Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * nbtsort.c
4 : * Build a btree from sorted input by loading leaf pages sequentially.
5 : *
6 : * NOTES
7 : *
8 : * We use tuplesort.c to sort the given index tuples into order.
9 : * Then we scan the index tuples in order and build the btree pages
10 : * for each level. We load source tuples into leaf-level pages.
11 : * Whenever we fill a page at one level, we add a link to it to its
12 : * parent level (starting a new parent level if necessary). When
13 : * done, we write out each final page on each level, adding it to
14 : * its parent level. When we have only one page on a level, it must be
15 : * the root -- it can be attached to the btree metapage and we are done.
16 : *
17 : * It is not wise to pack the pages entirely full, since then *any*
18 : * insertion would cause a split (and not only of the leaf page; the need
19 : * for a split would cascade right up the tree). The steady-state load
20 : * factor for btrees is usually estimated at 70%. We choose to pack leaf
21 : * pages to the user-controllable fill factor (default 90%) while upper pages
22 : * are always packed to 70%. This gives us reasonable density (there aren't
23 : * many upper pages if the keys are reasonable-size) without risking a lot of
24 : * cascading splits during early insertions.
25 : *
26 : * We use the bulk smgr loading facility to bypass the buffer cache and
27 : * WAL-log the pages efficiently.
28 : *
29 : * This code isn't concerned about the FSM at all. The caller is responsible
30 : * for initializing that.
31 : *
32 : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
33 : * Portions Copyright (c) 1994, Regents of the University of California
34 : *
35 : * IDENTIFICATION
36 : * src/backend/access/nbtree/nbtsort.c
37 : *
38 : *-------------------------------------------------------------------------
39 : */
40 :
41 : #include "postgres.h"
42 :
43 : #include "access/nbtree.h"
44 : #include "access/parallel.h"
45 : #include "access/relscan.h"
46 : #include "access/table.h"
47 : #include "access/tableam.h"
48 : #include "access/xact.h"
49 : #include "catalog/index.h"
50 : #include "commands/progress.h"
51 : #include "executor/instrument.h"
52 : #include "miscadmin.h"
53 : #include "pgstat.h"
54 : #include "storage/bulk_write.h"
55 : #include "tcop/tcopprot.h"
56 : #include "utils/rel.h"
57 : #include "utils/sortsupport.h"
58 : #include "utils/tuplesort.h"
59 :
60 :
61 : /* Magic numbers for parallel state sharing */
62 : #define PARALLEL_KEY_BTREE_SHARED UINT64CONST(0xA000000000000001)
63 : #define PARALLEL_KEY_TUPLESORT UINT64CONST(0xA000000000000002)
64 : #define PARALLEL_KEY_TUPLESORT_SPOOL2 UINT64CONST(0xA000000000000003)
65 : #define PARALLEL_KEY_QUERY_TEXT UINT64CONST(0xA000000000000004)
66 : #define PARALLEL_KEY_WAL_USAGE UINT64CONST(0xA000000000000005)
67 : #define PARALLEL_KEY_BUFFER_USAGE UINT64CONST(0xA000000000000006)
68 :
69 : /*
70 : * DISABLE_LEADER_PARTICIPATION disables the leader's participation in
71 : * parallel index builds. This may be useful as a debugging aid.
72 : #undef DISABLE_LEADER_PARTICIPATION
73 : */
74 :
75 : /*
76 : * Status record for spooling/sorting phase. (Note we may have two of
77 : * these due to the special requirements for uniqueness-checking with
78 : * dead tuples.)
79 : */
80 : typedef struct BTSpool
81 : {
82 : Tuplesortstate *sortstate; /* state data for tuplesort.c */
83 : Relation heap;
84 : Relation index;
85 : bool isunique;
86 : bool nulls_not_distinct;
87 : } BTSpool;
88 :
89 : /*
90 : * Status for index builds performed in parallel. This is allocated in a
91 : * dynamic shared memory segment. Note that there is a separate tuplesort TOC
92 : * entry, private to tuplesort.c but allocated by this module on its behalf.
93 : */
94 : typedef struct BTShared
95 : {
96 : /*
97 : * These fields are not modified during the sort. They primarily exist
98 : * for the benefit of worker processes that need to create BTSpool state
99 : * corresponding to that used by the leader.
100 : */
101 : Oid heaprelid;
102 : Oid indexrelid;
103 : bool isunique;
104 : bool nulls_not_distinct;
105 : bool isconcurrent;
106 : int scantuplesortstates;
107 :
108 : /* Query ID, for report in worker processes */
109 : int64 queryid;
110 :
111 : /*
112 : * workersdonecv is used to monitor the progress of workers. All parallel
113 : * participants must indicate that they are done before leader can use
114 : * mutable state that workers maintain during scan (and before leader can
115 : * proceed to tuplesort_performsort()).
116 : */
117 : ConditionVariable workersdonecv;
118 :
119 : /*
120 : * mutex protects all fields before heapdesc.
121 : *
122 : * These fields contain status information of interest to B-Tree index
123 : * builds that must work just the same when an index is built in parallel.
124 : */
125 : slock_t mutex;
126 :
127 : /*
128 : * Mutable state that is maintained by workers, and reported back to
129 : * leader at end of parallel scan.
130 : *
131 : * nparticipantsdone is number of worker processes finished.
132 : *
133 : * reltuples is the total number of input heap tuples.
134 : *
135 : * havedead indicates if RECENTLY_DEAD tuples were encountered during
136 : * build.
137 : *
138 : * indtuples is the total number of tuples that made it into the index.
139 : *
140 : * brokenhotchain indicates if any worker detected a broken HOT chain
141 : * during build.
142 : */
143 : int nparticipantsdone;
144 : double reltuples;
145 : bool havedead;
146 : double indtuples;
147 : bool brokenhotchain;
148 :
149 : /*
150 : * ParallelTableScanDescData data follows. Can't directly embed here, as
151 : * implementations of the parallel table scan desc interface might need
152 : * stronger alignment.
153 : */
154 : } BTShared;
155 :
156 : /*
157 : * Return pointer to a BTShared's parallel table scan.
158 : *
159 : * c.f. shm_toc_allocate as to why BUFFERALIGN is used, rather than just
160 : * MAXALIGN.
161 : */
162 : #define ParallelTableScanFromBTShared(shared) \
163 : (ParallelTableScanDesc) ((char *) (shared) + BUFFERALIGN(sizeof(BTShared)))
164 :
165 : /*
166 : * Status for leader in parallel index build.
167 : */
168 : typedef struct BTLeader
169 : {
170 : /* parallel context itself */
171 : ParallelContext *pcxt;
172 :
173 : /*
174 : * nparticipanttuplesorts is the exact number of worker processes
175 : * successfully launched, plus one leader process if it participates as a
176 : * worker (only DISABLE_LEADER_PARTICIPATION builds avoid leader
177 : * participating as a worker).
178 : */
179 : int nparticipanttuplesorts;
180 :
181 : /*
182 : * Leader process convenience pointers to shared state (leader avoids TOC
183 : * lookups).
184 : *
185 : * btshared is the shared state for entire build. sharedsort is the
186 : * shared, tuplesort-managed state passed to each process tuplesort.
187 : * sharedsort2 is the corresponding btspool2 shared state, used only when
188 : * building unique indexes. snapshot is the snapshot used by the scan iff
189 : * an MVCC snapshot is required.
190 : */
191 : BTShared *btshared;
192 : Sharedsort *sharedsort;
193 : Sharedsort *sharedsort2;
194 : Snapshot snapshot;
195 : WalUsage *walusage;
196 : BufferUsage *bufferusage;
197 : } BTLeader;
198 :
199 : /*
200 : * Working state for btbuild and its callback.
201 : *
202 : * When parallel CREATE INDEX is used, there is a BTBuildState for each
203 : * participant.
204 : */
205 : typedef struct BTBuildState
206 : {
207 : bool isunique;
208 : bool nulls_not_distinct;
209 : bool havedead;
210 : Relation heap;
211 : BTSpool *spool;
212 :
213 : /*
214 : * spool2 is needed only when the index is a unique index. Dead tuples are
215 : * put into spool2 instead of spool in order to avoid uniqueness check.
216 : */
217 : BTSpool *spool2;
218 : double indtuples;
219 :
220 : /*
221 : * btleader is only present when a parallel index build is performed, and
222 : * only in the leader process. (Actually, only the leader has a
223 : * BTBuildState. Workers have their own spool and spool2, though.)
224 : */
225 : BTLeader *btleader;
226 : } BTBuildState;
227 :
228 : /*
229 : * Status record for a btree page being built. We have one of these
230 : * for each active tree level.
231 : */
232 : typedef struct BTPageState
233 : {
234 : BulkWriteBuffer btps_buf; /* workspace for page building */
235 : BlockNumber btps_blkno; /* block # to write this page at */
236 : IndexTuple btps_lowkey; /* page's strict lower bound pivot tuple */
237 : OffsetNumber btps_lastoff; /* last item offset loaded */
238 : Size btps_lastextra; /* last item's extra posting list space */
239 : uint32 btps_level; /* tree level (0 = leaf) */
240 : Size btps_full; /* "full" if less than this much free space */
241 : struct BTPageState *btps_next; /* link to parent level, if any */
242 : } BTPageState;
243 :
244 : /*
245 : * Overall status record for index writing phase.
246 : */
247 : typedef struct BTWriteState
248 : {
249 : Relation heap;
250 : Relation index;
251 : BulkWriteState *bulkstate;
252 : BTScanInsert inskey; /* generic insertion scankey */
253 : BlockNumber btws_pages_alloced; /* # pages allocated */
254 : } BTWriteState;
255 :
256 :
257 : static double _bt_spools_heapscan(Relation heap, Relation index,
258 : BTBuildState *buildstate, IndexInfo *indexInfo);
259 : static void _bt_spooldestroy(BTSpool *btspool);
260 : static void _bt_spool(BTSpool *btspool, const ItemPointerData *self,
261 : const Datum *values, const bool *isnull);
262 : static void _bt_leafbuild(BTSpool *btspool, BTSpool *btspool2);
263 : static void _bt_build_callback(Relation index, ItemPointer tid, Datum *values,
264 : bool *isnull, bool tupleIsAlive, void *state);
265 : static BulkWriteBuffer _bt_blnewpage(BTWriteState *wstate, uint32 level);
266 : static BTPageState *_bt_pagestate(BTWriteState *wstate, uint32 level);
267 : static void _bt_slideleft(Page rightmostpage);
268 : static void _bt_sortaddtup(Page page, Size itemsize,
269 : const IndexTupleData *itup, OffsetNumber itup_off,
270 : bool newfirstdataitem);
271 : static void _bt_buildadd(BTWriteState *wstate, BTPageState *state,
272 : IndexTuple itup, Size truncextra);
273 : static void _bt_sort_dedup_finish_pending(BTWriteState *wstate,
274 : BTPageState *state,
275 : BTDedupState dstate);
276 : static void _bt_uppershutdown(BTWriteState *wstate, BTPageState *state);
277 : static void _bt_load(BTWriteState *wstate,
278 : BTSpool *btspool, BTSpool *btspool2);
279 : static void _bt_begin_parallel(BTBuildState *buildstate, bool isconcurrent,
280 : int request);
281 : static void _bt_end_parallel(BTLeader *btleader);
282 : static Size _bt_parallel_estimate_shared(Relation heap, Snapshot snapshot);
283 : static double _bt_parallel_heapscan(BTBuildState *buildstate,
284 : bool *brokenhotchain);
285 : static void _bt_leader_participate_as_worker(BTBuildState *buildstate);
286 : static void _bt_parallel_scan_and_sort(BTSpool *btspool, BTSpool *btspool2,
287 : BTShared *btshared, Sharedsort *sharedsort,
288 : Sharedsort *sharedsort2, int sortmem,
289 : bool progress);
290 :
291 :
292 : /*
293 : * btbuild() -- build a new btree index.
