forked from bloomberg/comdb2
-
Notifications
You must be signed in to change notification settings - Fork 0
/
genid.c
1079 lines (885 loc) · 29.8 KB
/
genid.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
Copyright 2015 Bloomberg Finance L.P.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
/*
* Genids
*
*/
#include <pthread.h>
#include <arpa/inet.h>
#include <errno.h>
#include <fcntl.h>
#include <limits.h>
#include <netinet/in.h>
#include <netinet/tcp.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <strings.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/uio.h>
#include <unistd.h>
#include <stddef.h>
#include <alloca.h>
#include <sys/poll.h>
#include <sys/select.h>
#include <sys/types.h>
#include <build/db.h>
#include <epochlib.h>
#include <ctrace.h>
#include <plhash.h>
#include <list.h>
#include <net.h>
#include "bdb_int.h"
#include "locks.h"
#include "locks_wrap.h"
#include "flibc.h"
#ifndef MAXSTACKDEPTH
#define MAXSTACKDEPTH 200
#endif
#include "genid.h"
#include "logmsg.h"
#define HI48(bdb_state) ((bdb_state)->gblcontext.genid48.hi48)
#define LO16(bdb_state) ((bdb_state)->gblcontext.genid48.lo16)
int gbl_block_set_commit_genid_trace = 0;
static unsigned long long commit_genid;
static DB_LSN commit_lsn;
static uint32_t commit_generation;
unsigned long long get_lowest_genid_for_datafile(int stripe)
{
unsigned long long g = stripe;
g = (g << GENID_STRIPE_SHIFT) & GENID_STRIPE_MASK;
return g;
}
int get_slotnum_from_genid(int modnum, unsigned long long genid)
{
uint16_t slot_part;
/* grab the 2 bytes containing slot part of the genid in net byte order */
slot_part = (uint16_t)((genid & GENID_SLOT_MASK) >> GENID_SLOT_PRE_SHIFT);
return (ntohs(slot_part) >> GENID_SLOT_POST_SHIFT) % modnum;
}
/* the search genid will have the updateid portion masked out */
unsigned long long get_search_genid(bdb_state_type *bdb_state,
unsigned long long genid)
{
int mask, bits, pidb = bdb_state->attr->participantid_bits;
if (!bdb_state->ondisk_header)
return genid;
genid = flibc_ntohll(genid);
bits = (12 - pidb);
mask = (0x1 << bits) - 1;
genid &= ~(mask << 4);
return flibc_htonll(genid);
}
/* The rep stream needs to operate without a bdb_state */
unsigned long long get_search_genid_rowlocks(unsigned long long genid)
{
int mask;
genid = flibc_ntohll(genid);
mask = (0x1 << 12) - 1;
genid &= ~(mask << 4);
return flibc_htonll(genid);
}
unsigned long long bdb_mask_updateid(bdb_state_type *bdb_state,
unsigned long long genid)
{
unsigned long long updateid;
unsigned long long noupdateid;
if (!bdb_state->ondisk_header && (genid & (~GENID_NORMALIZE_MASK))) {
/*fprintf(stderr, "fixing genid 0x%llx genid\n", genid);*/
/* no ondisk header, but updateid is set! this means its an old style
record and these are actually the rightmost bits of the previously
larger dupcount area. shift that area back into the new place to
fix it */
genid = flibc_ntohll(genid);
/* mask out the old dupcount area, which includes the updateid area */
noupdateid = genid & GENID_OLD_DUP_MASK;
/* extract out the old update id, shift it over to the new position */
updateid = genid & (~GENID_OLD_DUP_MASK);
updateid = updateid << 12;
/* or in the "new style" update id */
genid = noupdateid | updateid;
genid = flibc_htonll(genid);
} else {
/* Mask out the updateid field since it must be zero */
genid = get_search_genid(bdb_state, genid);
}
/*fprintf(stderr, "genid masked to 0x%llx genid\n", genid);*/
return genid;
}
/*
To support live schema change, we need to guarantee that the
genids will always be monotonically increasing. So we reset the counter
whenever the epoch time wraps. This has the neat side effect that we know
how many genids were allocated in a second just by inspecting the genids..
