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
|
/**
* Copyright (C) 2013-2014 MongoDB Inc.
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License, version 3,
* as published by the Free Software Foundation.
*
* 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
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
* 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 GNU Affero General 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::kQuery
#include "mongo/platform/basic.h"
#include "mongo/db/query/find.h"
#include "mongo/base/error_codes.h"
#include "mongo/client/dbclientinterface.h"
#include "mongo/db/auth/authorization_session.h"
#include "mongo/db/catalog/collection.h"
#include "mongo/db/catalog/database_holder.h"
#include "mongo/db/client.h"
#include "mongo/db/clientcursor.h"
#include "mongo/db/commands.h"
#include "mongo/db/curop.h"
#include "mongo/db/db_raii.h"
#include "mongo/db/exec/filter.h"
#include "mongo/db/exec/working_set_common.h"
#include "mongo/db/keypattern.h"
#include "mongo/db/matcher/extensions_callback_real.h"
#include "mongo/db/query/explain.h"
#include "mongo/db/query/find_common.h"
#include "mongo/db/query/get_executor.h"
#include "mongo/db/query/internal_plans.h"
#include "mongo/db/query/plan_summary_stats.h"
#include "mongo/db/query/query_planner_params.h"
#include "mongo/db/repl/replication_coordinator_global.h"
#include "mongo/db/s/collection_sharding_state.h"
#include "mongo/db/server_options.h"
#include "mongo/db/server_parameters.h"
#include "mongo/db/service_context.h"
#include "mongo/db/stats/top.h"
#include "mongo/db/storage/storage_options.h"
#include "mongo/s/chunk_version.h"
#include "mongo/s/stale_exception.h"
#include "mongo/stdx/memory.h"
#include "mongo/util/fail_point_service.h"
#include "mongo/util/log.h"
#include "mongo/util/mongoutils/str.h"
#include "mongo/util/scopeguard.h"
namespace mongo {
using std::unique_ptr;
using stdx::make_unique;
// Failpoint for checking whether we've received a getmore.
MONGO_FP_DECLARE(failReceivedGetmore);
bool isCursorTailable(const ClientCursor* cursor) {
return cursor->queryOptions() & QueryOption_CursorTailable;
}
bool isCursorAwaitData(const ClientCursor* cursor) {
return cursor->queryOptions() & QueryOption_AwaitData;
}
bool shouldSaveCursor(OperationContext* opCtx,
const Collection* collection,
PlanExecutor::ExecState finalState,
PlanExecutor* exec) {
if (PlanExecutor::FAILURE == finalState || PlanExecutor::DEAD == finalState) {
return false;
}
const QueryRequest& qr = exec->getCanonicalQuery()->getQueryRequest();
if (!qr.wantMore()) {
return false;
}
// We keep a tailable cursor around unless the collection we're tailing has no
// records.
//
// SERVER-13955: we should be able to create a tailable cursor that waits on
// an empty collection. Right now we do not keep a cursor if the collection
// has zero records.
if (qr.isTailable()) {
return collection && collection->numRecords(opCtx) != 0U;
}
return !exec->isEOF();
}
bool shouldSaveCursorGetMore(PlanExecutor::ExecState finalState,
PlanExecutor* exec,
bool isTailable) {
if (PlanExecutor::FAILURE == finalState || PlanExecutor::DEAD == finalState) {
return false;
}
if (isTailable) {
return true;
}
return !exec->isEOF();
}
void beginQueryOp(OperationContext* opCtx,
const NamespaceString& nss,
const BSONObj& queryObj,
long long ntoreturn,
long long ntoskip) {
auto curOp = CurOp::get(opCtx);
curOp->debug().ntoreturn = ntoreturn;
curOp->debug().ntoskip = ntoskip;
stdx::lock_guard<Client> lk(*opCtx->getClient());
curOp->setOpDescription_inlock(queryObj);
curOp->setNS_inlock(nss.ns());
}
void endQueryOp(OperationContext* opCtx,
Collection* collection,
const PlanExecutor& exec,
long long numResults,
CursorId cursorId) {
auto curOp = CurOp::get(opCtx);
// Fill out basic CurOp query exec properties.
curOp->debug().nreturned = numResults;
curOp->debug().cursorid = (0 == cursorId ? -1 : cursorId);
curOp->debug().cursorExhausted = (0 == cursorId);
// Fill out CurOp based on explain summary statistics.
