Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * nbtree.c
4 : * Implementation of Lehman and Yao's btree management algorithm for
5 : * Postgres.
6 : *
7 : * NOTES
8 : * This file contains only the public interface routines.
9 : *
10 : *
11 : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
12 : * Portions Copyright (c) 1994, Regents of the University of California
13 : *
14 : * IDENTIFICATION
15 : * src/backend/access/nbtree/nbtree.c
16 : *
17 : *-------------------------------------------------------------------------
18 : */
19 : #include "postgres.h"
20 :
21 : #include "access/nbtree.h"
22 : #include "access/relscan.h"
23 : #include "access/stratnum.h"
24 : #include "commands/progress.h"
25 : #include "commands/vacuum.h"
26 : #include "nodes/execnodes.h"
27 : #include "pgstat.h"
28 : #include "storage/bulk_write.h"
29 : #include "storage/condition_variable.h"
30 : #include "storage/indexfsm.h"
31 : #include "storage/ipc.h"
32 : #include "storage/lmgr.h"
33 : #include "storage/read_stream.h"
34 : #include "utils/datum.h"
35 : #include "utils/fmgrprotos.h"
36 : #include "utils/index_selfuncs.h"
37 : #include "utils/memutils.h"
38 :
39 :
40 : /*
41 : * BTPARALLEL_NOT_INITIALIZED indicates that the scan has not started.
42 : *
43 : * BTPARALLEL_NEED_PRIMSCAN indicates that some process must now seize the
44 : * scan to advance it via another call to _bt_first.
45 : *
46 : * BTPARALLEL_ADVANCING indicates that some process is advancing the scan to
47 : * a new page; others must wait.
48 : *
49 : * BTPARALLEL_IDLE indicates that no backend is currently advancing the scan
50 : * to a new page; some process can start doing that.
51 : *
52 : * BTPARALLEL_DONE indicates that the scan is complete (including error exit).
53 : */
54 : typedef enum
55 : {
56 : BTPARALLEL_NOT_INITIALIZED,
57 : BTPARALLEL_NEED_PRIMSCAN,
58 : BTPARALLEL_ADVANCING,
59 : BTPARALLEL_IDLE,
60 : BTPARALLEL_DONE,
61 : } BTPS_State;
62 :
63 : /*
64 : * BTParallelScanDescData contains btree specific shared information required
65 : * for parallel scan.
66 : */
67 : typedef struct BTParallelScanDescData
68 : {
69 : BlockNumber btps_nextScanPage; /* next page to be scanned */
70 : BlockNumber btps_lastCurrPage; /* page whose sibling link was copied into
71 : * btps_nextScanPage */
72 : BTPS_State btps_pageStatus; /* indicates whether next page is
73 : * available for scan. see above for
74 : * possible states of parallel scan. */
75 : LWLock btps_lock; /* protects shared parallel state */
76 : ConditionVariable btps_cv; /* used to synchronize parallel scan */
77 :
78 : /*
79 : * btps_arrElems is used when scans need to schedule another primitive
80 : * index scan with one or more SAOP arrays. Holds BTArrayKeyInfo.cur_elem
81 : * offsets for each = scan key associated with a ScalarArrayOp array.
82 : */
83 : int btps_arrElems[FLEXIBLE_ARRAY_MEMBER];
84 :
85 : /*
86 : * Additional space (at the end of the struct) is used when scans need to
87 : * schedule another primitive index scan with one or more skip arrays.
88 : * Holds a flattened datum representation for each = scan key associated
89 : * with a skip array.
90 : */
91 : } BTParallelScanDescData;
92 :
93 : typedef struct BTParallelScanDescData *BTParallelScanDesc;
94 :
95 :
96 : static bool _bt_start_prim_scan(IndexScanDesc scan);
97 : static void _bt_parallel_serialize_arrays(Relation rel, BTParallelScanDesc btscan,
98 : BTScanOpaque so);
99 : static void _bt_parallel_restore_arrays(Relation rel, BTParallelScanDesc btscan,
100 : BTScanOpaque so);
101 : static void btvacuumscan(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
102 : IndexBulkDeleteCallback callback, void *callback_state,
103 : BTCycleId cycleid);
104 : static BlockNumber btvacuumpage(BTVacState *vstate, Buffer buf);
105 : static BTVacuumPosting btreevacuumposting(BTVacState *vstate,
106 : IndexTuple posting,
107 : OffsetNumber updatedoffset,
108 : int *nremaining);
109 :
110 :
111 : /*
112 : * Btree handler function: return IndexAmRoutine with access method parameters
113 : * and callbacks.