To support ODH and the fast updates plan the new genid
time/32 + dupecount/16 + (particip&upd)/12 + stripe/4
the attribute variable, participantid_bits, specifies the number of bits
allocated to the participant stripe id. The remainder of those 12 bits
are used for the updateid
*/
struct call {
int ncalls;
int nframes;
intptr_t ip[1];
};
static u_int call_hashfunc(void *key, int len)
{
struct call *c = (struct call *)key;
int i;
u_int h = 0;
for (i = 0; i < c->nframes; i++)
h ^= (u_int)c->ip[i];
return h;
}
static int call_cmpfunc(void *key1, void *key2, int len)
{
struct call *c1, *c2;
int i;
c1 = (struct call *)key1;
c2 = (struct call *)key2;
if (c1->nframes != c2->nframes)
return -1;
for (i = 0; i < c1->nframes; i++)
if (c1->ip[i] != c2->ip[i])
return -1;
return 0;
}
static int dumpandclear(void *obj, void *arg)
{
struct call *c;
int i;
c = (struct call *)obj;
if (c->ncalls > 0) {
printf("----------------\n");
printf("%d: ", c->ncalls);
for (i = 0; i < c->nframes; i++) {
printf(" 0x%p", (void *)c->ip[i]);
}
printf("\n");
}
c->ncalls = 0;
return 0;
}
int bdb_genid_timestamp(unsigned long long genid)
{
int *iptr;
iptr = (int *)&genid;
/* genids (their components at least) are always "big-endian" */
return ntohl(iptr[0]);
}
/* Fabricated (recno) genids are retrieved from sqlite3BtreeNewRowid-
They will always have a timestamp value of 0 */
int bdb_genid_is_recno(bdb_state_type *bdb_state, unsigned long long genid)
{
if (bdb_state->genid_format == LLMETA_GENID_ORIGINAL) {
if(0 == bdb_genid_timestamp(genid))
{
return 1;
}
else
{
return 0;
}
}
else if (bdb_state->genid_format == LLMETA_GENID_48BIT) {
/* I don't think this is right - check. */
return is_genid_add(genid) || is_genid_upd(genid);
}
else {
logmsg(LOGMSG_FATAL, "Unknown genid format!\n");
abort();
}
}
/* this is for sqlite3BtreeNewRowid- I don't have to bother endianizing recno */
unsigned long long bdb_recno_to_genid(int recno)
{
unsigned long long genid = 0;
int *iptr = (int *)&genid;
iptr[1] = recno;
return genid;
}
static unsigned int get_dupecount_from_genid(unsigned long long genid)
{
unsigned short *sptr;
sptr = (unsigned short *)&genid;
/* genids (their components at least) are always "big-endian" */
return ntohs(sptr[2]);
}
static inline unsigned long long get_gblcontext_int(bdb_state_type *bdb_state)
{
return (bdb_state->genid_format == LLMETA_GENID_48BIT)
? flibc_htonll((HI48(bdb_state) << 16) | LO16(bdb_state))
: bdb_state->gblcontext.orig;
}
static inline void set_gblcontext_int(bdb_state_type *bdb_state,
unsigned long long gblcontext)
{
if (gblcontext == -1ULL) {
logmsg(LOGMSG_ERROR, "SETTING CONTEXT TO -1\n");
cheap_stack_trace();
}
if (bdb_cmp_genids(gblcontext, get_gblcontext_int(bdb_state)) < 0) {
if (gbl_block_set_commit_genid_trace) {
logmsg(LOGMSG_ERROR, "Blocked attempt to set lower gblcontext\n");
cheap_stack_trace();
}
} else if (bdb_state->genid_format != LLMETA_GENID_48BIT) {
bdb_state->gblcontext.orig = gblcontext;
} else {
LO16(bdb_state) = gblcontext >> 48;
HI48(bdb_state) = flibc_ntohll(gblcontext) >> 16;
}
}
/* setter/getters that grab a lock. */
void set_gblcontext(bdb_state_type *bdb_state, unsigned long long gblcontext)
{