PlanSummaryStats summaryStats;
Explain::getSummaryStats(exec, &summaryStats);
curOp->debug().setPlanSummaryMetrics(summaryStats);
if (collection) {
collection->infoCache()->notifyOfQuery(opCtx, summaryStats.indexesUsed);
}
if (curOp->shouldDBProfile()) {
BSONObjBuilder statsBob;
Explain::getWinningPlanStats(&exec, &statsBob);
curOp->debug().execStats = statsBob.obj();
}
}
namespace {
/**
* Uses 'cursor' to fill out 'bb' with the batch of result documents to
* be returned by this getMore.
*
* Returns the number of documents in the batch in 'numResults', which must be initialized to
* zero by the caller. Returns the final ExecState returned by the cursor in *state. Returns
* whether or not to save the ClientCursor in 'shouldSaveCursor'. Returns the slave's time to
* read until in 'slaveReadTill' (for master/slave).
*
* Returns an OK status if the batch was successfully generated, and a non-OK status if the
* PlanExecutor encounters a failure.
*/
void generateBatch(int ntoreturn,
ClientCursor* cursor,
BufBuilder* bb,
int* numResults,
Timestamp* slaveReadTill,
PlanExecutor::ExecState* state) {
PlanExecutor* exec = cursor->getExecutor();
BSONObj obj;
while (!FindCommon::enoughForGetMore(ntoreturn, *numResults) &&
PlanExecutor::ADVANCED == (*state = exec->getNext(&obj, NULL))) {
// If we can't fit this result inside the current batch, then we stash it for later.
if (!FindCommon::haveSpaceForNext(obj, *numResults, bb->len())) {
exec->enqueue(obj);
break;
}
// Add result to output buffer.
bb->appendBuf((void*)obj.objdata(), obj.objsize());
// Count the result.
(*numResults)++;
// Possibly note slave's position in the oplog.
if (cursor->queryOptions() & QueryOption_OplogReplay) {
BSONElement e = obj["ts"];
if (BSONType::Date == e.type() || BSONType::bsonTimestamp == e.type()) {
*slaveReadTill = e.timestamp();
}
}
}
// Propagate any errors to the caller.
if (PlanExecutor::FAILURE == *state) {
error() << "getMore executor error, stats: " << redact(Explain::getWinningPlanStats(exec));
uasserted(17406, "getMore executor error: " + WorkingSetCommon::toStatusString(obj));
} else if (PlanExecutor::DEAD == *state) {
uasserted(ErrorCodes::QueryPlanKilled,
str::stream() << "PlanExecutor killed: "
<< WorkingSetCommon::toStatusString(obj));
}
}
} // namespace
/**
* Called by db/instance.cpp. This is the getMore entry point.
*/
Message getMore(OperationContext* opCtx,
const char* ns,
int ntoreturn,
long long cursorid,
bool* exhaust,
bool* isCursorAuthorized) {
invariant(ntoreturn >= 0);
CurOp& curOp = *CurOp::get(opCtx);
curOp.ensureStarted();
// For testing, we may want to fail if we receive a getmore.
if (MONGO_FAIL_POINT(failReceivedGetmore)) {
invariant(0);
}
*exhaust = false;
const NamespaceString nss(ns);
// Cursors come in one of two flavors:
// - Cursors owned by the collection cursor manager, such as those generated via the find
// command. For these cursors, we hold the appropriate collection lock for the duration of the
// getMore using AutoGetCollectionForRead.
// - Cursors owned by the global cursor manager, such as those generated via the aggregate
// command. These cursors either hold no collection state or manage their collection state
// internally, so we acquire no locks.