294 : */
295 : IndexBuildResult *
296 50732 : btbuild(Relation heap, Relation index, IndexInfo *indexInfo)
297 : {
298 : IndexBuildResult *result;
299 : BTBuildState buildstate;
300 : double reltuples;
301 :
302 : #ifdef BTREE_BUILD_STATS
303 : if (log_btree_build_stats)
304 : ResetUsage();
305 : #endif /* BTREE_BUILD_STATS */
306 :
307 50732 : buildstate.isunique = indexInfo->ii_Unique;
308 50732 : buildstate.nulls_not_distinct = indexInfo->ii_NullsNotDistinct;
309 50732 : buildstate.havedead = false;
310 50732 : buildstate.heap = heap;
311 50732 : buildstate.spool = NULL;
312 50732 : buildstate.spool2 = NULL;
313 50732 : buildstate.indtuples = 0;
314 50732 : buildstate.btleader = NULL;
315 :
316 : /*
317 : * We expect to be called exactly once for any index relation. If that's
318 : * not the case, big trouble's what we have.
319 : */
320 50732 : if (RelationGetNumberOfBlocks(index) != 0)
321 0 : elog(ERROR, "index \"%s\" already contains data",
322 : RelationGetRelationName(index));
323 :
324 50732 : reltuples = _bt_spools_heapscan(heap, index, &buildstate, indexInfo);
325 :
326 : /*
327 : * Finish the build by (1) completing the sort of the spool file, (2)
328 : * inserting the sorted tuples into btree pages and (3) building the upper
329 : * levels. Finally, it may also be necessary to end use of parallelism.
330 : */
331 50720 : _bt_leafbuild(buildstate.spool, buildstate.spool2);
332 50636 : _bt_spooldestroy(buildstate.spool);
333 50636 : if (buildstate.spool2)
334 26 : _bt_spooldestroy(buildstate.spool2);
335 50636 : if (buildstate.btleader)
336 148 : _bt_end_parallel(buildstate.btleader);
337 :
338 50636 : result = (IndexBuildResult *) palloc(sizeof(IndexBuildResult));
339 :
340 50636 : result->heap_tuples = reltuples;
341 50636 : result->index_tuples = buildstate.indtuples;
342 :
343 : #ifdef BTREE_BUILD_STATS
344 : if (log_btree_build_stats)
345 : {
346 : ShowUsage("BTREE BUILD STATS");
347 : ResetUsage();
348 : }
349 : #endif /* BTREE_BUILD_STATS */
350 :
351 50636 : return result;
352 : }
353 :
354 : /*
355 : * Create and initialize one or two spool structures, and save them in caller's
356 : * buildstate argument. May also fill-in fields within indexInfo used by index
357 : * builds.
358 : *
359 : * Scans the heap, possibly in parallel, filling spools with IndexTuples. This
360 : * routine encapsulates all aspects of managing parallelism. Caller need only
361 : * call _bt_end_parallel() in parallel case after it is done with spool/spool2.
362 : *
363 : * Returns the total number of heap tuples scanned.
364 : */
365 : static double
366 50732 : _bt_spools_heapscan(Relation heap, Relation index, BTBuildState *buildstate,
367 : IndexInfo *indexInfo)
368 : {
369 50732 : BTSpool *btspool = (BTSpool *) palloc0(sizeof(BTSpool));
370 50732 : SortCoordinate coordinate = NULL;
371 50732 : double reltuples = 0;
372 :
373 : /*
374 : * We size the sort area as maintenance_work_mem rather than work_mem to
375 : * speed index creation. This should be OK since a single backend can't
376 : * run multiple index creations in parallel (see also: notes on
377 : * parallelism and maintenance_work_mem below).
378 : */
379 50732 : btspool->heap = heap;
380 50732 : btspool->index = index;
381 50732 : btspool->isunique = indexInfo->ii_Unique;
382 50732 : btspool->nulls_not_distinct = indexInfo->ii_NullsNotDistinct;
383 :
384 : /* Save as primary spool */
385 50732 : buildstate->spool = btspool;
386 :
387 : /* Report table scan phase started */
388 50732 : pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,
389 : PROGRESS_BTREE_PHASE_INDEXBUILD_TABLESCAN);
390 :
391 : /* Attempt to launch parallel worker scan when required */
392 50732 : if (indexInfo->ii_ParallelWorkers > 0)
393 148 : _bt_begin_parallel(buildstate, indexInfo->ii_Concurrent,
394 : indexInfo->ii_ParallelWorkers);
395 :
396 : /*
397 : * If parallel build requested and at least one worker process was
398 : * successfully launched, set up coordination state
399 : */
400 50732 : if (buildstate->btleader)
401 : {
402 148 : coordinate = (SortCoordinate) palloc0(sizeof(SortCoordinateData));
403 148 : coordinate->isWorker = false;
404 148 : coordinate->nParticipants =
405 148 : buildstate->btleader->nparticipanttuplesorts;
406 148 : coordinate->sharedsort = buildstate->btleader->sharedsort;
407 : }
408 :
409 : /*
410 : * Begin serial/leader tuplesort.
411 : *
412 : * In cases where parallelism is involved, the leader receives the same
413 : * share of maintenance_work_mem as a serial sort (it is generally treated
414 : * in the same way as a serial sort once we return). Parallel worker
415 : * Tuplesortstates will have received only a fraction of
416 : * maintenance_work_mem, though.
417 : *
418 : * We rely on the lifetime of the Leader Tuplesortstate almost not
419 : * overlapping with any worker Tuplesortstate's lifetime. There may be
420 : * some small overlap, but that's okay because we rely on leader
421 : * Tuplesortstate only allocating a small, fixed amount of memory here.
422 : * When its tuplesort_performsort() is called (by our caller), and
423 : * significant amounts of memory are likely to be used, all workers must
424 : * have already freed almost all memory held by their Tuplesortstates
425 : * (they are about to go away completely, too). The overall effect is
426 : * that maintenance_work_mem always represents an absolute high watermark
427 : * on the amount of memory used by a CREATE INDEX operation, regardless of
428 : * the use of parallelism or any other factor.
429 : */
430 101464 : buildstate->spool->sortstate =
431 50732 : tuplesort_begin_index_btree(heap, index, buildstate->isunique,
432 50732 : buildstate->nulls_not_distinct,
433 : maintenance_work_mem, coordinate,
434 : TUPLESORT_NONE);
435 :
436 : /*
437 : * If building a unique index, put dead tuples in a second spool to keep
438 : * them out of the uniqueness check. We expect that the second spool (for
439 : * dead tuples) won't get very full, so we give it only work_mem.
440 : */
441 50732 : if (indexInfo->ii_Unique)
442 : {
443 41430 : BTSpool *btspool2 = (BTSpool *) palloc0(sizeof(BTSpool));
444 41430 : SortCoordinate coordinate2 = NULL;
445 :
446 : /* Initialize secondary spool */
447 41430 : btspool2->heap = heap;
448 41430 : btspool2->index = index;
449 41430 : btspool2->isunique = false;
450 : /* Save as secondary spool */
451 41430 : buildstate->spool2 = btspool2;
452 :
453 41430 : if (buildstate->btleader)
454 : {
455 : /*
456 : * Set up non-private state that is passed to
457 : * tuplesort_begin_index_btree() about the basic high level
458 : * coordination of a parallel sort.
459 : */
460 64 : coordinate2 = (SortCoordinate) palloc0(sizeof(SortCoordinateData));
461 64 : coordinate2->isWorker = false;
462 64 : coordinate2->nParticipants =
463 64 : buildstate->btleader->nparticipanttuplesorts;
464 64 : coordinate2->sharedsort = buildstate->btleader->sharedsort2;
465 : }
466 :
467 : /*
468 : * We expect that the second one (for dead tuples) won't get very
469 : * full, so we give it only work_mem
470 : */
471 41430 : buildstate->spool2->sortstate =
472 41430 : tuplesort_begin_index_btree(heap, index, false, false, work_mem,
473 : coordinate2, TUPLESORT_NONE);
474 : }
475 :
476 : /* Fill spool using either serial or parallel heap scan */
477 50732 : if (!buildstate->btleader)
478 50584 : reltuples = table_index_build_scan(heap, index, indexInfo, true, true,
479 : _bt_build_callback, buildstate,
480 : NULL);
481 : else
482 148 : reltuples = _bt_parallel_heapscan(buildstate,
483 : &indexInfo->ii_BrokenHotChain);
484 :
485 : /*
486 : * Set the progress target for the next phase. Reset the block number
487 : * values set by table_index_build_scan
488 : */
489 : {
490 50720 : const int progress_index[] = {
491 : PROGRESS_CREATEIDX_TUPLES_TOTAL,
492 : PROGRESS_SCAN_BLOCKS_TOTAL,
493 : PROGRESS_SCAN_BLOCKS_DONE
494 : };
495 50720 : const int64 progress_vals[] = {
496 50720 : buildstate->indtuples,
497 : 0, 0
498 : };
499 :
500 50720 : pgstat_progress_update_multi_param(3, progress_index, progress_vals);
501 : }
502 :
503 : /* okay, all heap tuples are spooled */
504 50720 : if (buildstate->spool2 && !buildstate->havedead)
505 : {
506 : /* spool2 turns out to be unnecessary */
507 41404 : _bt_spooldestroy(buildstate->spool2);
508 41404 : buildstate->spool2 = NULL;
509 : }
510 :
511 50720 : return reltuples;
512 : }
513 :
514 : /*
515 : * clean up a spool structure and its substructures.