114 : */
115 : Datum
116 3573940 : bthandler(PG_FUNCTION_ARGS)
117 : {
118 3573940 : IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);
119 :
120 3573940 : amroutine->amstrategies = BTMaxStrategyNumber;
121 3573940 : amroutine->amsupport = BTNProcs;
122 3573940 : amroutine->amoptsprocnum = BTOPTIONS_PROC;
123 3573940 : amroutine->amcanorder = true;
124 3573940 : amroutine->amcanorderbyop = false;
125 3573940 : amroutine->amcanhash = false;
126 3573940 : amroutine->amconsistentequality = true;
127 3573940 : amroutine->amconsistentordering = true;
128 3573940 : amroutine->amcanbackward = true;
129 3573940 : amroutine->amcanunique = true;
130 3573940 : amroutine->amcanmulticol = true;
131 3573940 : amroutine->amoptionalkey = true;
132 3573940 : amroutine->amsearcharray = true;
133 3573940 : amroutine->amsearchnulls = true;
134 3573940 : amroutine->amstorage = false;
135 3573940 : amroutine->amclusterable = true;
136 3573940 : amroutine->ampredlocks = true;
137 3573940 : amroutine->amcanparallel = true;
138 3573940 : amroutine->amcanbuildparallel = true;
139 3573940 : amroutine->amcaninclude = true;
140 3573940 : amroutine->amusemaintenanceworkmem = false;
141 3573940 : amroutine->amsummarizing = false;
142 3573940 : amroutine->amparallelvacuumoptions =
143 : VACUUM_OPTION_PARALLEL_BULKDEL | VACUUM_OPTION_PARALLEL_COND_CLEANUP;
144 3573940 : amroutine->amkeytype = InvalidOid;
145 :
146 3573940 : amroutine->ambuild = btbuild;
147 3573940 : amroutine->ambuildempty = btbuildempty;
148 3573940 : amroutine->aminsert = btinsert;
149 3573940 : amroutine->aminsertcleanup = NULL;
150 3573940 : amroutine->ambulkdelete = btbulkdelete;
151 3573940 : amroutine->amvacuumcleanup = btvacuumcleanup;
152 3573940 : amroutine->amcanreturn = btcanreturn;
153 3573940 : amroutine->amcostestimate = btcostestimate;
154 3573940 : amroutine->amgettreeheight = btgettreeheight;
155 3573940 : amroutine->amoptions = btoptions;
156 3573940 : amroutine->amproperty = btproperty;
157 3573940 : amroutine->ambuildphasename = btbuildphasename;
158 3573940 : amroutine->amvalidate = btvalidate;
159 3573940 : amroutine->amadjustmembers = btadjustmembers;
160 3573940 : amroutine->ambeginscan = btbeginscan;
161 3573940 : amroutine->amrescan = btrescan;
162 3573940 : amroutine->amgettuple = btgettuple;
163 3573940 : amroutine->amgetbitmap = btgetbitmap;
164 3573940 : amroutine->amendscan = btendscan;
165 3573940 : amroutine->ammarkpos = btmarkpos;
166 3573940 : amroutine->amrestrpos = btrestrpos;
167 3573940 : amroutine->amestimateparallelscan = btestimateparallelscan;
168 3573940 : amroutine->aminitparallelscan = btinitparallelscan;
169 3573940 : amroutine->amparallelrescan = btparallelrescan;
170 3573940 : amroutine->amtranslatestrategy = bttranslatestrategy;
171 3573940 : amroutine->amtranslatecmptype = bttranslatecmptype;
172 :
173 3573940 : PG_RETURN_POINTER(amroutine);
174 : }
175 :
176 : /*
177 : * btbuildempty() -- build an empty btree index in the initialization fork
178 : */
179 : void
180 156 : btbuildempty(Relation index)
181 : {
182 156 : bool allequalimage = _bt_allequalimage(index, false);
183 : BulkWriteState *bulkstate;
184 : BulkWriteBuffer metabuf;
185 :
186 156 : bulkstate = smgr_bulk_start_rel(index, INIT_FORKNUM);
187 :
188 : /* Construct metapage. */
189 156 : metabuf = smgr_bulk_get_buf(bulkstate);
190 156 : _bt_initmetapage((Page) metabuf, P_NONE, 0, allequalimage);
191 156 : smgr_bulk_write(bulkstate, BTREE_METAPAGE, metabuf, true);
192 :
193 156 : smgr_bulk_finish(bulkstate);
194 156 : }
195 :
196 : /*
197 : * btinsert() -- insert an index tuple into a btree.
198 : *
199 : * Descend the tree recursively, find the appropriate location for our
200 : * new tuple, and put it there.
201 : */
202 : bool
203 7416952 : btinsert(Relation rel, Datum *values, bool *isnull,
204 : ItemPointer ht_ctid, Relation heapRel,
205 : IndexUniqueCheck checkUnique,
206 : bool indexUnchanged,
207 : IndexInfo *indexInfo)
208 : {
209 : bool result;
210 : IndexTuple itup;
211 :
212 : /* generate an index tuple */
213 7416952 : itup = index_form_tuple(RelationGetDescr(rel), values, isnull);
214 7416952 : itup->t_tid = *ht_ctid;
215 :
216 7416952 : result = _bt_doinsert(rel, itup, checkUnique, indexUnchanged, heapRel);
217 :
218 7416420 : pfree(itup);
219 :
220 7416420 : return result;
221 : }
222 :
223 : /*
224 : * btgettuple() -- Get the next tuple in the scan.
225 : */
226 : bool
227 36686826 : btgettuple(IndexScanDesc scan, ScanDirection dir)
228 : {
229 36686826 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
230 : bool res;
231 :
232 : Assert(scan->heapRelation != NULL);
233 :
234 : /* btree indexes are never lossy */
235 36686826 : scan->xs_recheck = false;
236 :
237 : /* Each loop iteration performs another primitive index scan */
238 : do
239 : {
240 : /*
241 : * If we've already initialized this scan, we can just advance it in
242 : * the appropriate direction. If we haven't done so yet, we call
243 : * _bt_first() to get the first item in the scan.
244 : */
245 36703772 : if (!BTScanPosIsValid(so->currPos))
246 16777938 : res = _bt_first(scan, dir);
247 : else
248 : {
249 : /*
250 : * Check to see if we should kill the previously-fetched tuple.
251 : */
252 19925834 : if (scan->kill_prior_tuple)
253 : {
254 : /*
255 : * Yes, remember it for later. (We'll deal with all such
256 : * tuples at once right before leaving the index page.) The
257 : * test for numKilled overrun is not just paranoia: if the
258 : * caller reverses direction in the indexscan then the same
259 : * item might get entered multiple times. It's not worth
260 : * trying to optimize that, so we don't detect it, but instead
261 : * just forget any excess entries.
262 : */
263 506568 : if (so->killedItems == NULL)
264 172536 : so->killedItems = palloc_array(int, MaxTIDsPerBTreePage);
265 506568 : if (so->numKilled < MaxTIDsPerBTreePage)
266 506568 : so->killedItems[so->numKilled++] = so->currPos.itemIndex;
267 : }
268 :
269 : /*
270 : * Now continue the scan.
271 : */
272 19925834 : res = _bt_next(scan, dir);
273 : }
274 :
275 : /* If we have a tuple, return it ... */
276 36703768 : if (res)
277 29394932 : break;
278 : /* ... otherwise see if we need another primitive index scan */
279 7308836 : } while (so->numArrayKeys && _bt_start_prim_scan(scan));
280 :
281 36686822 : return res;
282 : }
283 :
284 : /*
285 : * btgetbitmap() -- gets all matching tuples, and adds them to a bitmap
286 : */
287 : int64
288 18894 : btgetbitmap(IndexScanDesc scan, TIDBitmap *tbm)
289 : {
290 18894 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
291 18894 : int64 ntids = 0;
292 : ItemPointer heapTid;
293 :
294 : Assert(scan->heapRelation == NULL);
295 :
296 : /* Each loop iteration performs another primitive index scan */
297 : do
298 : {
299 : /* Fetch the first page & tuple */
300 19524 : if (_bt_first(scan, ForwardScanDirection))
301 : {
302 : /* Save tuple ID, and continue scanning */
303 15360 : heapTid = &scan->xs_heaptid;
304 15360 : tbm_add_tuples(tbm, heapTid, 1, false);
305 15360 : ntids++;
306 :
307 : for (;;)
308 : {
309 : /*
310 : * Advance to next tuple within page. This is the same as the
311 : * easy case in _bt_next().
