1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
|
/**
* Copyright (C) 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::kDefault
#include "mongo/platform/basic.h"
#include <vector>
#include "mongo/db/concurrency/lock_manager.h"
#include "mongo/config.h"
#include "mongo/db/concurrency/locker.h"
#include "mongo/util/assert_util.h"
#include "mongo/util/log.h"
#include "mongo/util/stringutils.h"
#include "mongo/util/timer.h"
namespace mongo {
using std::string;
namespace {
/**
* Map of conflicts. 'LockConflictsTable[newMode] & existingMode != 0' means that a new request
* with the given 'newMode' conflicts with an existing request with mode 'existingMode'.
*/
static const int LockConflictsTable[] = {
// MODE_NONE
0,
// MODE_IS
(1 << MODE_X),
// MODE_IX
(1 << MODE_S) | (1 << MODE_X),
// MODE_S
(1 << MODE_IX) | (1 << MODE_X),
// MODE_X
(1 << MODE_S) | (1 << MODE_X) | (1 << MODE_IS) | (1 << MODE_IX),
};
// Mask of modes
const uint64_t intentModes = (1 << MODE_IS) | (1<< MODE_IX);
// Ensure we do not add new modes without updating the conflicts table
BOOST_STATIC_ASSERT(
(sizeof(LockConflictsTable) / sizeof(LockConflictsTable[0])) == LockModesCount);
/**
* Maps the mode id to a string.
*/
static const char* LockModeNames[] = {
"NONE", "IS", "IX", "S", "X"
};
static const char* LegacyLockModeNames[] = {
"", "r", "w", "R", "W"
};
// Ensure we do not add new modes without updating the names array
BOOST_STATIC_ASSERT((sizeof(LockModeNames) / sizeof(LockModeNames[0])) == LockModesCount);
BOOST_STATIC_ASSERT(
(sizeof(LegacyLockModeNames) / sizeof(LegacyLockModeNames[0])) == LockModesCount);
// Helper functions for the lock modes
bool conflicts(LockMode newMode, uint32_t existingModesMask) {
return (LockConflictsTable[newMode] & existingModesMask) != 0;
}
uint32_t modeMask(LockMode mode) {
return 1 << mode;
}
/**
* Maps the resource id to a human-readable string.
*/
static const char* ResourceTypeNames[] = {
"Invalid",
"Global",
"MMAPV1Journal",
"Database",
"Collection",
"Metadata",
};
// Ensure we do not add new types without updating the names array
BOOST_STATIC_ASSERT(
(sizeof(ResourceTypeNames) / sizeof(ResourceTypeNames[0])) == ResourceTypesCount);
/**
* Maps the LockRequest status to a human-readable string.
*/
static const char* LockRequestStatusNames[] = {
"new",
"granted",
"waiting",
"converting",
};
// Ensure we do not add new status types without updating the names array
BOOST_STATIC_ASSERT(
(sizeof(LockRequestStatusNames) / sizeof(LockRequestStatusNames[0]))
== LockRequest::StatusCount);
} // namespace
/**
* There is one of these objects for each resource that has a lock request. Empty objects
* (i.e. LockHead with no requests) are allowed to exist on the lock manager's hash table.
*
* The memory and lifetime is controlled entirely by the LockManager class.
*
* Not thread-safe and should only be accessed under the LockManager's bucket lock.
* Must be locked before locking a partition, not after.
*/
struct LockHead {
/**
* Used for initialization of a LockHead, which might have been retrieved from cache and
* also in order to keep the LockHead structure a POD.
*/
void initNew(ResourceId resId) {
resourceId = resId;
grantedList.reset();
memset(grantedCounts, 0, sizeof(grantedCounts));
grantedModes = 0;
conflictList.reset();
memset(conflictCounts, 0, sizeof(conflictCounts));
conflictModes = 0;
conversionsCount = 0;
compatibleFirstCount = 0;
}
/**
* True iff there may be partitions with granted requests for this
* resource.
*/
bool partitioned() const {
return !partitions.empty();
}
/**
* Locates the request corresponding to the particular locker or returns NULL. Must be
* called with the bucket holding this lock head locked.
*/
LockRequest* findRequest(LockerId lockerId) const {
// Check the granted queue first
for (LockRequest* it = grantedList._front; it != NULL; it = it->next) {
if (it->locker->getId() == lockerId) {
return it;
}
}
// Check the conflict queue second
for (LockRequest* it = conflictList._front; it != NULL; it = it->next) {
if (it->locker->getId() == lockerId) {
return it;
}
}
return NULL;
}
/**
* Finish creation of request and put it on the lockhead's conflict or granted queues.