516 : */
517 : static void
518 92066 : _bt_spooldestroy(BTSpool *btspool)
519 : {
520 92066 : tuplesort_end(btspool->sortstate);
521 92066 : pfree(btspool);
522 92066 : }
523 :
524 : /*
525 : * spool an index entry into the sort file.
526 : */
527 : static void
528 12786100 : _bt_spool(BTSpool *btspool, const ItemPointerData *self, const Datum *values, const bool *isnull)
529 : {
530 12786100 : tuplesort_putindextuplevalues(btspool->sortstate, btspool->index,
531 : self, values, isnull);
532 12786100 : }
533 :
534 : /*
535 : * given a spool loaded by successive calls to _bt_spool,
536 : * create an entire btree.
537 : */
538 : static void
539 50720 : _bt_leafbuild(BTSpool *btspool, BTSpool *btspool2)
540 : {
541 : BTWriteState wstate;
542 :
543 : #ifdef BTREE_BUILD_STATS
544 : if (log_btree_build_stats)
545 : {
546 : ShowUsage("BTREE BUILD (Spool) STATISTICS");
547 : ResetUsage();
548 : }
549 : #endif /* BTREE_BUILD_STATS */
550 :
551 : /* Execute the sort */
552 50720 : pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,
553 : PROGRESS_BTREE_PHASE_PERFORMSORT_1);
554 50720 : tuplesort_performsort(btspool->sortstate);
555 50636 : if (btspool2)
556 : {
557 26 : pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,
558 : PROGRESS_BTREE_PHASE_PERFORMSORT_2);
559 26 : tuplesort_performsort(btspool2->sortstate);
560 : }
561 :
562 50636 : wstate.heap = btspool->heap;
563 50636 : wstate.index = btspool->index;
564 50636 : wstate.inskey = _bt_mkscankey(wstate.index, NULL);
565 : /* _bt_mkscankey() won't set allequalimage without metapage */
566 50636 : wstate.inskey->allequalimage = _bt_allequalimage(wstate.index, true);
567 :
568 : /* reserve the metapage */
569 50636 : wstate.btws_pages_alloced = BTREE_METAPAGE + 1;
570 :
571 50636 : pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,
572 : PROGRESS_BTREE_PHASE_LEAF_LOAD);
573 50636 : _bt_load(&wstate, btspool, btspool2);
574 50636 : }
575 :
576 : /*
577 : * Per-tuple callback for table_index_build_scan
578 : */
579 : static void
580 12786100 : _bt_build_callback(Relation index,
581 : ItemPointer tid,
582 : Datum *values,
583 : bool *isnull,
584 : bool tupleIsAlive,
585 : void *state)
586 : {
587 12786100 : BTBuildState *buildstate = (BTBuildState *) state;
588 :
589 : /*
590 : * insert the index tuple into the appropriate spool file for subsequent
591 : * processing
592 : */
593 12786100 : if (tupleIsAlive || buildstate->spool2 == NULL)
594 12785594 : _bt_spool(buildstate->spool, tid, values, isnull);
595 : else
596 : {
597 : /* dead tuples are put into spool2 */
598 506 : buildstate->havedead = true;
599 506 : _bt_spool(buildstate->spool2, tid, values, isnull);
600 : }
601 :
602 12786100 : buildstate->indtuples += 1;
603 12786100 : }
604 :
605 : /*
606 : * allocate workspace for a new, clean btree page, not linked to any siblings.
607 : */
608 : static BulkWriteBuffer
609 53000 : _bt_blnewpage(BTWriteState *wstate, uint32 level)
610 : {
611 : BulkWriteBuffer buf;
612 : Page page;
613 : BTPageOpaque opaque;
614 :
615 53000 : buf = smgr_bulk_get_buf(wstate->bulkstate);
616 53000 : page = (Page) buf;
617 :
618 : /* Zero the page and set up standard page header info */
619 53000 : _bt_pageinit(page, BLCKSZ);
620 :
621 : /* Initialize BT opaque state */
622 53000 : opaque = BTPageGetOpaque(page);
623 53000 : opaque->btpo_prev = opaque->btpo_next = P_NONE;
624 53000 : opaque->btpo_level = level;
625 53000 : opaque->btpo_flags = (level > 0) ? 0 : BTP_LEAF;
626 53000 : opaque->btpo_cycleid = 0;
627 :
628 : /* Make the P_HIKEY line pointer appear allocated */
629 53000 : ((PageHeader) page)->pd_lower += sizeof(ItemIdData);
630 :
631 53000 : return buf;
632 : }
633 :
634 : /*
635 : * emit a completed btree page, and release the working storage.
636 : */
637 : static void
638 103636 : _bt_blwritepage(BTWriteState *wstate, BulkWriteBuffer buf, BlockNumber blkno)
639 : {
640 103636 : smgr_bulk_write(wstate->bulkstate, blkno, buf, true);
641 : /* smgr_bulk_write took ownership of 'buf' */
642 103636 : }
643 :
644 : /*
645 : * allocate and initialize a new BTPageState. the returned structure
646 : * is suitable for immediate use by _bt_buildadd.
647 : */
648 : static BTPageState *
649 12890 : _bt_pagestate(BTWriteState *wstate, uint32 level)
650 : {
651 12890 : BTPageState *state = (BTPageState *) palloc0(sizeof(BTPageState));
652 :
653 : /* create initial page for level */
654 12890 : state->btps_buf = _bt_blnewpage(wstate, level);
655 :
656 : /* and assign it a page position */
657 12890 : state->btps_blkno = wstate->btws_pages_alloced++;
658 :
659 12890 : state->btps_lowkey = NULL;
660 : /* initialize lastoff so first item goes into P_FIRSTKEY */
661 12890 : state->btps_lastoff = P_HIKEY;
662 12890 : state->btps_lastextra = 0;
663 12890 : state->btps_level = level;
664 : /* set "full" threshold based on level. See notes at head of file. */
665 12890 : if (level > 0)
666 2692 : state->btps_full = (BLCKSZ * (100 - BTREE_NONLEAF_FILLFACTOR) / 100);
667 : else
668 10198 : state->btps_full = BTGetTargetPageFreeSpace(wstate->index);
669 :
670 : /* no parent level, yet */
671 12890 : state->btps_next = NULL;
672 :
673 12890 : return state;
674 : }
675 :
676 : /*
677 : * Slide the array of ItemIds from the page back one slot (from P_FIRSTKEY to
678 : * P_HIKEY, overwriting P_HIKEY).
679 : *
680 : * _bt_blnewpage() makes the P_HIKEY line pointer appear allocated, but the
681 : * rightmost page on its level is not supposed to get a high key. Now that
682 : * it's clear that this page is a rightmost page, remove the unneeded empty
683 : * P_HIKEY line pointer space.
684 : */
685 : static void
686 12890 : _bt_slideleft(Page rightmostpage)
687 : {
688 : OffsetNumber off;
689 : OffsetNumber maxoff;
690 : ItemId previi;
691 :
692 12890 : maxoff = PageGetMaxOffsetNumber(rightmostpage);
693 : Assert(maxoff >= P_FIRSTKEY);
694 12890 : previi = PageGetItemId(rightmostpage, P_HIKEY);
695 761300 : for (off = P_FIRSTKEY; off <= maxoff; off = OffsetNumberNext(off))
696 : {
697 748410 : ItemId thisii = PageGetItemId(rightmostpage, off);
698 :
699 748410 : *previi = *thisii;
700 748410 : previi = thisii;
701 : }
702 12890 : ((PageHeader) rightmostpage)->pd_lower -= sizeof(ItemIdData);
703 12890 : }
704 :
705 : /*
706 : * Add an item to a page being built.
707 : *
708 : * This is very similar to nbtinsert.c's _bt_pgaddtup(), but this variant
709 : * raises an error directly.
710 : *
711 : * Note that our nbtsort.c caller does not know yet if the page will be
712 : * rightmost. Offset P_FIRSTKEY is always assumed to be the first data key by
713 : * caller. Page that turns out to be the rightmost on its level is fixed by
714 : * calling _bt_slideleft().
715 : */
716 : static void
717 11498890 : _bt_sortaddtup(Page page,
718 : Size itemsize,
719 : const IndexTupleData *itup,
720 : OffsetNumber itup_off,
721 : bool newfirstdataitem)
722 : {
723 : IndexTupleData trunctuple;
724 :
725 11498890 : if (newfirstdataitem)
726 : {
727 2816 : trunctuple = *itup;
728 2816 : trunctuple.t_info = sizeof(IndexTupleData);
729 2816 : BTreeTupleSetNAtts(&trunctuple, 0, false);
730 2816 : itup = &trunctuple;
731 2816 : itemsize = sizeof(IndexTupleData);
732 : }
733 :
734 11498890 : if (PageAddItem(page, itup, itemsize, itup_off, false, false) == InvalidOffsetNumber)
735 0 : elog(ERROR, "failed to add item to the index page");
736 11498890 : }
737 :
738 : /*----------
739 : * Add an item to a disk page from the sort output (or add a posting list
740 : * item formed from the sort output).
741 : *
742 : * We must be careful to observe the page layout conventions of nbtsearch.c:
743 : * - rightmost pages start data items at P_HIKEY instead of at P_FIRSTKEY.
744 : * - on non-leaf pages, the key portion of the first item need not be
745 : * stored, we should store only the link.
746 : *
747 : * A leaf page being built looks like:
748 : *
749 : * +----------------+---------------------------------+
750 : * | PageHeaderData | linp0 linp1 linp2 ... |
751 : * +-----------+----+---------------------------------+
752 : * | ... linpN | |
753 : * +-----------+--------------------------------------+
754 : * | ^ last |
755 : * | |
756 : * +-------------+------------------------------------+
757 : * | | itemN ... |
758 : * +-------------+------------------+-----------------+
759 : * | ... item3 item2 item1 | "special space" |
760 : * +--------------------------------+-----------------+
761 : *
762 : * Contrast this with the diagram in bufpage.h; note the mismatch
763 : * between linps and items. This is because we reserve linp0 as a
764 : * placeholder for the pointer to the "high key" item; when we have
765 : * filled up the page, we will set linp0 to point to itemN and clear
766 : * linpN. On the other hand, if we find this is the last (rightmost)
767 : * page, we leave the items alone and slide the linp array over. If
768 : * the high key is to be truncated, offset 1 is deleted, and we insert
769 : * the truncated high key at offset 1.
770 : *
771 : * 'last' pointer indicates the last offset added to the page.