if (bdb_state->parent)
bdb_state = bdb_state->parent;
Pthread_mutex_lock(&(bdb_state->gblcontext_lock));
set_gblcontext_int(bdb_state, gblcontext);
Pthread_mutex_unlock(&(bdb_state->gblcontext_lock));
}
unsigned long long get_gblcontext(bdb_state_type *bdb_state)
{
unsigned long long gblcontext;
if (bdb_state->parent)
bdb_state = bdb_state->parent;
Pthread_mutex_lock(&(bdb_state->gblcontext_lock));
gblcontext = get_gblcontext_int(bdb_state);
Pthread_mutex_unlock(&(bdb_state->gblcontext_lock));
return gblcontext;
}
unsigned long long get_id(bdb_state_type *bdb_state)
{
unsigned int dupcount;
unsigned long long id;
unsigned int *iptr;
if (bdb_state->parent)
bdb_state = bdb_state->parent;
iptr = (unsigned int *)&id;
Pthread_mutex_lock(&(bdb_state->id_lock));
bdb_state->id++;
dupcount = bdb_state->id;
Pthread_mutex_unlock(&(bdb_state->id_lock));
iptr[0] = htonl(comdb2_time_epoch());
iptr[1] = htonl(dupcount);
return id;
}
int get_epoch_plusplus(bdb_state_type *bdb_state)
{
static int x = 0;
x++;
return x;
}
int genid_contains_time(bdb_state_type *bdb_state)
{
return bdb_state->genid_format == LLMETA_GENID_ORIGINAL;
}
static inline void set_commit_genid_lsn_gen(bdb_state_type *bdb_state,
unsigned long long genid,
const DB_LSN *lsn,
const uint32_t *generation)
{
if (bdb_state->parent)
bdb_state = bdb_state->parent;
if (bdb_cmp_genids(genid, commit_genid) < 0) {
/* This can occur legitamately after a rep_verify_match */
if (gbl_block_set_commit_genid_trace) {
logmsg(LOGMSG_ERROR, "Blocked attempt to set lower commit_genid\n");
cheap_stack_trace();
}
return;
}
commit_genid = genid;
if (lsn) commit_lsn = *lsn;
if (generation) commit_generation = *generation;
}
#define SEED48_MAX ((1ULL << 48) - 1)
static unsigned long long get_genid_48bit(bdb_state_type *bdb_state,
unsigned int dtafile, DB_LSN *lsn,
uint32_t generation, uint64_t seed)
{
unsigned long long genid;
unsigned long long seed48;
static time_t lastwarn = 0;
time_t now;
dtafile &= 0xf;
Pthread_mutex_lock(&(bdb_state->gblcontext_lock));
if (seed)
HI48(bdb_state) = seed;
while (HI48(bdb_state) >= SEED48_MAX) {
/* This database needs a clean dump & load (or we need to expand our
* genids */
logmsg(LOGMSG_ERROR, "%s: this database has run out of genids!\n",
__func__);
sleep(1);
}
LO16(bdb_state) = dtafile;
seed48 = ++HI48(bdb_state);
genid = flibc_htonll((seed48 << 16) | dtafile);
if (lsn) {
set_commit_genid_lsn_gen(bdb_state, genid, lsn, &generation);
}
Pthread_mutex_unlock(&(bdb_state->gblcontext_lock));
if (bdb_state->attr->genid48_warn_threshold &&
(SEED48_MAX - seed48) <= bdb_state->attr->genid48_warn_threshold &&
(now = time(NULL)) > lastwarn) {
logmsg(LOGMSG_WARN,
"%s: low-genid warning: this database has only "
"%llu genids remaining\n",
__func__, SEED48_MAX - seed48);
lastwarn = now;
}
return genid;
}
void seed_genid48(bdb_state_type *bdb_state, uint64_t seed)
{
get_genid_48bit(bdb_state, 0, NULL, 0, seed);
}
static unsigned long long get_genid_timebased(bdb_state_type *bdb_state,
unsigned int dtafile, DB_LSN *lsn,
uint32_t generation)
{
/* keep a stable copy of the gblcontext */
unsigned int *iptr;
unsigned long long genid;
unsigned long long gblcontext;
unsigned int epoch;