312 : */
313 2096484 : if (++so->currPos.itemIndex > so->currPos.lastItem)
314 : {
315 : /* let _bt_next do the heavy lifting */
316 20844 : if (!_bt_next(scan, ForwardScanDirection))
317 15360 : break;
318 : }
319 :
320 : /* Save tuple ID, and continue scanning */
321 2081124 : heapTid = &so->currPos.items[so->currPos.itemIndex].heapTid;
322 2081124 : tbm_add_tuples(tbm, heapTid, 1, false);
323 2081124 : ntids++;
324 : }
325 : }
326 : /* Now see if we need another primitive index scan */
327 19524 : } while (so->numArrayKeys && _bt_start_prim_scan(scan));
328 :
329 18894 : return ntids;
330 : }
331 :
332 : /*
333 : * btbeginscan() -- start a scan on a btree index
334 : */
335 : IndexScanDesc
336 16150886 : btbeginscan(Relation rel, int nkeys, int norderbys)
337 : {
338 : IndexScanDesc scan;
339 : BTScanOpaque so;
340 :
341 : /* no order by operators allowed */
342 : Assert(norderbys == 0);
343 :
344 : /* get the scan */
345 16150886 : scan = RelationGetIndexScan(rel, nkeys, norderbys);
346 :
347 : /* allocate private workspace */
348 16150886 : so = palloc_object(BTScanOpaqueData);
349 16150886 : BTScanPosInvalidate(so->currPos);
350 16150886 : BTScanPosInvalidate(so->markPos);
351 16150886 : if (scan->numberOfKeys > 0)
352 16137690 : so->keyData = (ScanKey) palloc(scan->numberOfKeys * sizeof(ScanKeyData));
353 : else
354 13196 : so->keyData = NULL;
355 :
356 16150886 : so->skipScan = false;
357 16150886 : so->needPrimScan = false;
358 16150886 : so->scanBehind = false;
359 16150886 : so->oppositeDirCheck = false;
360 16150886 : so->arrayKeys = NULL;
361 16150886 : so->orderProcs = NULL;
362 16150886 : so->arrayContext = NULL;
363 :
364 16150886 : so->killedItems = NULL; /* until needed */
365 16150886 : so->numKilled = 0;
366 :
367 : /*
368 : * We don't know yet whether the scan will be index-only, so we do not
369 : * allocate the tuple workspace arrays until btrescan. However, we set up
370 : * scan->xs_itupdesc whether we'll need it or not, since that's so cheap.
371 : */
372 16150886 : so->currTuples = so->markTuples = NULL;
373 :
374 16150886 : scan->xs_itupdesc = RelationGetDescr(rel);
375 :
376 16150886 : scan->opaque = so;
377 :
378 16150886 : return scan;
379 : }
380 :
381 : /*
382 : * btrescan() -- rescan an index relation
383 : */
384 : void
385 16781234 : btrescan(IndexScanDesc scan, ScanKey scankey, int nscankeys,
386 : ScanKey orderbys, int norderbys)
387 : {
388 16781234 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
389 :
390 : /* we aren't holding any read locks, but gotta drop the pins */
391 16781234 : if (BTScanPosIsValid(so->currPos))
392 : {
393 : /* Before leaving current page, deal with any killed items */
394 81068 : if (so->numKilled > 0)
395 1008 : _bt_killitems(scan);
396 81068 : BTScanPosUnpinIfPinned(so->currPos);
397 81068 : BTScanPosInvalidate(so->currPos);
398 : }
399 :
400 : /*
401 : * We prefer to eagerly drop leaf page pins before btgettuple returns.
402 : * This avoids making VACUUM wait to acquire a cleanup lock on the page.
403 : *
404 : * We cannot safely drop leaf page pins during index-only scans due to a
405 : * race condition involving VACUUM setting pages all-visible in the VM.
406 : * It's also unsafe for plain index scans that use a non-MVCC snapshot.
407 : *
408 : * When we drop pins eagerly, the mechanism that marks so->killedItems[]
409 : * index tuples LP_DEAD has to deal with concurrent TID recycling races.
410 : * The scheme used to detect unsafe TID recycling won't work when scanning
411 : * unlogged relations (since it involves saving an affected page's LSN).
412 : * Opt out of eager pin dropping during unlogged relation scans for now
413 : * (this is preferable to opting out of kill_prior_tuple LP_DEAD setting).
414 : *
415 : * Also opt out of dropping leaf page pins eagerly during bitmap scans.
416 : * Pins cannot be held for more than an instant during bitmap scans either
417 : * way, so we might as well avoid wasting cycles on acquiring page LSNs.
418 : *
419 : * See nbtree/README section on making concurrent TID recycling safe.
420 : *
421 : * Note: so->dropPin should never change across rescans.
422 : */
423 49987222 : so->dropPin = (!scan->xs_want_itup &&
424 16424754 : IsMVCCSnapshot(scan->xs_snapshot) &&
425 48421870 : RelationNeedsWAL(scan->indexRelation) &&
426 15215882 : scan->heapRelation != NULL);
427 :
428 16781234 : so->markItemIndex = -1;
429 16781234 : so->needPrimScan = false;
430 16781234 : so->scanBehind = false;
431 16781234 : so->oppositeDirCheck = false;
432 16781234 : BTScanPosUnpinIfPinned(so->markPos);
433 16781234 : BTScanPosInvalidate(so->markPos);
434 :
435 : /*
436 : * Allocate tuple workspace arrays, if needed for an index-only scan and
437 : * not already done in a previous rescan call. To save on palloc
438 : * overhead, both workspaces are allocated as one palloc block; only this
439 : * function and btendscan know that.
440 : *
441 : * NOTE: this data structure also makes it safe to return data from a
442 : * "name" column, even though btree name_ops uses an underlying storage
443 : * datatype of cstring. The risk there is that "name" is supposed to be
444 : * padded to NAMEDATALEN, but the actual index tuple is probably shorter.