772 : *
773 : * 'truncextra' is the size of the posting list in itup, if any. This
774 : * information is stashed for the next call here, when we may benefit
775 : * from considering the impact of truncating away the posting list on
776 : * the page before deciding to finish the page off. Posting lists are
777 : * often relatively large, so it is worth going to the trouble of
778 : * accounting for the saving from truncating away the posting list of
779 : * the tuple that becomes the high key (that may be the only way to
780 : * get close to target free space on the page). Note that this is
781 : * only used for the soft fillfactor-wise limit, not the critical hard
782 : * limit.
783 : *----------
784 : */
785 : static void
786 11458780 : _bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup,
787 : Size truncextra)
788 : {
789 : BulkWriteBuffer nbuf;
790 : Page npage;
791 : BlockNumber nblkno;
792 : OffsetNumber last_off;
793 : Size last_truncextra;
794 : Size pgspc;
795 : Size itupsz;
796 : bool isleaf;
797 :
798 : /*
799 : * This is a handy place to check for cancel interrupts during the btree
800 : * load phase of index creation.
801 : */
802 11458780 : CHECK_FOR_INTERRUPTS();
803 :
804 11458780 : nbuf = state->btps_buf;
805 11458780 : npage = (Page) nbuf;
806 11458780 : nblkno = state->btps_blkno;
807 11458780 : last_off = state->btps_lastoff;
808 11458780 : last_truncextra = state->btps_lastextra;
809 11458780 : state->btps_lastextra = truncextra;
810 :
811 11458780 : pgspc = PageGetFreeSpace(npage);
812 11458780 : itupsz = IndexTupleSize(itup);
813 11458780 : itupsz = MAXALIGN(itupsz);
814 : /* Leaf case has slightly different rules due to suffix truncation */
815 11458780 : isleaf = (state->btps_level == 0);
816 :
817 : /*
818 : * Check whether the new item can fit on a btree page on current level at
819 : * all.
820 : *
821 : * Every newly built index will treat heap TID as part of the keyspace,
822 : * which imposes the requirement that new high keys must occasionally have
823 : * a heap TID appended within _bt_truncate(). That may leave a new pivot
824 : * tuple one or two MAXALIGN() quantums larger than the original
825 : * firstright tuple it's derived from. v4 deals with the problem by
826 : * decreasing the limit on the size of tuples inserted on the leaf level
827 : * by the same small amount. Enforce the new v4+ limit on the leaf level,
828 : * and the old limit on internal levels, since pivot tuples may need to
829 : * make use of the reserved space. This should never fail on internal
830 : * pages.
831 : */
832 11458780 : if (unlikely(itupsz > BTMaxItemSize))
833 264 : _bt_check_third_page(wstate->index, wstate->heap, isleaf, npage,
834 : itup);
835 :
836 : /*
837 : * Check to see if current page will fit new item, with space left over to
838 : * append a heap TID during suffix truncation when page is a leaf page.
839 : *
840 : * It is guaranteed that we can fit at least 2 non-pivot tuples plus a
841 : * high key with heap TID when finishing off a leaf page, since we rely on
842 : * _bt_check_third_page() rejecting oversized non-pivot tuples. On
843 : * internal pages we can always fit 3 pivot tuples with larger internal
844 : * page tuple limit (includes page high key).
845 : *
846 : * Most of the time, a page is only "full" in the sense that the soft
847 : * fillfactor-wise limit has been exceeded. However, we must always leave
848 : * at least two items plus a high key on each page before starting a new
849 : * page. Disregard fillfactor and insert on "full" current page if we
850 : * don't have the minimum number of items yet. (Note that we deliberately
851 : * assume that suffix truncation neither enlarges nor shrinks new high key
852 : * when applying soft limit, except when last tuple has a posting list.)
853 : */
854 : Assert(last_truncextra == 0 || isleaf);
855 11458780 : if (pgspc < itupsz + (isleaf ? MAXALIGN(sizeof(ItemPointerData)) : 0) ||
856 11457624 : (pgspc + last_truncextra < state->btps_full && last_off > P_FIRSTKEY))
857 : {
858 : /*
859 : * Finish off the page and write it out.
860 : */
861 40110 : BulkWriteBuffer obuf = nbuf;
862 40110 : Page opage = npage;
863 40110 : BlockNumber oblkno = nblkno;
864 : ItemId ii;
865 : ItemId hii;
866 : IndexTuple oitup;
867 :
868 : /* Create new page of same level */
869 40110 : nbuf = _bt_blnewpage(wstate, state->btps_level);
870 40110 : npage = (Page) nbuf;
871 :
872 : /* and assign it a page position */
873 40110 : nblkno = wstate->btws_pages_alloced++;
874 :
875 : /*
876 : * We copy the last item on the page into the new page, and then
877 : * rearrange the old page so that the 'last item' becomes its high key
878 : * rather than a true data item. There had better be at least two
879 : * items on the page already, else the page would be empty of useful
880 : * data.
881 : */
882 : Assert(last_off > P_FIRSTKEY);
883 40110 : ii = PageGetItemId(opage, last_off);
884 40110 : oitup = (IndexTuple) PageGetItem(opage, ii);
885 40110 : _bt_sortaddtup(npage, ItemIdGetLength(ii), oitup, P_FIRSTKEY,
886 40110 : !isleaf);
887 :
888 : /*
889 : * Move 'last' into the high key position on opage. _bt_blnewpage()
890 : * allocated empty space for a line pointer when opage was first
891 : * created, so this is a matter of rearranging already-allocated space
892 : * on page, and initializing high key line pointer. (Actually, leaf
893 : * pages must also swap oitup with a truncated version of oitup, which
894 : * is sometimes larger than oitup, though never by more than the space
895 : * needed to append a heap TID.)
896 : */
897 40110 : hii = PageGetItemId(opage, P_HIKEY);
898 40110 : *hii = *ii;
899 40110 : ItemIdSetUnused(ii); /* redundant */
900 40110 : ((PageHeader) opage)->pd_lower -= sizeof(ItemIdData);
901 :
902 40110 : if (isleaf)
903 : {
904 : IndexTuple lastleft;
905 : IndexTuple truncated;
906 :
907 : /*
908 : * Truncate away any unneeded attributes from high key on leaf
909 : * level. This is only done at the leaf level because downlinks
910 : * in internal pages are either negative infinity items, or get
911 : * their contents from copying from one level down. See also:
912 : * _bt_split().
913 : *
914 : * We don't try to bias our choice of split point to make it more
915 : * likely that _bt_truncate() can truncate away more attributes,
916 : * whereas the split point used within _bt_split() is chosen much
917 : * more delicately. Even still, the lastleft and firstright
918 : * tuples passed to _bt_truncate() here are at least not fully
919 : * equal to each other when deduplication is used, unless there is
920 : * a large group of duplicates (also, unique index builds usually
921 : * have few or no spool2 duplicates). When the split point is
922 : * between two unequal tuples, _bt_truncate() will avoid including
923 : * a heap TID in the new high key, which is the most important
924 : * benefit of suffix truncation.
925 : *
926 : * Overwrite the old item with new truncated high key directly.
927 : * oitup is already located at the physical beginning of tuple
928 : * space, so this should directly reuse the existing tuple space.
929 : */
930 39986 : ii = PageGetItemId(opage, OffsetNumberPrev(last_off));
931 39986 : lastleft = (IndexTuple) PageGetItem(opage, ii);
932 :
933 : Assert(IndexTupleSize(oitup) > last_truncextra);
934 39986 : truncated = _bt_truncate(wstate->index, lastleft, oitup,
935 : wstate->inskey);
936 39986 : if (!PageIndexTupleOverwrite(opage, P_HIKEY, truncated, IndexTupleSize(truncated)))
937 0 : elog(ERROR, "failed to add high key to the index page");
938 39986 : pfree(truncated);
939 :
940 : /* oitup should continue to point to the page's high key */
941 39986 : hii = PageGetItemId(opage, P_HIKEY);
942 39986 : oitup = (IndexTuple) PageGetItem(opage, hii);
943 : }
944 :
945 : /*
946 : * Link the old page into its parent, using its low key. If we don't
947 : * have a parent, we have to create one; this adds a new btree level.
948 : */
949 40110 : if (state->btps_next == NULL)
950 2692 : state->btps_next = _bt_pagestate(wstate, state->btps_level + 1);
951 :
952 : Assert((BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) <=
953 : IndexRelationGetNumberOfKeyAttributes(wstate->index) &&
954 : BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) > 0) ||
955 : P_LEFTMOST(BTPageGetOpaque(opage)));
956 : Assert(BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) == 0 ||
957 : !P_LEFTMOST(BTPageGetOpaque(opage)));
958 40110 : BTreeTupleSetDownLink(state->btps_lowkey, oblkno);
959 40110 : _bt_buildadd(wstate, state->btps_next, state->btps_lowkey, 0);
960 40110 : pfree(state->btps_lowkey);
961 :
962 : /*
963 : * Save a copy of the high key from the old page. It is also the low
964 : * key for the new page.
965 : */
966 40110 : state->btps_lowkey = CopyIndexTuple(oitup);
967 :
968 : /*
969 : * Set the sibling links for both pages.
970 : */
971 : {
972 40110 : BTPageOpaque oopaque = BTPageGetOpaque(opage);
973 40110 : BTPageOpaque nopaque = BTPageGetOpaque(npage);
974 :
975 40110 : oopaque->btpo_next = nblkno;
976 40110 : nopaque->btpo_prev = oblkno;
977 40110 : nopaque->btpo_next = P_NONE; /* redundant */
978 : }
979 :
980 : /*
981 : * Write out the old page. _bt_blwritepage takes ownership of the
982 : * 'opage' buffer.
983 : */
984 40110 : _bt_blwritepage(wstate, obuf, oblkno);
985 :
986 : /*
987 : * Reset last_off to point to new page
988 : */
989 40110 : last_off = P_FIRSTKEY;
990 : }
991 :
992 : /*
993 : * By here, either original page is still the current page, or a new page
994 : * was created that became the current page. Either way, the current page
995 : * definitely has space for new item.
996 : *
997 : * If the new item is the first for its page, it must also be the first
998 : * item on its entire level. On later same-level pages, a low key for a
999 : * page will be copied from the prior page in the code above. Generate a
1000 : * minus infinity low key here instead.
1001 : */
1002 11458780 : if (last_off == P_HIKEY)
1003 : {
1004 : Assert(state->btps_lowkey == NULL);
1005 12890 : state->btps_lowkey = palloc0(sizeof(IndexTupleData));
1006 12890 : state->btps_lowkey->t_info = sizeof(IndexTupleData);
1007 12890 : BTreeTupleSetNAtts(state->btps_lowkey, 0, false);
1008 : }
1009 :
1010 : /*
1011 : * Add the new item into the current page.