* Returns LOCK_WAITING for conflict case and LOCK_OK otherwise.
*/
LockResult newRequest(LockRequest* request, LockMode mode) {
request->mode = mode;
request->lock = this;
request->partitionedLock = NULL;
if (!partitioned()) {
request->recursiveCount = 1;
}
// request->partitioned cannot be set to false, as this might be a migration, in
// which case access to that field is not protected. The 'partitioned' member instead
// indicates if a request was initially partitioned.
// New lock request. Queue after all granted modes and after any already requested
// conflicting modes.
if (conflicts(mode, grantedModes) ||
(!compatibleFirstCount && conflicts(mode, conflictModes))) {
request->status = LockRequest::STATUS_WAITING;
// Put it on the conflict queue. Conflicts are granted front to back.
if (request->enqueueAtFront) {
conflictList.push_front(request);
}
else {
conflictList.push_back(request);
}
incConflictModeCount(mode);
return LOCK_WAITING;
}
// No conflict, new request
request->status = LockRequest::STATUS_GRANTED;
grantedList.push_back(request);
incGrantedModeCount(mode);
if (request->compatibleFirst) {
compatibleFirstCount++;
}
return LOCK_OK;
}
/**
* Lock each partitioned LockHead in turn, and move any (granted) intent mode requests for
* lock->resourceId to lock, which must itself already be locked.
*/
void migratePartitionedLockHeads();
// Methods to maintain the granted queue
void incGrantedModeCount(LockMode mode) {
invariant(grantedCounts[mode] >= 0);
if (++grantedCounts[mode] == 1) {
invariant((grantedModes & modeMask(mode)) == 0);
grantedModes |= modeMask(mode);
}
}
void decGrantedModeCount(LockMode mode) {
invariant(grantedCounts[mode] >= 1);
if (--grantedCounts[mode] == 0) {
invariant((grantedModes & modeMask(mode)) == modeMask(mode));
grantedModes &= ~modeMask(mode);
}
}
// Methods to maintain the conflict queue
void incConflictModeCount(LockMode mode) {
invariant(conflictCounts[mode] >= 0);
if (++conflictCounts[mode] == 1) {
invariant((conflictModes & modeMask(mode)) == 0);
conflictModes |= modeMask(mode);
}
}
void decConflictModeCount(LockMode mode) {
invariant(conflictCounts[mode] >= 1);
if (--conflictCounts[mode] == 0) {
invariant((conflictModes & modeMask(mode)) == modeMask(mode));
conflictModes &= ~modeMask(mode);
}
}
// Id of the resource which this lock protects
ResourceId resourceId;
//
// Granted queue
//
// Doubly-linked list of requests, which have been granted. Newly granted requests go to
// the end of the queue. Conversion requests are granted from the beginning forward.
LockRequestList grantedList;
// Counts the grants and coversion counts for each of the supported lock modes. These
// counts should exactly match the aggregated modes on the granted list.
uint32_t grantedCounts[LockModesCount];
// Bit-mask of the granted + converting modes on the granted queue. Maintained in lock-step
// with the grantedCounts array.
uint32_t grantedModes;
//
// Conflict queue
//
// Doubly-linked list of requests, which have not been granted yet because they conflict
// with the set of granted modes. Requests are queued at the end of the queue and are
// granted from the beginning forward, which gives these locks FIFO ordering. Exceptions
// are high-priorty locks, such as the MMAP V1 flush lock.
LockRequestList conflictList;
// Counts the conflicting requests for each of the lock modes. These counts should exactly
// match the aggregated modes on the conflicts list.
uint32_t conflictCounts[LockModesCount];
// Bit-mask of the conflict modes on the conflict queue. Maintained in lock-step with the
// conflictCounts array.
uint32_t conflictModes;
// References partitions that may have PartitionedLockHeads for this LockHead.
// Non-empty implies the lock has no conflicts and only has intent modes as grantedModes.