445 : * However, since we only return data out of tuples sitting in the
446 : * currTuples array, a fetch of NAMEDATALEN bytes can at worst pull some
447 : * data out of the markTuples array --- running off the end of memory for
448 : * a SIGSEGV is not possible. Yeah, this is ugly as sin, but it beats
449 : * adding special-case treatment for name_ops elsewhere.
450 : */
451 16781234 : if (scan->xs_want_itup && so->currTuples == NULL)
452 : {
453 136108 : so->currTuples = (char *) palloc(BLCKSZ * 2);
454 136108 : so->markTuples = so->currTuples + BLCKSZ;
455 : }
456 :
457 : /*
458 : * Reset the scan keys
459 : */
460 16781234 : if (scankey && scan->numberOfKeys > 0)
461 16767808 : memcpy(scan->keyData, scankey, scan->numberOfKeys * sizeof(ScanKeyData));
462 16781234 : so->numberOfKeys = 0; /* until _bt_preprocess_keys sets it */
463 16781234 : so->numArrayKeys = 0; /* ditto */
464 16781234 : }
465 :
466 : /*
467 : * btendscan() -- close down a scan
468 : */
469 : void
470 16149262 : btendscan(IndexScanDesc scan)
471 : {
472 16149262 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
473 :
474 : /* we aren't holding any read locks, but gotta drop the pins */
475 16149262 : if (BTScanPosIsValid(so->currPos))
476 : {
477 : /* Before leaving current page, deal with any killed items */
478 9386874 : if (so->numKilled > 0)
479 90144 : _bt_killitems(scan);
480 9386874 : BTScanPosUnpinIfPinned(so->currPos);
481 : }
482 :
483 16149262 : so->markItemIndex = -1;
484 16149262 : BTScanPosUnpinIfPinned(so->markPos);
485 :
486 : /* No need to invalidate positions, the RAM is about to be freed. */
487 :
488 : /* Release storage */
489 16149262 : if (so->keyData != NULL)
490 16136096 : pfree(so->keyData);
491 : /* so->arrayKeys and so->orderProcs are in arrayContext */
492 16149262 : if (so->arrayContext != NULL)
493 4904 : MemoryContextDelete(so->arrayContext);
494 16149262 : if (so->killedItems != NULL)
495 172472 : pfree(so->killedItems);
496 16149262 : if (so->currTuples != NULL)
497 136064 : pfree(so->currTuples);
498 : /* so->markTuples should not be pfree'd, see btrescan */
499 16149262 : pfree(so);
500 16149262 : }
501 :
502 : /*
503 : * btmarkpos() -- save current scan position
504 : */
505 : void
506 130124 : btmarkpos(IndexScanDesc scan)
507 : {
508 130124 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
509 :
510 : /* There may be an old mark with a pin (but no lock). */
511 130124 : BTScanPosUnpinIfPinned(so->markPos);
512 :
513 : /*
514 : * Just record the current itemIndex. If we later step to next page
515 : * before releasing the marked position, _bt_steppage makes a full copy of
516 : * the currPos struct in markPos. If (as often happens) the mark is moved
517 : * before we leave the page, we don't have to do that work.
518 : */
519 130124 : if (BTScanPosIsValid(so->currPos))
520 130124 : so->markItemIndex = so->currPos.itemIndex;
521 : else
522 : {
523 0 : BTScanPosInvalidate(so->markPos);
524 0 : so->markItemIndex = -1;
525 : }
526 130124 : }
527 :
528 : /*
529 : * btrestrpos() -- restore scan to last saved position
530 : */
531 : void
532 54054 : btrestrpos(IndexScanDesc scan)
533 : {
534 54054 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
535 :
536 54054 : if (so->markItemIndex >= 0)
537 : {
538 : /*
539 : * The scan has never moved to a new page since the last mark. Just
540 : * restore the itemIndex.
541 : *
542 : * NB: In this case we can't count on anything in so->markPos to be
543 : * accurate.
544 : */
545 53946 : so->currPos.itemIndex = so->markItemIndex;
546 : }
547 : else
548 : {
549 : /*
550 : * The scan moved to a new page after last mark or restore, and we are
551 : * now restoring to the marked page. We aren't holding any read
552 : * locks, but if we're still holding the pin for the current position,
553 : * we must drop it.
554 : */
555 108 : if (BTScanPosIsValid(so->currPos))
556 : {
557 : /* Before leaving current page, deal with any killed items */
558 108 : if (so->numKilled > 0)
559 0 : _bt_killitems(scan);
560 108 : BTScanPosUnpinIfPinned(so->currPos);
561 : }
562 :
563 108 : if (BTScanPosIsValid(so->markPos))
564 : {
565 : /* bump pin on mark buffer for assignment to current buffer */
566 108 : if (BTScanPosIsPinned(so->markPos))
567 0 : IncrBufferRefCount(so->markPos.buf);
568 108 : memcpy(&so->currPos, &so->markPos,
569 : offsetof(BTScanPosData, items[1]) +
570 108 : so->markPos.lastItem * sizeof(BTScanPosItem));
571 108 : if (so->currTuples)
572 0 : memcpy(so->currTuples, so->markTuples,
573 0 : so->markPos.nextTupleOffset);
574 : /* Reset the scan's array keys (see _bt_steppage for why) */
575 108 : if (so->numArrayKeys)
576 : {
577 0 : _bt_start_array_keys(scan, so->currPos.dir);
578 0 : so->needPrimScan = false;
579 : }
580 : }
581 : else
582 0 : BTScanPosInvalidate(so->currPos);
583 : }
584 54054 : }
585 :
586 : /*
587 : * btestimateparallelscan -- estimate storage for BTParallelScanDescData
588 : */
589 : Size
590 64 : btestimateparallelscan(Relation rel, int nkeys, int norderbys)
591 : {
592 64 : int16 nkeyatts = IndexRelationGetNumberOfKeyAttributes(rel);
593 : Size estnbtreeshared,
594 : genericattrspace;
595 :
596 : /*
597 : * Pessimistically assume that every input scan key will be output with
598 : * its own SAOP array
599 : */
600 64 : estnbtreeshared = offsetof(BTParallelScanDescData, btps_arrElems) +
601 : sizeof(int) * nkeys;
602 :
603 : /* Single column indexes cannot possibly use a skip array */
604 64 : if (nkeyatts == 1)
605 46 : return estnbtreeshared;
606 :
607 : /*
608 : * Pessimistically assume that all attributes prior to the least
609 : * significant attribute require a skip array (and an associated key)
610 : */
611 18 : genericattrspace = datumEstimateSpace((Datum) 0, false, true,
612 : sizeof(Datum));
613 36 : for (int attnum = 1; attnum < nkeyatts; attnum++)
614 : {
615 : CompactAttribute *attr;
616 :
617 : /*
618 : * We make the conservative assumption that every index column will
619 : * also require a skip array.