1012 : */
1013 11458780 : last_off = OffsetNumberNext(last_off);
1014 11458780 : _bt_sortaddtup(npage, itupsz, itup, last_off,
1015 11458780 : !isleaf && last_off == P_FIRSTKEY);
1016 :
1017 11458780 : state->btps_buf = nbuf;
1018 11458780 : state->btps_blkno = nblkno;
1019 11458780 : state->btps_lastoff = last_off;
1020 11458780 : }
1021 :
1022 : /*
1023 : * Finalize pending posting list tuple, and add it to the index. Final tuple
1024 : * is based on saved base tuple, and saved list of heap TIDs.
1025 : *
1026 : * This is almost like _bt_dedup_finish_pending(), but it adds a new tuple
1027 : * using _bt_buildadd().
1028 : */
1029 : static void
1030 4741902 : _bt_sort_dedup_finish_pending(BTWriteState *wstate, BTPageState *state,
1031 : BTDedupState dstate)
1032 : {
1033 : Assert(dstate->nitems > 0);
1034 :
1035 4741902 : if (dstate->nitems == 1)
1036 4701432 : _bt_buildadd(wstate, state, dstate->base, 0);
1037 : else
1038 : {
1039 : IndexTuple postingtuple;
1040 : Size truncextra;
1041 :
1042 : /* form a tuple with a posting list */
1043 40470 : postingtuple = _bt_form_posting(dstate->base,
1044 40470 : dstate->htids,
1045 : dstate->nhtids);
1046 : /* Calculate posting list overhead */
1047 40470 : truncextra = IndexTupleSize(postingtuple) -
1048 40470 : BTreeTupleGetPostingOffset(postingtuple);
1049 :
1050 40470 : _bt_buildadd(wstate, state, postingtuple, truncextra);
1051 40470 : pfree(postingtuple);
1052 : }
1053 :
1054 4741902 : dstate->nmaxitems = 0;
1055 4741902 : dstate->nhtids = 0;
1056 4741902 : dstate->nitems = 0;
1057 4741902 : dstate->phystupsize = 0;
1058 4741902 : }
1059 :
1060 : /*
1061 : * Finish writing out the completed btree.
1062 : */
1063 : static void
1064 50636 : _bt_uppershutdown(BTWriteState *wstate, BTPageState *state)
1065 : {
1066 : BTPageState *s;
1067 50636 : BlockNumber rootblkno = P_NONE;
1068 50636 : uint32 rootlevel = 0;
1069 : BulkWriteBuffer metabuf;
1070 :
1071 : /*
1072 : * Each iteration of this loop completes one more level of the tree.
1073 : */
1074 63526 : for (s = state; s != NULL; s = s->btps_next)
1075 : {
1076 : BlockNumber blkno;
1077 : BTPageOpaque opaque;
1078 :
1079 12890 : blkno = s->btps_blkno;
1080 12890 : opaque = BTPageGetOpaque((Page) s->btps_buf);
1081 :
1082 : /*
1083 : * We have to link the last page on this level to somewhere.
1084 : *
1085 : * If we're at the top, it's the root, so attach it to the metapage.
1086 : * Otherwise, add an entry for it to its parent using its low key.
1087 : * This may cause the last page of the parent level to split, but
1088 : * that's not a problem -- we haven't gotten to it yet.
1089 : */
1090 12890 : if (s->btps_next == NULL)
1091 : {
1092 10198 : opaque->btpo_flags |= BTP_ROOT;
1093 10198 : rootblkno = blkno;
1094 10198 : rootlevel = s->btps_level;
1095 : }
1096 : else
1097 : {
1098 : Assert((BTreeTupleGetNAtts(s->btps_lowkey, wstate->index) <=
1099 : IndexRelationGetNumberOfKeyAttributes(wstate->index) &&
1100 : BTreeTupleGetNAtts(s->btps_lowkey, wstate->index) > 0) ||
1101 : P_LEFTMOST(opaque));
1102 : Assert(BTreeTupleGetNAtts(s->btps_lowkey, wstate->index) == 0 ||
1103 : !P_LEFTMOST(opaque));
1104 2692 : BTreeTupleSetDownLink(s->btps_lowkey, blkno);
1105 2692 : _bt_buildadd(wstate, s->btps_next, s->btps_lowkey, 0);
1106 2692 : pfree(s->btps_lowkey);
1107 2692 : s->btps_lowkey = NULL;
1108 : }
1109 :
1110 : /*
1111 : * This is the rightmost page, so the ItemId array needs to be slid
1112 : * back one slot. Then we can dump out the page.
1113 : */
1114 12890 : _bt_slideleft((Page) s->btps_buf);
1115 12890 : _bt_blwritepage(wstate, s->btps_buf, s->btps_blkno);
1116 12890 : s->btps_buf = NULL; /* writepage took ownership of the buffer */
1117 : }
1118 :
1119 : /*
1120 : * As the last step in the process, construct the metapage and make it
1121 : * point to the new root (unless we had no data at all, in which case it's
1122 : * set to point to "P_NONE"). This changes the index to the "valid" state
1123 : * by filling in a valid magic number in the metapage.
1124 : */
1125 50636 : metabuf = smgr_bulk_get_buf(wstate->bulkstate);
1126 50636 : _bt_initmetapage((Page) metabuf, rootblkno, rootlevel,
1127 50636 : wstate->inskey->allequalimage);
1128 50636 : _bt_blwritepage(wstate, metabuf, BTREE_METAPAGE);
1129 50636 : }
1130 :
1131 : /*
1132 : * Read tuples in correct sort order from tuplesort, and load them into
1133 : * btree leaves.
1134 : */
1135 : static void
1136 50636 : _bt_load(BTWriteState *wstate, BTSpool *btspool, BTSpool *btspool2)
1137 : {
1138 50636 : BTPageState *state = NULL;
1139 50636 : bool merge = (btspool2 != NULL);
1140 : IndexTuple itup,
1141 50636 : itup2 = NULL;
1142 : bool load1;
1143 50636 : TupleDesc tupdes = RelationGetDescr(wstate->index);
1144 : int i,
1145 50636 : keysz = IndexRelationGetNumberOfKeyAttributes(wstate->index);
1146 : SortSupport sortKeys;
1147 50636 : int64 tuples_done = 0;
1148 : bool deduplicate;
1149 :
1150 50636 : wstate->bulkstate = smgr_bulk_start_rel(wstate->index, MAIN_FORKNUM);
1151 :
1152 59670 : deduplicate = wstate->inskey->allequalimage && !btspool->isunique &&
1153 9034 : BTGetDeduplicateItems(wstate->index);
1154 :
1155 50636 : if (merge)
1156 : {
1157 : /*
1158 : * Another BTSpool for dead tuples exists. Now we have to merge
1159 : * btspool and btspool2.
1160 : */
1161 :
1162 : /* the preparation of merge */
1163 26 : itup = tuplesort_getindextuple(btspool->sortstate, true);
1164 26 : itup2 = tuplesort_getindextuple(btspool2->sortstate, true);
1165 :
1166 : /* Prepare SortSupport data for each column */
1167 26 : sortKeys = (SortSupport) palloc0(keysz * sizeof(SortSupportData));
1168 :
1169 54 : for (i = 0; i < keysz; i++)
1170 : {
1171 28 : SortSupport sortKey = sortKeys + i;
1172 28 : ScanKey scanKey = wstate->inskey->scankeys + i;
1173 : bool reverse;
1174 :
1175 28 : sortKey->ssup_cxt = CurrentMemoryContext;
1176 28 : sortKey->ssup_collation = scanKey->sk_collation;
1177 28 : sortKey->ssup_nulls_first =
1178 28 : (scanKey->sk_flags & SK_BT_NULLS_FIRST) != 0;
1179 28 : sortKey->ssup_attno = scanKey->sk_attno;
1180 : /* Abbreviation is not supported here */
1181 28 : sortKey->abbreviate = false;
1182 :
1183 : Assert(sortKey->ssup_attno != 0);
1184 :
1185 28 : reverse = (scanKey->sk_flags & SK_BT_DESC) != 0;
1186 :
1187 28 : PrepareSortSupportFromIndexRel(wstate->index, reverse, sortKey);
1188 : }
1189 :
1190 : for (;;)
1191 : {
1192 3348 : load1 = true; /* load BTSpool next ? */
1193 3348 : if (itup2 == NULL)
1194 : {
1195 158 : if (itup == NULL)
1196 26 : break;
1197 : }
1198 3190 : else if (itup != NULL)
1199 : {
1200 2992 : int32 compare = 0;
1201 :
1202 3240 : for (i = 1; i <= keysz; i++)
1203 : {
1204 : SortSupport entry;
1205 : Datum attrDatum1,
1206 : attrDatum2;
1207 : bool isNull1,
1208 : isNull2;
1209 :
1210 3044 : entry = sortKeys + i - 1;
1211 3044 : attrDatum1 = index_getattr(itup, i, tupdes, &isNull1);
1212 3044 : attrDatum2 = index_getattr(itup2, i, tupdes, &isNull2);
1213 :
1214 3044 : compare = ApplySortComparator(attrDatum1, isNull1,
1215 : attrDatum2, isNull2,
1216 : entry);
1217 3044 : if (compare > 0)
1218 : {
1219 278 : load1 = false;
1220 2796 : break;
1221 : }
1222 2766 : else if (compare < 0)
1223 2518 : break;
1224 : }
1225 :
1226 : /*
1227 : * If key values are equal, we sort on ItemPointer. This is
1228 : * required for btree indexes, since heap TID is treated as an
1229 : * implicit last key attribute in order to ensure that all
1230 : * keys in the index are physically unique.