// TODO: Remove this vector and make LockHead a POD
std::vector<LockManager::Partition *> partitions;
//
// Conversion
//
// Counts the number of requests on the granted queue, which have requested any kind of
// conflicting conversion and are blocked (i.e. all requests which are currently
// STATUS_CONVERTING). This is an optimization for unlocking in that we do not need to
// check the granted queue for requests in STATUS_CONVERTING if this count is zero. This
// saves cycles in the regular case and only burdens the less-frequent lock upgrade case.
uint32_t conversionsCount;
// Counts the number of requests on the granted queue, which have requested that the policy
// be switched to compatible-first. As long as this value is > 0, the policy will stay
// compatible-first.
uint32_t compatibleFirstCount;
};
/**
* The PartitionedLockHead allows optimizing the case where requests overwhelmingly use
* the intent lock modes MODE_IS and MODE_IX, which are compatible with each other.
* Having to use a single LockHead causes contention where none would be needed.
* So, each Locker is associated with a specific partition containing a mapping
* of resourceId to PartitionedLockHead.
*
* As long as all lock requests for a resource have an intent mode, as opposed to a conflicting
* mode, its LockHead may reference ParitionedLockHeads. A partitioned LockHead will not have
* any conflicts. The total set of granted requests (with intent mode) is the union of
* its grantedList and all grantedLists in PartitionedLockHeads.
*
* The existence of a PartitionedLockHead for a resource implies that its LockHead is
* partitioned. If a conflicting request is made on a LockHead, all requests from
* PartitionedLockHeads are migrated to that LockHead and the LockHead no longer partitioned.
*
* Not thread-safe, must be accessed under its partition lock.
* May not lock a LockManager bucket while holding a partition lock.
*/
struct PartitionedLockHead {
void initNew(ResourceId resId) {
grantedList.reset();
}
void newRequest(LockRequest* request, LockMode mode) {
request->lock = NULL;
request->partitionedLock = this;
request->recursiveCount = 1;
request->status = LockRequest::STATUS_GRANTED;
request->partitioned = true;
request->mode = mode;
grantedList.push_back(request);
}
//
// Granted queue
//
// Doubly-linked list of requests, which have been granted. Newly granted requests go to
// the end of the queue. The PartitionedLockHead never contains anything but granted
// requests with intent modes.
LockRequestList grantedList;
};
void LockHead::migratePartitionedLockHeads() {
invariant(partitioned());
// There can't be non-intent modes or conflicts when the lock is partitioned
invariant(!(grantedModes & ~intentModes) && !conflictModes);
// Migration time: lock each partition in turn and transfer its requests, if any
while(partitioned()) {
LockManager::Partition* partition = partitions.back();
SimpleMutex::scoped_lock scopedLock(partition->mutex);
LockManager::Partition::Map::iterator it = partition->data.find(resourceId);
if (it != partition->data.end()) {
PartitionedLockHead* partitionedLock = it->second;
while (!partitionedLock->grantedList.empty()) {
LockRequest* request = partitionedLock->grantedList._front;
partitionedLock->grantedList.remove(request);
// Ordering is important here, as the next/prev fields are shared.
// Note that newRequest() will preserve the recursiveCount in this case
LockResult res = newRequest(request, request->mode);
invariant(res == LOCK_OK); // Lock must still be granted
}
partition->data.erase(it);
delete partitionedLock;
}
// Don't pop-back to early as otherwise the lock will be considered not partioned in
// newRequest().
partitions.pop_back();
}
}
//
// LockManager
//
// Have more buckets than CPUs to reduce contention on lock and caches
const unsigned LockManager::_numLockBuckets(128);
// Balance scalability of intent locks against potential added cost of conflicting locks.
// The exact value doesn't appear very important, but should be power of two
const unsigned LockManager::_numPartitions = 32;
LockManager::LockManager() {
_lockBuckets = new LockBucket[_numLockBuckets];
_partitions = new Partition[_numPartitions];
}
LockManager::~LockManager() {
cleanupUnusedLocks();
for (unsigned i = 0; i < _numLockBuckets; i++) {
// TODO: dump more information about the non-empty bucket to see what locks were leaked
invariant(_lockBuckets[i].data.empty());
}
delete[] _lockBuckets;
delete[] _partitions;
}
LockResult LockManager::lock(ResourceId resId, LockRequest* request, LockMode mode) {
// Sanity check that requests are not being reused without proper cleanup
invariant(request->status == LockRequest::STATUS_NEW);
request->partitioned = (mode == MODE_IX || mode == MODE_IS);
// For intent modes, try the PartitionedLockHead
if (request->partitioned) {
Partition* partition = _getPartition(request);
SimpleMutex::scoped_lock scopedLock(partition->mutex);
// Fast path for intent locks
PartitionedLockHead* partitionedLock = partition->find(resId);
if (partitionedLock) {
partitionedLock->newRequest(request, mode);
return LOCK_OK;
}
// Unsuccessful: there was no PartitionedLockHead yet, so use regular LockHead.