unsigned int next_seed;
unsigned int munged_seed;
unsigned int munged_dtafile;
int epochtime;
int contexttime;
gblcontext = get_gblcontext(bdb_state);
if (!bdb_state->attr->genidplusplus) {
stall:
epochtime = comdb2_time_epoch();
contexttime = bdb_genid_timestamp(gblcontext);
if (contexttime > epochtime) {
logmsg(LOGMSG_WARN, "context is %d epoch is %d - stalling!!!\n",
contexttime, epochtime);
poll(NULL, 0, 100);
goto stall;
}
iptr = (unsigned int *)&genid;
try_again:
Pthread_mutex_lock(&(bdb_state->gblcontext_lock));
epoch = comdb2_time_epoch();
gblcontext = get_gblcontext_int(bdb_state);
contexttime = bdb_genid_timestamp(gblcontext);
if (contexttime == epoch) {
bdb_state->seed = get_dupecount_from_genid(gblcontext);
bdb_state->seed++;
} else {
bdb_state->seed = 1;
}
if (bdb_state->seed >= 0x10000) {
/* I can't conceive that this code will every execute - 64K
* insertions
* or updates a second would be rather good though. */
Pthread_mutex_unlock(&(bdb_state->gblcontext_lock));
/*fprintf(stderr, "WARNING: needed more than max genids per
* second (%d).\n", 0x10000); */
poll(NULL, 0, 10);
goto try_again;
}
} else {
iptr = (unsigned int *)&genid;
gblcontext = contexttime = epoch = get_epoch_plusplus(bdb_state);
bdb_state->seed = 1;
}
next_seed = bdb_state->seed;
/* shift over the seed then or in the dtafile */
munged_seed = next_seed << 16;
munged_dtafile = dtafile & 0x0000000f;
munged_seed = munged_seed | munged_dtafile;
/* genids (their components at least) are always "big-endian" */
iptr[0] = htonl(epoch);
iptr[1] = htonl(munged_seed);
set_gblcontext_int(bdb_state, genid);
/* this limps at a different speed compare to gblcontext */
if (lsn) {
set_commit_genid_lsn_gen(bdb_state, genid, lsn, &generation);
}
Pthread_mutex_unlock(&(bdb_state->gblcontext_lock));
return genid;
}
static inline
unsigned long long get_genid_int(bdb_state_type *bdb_state,
unsigned int dtafile, DB_LSN *lsn,
uint32_t generation)
{
/* if we were passed a child, find his parent */
/* this means we have a single genid allocator PER DB rather than PER TABLE
now. the nice part about this is maintaining distributed
"cursor contexts" becomes a much simpler problem rather than the broken
mess it currently is. */
extern int gbl_llmeta_open;
if (bdb_state->parent)
bdb_state = bdb_state->parent;
if (gbl_llmeta_open == 0)
return 0ULL;
#if defined FORCE_GENID48
if (1)
#else
if (bdb_state->genid_format == LLMETA_GENID_48BIT)
#endif
{
return get_genid_48bit(bdb_state, dtafile, lsn, generation, 0);
}
else {
return get_genid_timebased(bdb_state, dtafile, lsn, generation);
}
}
unsigned long long get_genid(bdb_state_type *bdb_state, unsigned int dtafile)
{
return get_genid_int(bdb_state, dtafile, NULL, 0);
}
unsigned long long bdb_get_a_genid(bdb_state_type *bdb_state)
{
return get_genid(bdb_state, 0);
}
/* set the genid's participant stripe value & return the new genid */
unsigned long long set_participant_stripeid(bdb_state_type *bdb_state,
int stripeid,
unsigned long long genid)
{
int mask, shift, pidb = (bdb_state->attr->participantid_bits);
if (pidb <= 0 || stripeid <= 0) {
return genid;
}
if (pidb > 12)
pidb = 12;
mask = (0x1 << pidb) - 1;
stripeid &= mask;