620 : *
621 : * Every skip array must have space to store its scan key's sk_flags.
622 : */
623 18 : estnbtreeshared = add_size(estnbtreeshared, sizeof(int));
624 :
625 : /* Consider space required to store a datum of opclass input type */
626 18 : attr = TupleDescCompactAttr(rel->rd_att, attnum - 1);
627 18 : if (attr->attbyval)
628 18 : {
629 : /* This index attribute stores pass-by-value datums */
630 18 : Size estfixed = datumEstimateSpace((Datum) 0, false,
631 18 : true, attr->attlen);
632 :
633 18 : estnbtreeshared = add_size(estnbtreeshared, estfixed);
634 18 : continue;
635 : }
636 :
637 : /*
638 : * This index attribute stores pass-by-reference datums.
639 : *
640 : * Assume that serializing this array will use just as much space as a
641 : * pass-by-value datum, in addition to space for the largest possible
642 : * whole index tuple (this is not just a per-datum portion of the
643 : * largest possible tuple because that'd be almost as large anyway).
644 : *
645 : * This is quite conservative, but it's not clear how we could do much
646 : * better. The executor requires an up-front storage request size
647 : * that reliably covers the scan's high watermark memory usage. We
648 : * can't be sure of the real high watermark until the scan is over.
649 : */
650 0 : estnbtreeshared = add_size(estnbtreeshared, genericattrspace);
651 0 : estnbtreeshared = add_size(estnbtreeshared, BTMaxItemSize);
652 : }
653 :
654 18 : return estnbtreeshared;
655 : }
656 :
657 : /*
658 : * _bt_start_prim_scan() -- start scheduled primitive index scan?
659 : *
660 : * Returns true if _bt_checkkeys scheduled another primitive index scan, just
661 : * as the last one ended. Otherwise returns false, indicating that the array
662 : * keys are now fully exhausted.
663 : *
664 : * Only call here during scans with one or more equality type array scan keys,
665 : * after _bt_first or _bt_next return false.
666 : */
667 : static bool
668 88934 : _bt_start_prim_scan(IndexScanDesc scan)
669 : {
670 88934 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
671 :
672 : Assert(so->numArrayKeys);
673 :
674 88934 : so->scanBehind = so->oppositeDirCheck = false; /* reset */
675 :
676 : /*
677 : * Array keys are advanced within _bt_checkkeys when the scan reaches the
678 : * leaf level (more precisely, they're advanced when the scan reaches the
679 : * end of each distinct set of array elements). This process avoids
680 : * repeat access to leaf pages (across multiple primitive index scans) by
681 : * advancing the scan's array keys when it allows the primitive index scan
682 : * to find nearby matching tuples (or when it eliminates ranges of array
683 : * key space that can't possibly be satisfied by any index tuple).
684 : *
685 : * _bt_checkkeys sets a simple flag variable to schedule another primitive
686 : * index scan. The flag tells us what to do.
687 : *
688 : * We cannot rely on _bt_first always reaching _bt_checkkeys. There are
689 : * various cases where that won't happen. For example, if the index is
690 : * completely empty, then _bt_first won't call _bt_readpage/_bt_checkkeys.
691 : * We also don't expect a call to _bt_checkkeys during searches for a
692 : * non-existent value that happens to be lower/higher than any existing
693 : * value in the index.
694 : *
695 : * We don't require special handling for these cases -- we don't need to
696 : * be explicitly instructed to _not_ perform another primitive index scan.
697 : * It's up to code under the control of _bt_first to always set the flag
698 : * when another primitive index scan will be required.
699 : *
700 : * This works correctly, even with the tricky cases listed above, which
701 : * all involve access to leaf pages "near the boundaries of the key space"
702 : * (whether it's from a leftmost/rightmost page, or an imaginary empty
703 : * leaf root page). If _bt_checkkeys cannot be reached by a primitive
704 : * index scan for one set of array keys, then it also won't be reached for
705 : * any later set ("later" in terms of the direction that we scan the index
706 : * and advance the arrays). The array keys won't have advanced in these
707 : * cases, but that's the correct behavior (even _bt_advance_array_keys
708 : * won't always advance the arrays at the point they become "exhausted").
709 : */
710 88934 : if (so->needPrimScan)
711 : {
712 : /*
713 : * Flag was set -- must call _bt_first again, which will reset the
714 : * scan's needPrimScan flag
715 : */
716 17576 : return true;
717 : }
718 :
719 : /* The top-level index scan ran out of tuples in this scan direction */
720 71358 : if (scan->parallel_scan != NULL)
721 30 : _bt_parallel_done(scan);
722 :
723 71358 : return false;
724 : }
725 :
726 : /*
727 : * _bt_parallel_serialize_arrays() -- Serialize parallel array state.
728 : *
729 : * Caller must have exclusively locked btscan->btps_lock when called.