//
// While we only need to acquire locks in the case of a cursor which is *not* globally owned, we
// need to create an AutoStatsTracker in either case. This is responsible for updating
// statistics in CurOp and Top. We avoid using AutoGetCollectionForReadCommand because we may
// need to drop and reacquire locks when the cursor is awaitData, but we don't want to update
// the stats twice.
//
// Note that we acquire our locks before our ClientCursorPin, in order to ensure that the pin's
// destructor is called before the lock's destructor (if there is one) so that the cursor
// cleanup can occur under the lock.
boost::optional<AutoGetCollectionForRead> readLock;
boost::optional<AutoStatsTracker> statsTracker;
CursorManager* cursorManager;
if (CursorManager::isGloballyManagedCursor(cursorid)) {
cursorManager = CursorManager::getGlobalCursorManager();
if (boost::optional<NamespaceString> nssForCurOp = nss.isGloballyManagedNamespace()
? nss.getTargetNSForGloballyManagedNamespace()
: nss) {
AutoGetDb autoDb(opCtx, nssForCurOp->db(), MODE_IS);
const auto profilingLevel = autoDb.getDb()
? boost::optional<int>{autoDb.getDb()->getProfilingLevel()}
: boost::none;
statsTracker.emplace(opCtx, *nssForCurOp, Top::LockType::NotLocked, profilingLevel);
auto view = autoDb.getDb()
? autoDb.getDb()->getViewCatalog()->lookup(opCtx, nssForCurOp->ns())
: nullptr;
uassert(
ErrorCodes::CommandNotSupportedOnView,
str::stream() << "Namespace " << nssForCurOp->ns()
<< " is a view. OP_GET_MORE operations are not supported on views. "
<< "Only clients which support the getMore command can be used to "
"query views.",
!view);
}
} else {
readLock.emplace(opCtx, nss);
const int doNotChangeProfilingLevel = 0;
statsTracker.emplace(opCtx,
nss,
Top::LockType::ReadLocked,
readLock->getDb() ? readLock->getDb()->getProfilingLevel()
: doNotChangeProfilingLevel);
Collection* collection = readLock->getCollection();
uassert(
ErrorCodes::OperationFailed, "collection dropped between getMore calls", collection);
cursorManager = collection->getCursorManager();
}
LOG(5) << "Running getMore, cursorid: " << cursorid;
// This checks to make sure the operation is allowed on a replicated node. Since we are not
// passing in a query object (necessary to check SlaveOK query option), the only state where
// reads are allowed is PRIMARY (or master in master/slave). This function uasserts if
// reads are not okay.
Status status =
repl::getGlobalReplicationCoordinator()->checkCanServeReadsFor_UNSAFE(opCtx, nss, true);
uassertStatusOK(status);
// A pin performs a CC lookup and if there is a CC, increments the CC's pin value so it
// doesn't time out. Also informs ClientCursor that there is somebody actively holding the
// CC, so don't delete it.
auto ccPin = cursorManager->pinCursor(opCtx, cursorid);
// These are set in the QueryResult msg we return.
int resultFlags = ResultFlag_AwaitCapable;
int numResults = 0;
int startingResult = 0;
const int InitialBufSize =
512 + sizeof(QueryResult::Value) + FindCommon::kMaxBytesToReturnToClientAtOnce;
BufBuilder bb(InitialBufSize);
bb.skip(sizeof(QueryResult::Value));
if (!ccPin.isOK()) {
if (ccPin == ErrorCodes::CursorNotFound) {
cursorid = 0;
resultFlags = ResultFlag_CursorNotFound;
} else {
invariant(ccPin == ErrorCodes::QueryPlanKilled);
uassertStatusOK(ccPin.getStatus());
}
} else {
ClientCursor* cc = ccPin.getValue().getCursor();
// Check for spoofing of the ns such that it does not match the one originally
// there for the cursor.