1231 : */
1232 2992 : if (compare == 0)
1233 : {
1234 196 : compare = ItemPointerCompare(&itup->t_tid, &itup2->t_tid);
1235 : Assert(compare != 0);
1236 196 : if (compare > 0)
1237 30 : load1 = false;
1238 : }
1239 : }
1240 : else
1241 198 : load1 = false;
1242 :
1243 : /* When we see first tuple, create first index page */
1244 3322 : if (state == NULL)
1245 26 : state = _bt_pagestate(wstate, 0);
1246 :
1247 3322 : if (load1)
1248 : {
1249 2816 : _bt_buildadd(wstate, state, itup, 0);
1250 2816 : itup = tuplesort_getindextuple(btspool->sortstate, true);
1251 : }
1252 : else
1253 : {
1254 506 : _bt_buildadd(wstate, state, itup2, 0);
1255 506 : itup2 = tuplesort_getindextuple(btspool2->sortstate, true);
1256 : }
1257 :
1258 : /* Report progress */
1259 3322 : pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE,
1260 : ++tuples_done);
1261 : }
1262 26 : pfree(sortKeys);
1263 : }
1264 50610 : else if (deduplicate)
1265 : {
1266 : /* merge is unnecessary, deduplicate into posting lists */
1267 : BTDedupState dstate;
1268 :
1269 9034 : dstate = (BTDedupState) palloc(sizeof(BTDedupStateData));
1270 9034 : dstate->deduplicate = true; /* unused */
1271 9034 : dstate->nmaxitems = 0; /* unused */
1272 9034 : dstate->maxpostingsize = 0; /* set later */
1273 : /* Metadata about base tuple of current pending posting list */
1274 9034 : dstate->base = NULL;
1275 9034 : dstate->baseoff = InvalidOffsetNumber; /* unused */
1276 9034 : dstate->basetupsize = 0;
1277 : /* Metadata about current pending posting list TIDs */
1278 9034 : dstate->htids = NULL;
1279 9034 : dstate->nhtids = 0;
1280 9034 : dstate->nitems = 0;
1281 9034 : dstate->phystupsize = 0; /* unused */
1282 9034 : dstate->nintervals = 0; /* unused */
1283 :
1284 6120680 : while ((itup = tuplesort_getindextuple(btspool->sortstate,
1285 6120680 : true)) != NULL)
1286 : {
1287 : /* When we see first tuple, create first index page */
1288 6111646 : if (state == NULL)
1289 : {
1290 2460 : state = _bt_pagestate(wstate, 0);
1291 :
1292 : /*
1293 : * Limit size of posting list tuples to 1/10 space we want to
1294 : * leave behind on the page, plus space for final item's line
1295 : * pointer. This is equal to the space that we'd like to
1296 : * leave behind on each leaf page when fillfactor is 90,
1297 : * allowing us to get close to fillfactor% space utilization
1298 : * when there happen to be a great many duplicates. (This
1299 : * makes higher leaf fillfactor settings ineffective when
1300 : * building indexes that have many duplicates, but packing
1301 : * leaf pages full with few very large tuples doesn't seem
1302 : * like a useful goal.)
1303 : */
1304 2460 : dstate->maxpostingsize = MAXALIGN_DOWN((BLCKSZ * 10 / 100)) -
1305 : sizeof(ItemIdData);
1306 : Assert(dstate->maxpostingsize <= BTMaxItemSize &&
1307 : dstate->maxpostingsize <= INDEX_SIZE_MASK);
1308 2460 : dstate->htids = palloc(dstate->maxpostingsize);
1309 :
1310 : /* start new pending posting list with itup copy */
1311 2460 : _bt_dedup_start_pending(dstate, CopyIndexTuple(itup),
1312 : InvalidOffsetNumber);
1313 : }
1314 6109186 : else if (_bt_keep_natts_fast(wstate->index, dstate->base,
1315 1377682 : itup) > keysz &&
1316 1377682 : _bt_dedup_save_htid(dstate, itup))
1317 : {
1318 : /*
1319 : * Tuple is equal to base tuple of pending posting list. Heap
1320 : * TID from itup has been saved in state.
1321 : */
1322 : }
1323 : else
1324 : {
1325 : /*
1326 : * Tuple is not equal to pending posting list tuple, or
1327 : * _bt_dedup_save_htid() opted to not merge current item into
1328 : * pending posting list.
1329 : */
1330 4739442 : _bt_sort_dedup_finish_pending(wstate, state, dstate);
1331 4739442 : pfree(dstate->base);
1332 :
1333 : /* start new pending posting list with itup copy */
1334 4739442 : _bt_dedup_start_pending(dstate, CopyIndexTuple(itup),
1335 : InvalidOffsetNumber);
1336 : }
1337 :
1338 : /* Report progress */
1339 6111646 : pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE,
1340 : ++tuples_done);
1341 : }
1342 :
1343 9034 : if (state)
1344 : {
1345 : /*
1346 : * Handle the last item (there must be a last item when the
1347 : * tuplesort returned one or more tuples)
1348 : */
1349 2460 : _bt_sort_dedup_finish_pending(wstate, state, dstate);
1350 2460 : pfree(dstate->base);
1351 2460 : pfree(dstate->htids);
1352 : }
1353 :
1354 9034 : pfree(dstate);
1355 : }
1356 : else
1357 : {
1358 : /* merging and deduplication are both unnecessary */
1359 6712330 : while ((itup = tuplesort_getindextuple(btspool->sortstate,
1360 6712330 : true)) != NULL)
1361 : {
1362 : /* When we see first tuple, create first index page */
1363 6670754 : if (state == NULL)
1364 7712 : state = _bt_pagestate(wstate, 0);
1365 :
1366 6670754 : _bt_buildadd(wstate, state, itup, 0);
1367 :
1368 : /* Report progress */
1369 6670754 : pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE,
1370 : ++tuples_done);
1371 : }
1372 : }
1373 :
1374 : /* Close down final pages and write the metapage */
1375 50636 : _bt_uppershutdown(wstate, state);
1376 50636 : smgr_bulk_finish(wstate->bulkstate);
1377 50636 : }
1378 :
1379 : /*
1380 : * Create parallel context, and launch workers for leader.
1381 : *
1382 : * buildstate argument should be initialized (with the exception of the
1383 : * tuplesort state in spools, which may later be created based on shared
1384 : * state initially set up here).
1385 : *
1386 : * isconcurrent indicates if operation is CREATE INDEX CONCURRENTLY.
1387 : *
1388 : * request is the target number of parallel worker processes to launch.
1389 : *
1390 : * Sets buildstate's BTLeader, which caller must use to shut down parallel
1391 : * mode by passing it to _bt_end_parallel() at the very end of its index
1392 : * build. If not even a single worker process can be launched, this is
1393 : * never set, and caller should proceed with a serial index build.
1394 : */
1395 : static void
1396 148 : _bt_begin_parallel(BTBuildState *buildstate, bool isconcurrent, int request)
1397 : {
1398 : ParallelContext *pcxt;
1399 : int scantuplesortstates;
1400 : Snapshot snapshot;
1401 : Size estbtshared;
1402 : Size estsort;
1403 : BTShared *btshared;
1404 : Sharedsort *sharedsort;
1405 : Sharedsort *sharedsort2;
1406 148 : BTSpool *btspool = buildstate->spool;
1407 148 : BTLeader *btleader = (BTLeader *) palloc0(sizeof(BTLeader));
1408 : WalUsage *walusage;
1409 : BufferUsage *bufferusage;
1410 148 : bool leaderparticipates = true;
1411 : int querylen;
1412 :
1413 : #ifdef DISABLE_LEADER_PARTICIPATION
1414 : leaderparticipates = false;
1415 : #endif
1416 :
1417 : /*
1418 : * Enter parallel mode, and create context for parallel build of btree
1419 : * index
1420 : */
1421 148 : EnterParallelMode();
1422 : Assert(request > 0);
1423 148 : pcxt = CreateParallelContext("postgres", "_bt_parallel_build_main",
1424 : request);
1425 :
1426 148 : scantuplesortstates = leaderparticipates ? request + 1 : request;
1427 :
1428 : /*
1429 : * Prepare for scan of the base relation. In a normal index build, we use
1430 : * SnapshotAny because we must retrieve all tuples and do our own time
1431 : * qual checks (because we have to index RECENTLY_DEAD tuples). In a
1432 : * concurrent build, we take a regular MVCC snapshot and index whatever's
1433 : * live according to that.
1434 : */
1435 148 : if (!isconcurrent)
1436 148 : snapshot = SnapshotAny;
1437 : else
1438 0 : snapshot = RegisterSnapshot(GetTransactionSnapshot());
1439 :
1440 : /*
1441 : * Estimate size for our own PARALLEL_KEY_BTREE_SHARED workspace, and
1442 : * PARALLEL_KEY_TUPLESORT tuplesort workspace
1443 : */
1444 148 : estbtshared = _bt_parallel_estimate_shared(btspool->heap, snapshot);
1445 148 : shm_toc_estimate_chunk(&pcxt->estimator, estbtshared);
1446 148 : estsort = tuplesort_estimate_shared(scantuplesortstates);
1447 148 : shm_toc_estimate_chunk(&pcxt->estimator, estsort);
1448 :
1449 : /*
1450 : * Unique case requires a second spool, and so we may have to account for
1451 : * another shared workspace for that -- PARALLEL_KEY_TUPLESORT_SPOOL2
1452 : */
1453 148 : if (!btspool->isunique)
1454 84 : shm_toc_estimate_keys(&pcxt->estimator, 2);
1455 : else
1456 : {
1457 64 : shm_toc_estimate_chunk(&pcxt->estimator, estsort);
1458 64 : shm_toc_estimate_keys(&pcxt->estimator, 3);
1459 : }
1460 :
1461 : /*
1462 : * Estimate space for WalUsage and BufferUsage -- PARALLEL_KEY_WAL_USAGE
1463 : * and PARALLEL_KEY_BUFFER_USAGE.
1464 : *
1465 : * If there are no extensions loaded that care, we could skip this. We
1466 : * have no way of knowing whether anyone's looking at pgWalUsage or
1467 : * pgBufferUsage, so do it unconditionally.
1468 : */
1469 148 : shm_toc_estimate_chunk(&pcxt->estimator,
1470 : mul_size(sizeof(WalUsage), pcxt->nworkers));
1471 148 : shm_toc_estimate_keys(&pcxt->estimator, 1);
1472 148 : shm_toc_estimate_chunk(&pcxt->estimator,
1473 : mul_size(sizeof(BufferUsage), pcxt->nworkers));
1474 148 : shm_toc_estimate_keys(&pcxt->estimator, 1);
1475 :
1476 : /* Finally, estimate PARALLEL_KEY_QUERY_TEXT space */
1477 148 : if (debug_query_string)
1478 : {
1479 148 : querylen = strlen(debug_query_string);
1480 148 : shm_toc_estimate_chunk(&pcxt->estimator, querylen + 1);
1481 148 : shm_toc_estimate_keys(&pcxt->estimator, 1);
1482 : }
1483 : else
1484 0 : querylen = 0; /* keep compiler quiet */
1485 :
1486 : /* Everyone's had a chance to ask for space, so now create the DSM */
1487 148 : InitializeParallelDSM(pcxt);
1488 :
1489 : /* If no DSM segment was available, back out (do serial build) */
1490 148 : if (pcxt->seg == NULL)
1491 : {
1492 0 : if (IsMVCCSnapshot(snapshot))
1493 0 : UnregisterSnapshot(snapshot);
1494 0 : DestroyParallelContext(pcxt);
1495 0 : ExitParallelMode();
1496 0 : return;
1497 : }
1498 :
1499 : /* Store shared build state, for which we reserved space */
1500 148 : btshared = (BTShared *) shm_toc_allocate(pcxt->toc, estbtshared);
1501 : /* Initialize immutable state */
1502 148 : btshared->heaprelid = RelationGetRelid(btspool->heap);
1503 148 : btshared->indexrelid = RelationGetRelid(btspool->index);
1504 148 : btshared->isunique = btspool->isunique;
1505 148 : btshared->nulls_not_distinct = btspool->nulls_not_distinct;
1506 148 : btshared->isconcurrent = isconcurrent;
1507 148 : btshared->scantuplesortstates = scantuplesortstates;
1508 148 : btshared->queryid = pgstat_get_my_query_id();
1509 148 : ConditionVariableInit(&btshared->workersdonecv);
1510 148 : SpinLockInit(&btshared->mutex);
1511 : /* Initialize mutable state */
1512 148 : btshared->nparticipantsdone = 0;
1513 148 : btshared->reltuples = 0.0;
1514 148 : btshared->havedead = false;
1515 148 : btshared->indtuples = 0.0;
1516 148 : btshared->brokenhotchain = false;
1517 148 : table_parallelscan_initialize(btspool->heap,
1518 : ParallelTableScanFromBTShared(btshared),
1519 : snapshot);
1520 :
1521 : /*
1522 : * Store shared tuplesort-private state, for which we reserved space.