// Must not hold any locks. It is OK for requests with intent modes to be on
// both a PartitionedLockHead and a regular LockHead, so the race here is benign.
}
// Use regular LockHead, maybe start partitioning
LockBucket* bucket = _getBucket(resId);
SimpleMutex::scoped_lock scopedLock(bucket->mutex);
LockHead* lock = bucket->findOrInsert(resId);
// Start a partitioned lock if possible
if (request->partitioned && !(lock->grantedModes & (~intentModes))
&& !lock->conflictModes) {
Partition* partition = _getPartition(request);
SimpleMutex::scoped_lock scopedLock(partition->mutex);
PartitionedLockHead* partitionedLock = partition->findOrInsert(resId);
invariant(partitionedLock);
lock->partitions.push_back(partition);
partitionedLock->newRequest(request, mode);
return LOCK_OK;
}
// For the first lock with a non-intent mode, migrate requests from partitioned lock heads
if (lock->partitioned()) {
lock->migratePartitionedLockHeads();
}
request->partitioned = false;
return lock->newRequest(request, mode);
}
LockResult LockManager::convert(ResourceId resId, LockRequest* request, LockMode newMode) {
// If we are here, we already hold the lock in some mode. In order to keep it simple, we do
// not allow requesting a conversion while a lock is already waiting or pending conversion.
invariant(request->status == LockRequest::STATUS_GRANTED);
invariant(request->recursiveCount > 0);
request->recursiveCount++;
// Fast path for acquiring the same lock multiple times in modes, which are already covered
// by the current mode. It is safe to do this without locking, because 1) all calls for the
// same lock request must be done on the same thread and 2) if there are lock requests
// hanging off a given LockHead, then this lock will never disappear.
if ((LockConflictsTable[request->mode] | LockConflictsTable[newMode]) ==
LockConflictsTable[request->mode]) {
return LOCK_OK;
}
// TODO: For the time being we do not need conversions between unrelated lock modes (i.e.,
// modes which both add and remove to the conflicts set), so these are not implemented yet
// (e.g., S -> IX).
invariant((LockConflictsTable[request->mode] | LockConflictsTable[newMode]) ==
LockConflictsTable[newMode]);
LockBucket* bucket = _getBucket(resId);
SimpleMutex::scoped_lock scopedLock(bucket->mutex);
LockBucket::Map::iterator it = bucket->data.find(resId);
invariant(it != bucket->data.end());
LockHead* const lock = it->second;
if (lock->partitioned()) {
lock->migratePartitionedLockHeads();
}
// Construct granted mask without our current mode, so that it is not counted as
// conflicting
uint32_t grantedModesWithoutCurrentRequest = 0;
// We start the counting at 1 below, because LockModesCount also includes MODE_NONE
// at position 0, which can never be acquired/granted.
for (uint32_t i = 1; i < LockModesCount; i++) {
const uint32_t currentRequestHolds =
(request->mode == static_cast<LockMode>(i) ? 1 : 0);
if (lock->grantedCounts[i] > currentRequestHolds) {
grantedModesWithoutCurrentRequest |= modeMask(static_cast<LockMode>(i));
}
}
// This check favours conversion requests over pending requests. For example:
//
// T1 requests lock L in IS
// T2 requests lock L in X
// T1 then upgrades L from IS -> S
//
// Because the check does not look into the conflict modes bitmap, it will grant L to
// T1 in S mode, instead of block, which would otherwise cause deadlock.
if (conflicts(newMode, grantedModesWithoutCurrentRequest)) {
request->status = LockRequest::STATUS_CONVERTING;
invariant(request->recursiveCount > 1);
request->convertMode = newMode;
lock->conversionsCount++;
lock->incGrantedModeCount(request->convertMode);
return LOCK_WAITING;
}
else { // No conflict, existing request
lock->incGrantedModeCount(newMode);
lock->decGrantedModeCount(request->mode);
request->mode = newMode;
return LOCK_OK;
}
}
bool LockManager::unlock(LockRequest* request) {
// Fast path for decrementing multiple references of the same lock. It is safe to do this
// without locking, because 1) all calls for the same lock request must be done on the same
// thread and 2) if there are lock requests hanging of a given LockHead, then this lock
// will never disappear.
