabled.
RetryMode *RetryMode
// Type specifies the kind of operation this is. There is only one mode that enables retry: Write.
// For more information about what this mode does, please refer to it's definition. Both Type and
// RetryMode must be set for retryability to be enabled.
Type Type
// Batches contains the documents that are split when executing a write command that potentially
// has more documents than can fit in a single command. This should only be specified for
// commands that are batch compatible. For more information, please refer to the definition of
// Batches.
Batches *Batches
// Legacy sets the legacy type for this operation. There are only 3 types that require legacy
// support: find, getMore, and killCursors. For more information about LegacyOperationKind,
// please refer to it's definition.
Legacy LegacyOperationKind
// CommandMonitor specifies the monitor to use for APM events. If this field is not set,
// no events will be reported.
CommandMonitor *event.CommandMonitor
// Crypt specifies a Crypt object to use for automatic client side encryption and decryption.
Crypt *Crypt
}
// shouldEncrypt returns true if this operation should automatically be encrypted.
func (op Operation) shouldEncrypt() bool {
return op.Crypt != nil && !op.Crypt.BypassAutoEncryption
}
// selectServer handles performing server selection for an operation.
func (op Operation) selectServer(ctx context.Context) (Server, error) {
if err := op.Validate(); err != nil {
return nil, err
}
select {
case <-ctx.Done():
return nil, ctx.Err()
default:
}
selector := op.Selector
if selector == nil {
rp := op.ReadPreference
if rp == nil {
rp = readpref.Primary()
}
selector = description.CompositeSelector([]description.ServerSelector{
description.ReadPrefSelector(rp),
description.LatencySelector(defaultLocalThreshold),
})
}
return op.Deployment.SelectServer(ctx, selector)
}
// Validate validates this operation, ensuring the fields are set properly.
func (op Operation) Validate() error {
if op.CommandFn == nil {
return InvalidOperationError{MissingField: "CommandFn"}
}
if op.Deployment == nil {
return InvalidOperationError{MissingField: "Deployment"}
}
if op.Database == "" {
return InvalidOperationError{MissingField: "Database"}
}
if op.Client != nil && !writeconcern.AckWrite(op.WriteConcern) {
return errors.New("session provided for an unacknowledged write")
}
return nil
}
// Execute runs this operation. The scratch parameter will be used and overwritten (potentially many
// times), this should mainly be used to enable pooling of byte slices.
func (op Operation) Execute(ctx context.Context, scratch []byte) error {
err := op.Validate()
if err != nil {
return err
}
srvr, err := op.selectServer(ctx)
if err != nil {
return err
}
conn, err := srvr.Connection(ctx)
if err != nil {
return err
}
defer conn.Close()
desc := description.SelectedServer{Server: conn.Description(), Kind: op.Deployment.Kind()}
scratch = scratch[:0]
if desc.WireVersion == nil || desc.WireVersion.Max < 4 {
switch op.Legacy {
case LegacyFind:
return op.legacyFind(ctx, scratch, srvr, conn, desc)
case LegacyGetMore:
return op.legacyGetMore(ctx, scratch, srvr, conn, desc)
case LegacyKillCursors:
return op.legacyKillCursors(ctx, scratch, srvr, conn, desc)
}
}
if desc.WireVersion == nil || desc.WireVersion.Max < 3 {
switch op.Legacy {
case LegacyListCollections:
return op.legacyListCollections(ctx, scratch, srvr, conn, desc)
case LegacyListIndexes:
return op.legacyListIndexes(ctx, scratch, srvr, conn, desc)
}
}
var res bsoncore.Document
var operationErr WriteCommandError
var original error
var retries int
retryable := op.retryable(desc.Server)
if retryable && op.RetryMode != nil {
switch op.Type {
case Write:
if op.Client == nil {
break
}
switch *op.RetryMode {
case RetryOnce, RetryOncePerCommand:
retries = 1
case RetryContext:
retries = -1
}
op.Client.RetryWrite = false
if *op.RetryMode > RetryNone {
op.Client.RetryWrite = true
if !op.Client.Committing && !op.Client.Aborting {
op.Client.IncrementTxnNumber()
}
}
case Read:
switch *op.RetryMode {
case RetryOnce, RetryOncePerCommand:
retries = 1
case RetryContext:
retries = -1
}
}
}
batching := op.Batches.Valid()
for {
if batching {
targetBatchSize := desc.MaxDocumentSize
maxDocSize := desc.MaxDocumentSize
if op.shouldEncrypt() {
// For client-side encryption, we want the batch to be split at 2 MiB instead of 16MiB.