shift = (16 - pidb);
genid &= ~(mask << shift);
genid |= (stripeid << shift);
return genid;
}
/* set the genid's update id value & return the new genid */
unsigned long long set_updateid(bdb_state_type *bdb_state, int updateid,
unsigned long long genid)
{
int mask, bits, pidb = bdb_state->attr->participantid_bits;
if (pidb >= 12) {
return genid;
}
genid = flibc_htonll(genid);
if (pidb < 0)
pidb = 0;
bits = (12 - pidb);
mask = (0x1 << bits) - 1;
updateid &= mask;
genid &= ~(mask << 4);
genid |= (updateid << 4);
return flibc_ntohll(genid);
}
/* return the maximum updateid for this bdb_state */
int max_updateid(bdb_state_type *bdb_state)
{
int mask, bits, pidb = bdb_state->attr->participantid_bits;
if (pidb >= 12) {
return 0;
}
if (pidb < 0)
pidb = 0;
bits = (12 - pidb);
mask = (0x1 << bits) - 1;
return mask;
}
/* return the maximum participant stripe-id for this bdb_state */
int max_participant_stripeid(bdb_state_type *bdb_state)
{
int mask, pidb = bdb_state->attr->participantid_bits;
if (pidb <= 0)
return 0;
if (pidb > 12)
pidb = 12;
mask = (0x1 << pidb) - 1;
return mask;
}
/* the bdb_state object tells how many bits are in the participant stripe */
int bdb_get_participant_stripe_from_genid(bdb_state_type *bdb_state,
unsigned long long genid)
{
int mask, pidb = bdb_state->attr->participantid_bits;
if (pidb <= 0)
return 0;
if (pidb > 12)
pidb = 12;
mask = (0x1 << pidb) - 1;
genid >>= (16 - pidb);
return (genid & mask);
}
/* using the bdb_state object, return the updateid for this genid */
int get_updateid_from_genid(bdb_state_type *bdb_state, unsigned long long genid)
{
int mask, bits, pidb = bdb_state->attr->participantid_bits;
if (pidb >= 12)
return 0;
if (pidb < 0)
pidb = 0;
bits = (12 - pidb);
mask = (0x1 << bits) - 1;
genid = flibc_htonll(genid);
genid >>= 4;
return (genid & mask);
}
/* externally callable */
unsigned long long bdb_get_timestamp(bdb_state_type *bdb_state)
{
return bdb_genid_to_host_order(get_genid(bdb_state, 0));
}
/* convert a genid to host byte order. this should rarely be used, throughout
* the codebase genids should remain in network byte order; this is really only
* used so that genids can be properly printed to a string */
unsigned long long bdb_genid_to_host_order(unsigned long long genid)
{
return flibc_ntohll(genid);
}
unsigned long long bdb_genid_to_net_order(unsigned long long genid)
{
return flibc_htonll(genid);
}
/* when the database comes up, we set a master_cmpcontext value by incrementing
* the slot-component of the maximum genid.
*/
unsigned long long bdb_increment_slot(bdb_state_type *bdb_state,
unsigned long long a)
{
uint16_t slot;
/* grab the slot */
slot = (uint16_t)((a & GENID_SLOT_MASK) >> GENID_SLOT_PRE_SHIFT);
/* increment */
slot = (ntohs(slot) >> GENID_SLOT_POST_SHIFT) + 1;
/* mask out in argument genid */
a &= ~(GENID_SLOT_MASK);
/* back to big-endian */
slot = htons(slot << GENID_SLOT_POST_SHIFT);
/* replace slot */
a |= (unsigned long long)(slot) << GENID_SLOT_PRE_SHIFT;
return a;
}
unsigned long long bdb_mask_stripe(bdb_state_type *bdb_state,
unsigned long long a)
{
return (a & ~(GENID_STRIPE_MASK));
}
/* Compare two genids to determine which one would have been allocated first.