1523 : * Then, initialize opaque state using tuplesort routine.
1524 : */
1525 148 : sharedsort = (Sharedsort *) shm_toc_allocate(pcxt->toc, estsort);
1526 148 : tuplesort_initialize_shared(sharedsort, scantuplesortstates,
1527 : pcxt->seg);
1528 :
1529 148 : shm_toc_insert(pcxt->toc, PARALLEL_KEY_BTREE_SHARED, btshared);
1530 148 : shm_toc_insert(pcxt->toc, PARALLEL_KEY_TUPLESORT, sharedsort);
1531 :
1532 : /* Unique case requires a second spool, and associated shared state */
1533 148 : if (!btspool->isunique)
1534 84 : sharedsort2 = NULL;
1535 : else
1536 : {
1537 : /*
1538 : * Store additional shared tuplesort-private state, for which we
1539 : * reserved space. Then, initialize opaque state using tuplesort
1540 : * routine.
1541 : */
1542 64 : sharedsort2 = (Sharedsort *) shm_toc_allocate(pcxt->toc, estsort);
1543 64 : tuplesort_initialize_shared(sharedsort2, scantuplesortstates,
1544 : pcxt->seg);
1545 :
1546 64 : shm_toc_insert(pcxt->toc, PARALLEL_KEY_TUPLESORT_SPOOL2, sharedsort2);
1547 : }
1548 :
1549 : /* Store query string for workers */
1550 148 : if (debug_query_string)
1551 : {
1552 : char *sharedquery;
1553 :
1554 148 : sharedquery = (char *) shm_toc_allocate(pcxt->toc, querylen + 1);
1555 148 : memcpy(sharedquery, debug_query_string, querylen + 1);
1556 148 : shm_toc_insert(pcxt->toc, PARALLEL_KEY_QUERY_TEXT, sharedquery);
1557 : }
1558 :
1559 : /*
1560 : * Allocate space for each worker's WalUsage and BufferUsage; no need to
1561 : * initialize.
1562 : */
1563 148 : walusage = shm_toc_allocate(pcxt->toc,
1564 148 : mul_size(sizeof(WalUsage), pcxt->nworkers));
1565 148 : shm_toc_insert(pcxt->toc, PARALLEL_KEY_WAL_USAGE, walusage);
1566 148 : bufferusage = shm_toc_allocate(pcxt->toc,
1567 148 : mul_size(sizeof(BufferUsage), pcxt->nworkers));
1568 148 : shm_toc_insert(pcxt->toc, PARALLEL_KEY_BUFFER_USAGE, bufferusage);
1569 :
1570 : /* Launch workers, saving status for leader/caller */
1571 148 : LaunchParallelWorkers(pcxt);
1572 148 : btleader->pcxt = pcxt;
1573 148 : btleader->nparticipanttuplesorts = pcxt->nworkers_launched;
1574 148 : if (leaderparticipates)
1575 148 : btleader->nparticipanttuplesorts++;
1576 148 : btleader->btshared = btshared;
1577 148 : btleader->sharedsort = sharedsort;
1578 148 : btleader->sharedsort2 = sharedsort2;
1579 148 : btleader->snapshot = snapshot;
1580 148 : btleader->walusage = walusage;
1581 148 : btleader->bufferusage = bufferusage;
1582 :
1583 : /* If no workers were successfully launched, back out (do serial build) */
1584 148 : if (pcxt->nworkers_launched == 0)
1585 : {
1586 0 : _bt_end_parallel(btleader);
1587 0 : return;
1588 : }
1589 :
1590 : /* Save leader state now that it's clear build will be parallel */
1591 148 : buildstate->btleader = btleader;
1592 :
1593 : /* Join heap scan ourselves */
1594 148 : if (leaderparticipates)
1595 148 : _bt_leader_participate_as_worker(buildstate);
1596 :
1597 : /*
1598 : * Caller needs to wait for all launched workers when we return. Make
1599 : * sure that the failure-to-start case will not hang forever.
1600 : */
1601 148 : WaitForParallelWorkersToAttach(pcxt);
1602 : }
1603 :
1604 : /*
1605 : * Shut down workers, destroy parallel context, and end parallel mode.
1606 : */
1607 : static void
1608 148 : _bt_end_parallel(BTLeader *btleader)
1609 : {
1610 : int i;
1611 :
1612 : /* Shutdown worker processes */
1613 148 : WaitForParallelWorkersToFinish(btleader->pcxt);
1614 :
1615 : /*
1616 : * Next, accumulate WAL usage. (This must wait for the workers to finish,
1617 : * or we might get incomplete data.)
1618 : */
1619 296 : for (i = 0; i < btleader->pcxt->nworkers_launched; i++)
1620 148 : InstrAccumParallelQuery(&btleader->bufferusage[i], &btleader->walusage[i]);
1621 :
1622 : /* Free last reference to MVCC snapshot, if one was used */
1623 148 : if (IsMVCCSnapshot(btleader->snapshot))
1624 0 : UnregisterSnapshot(btleader->snapshot);
1625 148 : DestroyParallelContext(btleader->pcxt);
1626 148 : ExitParallelMode();
1627 148 : }
1628 :
1629 : /*
1630 : * Returns size of shared memory required to store state for a parallel
1631 : * btree index build based on the snapshot its parallel scan will use.
1632 : */
1633 : static Size
1634 148 : _bt_parallel_estimate_shared(Relation heap, Snapshot snapshot)
1635 : {
1636 : /* c.f. shm_toc_allocate as to why BUFFERALIGN is used */
1637 148 : return add_size(BUFFERALIGN(sizeof(BTShared)),
1638 : table_parallelscan_estimate(heap, snapshot));
1639 : }
1640 :
1641 : /*
1642 : * Within leader, wait for end of heap scan.
1643 : *
1644 : * When called, parallel heap scan started by _bt_begin_parallel() will
1645 : * already be underway within worker processes (when leader participates
1646 : * as a worker, we should end up here just as workers are finishing).
1647 : *
1648 : * Fills in fields needed for ambuild statistics, and lets caller set
1649 : * field indicating that some worker encountered a broken HOT chain.
1650 : *
1651 : * Returns the total number of heap tuples scanned.
1652 : */
1653 : static double
1654 148 : _bt_parallel_heapscan(BTBuildState *buildstate, bool *brokenhotchain)
1655 : {
1656 148 : BTShared *btshared = buildstate->btleader->btshared;
1657 : int nparticipanttuplesorts;
1658 : double reltuples;
1659 :
1660 148 : nparticipanttuplesorts = buildstate->btleader->nparticipanttuplesorts;
1661 : for (;;)
1662 : {
1663 336 : SpinLockAcquire(&btshared->mutex);
1664 336 : if (btshared->nparticipantsdone == nparticipanttuplesorts)
1665 : {
1666 148 : buildstate->havedead = btshared->havedead;
1667 148 : buildstate->indtuples = btshared->indtuples;
1668 148 : *brokenhotchain = btshared->brokenhotchain;
1669 148 : reltuples = btshared->reltuples;
1670 148 : SpinLockRelease(&btshared->mutex);
1671 148 : break;
1672 : }
1673 188 : SpinLockRelease(&btshared->mutex);
1674 :
1675 188 : ConditionVariableSleep(&btshared->workersdonecv,
1676 : WAIT_EVENT_PARALLEL_CREATE_INDEX_SCAN);
1677 : }
1678 :
1679 148 : ConditionVariableCancelSleep();
1680 :
1681 148 : return reltuples;
1682 : }
1683 :
1684 : /*
1685 : * Within leader, participate as a parallel worker.
1686 : */
1687 : static void
1688 148 : _bt_leader_participate_as_worker(BTBuildState *buildstate)
1689 : {
1690 148 : BTLeader *btleader = buildstate->btleader;
1691 : BTSpool *leaderworker;
1692 : BTSpool *leaderworker2;
1693 : int sortmem;
1694 :
1695 : /* Allocate memory and initialize private spool */
1696 148 : leaderworker = (BTSpool *) palloc0(sizeof(BTSpool));
1697 148 : leaderworker->heap = buildstate->spool->heap;
1698 148 : leaderworker->index = buildstate->spool->index;
1699 148 : leaderworker->isunique = buildstate->spool->isunique;
1700 148 : leaderworker->nulls_not_distinct = buildstate->spool->nulls_not_distinct;
1701 :
1702 : /* Initialize second spool, if required */
1703 148 : if (!btleader->btshared->isunique)
1704 84 : leaderworker2 = NULL;
1705 : else
1706 : {
1707 : /* Allocate memory for worker's own private secondary spool */
1708 64 : leaderworker2 = (BTSpool *) palloc0(sizeof(BTSpool));
1709 :
1710 : /* Initialize worker's own secondary spool */
1711 64 : leaderworker2->heap = leaderworker->heap;
1712 64 : leaderworker2->index = leaderworker->index;
1713 64 : leaderworker2->isunique = false;
1714 : }
1715 :
1716 : /*
1717 : * Might as well use reliable figure when doling out maintenance_work_mem
1718 : * (when requested number of workers were not launched, this will be
1719 : * somewhat higher than it is for other workers).
1720 : */
1721 148 : sortmem = maintenance_work_mem / btleader->nparticipanttuplesorts;
1722 :
1723 : /* Perform work common to all participants */
1724 148 : _bt_parallel_scan_and_sort(leaderworker, leaderworker2, btleader->btshared,
1725 : btleader->sharedsort, btleader->sharedsort2,
1726 : sortmem, true);
1727 :
1728 : #ifdef BTREE_BUILD_STATS
1729 : if (log_btree_build_stats)
1730 : {
1731 : ShowUsage("BTREE BUILD (Leader Partial Spool) STATISTICS");
1732 : ResetUsage();
1733 : }
1734 : #endif /* BTREE_BUILD_STATS */
1735 148 : }
1736 :
1737 : /*
1738 : * Perform work within a launched parallel process.