// If there's only one document in the batch, it can be up to 16MiB, so we set target batch size to
// 2MiB but max document size to 16MiB. This will allow the AdvanceBatch call to create a batch
// with a single large document.
targetBatchSize = cryptMaxBsonObjectSize
}
err = op.Batches.AdvanceBatch(int(desc.MaxBatchCount), int(targetBatchSize), int(maxDocSize))
if err != nil {
// TODO(GODRIVER-982): Should we also be returning operationErr?
return err
}
}
// convert to wire message
if len(scratch) > 0 {
scratch = scratch[:0]
}
wm, startedInfo, err := op.createWireMessage(ctx, scratch, desc)
if err != nil {
return err
}
// set extra data and send event if possible
startedInfo.connID = conn.ID()
startedInfo.cmdName = op.getCommandName(startedInfo.cmd)
op.publishStartedEvent(ctx, startedInfo)
// get the moreToCome flag information before we compress
moreToCome := wiremessage.IsMsgMoreToCome(wm)
// compress wiremessage if allowed
if compressor, ok := conn.(Compressor); ok && op.canCompress(startedInfo.cmdName) {
wm, err = compressor.CompressWireMessage(wm, nil)
if err != nil {
return err
}
}
finishedInfo := finishedInformation{
cmdName: startedInfo.cmdName,
requestID: startedInfo.requestID,
startTime: time.Now(),
connID: startedInfo.connID,
}
// roundtrip using either the full roundTripper or a special one for when the moreToCome
// flag is set
var roundTrip = op.roundTrip
if moreToCome {
roundTrip = op.moreToComeRoundTrip
}
res, err = roundTrip(ctx, conn, wm)
if ep, ok := srvr.(ErrorProcessor); ok {
ep.ProcessError(err)
}
finishedInfo.response = res
finishedInfo.cmdErr = err
op.publishFinishedEvent(ctx, finishedInfo)
// Pull out $clusterTime and operationTime and update session and clock. We handle this before
// handling the error to ensure we are properly gossiping the cluster time.
op.updateClusterTimes(res)
op.updateOperationTime(res)
op.Client.UpdateRecoveryToken(bson.Raw(res))
// automatically attempt to decrypt all results if client side encryption enabled
if op.Crypt != nil {
// use decryptErr isntead of err because err is used below for retrying
var decryptErr error
res, decryptErr = op.Crypt.Decrypt(ctx, res)
if decryptErr != nil {
return decryptErr
}
}
var perr error
if op.ProcessResponseFn != nil {
perr = op.ProcessResponseFn(res, srvr, desc.Server)
}
switch tt := err.(type) {
case WriteCommandError:
if e := err.(WriteCommandError); retryable && op.Type == Write && e.UnsupportedStorageEngine() {
return ErrUnsupportedStorageEngine
}
if retryable && tt.Retryable() && retries != 0 {
retries--
original, err = err, nil
conn.Close() // Avoid leaking the connection.
srvr, err = op.selectServer(ctx)
if err != nil {
return original
}
conn, err = srvr.Connection(ctx)
if err != nil || conn == nil || !op.retryable(conn.Description()) {
if conn != nil {
conn.Close()
}
return original
}
defer conn.Close() // Avoid leaking the new connection.
if op.Client != nil && op.Client.Committing {
// Apply majority write concern for retries
op.Client.UpdateCommitTransactionWriteConcern()
op.WriteConcern = op.Client.CurrentWc
}
continue
}
// If batching is enabled and either ordered is the default (which is true) or
// explicitly set to true and we have write errors, return the errors.
if batching && (op.Batches.Ordered == nil || *op.Batches.Ordered == true) && len(tt.WriteErrors) > 0 {
return tt
}
if op.Client != nil && op.Client.Committing && tt.WriteConcernError != nil {
// When running commitTransaction we return WriteConcernErrors as an Error.