invariant(request->recursiveCount > 0);
request->recursiveCount--;
if ((request->status == LockRequest::STATUS_GRANTED) && (request->recursiveCount > 0)) {
return false;
}
if (request->partitioned) {
// Unlocking a lock that was acquired as partitioned. The lock request may since have
// moved to the lock head, but there is no safe way to find out without synchronizing
// thorough the partition mutex. Migrations are expected to be rare.
invariant(request->status == LockRequest::STATUS_GRANTED
|| request->status == LockRequest::STATUS_CONVERTING);
Partition* partition = _getPartition(request);
SimpleMutex::scoped_lock scopedLock(partition->mutex);
// Fast path: still partitioned.
if (request->partitionedLock) {
request->partitionedLock->grantedList.remove(request);
return true;
}
// not partitioned anymore, fall through to regular case
}
invariant(request->lock);
LockHead* lock = request->lock;
LockBucket* bucket = _getBucket(lock->resourceId);
SimpleMutex::scoped_lock scopedLock(bucket->mutex);
if (request->status == LockRequest::STATUS_GRANTED) {
// This releases a currently held lock and is the most common path, so it should be
// as efficient as possible. The fast path for decrementing multiple references did
// already ensure request->recursiveCount == 0.
// Remove from the granted list
lock->grantedList.remove(request);
lock->decGrantedModeCount(request->mode);
if (request->compatibleFirst) {
lock->compatibleFirstCount--;
}
_onLockModeChanged(lock, lock->grantedCounts[request->mode] == 0);
}
else if (request->status == LockRequest::STATUS_WAITING) {
// This cancels a pending lock request
invariant(request->recursiveCount == 0);
lock->conflictList.remove(request);
lock->decConflictModeCount(request->mode);
}
else if (request->status == LockRequest::STATUS_CONVERTING) {
// This cancels a pending convert request
invariant(request->recursiveCount > 0);
// Lock only goes from GRANTED to CONVERTING, so cancelling the conversion request
// brings it back to the previous granted mode.
request->status = LockRequest::STATUS_GRANTED;
lock->conversionsCount--;
lock->decGrantedModeCount(request->convertMode);
request->convertMode = MODE_NONE;
_onLockModeChanged(lock, lock->grantedCounts[request->convertMode] == 0);
}
else {
// Invalid request status
invariant(false);
}
return (request->recursiveCount == 0);
}
void LockManager::downgrade(LockRequest* request, LockMode newMode) {
invariant(request->lock);
invariant(request->status == LockRequest::STATUS_GRANTED);
invariant(request->recursiveCount > 0);
// The conflict set of the newMode should be a subset of the conflict set of the old mode.
// Can't downgrade from S -> IX for example.
invariant((LockConflictsTable[request->mode] | LockConflictsTable[newMode])
== LockConflictsTable[request->mode]);
LockHead* lock = request->lock;
LockBucket* bucket = _getBucket(lock->resourceId);
SimpleMutex::scoped_lock scopedLock(bucket->mutex);
lock->incGrantedModeCount(newMode);
lock->decGrantedModeCount(request->mode);
request->mode = newMode;
_onLockModeChanged(lock, true);
}
void LockManager::cleanupUnusedLocks() {
size_t deletedLockHeads = 0;
for (unsigned i = 0; i < _numLockBuckets; i++) {
LockBucket* bucket = &_lockBuckets[i];
SimpleMutex::scoped_lock scopedLock(bucket->mutex);
LockBucket::Map::iterator it = bucket->data.begin();
while (it != bucket->data.end()) {
LockHead* lock = it->second;
if (lock->partitioned()) {
lock->migratePartitionedLockHeads();
}
if (lock->grantedModes == 0) {
invariant(lock->grantedModes == 0);
invariant(lock->grantedList._front == NULL);
invariant(lock->grantedList._back == NULL);
invariant(lock->conflictModes == 0);
invariant(lock->conflictList._front == NULL);
invariant(lock->conflictList._back == NULL);
invariant(lock->conversionsCount == 0);
invariant(lock->compatibleFirstCount == 0);
bucket->data.erase(it++);
deletedLockHeads++;
delete lock;
}
else {
it++;
}
}
}
}
void LockManager::_onLockModeChanged(LockHead* lock, bool checkConflictQueue) {
// Unblock any converting requests (because conversions are still counted as granted and
// are on the granted queue).