1739 : */
1740 : void
1741 148 : _bt_parallel_build_main(dsm_segment *seg, shm_toc *toc)
1742 : {
1743 : char *sharedquery;
1744 : BTSpool *btspool;
1745 : BTSpool *btspool2;
1746 : BTShared *btshared;
1747 : Sharedsort *sharedsort;
1748 : Sharedsort *sharedsort2;
1749 : Relation heapRel;
1750 : Relation indexRel;
1751 : LOCKMODE heapLockmode;
1752 : LOCKMODE indexLockmode;
1753 : WalUsage *walusage;
1754 : BufferUsage *bufferusage;
1755 : int sortmem;
1756 :
1757 : #ifdef BTREE_BUILD_STATS
1758 : if (log_btree_build_stats)
1759 : ResetUsage();
1760 : #endif /* BTREE_BUILD_STATS */
1761 :
1762 : /*
1763 : * The only possible status flag that can be set to the parallel worker is
1764 : * PROC_IN_SAFE_IC.
1765 : */
1766 : Assert((MyProc->statusFlags == 0) ||
1767 : (MyProc->statusFlags == PROC_IN_SAFE_IC));
1768 :
1769 : /* Set debug_query_string for individual workers first */
1770 148 : sharedquery = shm_toc_lookup(toc, PARALLEL_KEY_QUERY_TEXT, true);
1771 148 : debug_query_string = sharedquery;
1772 :
1773 : /* Report the query string from leader */
1774 148 : pgstat_report_activity(STATE_RUNNING, debug_query_string);
1775 :
1776 : /* Look up nbtree shared state */
1777 148 : btshared = shm_toc_lookup(toc, PARALLEL_KEY_BTREE_SHARED, false);
1778 :
1779 : /* Open relations using lock modes known to be obtained by index.c */
1780 148 : if (!btshared->isconcurrent)
1781 : {
1782 148 : heapLockmode = ShareLock;
1783 148 : indexLockmode = AccessExclusiveLock;
1784 : }
1785 : else
1786 : {
1787 0 : heapLockmode = ShareUpdateExclusiveLock;
1788 0 : indexLockmode = RowExclusiveLock;
1789 : }
1790 :
1791 : /* Track query ID */
1792 148 : pgstat_report_query_id(btshared->queryid, false);
1793 :
1794 : /* Open relations within worker */
1795 148 : heapRel = table_open(btshared->heaprelid, heapLockmode);
1796 148 : indexRel = index_open(btshared->indexrelid, indexLockmode);
1797 :
1798 : /* Initialize worker's own spool */
1799 148 : btspool = (BTSpool *) palloc0(sizeof(BTSpool));
1800 148 : btspool->heap = heapRel;
1801 148 : btspool->index = indexRel;
1802 148 : btspool->isunique = btshared->isunique;
1803 148 : btspool->nulls_not_distinct = btshared->nulls_not_distinct;
1804 :
1805 : /* Look up shared state private to tuplesort.c */
1806 148 : sharedsort = shm_toc_lookup(toc, PARALLEL_KEY_TUPLESORT, false);
1807 148 : tuplesort_attach_shared(sharedsort, seg);
1808 148 : if (!btshared->isunique)
1809 : {
1810 84 : btspool2 = NULL;
1811 84 : sharedsort2 = NULL;
1812 : }
1813 : else
1814 : {
1815 : /* Allocate memory for worker's own private secondary spool */
1816 64 : btspool2 = (BTSpool *) palloc0(sizeof(BTSpool));
1817 :
1818 : /* Initialize worker's own secondary spool */
1819 64 : btspool2->heap = btspool->heap;
1820 64 : btspool2->index = btspool->index;
1821 64 : btspool2->isunique = false;
1822 : /* Look up shared state private to tuplesort.c */
1823 64 : sharedsort2 = shm_toc_lookup(toc, PARALLEL_KEY_TUPLESORT_SPOOL2, false);
1824 64 : tuplesort_attach_shared(sharedsort2, seg);
1825 : }
1826 :
1827 : /* Prepare to track buffer usage during parallel execution */
1828 148 : InstrStartParallelQuery();
1829 :
1830 : /* Perform sorting of spool, and possibly a spool2 */
1831 148 : sortmem = maintenance_work_mem / btshared->scantuplesortstates;
1832 148 : _bt_parallel_scan_and_sort(btspool, btspool2, btshared, sharedsort,
1833 : sharedsort2, sortmem, false);
1834 :
1835 : /* Report WAL/buffer usage during parallel execution */
1836 148 : bufferusage = shm_toc_lookup(toc, PARALLEL_KEY_BUFFER_USAGE, false);
1837 148 : walusage = shm_toc_lookup(toc, PARALLEL_KEY_WAL_USAGE, false);
1838 148 : InstrEndParallelQuery(&bufferusage[ParallelWorkerNumber],
1839 148 : &walusage[ParallelWorkerNumber]);
1840 :
1841 : #ifdef BTREE_BUILD_STATS
1842 : if (log_btree_build_stats)
1843 : {
1844 : ShowUsage("BTREE BUILD (Worker Partial Spool) STATISTICS");
1845 : ResetUsage();
1846 : }
1847 : #endif /* BTREE_BUILD_STATS */
1848 :
1849 148 : index_close(indexRel, indexLockmode);
1850 148 : table_close(heapRel, heapLockmode);
1851 148 : }
1852 :
1853 : /*
1854 : * Perform a worker's portion of a parallel sort.
1855 : *
1856 : * This generates a tuplesort for passed btspool, and a second tuplesort
1857 : * state if a second btspool is need (i.e. for unique index builds). All
1858 : * other spool fields should already be set when this is called.
1859 : *
1860 : * sortmem is the amount of working memory to use within each worker,
1861 : * expressed in KBs.
1862 : *
1863 : * When this returns, workers are done, and need only release resources.
1864 : */
1865 : static void
1866 296 : _bt_parallel_scan_and_sort(BTSpool *btspool, BTSpool *btspool2,
1867 : BTShared *btshared, Sharedsort *sharedsort,
1868 : Sharedsort *sharedsort2, int sortmem, bool progress)
1869 : {
1870 : SortCoordinate coordinate;
1871 : BTBuildState buildstate;
1872 : TableScanDesc scan;
1873 : double reltuples;
1874 : IndexInfo *indexInfo;
1875 :
1876 : /* Initialize local tuplesort coordination state */
1877 296 : coordinate = palloc0(sizeof(SortCoordinateData));
1878 296 : coordinate->isWorker = true;
1879 296 : coordinate->nParticipants = -1;
1880 296 : coordinate->sharedsort = sharedsort;
1881 :
1882 : /* Begin "partial" tuplesort */
1883 592 : btspool->sortstate = tuplesort_begin_index_btree(btspool->heap,
1884 : btspool->index,
1885 296 : btspool->isunique,
1886 296 : btspool->nulls_not_distinct,
1887 : sortmem, coordinate,
1888 : TUPLESORT_NONE);
1889 :
1890 : /*
1891 : * Just as with serial case, there may be a second spool. If so, a
1892 : * second, dedicated spool2 partial tuplesort is required.
1893 : */
1894 296 : if (btspool2)
1895 : {
1896 : SortCoordinate coordinate2;
1897 :
1898 : /*
1899 : * We expect that the second one (for dead tuples) won't get very
1900 : * full, so we give it only work_mem (unless sortmem is less for
1901 : * worker). Worker processes are generally permitted to allocate
1902 : * work_mem independently.
1903 : */
1904 128 : coordinate2 = palloc0(sizeof(SortCoordinateData));
1905 128 : coordinate2->isWorker = true;
1906 128 : coordinate2->nParticipants = -1;
1907 128 : coordinate2->sharedsort = sharedsort2;
1908 128 : btspool2->sortstate =
1909 128 : tuplesort_begin_index_btree(btspool->heap, btspool->index, false, false,
1910 : Min(sortmem, work_mem), coordinate2,
1911 : false);
1912 : }
1913 :
1914 : /* Fill in buildstate for _bt_build_callback() */
1915 296 : buildstate.isunique = btshared->isunique;
1916 296 : buildstate.nulls_not_distinct = btshared->nulls_not_distinct;
1917 296 : buildstate.havedead = false;
1918 296 : buildstate.heap = btspool->heap;
1919 296 : buildstate.spool = btspool;
1920 296 : buildstate.spool2 = btspool2;
1921 296 : buildstate.indtuples = 0;
1922 296 : buildstate.btleader = NULL;
1923 :
1924 : /* Join parallel scan */
1925 296 : indexInfo = BuildIndexInfo(btspool->index);
1926 296 : indexInfo->ii_Concurrent = btshared->isconcurrent;
1927 296 : scan = table_beginscan_parallel(btspool->heap,
1928 : ParallelTableScanFromBTShared(btshared));
1929 296 : reltuples = table_index_build_scan(btspool->heap, btspool->index, indexInfo,
1930 : true, progress, _bt_build_callback,
1931 : &buildstate, scan);
1932 :
1933 : /* Execute this worker's part of the sort */
1934 296 : if (progress)
1935 148 : pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,
1936 : PROGRESS_BTREE_PHASE_PERFORMSORT_1);
1937 296 : tuplesort_performsort(btspool->sortstate);
1938 296 : if (btspool2)
1939 : {
1940 128 : if (progress)
1941 64 : pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,
1942 : PROGRESS_BTREE_PHASE_PERFORMSORT_2);
1943 128 : tuplesort_performsort(btspool2->sortstate);
1944 : }
1945 :
1946 : /*
1947 : * Done. Record ambuild statistics, and whether we encountered a broken
1948 : * HOT chain.
1949 : */
1950 296 : SpinLockAcquire(&btshared->mutex);
1951 296 : btshared->nparticipantsdone++;
1952 296 : btshared->reltuples += reltuples;
1953 296 : if (buildstate.havedead)
1954 0 : btshared->havedead = true;
1955 296 : btshared->indtuples += buildstate.indtuples;
1956 296 : if (indexInfo->ii_BrokenHotChain)
1957 0 : btshared->brokenhotchain = true;
1958 296 : SpinLockRelease(&btshared->mutex);
1959 :
1960 : /* Notify leader */
1961 296 : ConditionVariableSignal(&btshared->workersdonecv);
1962 :
1963 : /* We can end tuplesorts immediately */
1964 296 : tuplesort_end(btspool->sortstate);
1965 296 : if (btspool2)
1966 128 : tuplesort_end(btspool2->sortstate);
1967 296 : }
|