err := Error{
Name: tt.WriteConcernError.Name,
Code: int32(tt.WriteConcernError.Code),
Message: tt.WriteConcernError.Message,
}
if err.Code == 64 || err.Code == 50 || tt.WriteConcernError.Retryable() {
err.Labels = []string{UnknownTransactionCommitResult}
}
return err
}
operationErr.WriteConcernError = tt.WriteConcernError
operationErr.WriteErrors = append(operationErr.WriteErrors, tt.WriteErrors...)
case Error:
if tt.HasErrorLabel(TransientTransactionError) || tt.HasErrorLabel(UnknownTransactionCommitResult) {
op.Client.ClearPinnedServer()
}
if e := err.(Error); retryable && op.Type == Write && e.UnsupportedStorageEngine() {
return ErrUnsupportedStorageEngine
}
if retryable && tt.Retryable() && retries != 0 {
retries--
original, err = err, nil
conn.Close() // Avoid leaking the connection.
srvr, err = op.selectServer(ctx)
if err != nil {
return original
}
conn, err = srvr.Connection(ctx)
if err != nil || conn == nil || !op.retryable(conn.Description()) {
if conn != nil {
conn.Close()
}
return original
}
defer conn.Close() // Avoid leaking the new connection.
if op.Client != nil && op.Client.Committing {
// Apply majority write concern for retries
op.Client.UpdateCommitTransactionWriteConcern()
op.WriteConcern = op.Client.CurrentWc
}
continue
}
if op.Client != nil && op.Client.Committing && (tt.Retryable() || tt.Code == 50) {
// If we got a retryable error or MaxTimeMSExpired error, we add UnknownTransactionCommitResult.
tt.Labels = append(tt.Labels, UnknownTransactionCommitResult)
}
return tt
case nil:
if moreToCome {
return ErrUnacknowledgedWrite
}
if perr != nil {
return perr
}
default:
return err
}
if batching && len(op.Batches.Documents) > 0 {
if retryable && op.Client != nil && op.RetryMode != nil {
if *op.RetryMode > RetryNone {
op.Client.IncrementTxnNumber()
}
if *op.RetryMode == RetryOncePerCommand {
retries = 1
}
}
op.Batches.ClearBatch()
continue
}
break
}
if len(operationErr.WriteErrors) > 0 || operationErr.WriteConcernError != nil {
return operationErr
}
return nil
}
// Retryable writes are supported if the server supports sessions, the operation is not
// within a transaction, and the write is acknowledged
func (op Operation) retryable(desc description.Server) bool {
switch op.Type {
case Write:
if op.Client != nil && (op.Client.Committing || op.Client.Aborting) {
return true
}
if op.Deployment.SupportsRetryWrites() &&
desc.WireVersion != nil && desc.WireVersion.Max >= 6 &&
op.Client != nil && !(op.Client.TransactionInProgress() || op.Client.TransactionStarting()) &&
writeconcern.AckWrite(op.WriteConcern) {
return true
}
case Read:
if op.Client != nil && (op.Client.Committing || op.Client.Aborting) {
return true
}
if desc.WireVersion != nil && desc.WireVersion.Max >= 6 &&
(op.Client == nil || !(op.Client.TransactionInProgress() || op.Client.TransactionStarting())) {
return true
}
}
return false
}
// roundTrip writes a wiremessage to the connection and then reads a wiremessage. The wm parameter
// is reused when reading the wiremessage.