* Return codes:
* -1 a < b
* 0 a == b
* 1 a > b
*/
int bdb_cmp_genids(unsigned long long a, unsigned long long b)
{
unsigned int *a_iptr;
unsigned int *b_iptr;
unsigned int a_epoch;
unsigned int b_epoch;
unsigned int a_rhs;
unsigned int b_rhs;
a_iptr = (unsigned int *)&a;
b_iptr = (unsigned int *)&b;
/* first compare the epoch component in host byte order */
a_epoch = ntohl(a_iptr[0]);
b_epoch = ntohl(b_iptr[0]);
if (a_epoch < b_epoch)
return -1;
if (a_epoch > b_epoch)
return 1;
/* take the stripe out then get the right hand side in host byte order */
a &= ~GENID_STRIPE_MASK;
b &= ~GENID_STRIPE_MASK;
a_rhs = ntohl(a_iptr[1]);
b_rhs = ntohl(b_iptr[1]);
if (a_rhs < b_rhs)
return -1;
if (a_rhs > b_rhs)
return 1;
return 0;
}
/* return 0 if g1 and g2 are in-place equivalent, non-0 otherwise */
int bdb_inplace_cmp_genids(bdb_state_type *bdb_state, unsigned long long g1,
unsigned long long g2)
{
unsigned long long s1, s2;
/*
s1=get_search_genid(bdb_state, g1);
s2=get_search_genid(bdb_state, g2);
*/
s1 = bdb_mask_updateid(bdb_state, g1);
s2 = bdb_mask_updateid(bdb_state, g2);
return bdb_cmp_genids(s1, s2);
}
unsigned long long bdb_get_cmp_context_int(bdb_state_type *bdb_state)
{
if (bdb_state->parent)
bdb_state = bdb_state->parent;
/* if we're the master, get a context, otherwise use the last one we
recieved from the master */
if (bdb_state->repinfo->master_host ==
net_get_mynode(bdb_state->repinfo->netinfo)) {
return get_genid(bdb_state, 0);
} else {
return get_gblcontext(bdb_state);
}
}
unsigned long long bdb_get_cmp_context(bdb_state_type *bdb_state)
{
unsigned long long context;
context = bdb_get_cmp_context_int(bdb_state);
/*fprintf(stderr, "bdb_get_cmp_context %p 0x%llx\n", bdb_state, context);*/
return context;
}
int bdb_next_dtafile(bdb_state_type *bdb_state)
{
int stripe;
if (!bdb_state->parent)
return 0;
Pthread_mutex_lock(&bdb_state->last_dta_lk);
stripe = bdb_state->last_dta =
(bdb_state->last_dta + 1) % bdb_state->attr->dtastripe;
Pthread_mutex_unlock(&bdb_state->last_dta_lk);
return stripe;
}
/* Check that the given genid is older than the compare context being given.
* Returns: 1 genid is older, 0 genid is newer */
int bdb_check_genid_is_older(bdb_state_type *bdb_state,
unsigned long long genid,
unsigned long long context)
{
unsigned int *genid_iptr, *context_iptr;
unsigned int genid_epoch;
unsigned int context_epoch;
/*
fprintf(stderr, "bdb_check_genid_is_older genid: %llx context: %llx\n",
genid, context);
*/
if (!context || !genid)
return 1;
genid_iptr = (unsigned int *)&genid;
context_iptr = (unsigned int *)&context;
/* first compare epoch times */
genid_epoch = ntohl(genid_iptr[0]);
context_epoch = ntohl(context_iptr[0]);
if (genid_epoch > context_epoch)
return 0;
else if (genid_epoch < context_epoch)
return 1;
else {
/* We're in the same second. compare seeds. seeds are not reset each
* second - they wrap. i found that drawing circles helped me
*understand
* this logic.