for (LockRequest* iter = lock->grantedList._front;
(iter != NULL) && (lock->conversionsCount > 0);
iter = iter->next) {
// Conversion requests are going in a separate queue
if (iter->status == LockRequest::STATUS_CONVERTING) {
invariant(iter->convertMode != 0);
// Construct granted mask without our current mode, so that it is not accounted as
// a conflict
uint32_t grantedModesWithoutCurrentRequest = 0;
// We start the counting at 1 below, because LockModesCount also includes
// MODE_NONE at position 0, which can never be acquired/granted.
for (uint32_t i = 1; i < LockModesCount; i++) {
const uint32_t currentRequestHolds =
(iter->mode == static_cast<LockMode>(i) ? 1 : 0);
const uint32_t currentRequestWaits =
(iter->convertMode == static_cast<LockMode>(i) ? 1 : 0);
// We cannot both hold and wait on the same lock mode
invariant(currentRequestHolds + currentRequestWaits <= 1);
if (lock->grantedCounts[i] > (currentRequestHolds + currentRequestWaits)) {
grantedModesWithoutCurrentRequest |= modeMask(static_cast<LockMode>(i));
}
}
if (!conflicts(iter->convertMode, grantedModesWithoutCurrentRequest)) {
lock->conversionsCount--;
lock->decGrantedModeCount(iter->mode);
iter->status = LockRequest::STATUS_GRANTED;
iter->mode = iter->convertMode;
iter->convertMode = MODE_NONE;
iter->notify->notify(lock->resourceId, LOCK_OK);
}
}
}
// Grant any conflicting requests, which might now be unblocked. Note that the loop below
// slightly violates fairness in that it will grant *all* compatible requests on the line
// even though there might be conflicting ones interspersed between them. For example,
// consider an X lock was just freed and the conflict queue looked like this:
//
// IS -> IS -> X -> X -> S -> IS
//
// In strict FIFO, we should grant the first two IS modes and then stop when we reach the
// first X mode (the third request on the queue). However, the loop below would actually
// grant all IS + S modes and once they all drain it will grant X.
LockRequest* iterNext = NULL;
for (LockRequest* iter = lock->conflictList._front;
(iter != NULL) && checkConflictQueue;
iter = iterNext) {
invariant(iter->status == LockRequest::STATUS_WAITING);
// Store the actual next pointer, because we muck with the iter below and move it to
// the granted queue.
iterNext = iter->next;
if (conflicts(iter->mode, lock->grantedModes)) {
continue;
}
iter->status = LockRequest::STATUS_GRANTED;
lock->conflictList.remove(iter);
lock->grantedList.push_back(iter);
lock->incGrantedModeCount(iter->mode);
lock->decConflictModeCount(iter->mode);
if (iter->compatibleFirst) {
lock->compatibleFirstCount++;
}
iter->notify->notify(lock->resourceId, LOCK_OK);
// Small optimization - nothing is compatible with MODE_X, so no point in looking
// further in the conflict queue.
if (iter->mode == MODE_X) {
break;
}
}
// This is a convenient place to check that the state of the two request queues is in sync
// with the bitmask on the modes.