func (op Operation) roundTrip(ctx context.Context, conn Connection, wm []byte) ([]byte, error) {
err := conn.WriteWireMessage(ctx, wm)
if err != nil {
labels := []string{NetworkError}
if op.Client != nil {
op.Client.MarkDirty()
}
if op.Client != nil && op.Client.TransactionRunning() && !op.Client.Committing {
labels = append(labels, TransientTransactionError)
}
if op.Client != nil && op.Client.Committing {
labels = append(labels, UnknownTransactionCommitResult)
}
return nil, Error{Message: err.Error(), Labels: labels, Wrapped: err}
}
wm, err = conn.ReadWireMessage(ctx, wm[:0])
if err != nil {
labels := []string{NetworkError}
if op.Client != nil {
op.Client.MarkDirty()
}
if op.Client != nil && op.Client.TransactionRunning() && !op.Client.Committing {
labels = append(labels, TransientTransactionError)
}
if op.Client != nil && op.Client.Committing {
labels = append(labels, UnknownTransactionCommitResult)
}
return nil, Error{Message: err.Error(), Labels: labels, Wrapped: err}
}
// decompress wiremessage
wm, err = op.decompressWireMessage(wm)
if err != nil {
return nil, err
}
// decode
res, err := op.decodeResult(wm)
// Pull out $clusterTime and operationTime and update session and clock. We handle this before
// handling the error to ensure we are properly gossiping the cluster time.
op.updateClusterTimes(res)
op.updateOperationTime(res)
return res, err
}
// moreToComeRoundTrip writes a wiremessage to the provided connection. This is used when an OP_MSG is
// being sent with the moreToCome bit set.
func (op *Operation) moreToComeRoundTrip(ctx context.Context, conn Connection, wm []byte) ([]byte, error) {
err := conn.WriteWireMessage(ctx, wm)
if err != nil {
if op.Client != nil {
op.Client.MarkDirty()
}
err = Error{Message: err.Error(), Labels: []string{TransientTransactionError, NetworkError}, Wrapped: err}
}
return bsoncore.BuildDocument(nil, bsoncore.AppendInt32Element(nil, "ok", 1)), err
}
// decompressWireMessage handles decompressing a wiremessage. If the wiremessage
// is not compressed, this method will return the wiremessage.
func (Operation) decompressWireMessage(wm []byte) ([]byte, error) {
// read the header and ensure this is a compressed wire message
length, reqid, respto, opcode, rem, ok := wiremessage.ReadHeader(wm)
if !ok || len(wm) < int(length) {
return nil, errors.New("malformed wire message: insufficient bytes")
}
if opcode != wiremessage.OpCompressed {
return wm, nil
}
// get the original opcode and uncompressed size
opcode, rem, ok = wiremessage.ReadCompressedOriginalOpCode(rem)
if !ok {
return nil, errors.New("malformed OP_COMPRESSED: missing original opcode")
}
uncompressedSize, rem, ok := wiremessage.ReadCompressedUncompressedSize(rem)
if !ok {
return nil, errors.New("malformed OP_COMPRESSED: missing uncompressed size")
}
// get the compressor ID and decompress the message
compressorID, rem, ok := wiremessage.ReadCompressedCompressorID(rem)
if !ok {
return nil, errors.New("malformed OP_COMPRESSED: missing compressor ID")
}
compressedSize := length - 25 // header (16) + original opcode (4) + uncompressed size (4) + compressor ID (1)
// return the original wiremessage
msg, rem, ok := wiremessage.ReadCompressedCompressedMessage(rem, compressedSize)
if !ok {
return nil, errors.New("malformed OP_COMPRESSED: insufficient bytes for compressed wiremessage")
}
header := make([]byte, 0, uncompressedSize+16)
header = wiremessage.AppendHeader(header, uncompressedSize+16, reqid, respto, opcode)
opts := CompressionOpts{
Compressor: compressorID,
UncompressedSize: uncompressedSize,
}
uncompressed, err := DecompressPayload(msg, opts)
if err != nil {
return nil, err
}
return append(header, uncompressed...), nil
}
func (op Operation) createWireMessage(ctx context.Context, dst []byte,
desc description.SelectedServer) ([]byte, startedInformation, error) {
if desc.WireVersion == nil || desc.WireVersion.Max < wiremessage.OpmsgWireVersion {
return op.createQueryWireMessage(dst, desc)
}
return op.createMsgWireMessage(ctx, dst, desc)
}
func (op Operation) addBatchArray(dst []byte) []byte {
aidx, dst := bsoncore.AppendArrayElementStart(dst, op.Batches.Identifier)
for i, doc := range op.Batches.Current {
dst = bsoncore.AppendDocumentElement(dst, strconv.Itoa(i), doc)
}
dst, _ = bsoncore.AppendArrayEnd(dst, aidx)
return dst
}
func (op Operation) createQueryWireMessage(dst []byte, desc description.SelectedServer) ([]byte, startedInformation, error) {
var info startedInformation
flags := op.slaveOK(desc)
var wmindex int32
info.requestID = wiremessage.NextRequestID()
wmindex, dst = wiremessage.AppendHeaderStart(dst, info.requestID, 0, wiremessage.OpQuery)
dst = wiremessage.AppendQueryFlags(dst, flags)
// FullCollectionName
dst = append(dst, op.Database...)