uassert(ErrorCodes::Unauthorized,
str::stream() << "Requested getMore on namespace " << ns << ", but cursor "
<< cursorid
<< " belongs to namespace "
<< cc->nss().ns(),
nss == cc->nss());
// A user can only call getMore on their own cursor. If there were multiple users
// authenticated when the cursor was created, then at least one of them must be
// authenticated in order to run getMore on the cursor.
uassert(ErrorCodes::Unauthorized,
str::stream() << "cursor id " << cursorid
<< " was not created by the authenticated user",
AuthorizationSession::get(opCtx->getClient())
->isCoauthorizedWith(cc->getAuthenticatedUsers()));
*isCursorAuthorized = true;
if (cc->isReadCommitted())
uassertStatusOK(opCtx->recoveryUnit()->setReadFromMajorityCommittedSnapshot());
uassert(40548,
"OP_GET_MORE operations are not supported on tailable aggregations. Only clients "
"which support the getMore command can be used on tailable aggregations.",
readLock || !isCursorAwaitData(cc));
// If the operation that spawned this cursor had a time limit set, apply leftover
// time to this getmore.
if (cc->getLeftoverMaxTimeMicros() < Microseconds::max()) {
uassert(40136,
"Illegal attempt to set operation deadline within DBDirectClient",
!opCtx->getClient()->isInDirectClient());
opCtx->setDeadlineAfterNowBy(cc->getLeftoverMaxTimeMicros());
}
opCtx->checkForInterrupt(); // May trigger maxTimeAlwaysTimeOut fail point.
cc->updateSlaveLocation(opCtx);
// If we're replaying the oplog, we save the last time that we read.
Timestamp slaveReadTill;
// What number result are we starting at? Used to fill out the reply.
startingResult = cc->pos();
uint64_t notifierVersion = 0;
std::shared_ptr<CappedInsertNotifier> notifier;
if (isCursorAwaitData(cc)) {
invariant(readLock->getCollection()->isCapped());
// Retrieve the notifier which we will wait on until new data arrives. We make sure
// to do this in the lock because once we drop the lock it is possible for the
// collection to become invalid. The notifier itself will outlive the collection if
// the collection is dropped, as we keep a shared_ptr to it.
notifier = readLock->getCollection()->getCappedInsertNotifier();
// Must get the version before we call generateBatch in case a write comes in after
// that call and before we call wait on the notifier.
notifierVersion = notifier->getVersion();
}
PlanExecutor* exec = cc->getExecutor();
exec->reattachToOperationContext(opCtx);
exec->restoreState();
auto planSummary = Explain::getPlanSummary(exec);
{
stdx::lock_guard<Client> lk(*opCtx->getClient());
curOp.setPlanSummary_inlock(planSummary);
// Ensure that the original query object is available in the slow query log, profiler
// and currentOp. Upconvert _query to resemble a getMore command, and set the original
// command or upconverted legacy query in the originatingCommand field.
curOp.setOpDescription_inlock(upconvertGetMoreEntry(nss, cursorid, ntoreturn));
curOp.setOriginatingCommand_inlock(cc->getOriginatingCommandObj());
}
PlanExecutor::ExecState state;
// We report keysExamined and docsExamined to OpDebug for a given getMore operation. To
// obtain these values we need to take a diff of the pre-execution and post-execution
// metrics, as they accumulate over the course of a cursor's lifetime.
PlanSummaryStats preExecutionStats;
Explain::getSummaryStats(*exec, &preExecutionStats);
generateBatch(ntoreturn, cc, &bb, &numResults, &slaveReadTill, &state);
// If this is an await data cursor, and we hit EOF without generating any results, then
// we block waiting for new data to arrive.
if (isCursorAwaitData(cc) && state == PlanExecutor::IS_EOF && numResults == 0) {
// Save the PlanExecutor and drop our locks.
exec->saveState();
readLock.reset();
// Block waiting for data for up to 1 second. Time spent blocking is not counted towards
// the total operation latency.