invariant((lock->grantedModes == 0) ^ (lock->grantedList._front != NULL));
invariant((lock->conflictModes == 0) ^ (lock->conflictList._front != NULL));
}
LockManager::LockBucket* LockManager::_getBucket(ResourceId resId) const {
return &_lockBuckets[resId % _numLockBuckets];
}
LockManager::Partition* LockManager::_getPartition(LockRequest* request) const {
return &_partitions[request->locker->getId() % _numPartitions];
}
void LockManager::dump() const {
log() << "Dumping LockManager @ " << static_cast<const void*>(this) << '\n';
for (unsigned i = 0; i < _numLockBuckets; i++) {
LockBucket* bucket = &_lockBuckets[i];
SimpleMutex::scoped_lock scopedLock(bucket->mutex);
if (!bucket->data.empty()) {
_dumpBucket(bucket);
}
}
}
void LockManager::_dumpBucket(const LockBucket* bucket) const {
for (LockBucket::Map::const_iterator it = bucket->data.begin();
it != bucket->data.end();
it++) {
const LockHead* lock = it->second;
if (lock->grantedList.empty()) {
// If there are no granted requests, this lock is empty, so no need to print it
continue;
}
StringBuilder sb;
sb << "Lock @ " << lock << ": " << lock->resourceId.toString() << '\n';
sb << "GRANTED:\n";
for (const LockRequest* iter = lock->grantedList._front;
iter != NULL;
iter = iter->next) {
sb << '\t'
<< "LockRequest " << iter->locker->getId() << " @ " << iter->locker << ": "
<< "Mode = " << modeName(iter->mode) << "; "
<< "ConvertMode = " << modeName(iter->convertMode) << "; "
<< "EnqueueAtFront = " << iter->enqueueAtFront << "; "
<< "CompatibleFirst = " << iter->compatibleFirst << "; "
<< '\n';
}
sb << '\n';
sb << "PENDING:\n";
for (const LockRequest* iter = lock->conflictList._front;
iter != NULL;
iter = iter->next) {
sb << '\t'
<< "LockRequest " << iter->locker->getId() << " @ " << iter->locker << ": "
<< "Mode = " << modeName(iter->mode) << "; "
<< "ConvertMode = " << modeName(iter->convertMode) << "; "
<< "EnqueueAtFront = " << iter->enqueueAtFront << "; "
<< "CompatibleFirst = " << iter->compatibleFirst << "; "
<< '\n';
}
log() << sb.str();
}
}
PartitionedLockHead* LockManager::Partition::find(ResourceId resId) {
Map::iterator it = data.find(resId);
return it == data.end() ? NULL : it->second;
}
PartitionedLockHead* LockManager::Partition::findOrInsert(ResourceId resId) {
PartitionedLockHead* lock;
Map::iterator it = data.find(resId);
if (it == data.end()) {
lock = new PartitionedLockHead();
lock->initNew(resId);
data.insert(Map::value_type(resId, lock));
}
else {
lock = it->second;
}
return lock;
}
LockHead* LockManager::LockBucket::findOrInsert(ResourceId resId) {
LockHead* lock;
Map::iterator it = data.find(resId);
if (it == data.end()) {
lock = new LockHead();
lock->initNew(resId);
data.insert(Map::value_type(resId, lock));
}
else {
lock = it->second;
}
return lock;
}
//
// DeadlockDetector
//
DeadlockDetector::DeadlockDetector(const LockManager& lockMgr, const Locker* initialLocker)
: _lockMgr(lockMgr),
_initialLockerId(initialLocker->getId()),
_foundCycle(false) {
const ResourceId resId = initialLocker->getWaitingResource();
// If there is no resource waiting there is nothing to do
if (resId.isValid()) {
_queue.push_front(UnprocessedNode(_initialLockerId, resId));
}
}
bool DeadlockDetector::next() {
if (_queue.empty()) return false;
UnprocessedNode front = _queue.front();
_queue.pop_front();
_processNextNode(front);
return !_queue.empty();
}
bool DeadlockDetector::hasCycle() const {
invariant(_queue.empty());
return _foundCycle;
}
string DeadlockDetector::toString() const {
StringBuilder sb;
for (WaitForGraph::const_iterator it = _graph.begin(); it != _graph.end(); it++) {
sb << "Locker " << it->first << " waits for resource " << it->second.resId.toString()
<< " held by [";
const ConflictingOwnersList owners = it->second.owners;
for (ConflictingOwnersList::const_iterator itW = owners.begin();
itW != owners.end();
itW++) {
sb << *itW << ", ";
}
sb << "]\n";
}
return sb.str();
}
void DeadlockDetector::_processNextNode(const UnprocessedNode& node) {
// Locate the request
LockManager::LockBucket* bucket = _lockMgr._getBucket(node.resId);
SimpleMutex::scoped_lock scopedLock(bucket->mutex);
LockManager::LockBucket::Map::const_iterator iter = bucket->data.find(node.resId);
if (iter == bucket->data.end()) {
return;
}
const LockHead* lock = iter->second;
LockRequest* request = lock->findRequest(node.lockerId);
// It is possible that a request which was thought to be waiting suddenly became
// granted, so check that before proceeding
if (!request || (request->status == LockRequest::STATUS_GRANTED)) {
return;
}
std::pair<WaitForGraph::iterator, bool> val =
_graph.insert(WaitForGraphPair(node.lockerId, Edges(node.resId)));
if (!val.second) {
// We already saw this locker id, which means we have a cycle.