dst = append(dst, dollarCmd[:]...)
dst = append(dst, 0x00)
dst = wiremessage.AppendQueryNumberToSkip(dst, 0)
dst = wiremessage.AppendQueryNumberToReturn(dst, -1)
wrapper := int32(-1)
rp, err := op.createReadPref(desc.Server.Kind, desc.Kind, true)
if err != nil {
return dst, info, err
}
if len(rp) > 0 {
wrapper, dst = bsoncore.AppendDocumentStart(dst)
dst = bsoncore.AppendHeader(dst, bsontype.EmbeddedDocument, "$query")
}
idx, dst := bsoncore.AppendDocumentStart(dst)
dst, err = op.CommandFn(dst, desc)
if err != nil {
return dst, info, err
}
if op.Batches != nil && len(op.Batches.Current) > 0 {
dst = op.addBatchArray(dst)
}
dst, err = op.addReadConcern(dst, desc)
if err != nil {
return dst, info, err
}
dst, err = op.addWriteConcern(dst, desc)
if err != nil {
return dst, info, err
}
dst, err = op.addSession(dst, desc)
if err != nil {
return dst, info, err
}
dst = op.addClusterTime(dst, desc)
dst, _ = bsoncore.AppendDocumentEnd(dst, idx)
// Command monitoring only reports the document inside $query
info.cmd = dst[idx:]
if len(rp) > 0 {
var err error
dst = bsoncore.AppendDocumentElement(dst, "$readPreference", rp)
dst, err = bsoncore.AppendDocumentEnd(dst, wrapper)
if err != nil {
return dst, info, err
}
}
return bsoncore.UpdateLength(dst, wmindex, int32(len(dst[wmindex:]))), info, nil
}
func (op Operation) createMsgWireMessage(ctx context.Context, dst []byte, desc description.SelectedServer) ([]byte, startedInformation, error) {
var info startedInformation
var flags wiremessage.MsgFlag
var wmindex int32
// We set the MoreToCome bit if we have a write concern, it's unacknowledged, and we either
// aren't batching or we are encoding the last batch.
if op.WriteConcern != nil && !writeconcern.AckWrite(op.WriteConcern) && (op.Batches == nil || len(op.Batches.Documents) == 0) {
flags = wiremessage.MoreToCome
}
info.requestID = wiremessage.NextRequestID()
wmindex, dst = wiremessage.AppendHeaderStart(dst, info.requestID, 0, wiremessage.OpMsg)
dst = wiremessage.AppendMsgFlags(dst, flags)
// Body
dst = wiremessage.AppendMsgSectionType(dst, wiremessage.SingleDocument)
idx, dst := bsoncore.AppendDocumentStart(dst)
dst, err := op.addCommandFields(ctx, dst, desc)
if err != nil {
return dst, info, err
}
dst, err = op.addReadConcern(dst, desc)
if err != nil {
return dst, info, err
}
dst, err = op.addWriteConcern(dst, desc)
if err != nil {
return dst, info, err
}
dst, err = op.addSession(dst, desc)
if err != nil {
return dst, info, err
}
dst = op.addClusterTime(dst, desc)
dst = bsoncore.AppendStringElement(dst, "$db", op.Database)
rp, err := op.createReadPref(desc.Server.Kind, desc.Kind, false)
if err != nil {
return dst, info, err
}
if len(rp) > 0 {
dst = bsoncore.AppendDocumentElement(dst, "$readPreference", rp)
}
dst, _ = bsoncore.AppendDocumentEnd(dst, idx)
// The command document for monitoring shouldn't include the type 1 payload as a document sequence
info.cmd = dst[idx:]
// add batch as a document sequence if auto encryption is not enabled
// if auto encryption is enabled, the batch will already be an array in the command document
if !op.shouldEncrypt() && op.Batches != nil && len(op.Batches.Current) > 0 {
info.documentSequenceIncluded = true
dst = wiremessage.AppendMsgSectionType(dst, wiremessage.DocumentSequence)
idx, dst = bsoncore.ReserveLength(dst)
dst = append(dst, op.Batches.Identifier...)