curOp.pauseTimer();
Seconds timeout(1);
notifier->wait(notifierVersion, timeout);
notifier.reset();
curOp.resumeTimer();
// Reacquiring locks.
readLock.emplace(opCtx, nss);
exec->restoreState();
// We woke up because either the timed_wait expired, or there was more data. Either
// way, attempt to generate another batch of results.
generateBatch(ntoreturn, cc, &bb, &numResults, &slaveReadTill, &state);
}
PlanSummaryStats postExecutionStats;
Explain::getSummaryStats(*exec, &postExecutionStats);
postExecutionStats.totalKeysExamined -= preExecutionStats.totalKeysExamined;
postExecutionStats.totalDocsExamined -= preExecutionStats.totalDocsExamined;
curOp.debug().setPlanSummaryMetrics(postExecutionStats);
// We do not report 'execStats' for aggregation or other globally managed cursors, both in
// the original request and subsequent getMore. It would be useful to have this information
// for an aggregation, but the source PlanExecutor could be destroyed before we know whether
// we need execStats and we do not want to generate for all operations due to cost.
if (!CursorManager::isGloballyManagedCursor(cursorid) && curOp.shouldDBProfile()) {
BSONObjBuilder execStatsBob;
Explain::getWinningPlanStats(exec, &execStatsBob);
curOp.debug().execStats = execStatsBob.obj();
}
// Our two possible ClientCursorPin cleanup paths are:
// 1) If the cursor is not going to be saved, we call deleteUnderlying() on the pin.
// 2) If the cursor is going to be saved, we simply let the pin go out of scope. In this
// case, the pin's destructor will be invoked, which will call release() on the pin.
// Because our ClientCursorPin is declared after our lock is declared, this will happen
// under the lock if any locking was necessary.
if (!shouldSaveCursorGetMore(state, exec, isCursorTailable(cc))) {
ccPin.getValue().deleteUnderlying();
// cc is now invalid, as is the executor
cursorid = 0;
cc = nullptr;
curOp.debug().cursorExhausted = true;
LOG(5) << "getMore NOT saving client cursor, ended with state "
<< PlanExecutor::statestr(state);
} else {
// Continue caching the ClientCursor.
cc->incPos(numResults);
exec->saveState();
exec->detachFromOperationContext();
LOG(5) << "getMore saving client cursor ended with state "
<< PlanExecutor::statestr(state);
// Possibly note slave's position in the oplog.
if ((cc->queryOptions() & QueryOption_OplogReplay) && !slaveReadTill.isNull()) {
cc->slaveReadTill(slaveReadTill);
}
*exhaust = cc->queryOptions() & QueryOption_Exhaust;
// We assume that cursors created through a DBDirectClient are always used from their
// original OperationContext, so we do not need to move time to and from the cursor.
if (!opCtx->getClient()->isInDirectClient()) {
// If the getmore had a time limit, remaining time is "rolled over" back to the
// cursor (for use by future getmore ops).
cc->setLeftoverMaxTimeMicros(opCtx->getRemainingMaxTimeMicros());
}
}
}
QueryResult::View qr = bb.buf();
qr.msgdata().setLen(bb.len());
qr.msgdata().setOperation(opReply);
qr.setResultFlags(resultFlags);
qr.setCursorId(cursorid);
qr.setStartingFrom(startingResult);
qr.setNReturned(numResults);
LOG(5) << "getMore returned " << numResults << " results\n";
return Message(bb.release());
}
std::string runQuery(OperationContext* opCtx,
QueryMessage& q,
const NamespaceString& nss,
Message& result) {
CurOp& curOp = *CurOp::get(opCtx);
curOp.ensureStarted();
uassert(ErrorCodes::InvalidNamespace,
str::stream() << "Invalid ns [" << nss.ns() << "]",
nss.isValid());
invariant(!nss.isCommand());
// Set CurOp information.
beginQueryOp(opCtx, nss, q.query, q.ntoreturn, q.ntoskip);
// Parse the qm into a CanonicalQuery.