if (!_foundCycle) {
_foundCycle = (node.lockerId == _initialLockerId);
}
return;
}
Edges& edges = val.first->second;
bool seen = false;
for (LockRequest* it = lock->grantedList._back; it != NULL; it = it->prev) {
// We can't conflict with ourselves
if (it == request) {
seen = true;
continue;
}
// If we are a regular conflicting request, both granted and conversion modes need to
// be checked for conflict, since conversions will be granted first.
if (request->status == LockRequest::STATUS_WAITING) {
if (conflicts(request->mode, modeMask(it->mode)) ||
conflicts(request->mode, modeMask(it->convertMode))) {
const LockerId lockerId = it->locker->getId();
const ResourceId waitResId = it->locker->getWaitingResource();
if (waitResId.isValid()) {
_queue.push_front(UnprocessedNode(lockerId, waitResId));
edges.owners.push_back(lockerId);
}
}
continue;
}
// If we are a conversion request, only requests, which are before us need to be
// accounted for.
invariant(request->status == LockRequest::STATUS_CONVERTING);
if (conflicts(request->convertMode, modeMask(it->mode)) ||
(seen && conflicts(request->convertMode, modeMask(it->convertMode)))) {
const LockerId lockerId = it->locker->getId();
const ResourceId waitResId = it->locker->getWaitingResource();
if (waitResId.isValid()) {
_queue.push_front(UnprocessedNode(lockerId, waitResId));
edges.owners.push_back(lockerId);
}
}
}
// All conflicting waits, which would be granted before us
for (LockRequest* it = request->prev;
(request->status == LockRequest::STATUS_WAITING) && (it != NULL);
it = it->prev) {
// We started from the previous element, so we should never see ourselves
invariant(it != request);
if (conflicts(request->mode, modeMask(it->mode))) {
const LockerId lockerId = it->locker->getId();
const ResourceId waitResId = it->locker->getWaitingResource();
if (waitResId.isValid()) {
_queue.push_front(UnprocessedNode(lockerId, waitResId));
edges.owners.push_back(lockerId);
}
}
}
}
//
// ResourceId
//
static const StringData::Hasher stringDataHashFunction = StringData::Hasher();
uint64_t ResourceId::fullHash(ResourceType type, uint64_t hashId) {
return (static_cast<uint64_t>(type) << (64 - resourceTypeBits))
+ (hashId & (std::numeric_limits<uint64_t>::max() >> resourceTypeBits));
}
ResourceId::ResourceId(ResourceType type, StringData ns)
: _fullHash(fullHash(type, stringDataHashFunction(ns))) {
#ifdef MONGO_CONFIG_DEBUG_BUILD
_nsCopy = ns.toString();
#endif
}
ResourceId::ResourceId(ResourceType type, const string& ns)
: _fullHash(fullHash(type, stringDataHashFunction(ns))) {
#ifdef MONGO_CONFIG_DEBUG_BUILD
_nsCopy = ns;
#endif
}
ResourceId::ResourceId(ResourceType type, uint64_t hashId)
: _fullHash(fullHash(type, hashId)) { }
string ResourceId::toString() const {
StringBuilder ss;
ss << "{" << _fullHash << ": " << resourceTypeName(getType())
<< ", " << getHashId();
#ifdef MONGO_CONFIG_DEBUG_BUILD
ss << ", " << _nsCopy;
#endif
ss << "}";
return ss.str();
}
//
// LockRequest
//
void LockRequest::initNew(Locker* locker, LockGrantNotification* notify) {
this->locker = locker;
this->notify = notify;
enqueueAtFront = false;
compatibleFirst = false;
recursiveCount = 0;
lock = NULL;
prev = NULL;
next = NULL;
status = STATUS_NEW;
partitioned = false;
mode = MODE_NONE;
convertMode = MODE_NONE;
}
//
// Helper calls
//
const char* modeName(LockMode mode) {
return LockModeNames[mode];
}
const char* legacyModeName(LockMode mode) {
return LegacyLockModeNames[mode];
}
bool isModeCovered(LockMode mode, LockMode coveringMode) {
return (LockConflictsTable[coveringMode] | LockConflictsTable[mode]) ==
LockConflictsTable[coveringMode];
}
const char* resourceTypeName(ResourceType resourceType) {
return ResourceTypeNames[resourceType];
}
const char* lockRequestStatusName(LockRequest::Status status) {
return LockRequestStatusNames[status];
}
} // namespace mongo
|