*
* For live schema change the genid allocation system was changed to
* always hand out monotonically increasing genids. But I still keep
* this strange circular logic. Why? Just in case we get scenarios
* like old server is master, new server is replicant receiving genids
* that have wrapped like this.. This logic is complicated but should
* be harmless for the new scheme (unless we allocate a thumping large
* number of records each second). */
unsigned int genid_seed = ntohl(genid_iptr[1]) >> 4;
unsigned int context_seed = ntohl(context_iptr[1]) >> 4;
unsigned int normalised = (genid_seed - context_seed) & 0xfffffff;
/* once we have normalised the genid seed against the context seed using
* clock arithmetic, anything in the first hemisphere of the clock is
* bad (bad = newer). */
if (normalised < 0x7ffffff)
return 0;
else
return 1;
}
}
/* just like above, but reverse logic */
int bdb_check_genid_is_newer(bdb_state_type *bdb_state,
unsigned long long genid,
unsigned long long context)
{
unsigned int *genid_iptr, *context_iptr;
unsigned int genid_epoch;
unsigned int context_epoch;
if (!context || !genid)
return 0;
genid_iptr = (unsigned int *)&genid;
context_iptr = (unsigned int *)&context;
genid_epoch = ntohl(genid_iptr[0]);
context_epoch = ntohl(context_iptr[0]);
if (genid_epoch <= context_epoch)
return 0;
else if (genid_epoch > context_epoch)
return 1;
else {
unsigned int genid_seed = ntohl(genid_iptr[1]) >> 4;
unsigned int context_seed = ntohl(context_iptr[1]) >> 4;
unsigned int normalized = (context_seed - genid_seed) & 0xfffffff;
if (normalized < 0x7ffffff)
return 1;
else
return 0;
}
}
DB *get_dbp_from_genid(bdb_state_type *bdb_state, int dtanum,
unsigned long long genid, int *out_dtafile)
{
int dtafile;
if (bdb_state->attr->dtastripe) {
if (0 == dtanum || bdb_state->attr->blobstripe) {
dtafile = get_dtafile_from_genid(genid);
if (dtanum > 0 && bdb_state->blobstripe_convert_genid) {
/* kludge for records that were inserted/updated before we
* converted the db to blobstripe. */
if (bdb_check_genid_is_older(
bdb_state, genid, bdb_state->blobstripe_convert_genid))
dtafile = 0;
}
} else
dtafile = 0;
} else {
dtafile = 0;
}
if (out_dtafile)
*out_dtafile = dtafile;
return bdb_state->dbp_data[dtanum][dtafile];
}
unsigned long long bdb_normalise_genid(bdb_state_type *bdb_state,
unsigned long long genid)
{
if (!bdb_state->ondisk_header || !bdb_state->inplace_updates) {
return bdb_mask_updateid(bdb_state, genid);
} else {
return genid;
}
}
unsigned long long increment_seq(unsigned long long crt)
{
crt = bdb_genid_to_host_order(crt);
return bdb_genid_to_net_order(++crt);
}
/* gets the datafile/stripe that an add from this thread would write to
*/
int bdb_get_active_stripe_int(bdb_state_type *bdb_state)
{
bdb_state_type *parent;
size_t id;
if (bdb_state->parent)
parent = bdb_state->parent;
else
parent = bdb_state;
if (parent->attr->round_robin_stripes)
return bdb_next_dtafile(bdb_state);
id = (size_t)pthread_getspecific(parent->tid_key);
return id % bdb_state->attr->dtastripe;
}
int bdb_get_active_stripe(bdb_state_type *bdb_state)
{
int dtafile;
BDB_READLOCK("bdb_get_active_dtafile");
dtafile = bdb_get_active_stripe_int(bdb_state);
BDB_RELLOCK();
return dtafile;
}
/* this is called by writers on the master */
unsigned long long bdb_gen_commit_genid(bdb_state_type *bdb_state,
const void *plsn, uint32_t generation)
{
unsigned long long ret;
/* locks embedded */
ret = get_genid_int(bdb_state, 0, (DB_LSN *)plsn, generation);
/*
fprintf( stderr, "%s:%d got context %llx\n", __FILE__, __LINE__, ret);
*/
return ret;
}
/* this is called by readers that need starting context */
unsigned long long bdb_get_commit_genid_generation( bdb_state_type *bdb_state, void *plsn, uint32_t *generation)
{
unsigned long long ret = 0;
if (bdb_state->parent)
bdb_state = bdb_state->parent;
Pthread_mutex_lock(&(bdb_state->gblcontext_lock));
ret = commit_genid;
if (plsn)
*(DB_LSN *)plsn = commit_lsn;