const boost::intrusive_ptr<ExpressionContext> expCtx;
auto statusWithCQ =
CanonicalQuery::canonicalize(opCtx,
q,
expCtx,
ExtensionsCallbackReal(opCtx, &nss),
MatchExpressionParser::kAllowAllSpecialFeatures &
~MatchExpressionParser::AllowedFeatures::kExpr);
if (!statusWithCQ.isOK()) {
uasserted(17287,
str::stream() << "Can't canonicalize query: "
<< statusWithCQ.getStatus().toString());
}
unique_ptr<CanonicalQuery> cq = std::move(statusWithCQ.getValue());
invariant(cq.get());
LOG(5) << "Running query:\n" << redact(cq->toString());
LOG(2) << "Running query: " << redact(cq->toStringShort());
// Parse, canonicalize, plan, transcribe, and get a plan executor.
AutoGetCollectionOrViewForReadCommand ctx(opCtx, nss);
Collection* collection = ctx.getCollection();
if (ctx.getView()) {
uasserted(ErrorCodes::CommandNotSupportedOnView,
str::stream()
<< "Namespace "
<< nss.ns()
<< " is a view. Legacy find operations are not supported on views. "
<< "Only clients which support the find command can be used to query views.");
}
// We have a parsed query. Time to get the execution plan for it.
auto exec = uassertStatusOK(
getExecutorFind(opCtx, collection, nss, std::move(cq), PlanExecutor::YIELD_AUTO));
const QueryRequest& qr = exec->getCanonicalQuery()->getQueryRequest();
// If it's actually an explain, do the explain and return rather than falling through
// to the normal query execution loop.
if (qr.isExplain()) {
BufBuilder bb;
bb.skip(sizeof(QueryResult::Value));
BSONObjBuilder explainBob;
Explain::explainStages(
exec.get(), collection, ExplainOptions::Verbosity::kExecAllPlans, &explainBob);
// Add the resulting object to the return buffer.
BSONObj explainObj = explainBob.obj();
bb.appendBuf((void*)explainObj.objdata(), explainObj.objsize());
// Set query result fields.
QueryResult::View qr = bb.buf();
qr.setResultFlagsToOk();
qr.msgdata().setLen(bb.len());
curOp.debug().responseLength = bb.len();
qr.msgdata().setOperation(opReply);
qr.setCursorId(0);
qr.setStartingFrom(0);
qr.setNReturned(1);
result.setData(bb.release());
return "";
}
// Handle query option $maxTimeMS (not used with commands).
if (qr.getMaxTimeMS() > 0) {
uassert(40116,
"Illegal attempt to set operation deadline within DBDirectClient",
!opCtx->getClient()->isInDirectClient());
opCtx->setDeadlineAfterNowBy(Milliseconds{qr.getMaxTimeMS()});
}
opCtx->checkForInterrupt(); // May trigger maxTimeAlwaysTimeOut fail point.
// uassert if we are not on a primary, and not a secondary with SlaveOk query parameter set.
bool slaveOK = qr.isSlaveOk() || qr.hasReadPref();
Status serveReadsStatus =
repl::getGlobalReplicationCoordinator()->checkCanServeReadsFor_UNSAFE(opCtx, nss, slaveOK);
uassertStatusOK(serveReadsStatus);
// Run the query.
// bb is used to hold query results
// this buffer should contain either requested documents per query or
// explain information, but not both
BufBuilder bb(FindCommon::kInitReplyBufferSize);
bb.skip(sizeof(QueryResult::Value));
// How many results have we obtained from the executor?
int numResults = 0;
// If we're replaying the oplog, we save the last time that we read.
Timestamp slaveReadTill;
BSONObj obj;
PlanExecutor::ExecState state;
// Get summary info about which plan the executor is using.