dst = append(dst, 0x00)
for _, doc := range op.Batches.Current {
dst = append(dst, doc...)
}
dst = bsoncore.UpdateLength(dst, idx, int32(len(dst[idx:])))
}
return bsoncore.UpdateLength(dst, wmindex, int32(len(dst[wmindex:]))), info, nil
}
// addCommandFields adds the fields for a command to the wire message in dst. This assumes that the start of the document
// has already been added and does not add the final 0 byte.
func (op Operation) addCommandFields(ctx context.Context, dst []byte, desc description.SelectedServer) ([]byte, error) {
if !op.shouldEncrypt() {
return op.CommandFn(dst, desc)
}
if desc.WireVersion.Max < cryptMinWireVersion {
return dst, errors.New("auto-encryption requires a MongoDB version of 4.2")
}
// create temporary command document
cidx, cmdDst := bsoncore.AppendDocumentStart(nil)
var err error
cmdDst, err = op.CommandFn(cmdDst, desc)
if err != nil {
return dst, err
}
// use a BSON array instead of a type 1 payload because mongocryptd will convert to arrays regardless
if op.Batches != nil && len(op.Batches.Current) > 0 {
cmdDst = op.addBatchArray(cmdDst)
}
cmdDst, _ = bsoncore.AppendDocumentEnd(cmdDst, cidx)
// encrypt the command
encrypted, err := op.Crypt.Encrypt(ctx, op.Database, cmdDst)
if err != nil {
return dst, err
}
// append encrypted command to original destination, removing the first 4 bytes (length) and final byte (terminator)
dst = append(dst, encrypted[4:len(encrypted)-1]...)
return dst, nil
}
func (op Operation) addReadConcern(dst []byte, desc description.SelectedServer) ([]byte, error) {
if op.MinimumReadConcernWireVersion > 0 && (desc.WireVersion == nil || !desc.WireVersion.Includes(op.MinimumReadConcernWireVersion)) {
return dst, nil
}
rc := op.ReadConcern
client := op.Client
// Starting transaction's read concern overrides all others
if client != nil && client.TransactionStarting() && client.CurrentRc != nil {
rc = client.CurrentRc
}
// start transaction must append afterclustertime IF causally consistent and operation time exists
if rc == nil && client != nil && client.TransactionStarting() && client.Consistent && client.OperationTime != nil {
rc = readconcern.New()
}
if rc == nil {
return dst, nil
}
_, data, err := rc.MarshalBSONValue() // always returns a document
if err != nil {
return dst, err
}
if description.SessionsSupported(desc.WireVersion) && client != nil && client.Consistent && client.OperationTime != nil {
data = data[:len(data)-1] // remove the null byte
data = bsoncore.AppendTimestampElement(data, "afterClusterTime", client.OperationTime.T, client.OperationTime.I)
data, _ = bsoncore.AppendDocumentEnd(data, 0)
}
if len(data) == bsoncore.EmptyDocumentLength {
return dst, nil
}
return bsoncore.AppendDocumentElement(dst, "readConcern", data), nil
}
func (op Operation) addWriteConcern(dst []byte, desc description.SelectedServer) ([]byte, error) {
if op.MinimumWriteConcernWireVersion > 0 && (desc.WireVersion == nil || !desc.WireVersion.Includes(op.MinimumWriteConcernWireVersion)) {
return dst, nil
}
wc := op.WriteConcern
if wc == nil {
return dst, nil
}
t, data, err := wc.MarshalBSONValue()
if err == writeconcern.ErrEmptyWriteConcern {
return dst, nil
}
if err != nil {
return dst, err
}
return append(bsoncore.AppendHeader(dst, t, "writeConcern"), data...), nil
}
func (op Operation) addSession(dst []byte, desc description.SelectedServer) ([]byte, error) {
client := op.Client
if client == nil || !description.SessionsSupported(desc.WireVersion) || desc.SessionTimeoutMinutes == 0 {
return dst, nil
}
if client.Terminated {