730 : */
731 : static void
732 36 : _bt_parallel_serialize_arrays(Relation rel, BTParallelScanDesc btscan,
733 : BTScanOpaque so)
734 : {
735 : char *datumshared;
736 :
737 : /* Space for serialized datums begins immediately after btps_arrElems[] */
738 36 : datumshared = ((char *) &btscan->btps_arrElems[so->numArrayKeys]);
739 72 : for (int i = 0; i < so->numArrayKeys; i++)
740 : {
741 36 : BTArrayKeyInfo *array = &so->arrayKeys[i];
742 36 : ScanKey skey = &so->keyData[array->scan_key];
743 :
744 36 : if (array->num_elems != -1)
745 : {
746 : /* Save SAOP array's cur_elem (no need to copy key/datum) */
747 : Assert(!(skey->sk_flags & SK_BT_SKIP));
748 36 : btscan->btps_arrElems[i] = array->cur_elem;
749 36 : continue;
750 : }
751 :
752 : /* Save all mutable state associated with skip array's key */
753 : Assert(skey->sk_flags & SK_BT_SKIP);
754 0 : memcpy(datumshared, &skey->sk_flags, sizeof(int));
755 0 : datumshared += sizeof(int);
756 :
757 0 : if (skey->sk_flags & (SK_BT_MINVAL | SK_BT_MAXVAL))
758 : {
759 : /* No sk_argument datum to serialize */
760 : Assert(skey->sk_argument == 0);
761 0 : continue;
762 : }
763 :
764 0 : datumSerialize(skey->sk_argument, (skey->sk_flags & SK_ISNULL) != 0,
765 0 : array->attbyval, array->attlen, &datumshared);
766 : }
767 36 : }
768 :
769 : /*
770 : * _bt_parallel_restore_arrays() -- Restore serialized parallel array state.
771 : *
772 : * Caller must have exclusively locked btscan->btps_lock when called.
773 : */
774 : static void
775 36 : _bt_parallel_restore_arrays(Relation rel, BTParallelScanDesc btscan,
776 : BTScanOpaque so)
777 : {
778 : char *datumshared;
779 :
780 : /* Space for serialized datums begins immediately after btps_arrElems[] */
781 36 : datumshared = ((char *) &btscan->btps_arrElems[so->numArrayKeys]);
782 72 : for (int i = 0; i < so->numArrayKeys; i++)
783 : {
784 36 : BTArrayKeyInfo *array = &so->arrayKeys[i];
785 36 : ScanKey skey = &so->keyData[array->scan_key];
786 : bool isnull;
787 :
788 36 : if (array->num_elems != -1)
789 : {
790 : /* Restore SAOP array using its saved cur_elem */
791 : Assert(!(skey->sk_flags & SK_BT_SKIP));
792 36 : array->cur_elem = btscan->btps_arrElems[i];
793 36 : skey->sk_argument = array->elem_values[array->cur_elem];
794 36 : continue;
795 : }
796 :
797 : /* Restore skip array by restoring its key directly */
798 0 : if (!array->attbyval && skey->sk_argument)
799 0 : pfree(DatumGetPointer(skey->sk_argument));
800 0 : skey->sk_argument = (Datum) 0;
801 0 : memcpy(&skey->sk_flags, datumshared, sizeof(int));
802 0 : datumshared += sizeof(int);
803 :
804 : Assert(skey->sk_flags & SK_BT_SKIP);
805 :
806 0 : if (skey->sk_flags & (SK_BT_MINVAL | SK_BT_MAXVAL))
807 : {
808 : /* No sk_argument datum to restore */
809 0 : continue;
810 : }
811 :
812 0 : skey->sk_argument = datumRestore(&datumshared, &isnull);
813 0 : if (isnull)
814 : {
815 : Assert(skey->sk_argument == 0);
816 : Assert(skey->sk_flags & SK_SEARCHNULL);
817 : Assert(skey->sk_flags & SK_ISNULL);
818 : }
819 : }
820 36 : }
821 :
822 : /*
823 : * btinitparallelscan -- initialize BTParallelScanDesc for parallel btree scan
824 : */
825 : void
826 64 : btinitparallelscan(void *target)
827 : {
828 64 : BTParallelScanDesc bt_target = (BTParallelScanDesc) target;
829 :
830 64 : LWLockInitialize(&bt_target->btps_lock,
831 : LWTRANCHE_PARALLEL_BTREE_SCAN);
832 64 : bt_target->btps_nextScanPage = InvalidBlockNumber;
833 64 : bt_target->btps_lastCurrPage = InvalidBlockNumber;
834 64 : bt_target->btps_pageStatus = BTPARALLEL_NOT_INITIALIZED;
835 64 : ConditionVariableInit(&bt_target->btps_cv);
836 64 : }
837 :
838 : /*
839 : * btparallelrescan() -- reset parallel scan
840 : */
841 : void
842 24 : btparallelrescan(IndexScanDesc scan)
843 : {
844 : BTParallelScanDesc btscan;
845 24 : ParallelIndexScanDesc parallel_scan = scan->parallel_scan;
846 :
847 : Assert(parallel_scan);
848 :
849 24 : btscan = (BTParallelScanDesc) OffsetToPointer(parallel_scan,
850 : parallel_scan->ps_offset_am);
851 :
852 : /*
853 : * In theory, we don't need to acquire the LWLock here, because there
854 : * shouldn't be any other workers running at this point, but we do so for
855 : * consistency.
856 : */
857 24 : LWLockAcquire(&btscan->btps_lock, LW_EXCLUSIVE);
858 24 : btscan->btps_nextScanPage = InvalidBlockNumber;
859 24 : btscan->btps_lastCurrPage = InvalidBlockNumber;
860 24 : btscan->btps_pageStatus = BTPARALLEL_NOT_INITIALIZED;
861 24 : LWLockRelease(&btscan->btps_lock);
862 24 : }
863 :
864 : /*
865 : * _bt_parallel_seize() -- Begin the process of advancing the scan to a new
866 : * page. Other scans must wait until we call _bt_parallel_release()
867 : * or _bt_parallel_done().
868 : *
869 : * The return value is true if we successfully seized the scan and false
870 : * if we did not. The latter case occurs when no pages remain, or when
871 : * another primitive index scan is scheduled that caller's backend cannot
872 : * start just yet (only backends that call from _bt_first are capable of
873 : * starting primitive index scans, which they indicate by passing first=true).
874 : *
875 : * If the return value is true, *next_scan_page returns the next page of the
876 : * scan, and *last_curr_page returns the page that *next_scan_page came from.
877 : * An invalid *next_scan_page means the scan hasn't yet started, or that
878 : * caller needs to start the next primitive index scan (if it's the latter
879 : * case we'll set so.needPrimScan).
880 : *
881 : * Callers should ignore the value of *next_scan_page and *last_curr_page if
882 : * the return value is false.