{
stdx::lock_guard<Client> lk(*opCtx->getClient());
curOp.setPlanSummary_inlock(Explain::getPlanSummary(exec.get()));
}
while (PlanExecutor::ADVANCED == (state = exec->getNext(&obj, NULL))) {
// If we can't fit this result inside the current batch, then we stash it for later.
if (!FindCommon::haveSpaceForNext(obj, numResults, bb.len())) {
exec->enqueue(obj);
break;
}
// Add result to output buffer.
bb.appendBuf((void*)obj.objdata(), obj.objsize());
// Count the result.
++numResults;
// Possibly note slave's position in the oplog.
if (qr.isOplogReplay()) {
BSONElement e = obj["ts"];
if (Date == e.type() || bsonTimestamp == e.type()) {
slaveReadTill = e.timestamp();
}
}
if (FindCommon::enoughForFirstBatch(qr, numResults)) {
LOG(5) << "Enough for first batch, wantMore=" << qr.wantMore()
<< " ntoreturn=" << qr.getNToReturn().value_or(0)
<< " numResults=" << numResults;
break;
}
}
// Caller expects exceptions thrown in certain cases.
if (PlanExecutor::FAILURE == state || PlanExecutor::DEAD == state) {
error() << "Plan executor error during find: " << PlanExecutor::statestr(state)
<< ", stats: " << redact(Explain::getWinningPlanStats(exec.get()));
uasserted(17144, "Executor error: " + WorkingSetCommon::toStatusString(obj));
}
// Before saving the cursor, ensure that whatever plan we established happened with the expected
// collection version
auto css = CollectionShardingState::get(opCtx, nss);
css->checkShardVersionOrThrow(opCtx);
// Fill out CurOp based on query results. If we have a cursorid, we will fill out CurOp with
// this cursorid later.
long long ccId = 0;
if (shouldSaveCursor(opCtx, collection, state, exec.get())) {
// We won't use the executor until it's getMore'd.
exec->saveState();
exec->detachFromOperationContext();
// Allocate a new ClientCursor and register it with the cursor manager.
ClientCursorPin pinnedCursor = collection->getCursorManager()->registerCursor(
opCtx,
{std::move(exec),
nss,
AuthorizationSession::get(opCtx->getClient())->getAuthenticatedUserNames(),
opCtx->recoveryUnit()->isReadingFromMajorityCommittedSnapshot(),
upconvertQueryEntry(q.query, qr.nss(), q.ntoreturn, q.ntoskip)});
ccId = pinnedCursor.getCursor()->cursorid();
LOG(5) << "caching executor with cursorid " << ccId << " after returning " << numResults
<< " results";
// TODO document
if (qr.isOplogReplay() && !slaveReadTill.isNull()) {
pinnedCursor.getCursor()->slaveReadTill(slaveReadTill);
}
// TODO document
if (qr.isExhaust()) {
curOp.debug().exhaust = true;
}
pinnedCursor.getCursor()->setPos(numResults);
// We assume that cursors created through a DBDirectClient are always used from their
// original OperationContext, so we do not need to move time to and from the cursor.
if (!opCtx->getClient()->isInDirectClient()) {
// If the query had a time limit, remaining time is "rolled over" to the cursor (for
// use by future getmore ops).
pinnedCursor.getCursor()->setLeftoverMaxTimeMicros(opCtx->getRemainingMaxTimeMicros());
}
endQueryOp(opCtx, collection, *pinnedCursor.getCursor()->getExecutor(), numResults, ccId);
} else {
LOG(5) << "Not caching executor but returning " << numResults << " results.";
endQueryOp(opCtx, collection, *exec, numResults, ccId);
}
// Fill out the output buffer's header.
QueryResult::View queryResultView = bb.buf();
queryResultView.setCursorId(ccId);
queryResultView.setResultFlagsToOk();
queryResultView.msgdata().setLen(bb.len());
queryResultView.msgdata().setOperation(opReply);
queryResultView.setStartingFrom(0);
queryResultView.setNReturned(numResults);
// Add the results from the query into the output buffer.
result.setData(bb.release());
// curOp.debug().exhaust is set above.
return curOp.debug().exhaust ? nss.ns() : "";
}
} // namespace mongo
|