return dst, session.ErrSessionEnded
}
lsid, _ := client.SessionID.MarshalBSON()
dst = bsoncore.AppendDocumentElement(dst, "lsid", lsid)
var addedTxnNumber bool
if op.Type == Write && client.RetryWrite {
addedTxnNumber = true
dst = bsoncore.AppendInt64Element(dst, "txnNumber", op.Client.TxnNumber)
}
if client.TransactionRunning() || client.RetryingCommit {
if !addedTxnNumber {
dst = bsoncore.AppendInt64Element(dst, "txnNumber", op.Client.TxnNumber)
}
if client.TransactionStarting() {
dst = bsoncore.AppendBooleanElement(dst, "startTransaction", true)
}
dst = bsoncore.AppendBooleanElement(dst, "autocommit", false)
}
client.ApplyCommand(desc.Server)
return dst, nil
}
func (op Operation) addClusterTime(dst []byte, desc description.SelectedServer) []byte {
client, clock := op.Client, op.Clock
if (clock == nil && client == nil) || !description.SessionsSupported(desc.WireVersion) {
return dst
}
clusterTime := clock.GetClusterTime()
if client != nil {
clusterTime = session.MaxClusterTime(clusterTime, client.ClusterTime)
}
if clusterTime == nil {
return dst
}
val, err := clusterTime.LookupErr("$clusterTime")
if err != nil {
return dst
}
return append(bsoncore.AppendHeader(dst, val.Type, "$clusterTime"), val.Value...)
// return bsoncore.AppendDocumentElement(dst, "$clusterTime", clusterTime)
}
// updateClusterTimes updates the cluster times for the session and cluster clock attached to this
// operation. While the session's AdvanceClusterTime may return an error, this method does not
// because an error being returned from this method will not be returned further up.
func (op Operation) updateClusterTimes(response bsoncore.Document) {
// Extract cluster time.
value, err := response.LookupErr("$clusterTime")
if err != nil {
// $clusterTime not included by the server
return
}
clusterTime := bsoncore.BuildDocumentFromElements(nil, bsoncore.AppendValueElement(nil, "$clusterTime", value))
sess, clock := op.Client, op.Clock
if sess != nil {
_ = sess.AdvanceClusterTime(bson.Raw(clusterTime))
}
if clock != nil {
clock.AdvanceClusterTime(bson.Raw(clusterTime))
}
}
// updateOperationTime updates the operation time on the session attached to this operation. While
// the session's AdvanceOperationTime method may return an error, this method does not because an
// error being returned from this method will not be returned further up.
func (op Operation) updateOperationTime(response bsoncore.Document) {
sess := op.Client
if sess == nil {
return
}
opTimeElem, err := response.LookupErr("operationTime")
if err != nil {
// operationTime not included by the server
return
}
t, i := opTimeElem.Timestamp()
_ = sess.AdvanceOperationTime(&primitive.Timestamp{
T: t,
I: i,
})
}
func (op Operation) getReadPrefBasedOnTransaction() (*readpref.ReadPref, error) {
if op.Client != nil && op.Client.TransactionRunning() {
// Transaction's read preference always takes priority
rp := op.Client.CurrentRp
// Reads in a transaction must have read preference primary
// This must not be checked in startTransaction
if rp != nil && !op.Client.TransactionStarting() && rp.Mode() != readpref.PrimaryMode {
return nil, ErrNonPrimaryReadPref
}
return rp, nil
}
return op.ReadPreference, nil
}
func (op Operation) createReadPref(serverKind description.ServerKind, topologyKind description.TopologyKind, isOpQuery bool) (bsoncore.Document, error) {
if serverKind == description.Standalone || (isOpQuery && serverKind != description.Mongos) || op.Type == Write {
// Don't send read preference for non-mongos when using OP_QUERY or for all standalones
return nil, nil