883 : */
884 : bool
885 1658 : _bt_parallel_seize(IndexScanDesc scan, BlockNumber *next_scan_page,
886 : BlockNumber *last_curr_page, bool first)
887 : {
888 1658 : Relation rel = scan->indexRelation;
889 1658 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
890 1658 : bool exit_loop = false,
891 1658 : status = true,
892 1658 : endscan = false;
893 1658 : ParallelIndexScanDesc parallel_scan = scan->parallel_scan;
894 : BTParallelScanDesc btscan;
895 :
896 1658 : *next_scan_page = InvalidBlockNumber;
897 1658 : *last_curr_page = InvalidBlockNumber;
898 :
899 : /*
900 : * Reset so->currPos, and initialize moreLeft/moreRight such that the next
901 : * call to _bt_readnextpage treats this backend similarly to a serial
902 : * backend that steps from *last_curr_page to *next_scan_page (unless this
903 : * backend's so->currPos is initialized by _bt_readfirstpage before then).
904 : */
905 1658 : BTScanPosInvalidate(so->currPos);
906 1658 : so->currPos.moreLeft = so->currPos.moreRight = true;
907 :
908 1658 : if (first)
909 : {
910 : /*
911 : * Initialize array related state when called from _bt_first, assuming
912 : * that this will be the first primitive index scan for the scan
913 : */
914 446 : so->needPrimScan = false;
915 446 : so->scanBehind = false;
916 446 : so->oppositeDirCheck = false;
917 : }
918 : else
919 : {
920 : /*
921 : * Don't attempt to seize the scan when it requires another primitive
922 : * index scan, since caller's backend cannot start it right now
923 : */
924 1212 : if (so->needPrimScan)
925 0 : return false;
926 : }
927 :
928 1658 : btscan = (BTParallelScanDesc) OffsetToPointer(parallel_scan,
929 : parallel_scan->ps_offset_am);
930 :
931 : while (1)
932 : {
933 1686 : LWLockAcquire(&btscan->btps_lock, LW_EXCLUSIVE);
934 :
935 1686 : if (btscan->btps_pageStatus == BTPARALLEL_DONE)
936 : {
937 : /* We're done with this parallel index scan */
938 312 : status = false;
939 : }
940 1374 : else if (btscan->btps_pageStatus == BTPARALLEL_IDLE &&
941 1222 : btscan->btps_nextScanPage == P_NONE)
942 : {
943 : /* End this parallel index scan */
944 10 : status = false;
945 10 : endscan = true;
946 : }
947 1364 : else if (btscan->btps_pageStatus == BTPARALLEL_NEED_PRIMSCAN)
948 : {
949 : Assert(so->numArrayKeys);
950 :
951 36 : if (first)
952 : {
953 : /* Can start scheduled primitive scan right away, so do so */
954 36 : btscan->btps_pageStatus = BTPARALLEL_ADVANCING;
955 :
956 : /* Restore scan's array keys from serialized values */
957 36 : _bt_parallel_restore_arrays(rel, btscan, so);
958 36 : exit_loop = true;
959 : }
960 : else
961 : {
962 : /*
963 : * Don't attempt to seize the scan when it requires another
964 : * primitive index scan, since caller's backend cannot start
965 : * it right now
966 : */
967 0 : status = false;
968 : }
969 :
970 : /*
971 : * Either way, update backend local state to indicate that a
972 : * pending primitive scan is required
973 : */
974 36 : so->needPrimScan = true;
975 36 : so->scanBehind = false;
976 36 : so->oppositeDirCheck = false;
977 : }
978 1328 : else if (btscan->btps_pageStatus != BTPARALLEL_ADVANCING)
979 : {
980 : /*
981 : * We have successfully seized control of the scan for the purpose
982 : * of advancing it to a new page!
983 : */
984 1300 : btscan->btps_pageStatus = BTPARALLEL_ADVANCING;
985 : Assert(btscan->btps_nextScanPage != P_NONE);
986 1300 : *next_scan_page = btscan->btps_nextScanPage;
987 1300 : *last_curr_page = btscan->btps_lastCurrPage;
988 1300 : exit_loop = true;
989 : }
990 1686 : LWLockRelease(&btscan->btps_lock);
991 1686 : if (exit_loop || !status)
992 : break;
993 28 : ConditionVariableSleep(&btscan->btps_cv, WAIT_EVENT_BTREE_PAGE);
994 : }
995 1658 : ConditionVariableCancelSleep();
996 :
997 : /* When the scan has reached the rightmost (or leftmost) page, end it */
998 1658 : if (endscan)
999 10 : _bt_parallel_done(scan);
1000 :
1001 1658 : return status;
1002 : }
1003 :
1004 : /*
1005 : * _bt_parallel_release() -- Complete the process of advancing the scan to a
1006 : * new page. We now have the new value btps_nextScanPage; another backend
1007 : * can now begin advancing the scan.
1008 : *
1009 : * Callers whose scan uses array keys must save their curr_page argument so
1010 : * that it can be passed to _bt_parallel_primscan_schedule, should caller
1011 : * determine that another primitive index scan is required.
1012 : *
1013 : * If caller's next_scan_page is P_NONE, the scan has reached the index's
1014 : * rightmost/leftmost page. This is treated as reaching the end of the scan
1015 : * within _bt_parallel_seize.
1016 : *
1017 : * Note: unlike the serial case, parallel scans don't need to remember both
1018 : * sibling links. next_scan_page is whichever link is next given the scan's
1019 : * direction. That's all we'll ever need, since the direction of a parallel
1020 : * scan can never change.
1021 : */
1022 : void
1023 1336 : _bt_parallel_release(IndexScanDesc scan, BlockNumber next_scan_page,
1024 : BlockNumber curr_page)
1025 : {
1026 1336 : ParallelIndexScanDesc parallel_scan = scan->parallel_scan;
1027 : BTParallelScanDesc btscan;
1028 :
1029 : Assert(BlockNumberIsValid(next_scan_page));
1030 :
1031 1336 : btscan = (BTParallelScanDesc) OffsetToPointer(parallel_scan,
1032 : parallel_scan->ps_offset_am);
1033 :
1034 1336 : LWLockAcquire(&btscan->btps_lock, LW_EXCLUSIVE);
1035 1336 : btscan->btps_nextScanPage = next_scan_page;
1036 1336 : btscan->btps_lastCurrPage = curr_page;
1037 1336 : btscan->btps_pageStatus = BTPARALLEL_IDLE;
1038 1336 : LWLockRelease(&btscan->btps_lock);
1039 1336 : ConditionVariableSignal(&btscan->btps_cv);
1040 1336 : }
1041 :
1042 : /*
1043 : * _bt_parallel_done() -- Mark the parallel scan as complete.
1044 : *
1045 : * When there are no pages left to scan, this function should be called to
1046 : * notify other workers. Otherwise, they might wait forever for the scan to
1047 : * advance to the next page.
1048 : */
1049 : void
1050 7328078 : _bt_parallel_done(IndexScanDesc scan)
1051 : {
1052 7328078 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
1053 7328078 : ParallelIndexScanDesc parallel_scan = scan->parallel_scan;
1054 : BTParallelScanDesc btscan;
1055 7328078 : bool status_changed = false;
1056 :
1057 : Assert(!BTScanPosIsValid(so->currPos));
1058 :
1059 : /* Do nothing, for non-parallel scans */
1060 7328078 : if (parallel_scan == NULL)
1061 7327920 : return;
1062 :
1063 : /*
1064 : * Should not mark parallel scan done when there's still a pending
1065 : * primitive index scan
1066 : */
1067 158 : if (so->needPrimScan)
1068 36 : return;
1069 :
1070 122 : btscan = (BTParallelScanDesc) OffsetToPointer(parallel_scan,
1071 : parallel_scan->ps_offset_am);
1072 :
1073 : /*
1074 : * Mark the parallel scan as done, unless some other process did so
1075 : * already
1076 : */
1077 122 : LWLockAcquire(&btscan->btps_lock, LW_EXCLUSIVE);
1078 : Assert(btscan->btps_pageStatus != BTPARALLEL_NEED_PRIMSCAN);
1079 122 : if (btscan->btps_pageStatus != BTPARALLEL_DONE)
1080 : {
1081 88 : btscan->btps_pageStatus = BTPARALLEL_DONE;
1082 88 : status_changed = true;
1083 : }
1084 122 : LWLockRelease(&btscan->btps_lock);
1085 :
1086 : /* wake up all the workers associated with this parallel scan */
1087 122 : if (status_changed)
1088 88 : ConditionVariableBroadcast(&btscan->btps_cv);
1089 : }
1090 :
1091 : /*
1092 : * _bt_parallel_primscan_schedule() -- Schedule another primitive index scan.
1093 : *
1094 : * Caller passes the curr_page most recently passed to _bt_parallel_release
1095 : * by its backend. Caller successfully schedules the next primitive index scan
1096 : * if the shared parallel state hasn't been seized since caller's backend last
1097 : * advanced the scan.
1098 : */
1099 : void
1100 36 : _bt_parallel_primscan_schedule(IndexScanDesc scan, BlockNumber curr_page)
1101 : {
1102 36 : Relation rel = scan->indexRelation;
1103 36 : BTScanOpaque so = (BTScanOpaque) scan->opaque;
1104 36 : ParallelIndexScanDesc parallel_scan = scan->parallel_scan;
1105 : BTParallelScanDesc btscan;
1106 :
1107 : Assert(so->numArrayKeys);
1108 :
1109 36 : btscan = (BTParallelScanDesc) OffsetToPointer(parallel_scan,
1110 : parallel_scan->ps_offset_am);
1111 :
1112 36 : LWLockAcquire(&btscan->btps_lock, LW_EXCLUSIVE);
1113 36 : if (btscan->btps_lastCurrPage == curr_page &&
1114 36 : btscan->btps_pageStatus == BTPARALLEL_IDLE)
1115 : {
1116 36 : btscan->btps_nextScanPage = InvalidBlockNumber;
1117 36 : btscan->btps_lastCurrPage = InvalidBlockNumber;
1118 36 : btscan->btps_pageStatus = BTPARALLEL_NEED_PRIMSCAN;
1119 :
1120 : /* Serialize scan's current array keys */
1121 36 : _bt_parallel_serialize_arrays(rel, btscan, so);
1122 : }
1123 36 : LWLockRelease(&btscan->btps_lock);
1124 36 : }
1125 :
1126 : /*
1127 : * Bulk deletion of all index entries pointing to a set of heap tuples.
1128 : * The set of target tuples is specified via a callback routine that tells
1129 : * whether any given heap tuple (identified by ItemPointer) is being deleted.
1130 : *
1131 : * Result: a palloc'd struct containing statistical info for VACUUM displays.
1132 : */
1133 : IndexBulkDeleteResult *
1134 2902 : btbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
1135 : IndexBulkDeleteCallback callback, void *callback_state)
1136 : {
1137 2902 : Relation rel = info->index;
1138 : BTCycleId cycleid;
1139 :
1140 : /* allocate stats if first time through, else re-use existing struct */
1141 2902 : if (stats == NULL)
1142 2898 : stats = palloc0_object(IndexBulkDeleteResult);
1143 :
1144 : /* Establish the vacuum cycle ID to use for this scan */
1145 : /* The ENSURE stuff ensures we clean up shared memory on failure */
1146 2902 : PG_ENSURE_ERROR_CLEANUP(_bt_end_vacuum_callback, PointerGetDatum(rel));
1147 : {
1148 2902 : cycleid = _bt_start_vacuum(rel);
1149 :
1150 2902 : btvacuumscan(info, stats, callback, callback_state, cycleid);
1151 : }
1152 2902 : PG_END_ENSURE_ERROR_CLEANUP(_bt_end_vacuum_callback, PointerGetDatum(rel));
1153 2902 : _bt_end_vacuum(rel);
1154 :
1155 2902 : return stats;
1156 : }
1157 :
1158 : /*
1159 : * Post-VACUUM cleanup.
1160 : *
1161 : * Result: a palloc'd struct containing statistical info for VACUUM displays.
1162 : */
1163 : IndexBulkDeleteResult *
1164 261716 : btvacuumcleanup(IndexVacuumInfo *info, IndexBulkDeleteResult *stats)
1165 : {
1166 : BlockNumber num_delpages;
1167 :
1168 : /* No-op in ANALYZE ONLY mode */
1169 261716 : if (info->analyze_only)
1170 18164 : return stats;
1171 :
1172 : /*
1173 : * If btbulkdelete was called, we need not do anything (we just maintain
1174 : * the information used within _bt_vacuum_needs_cleanup() by calling
1175 : * _bt_set_cleanup_info() below).
1176 : *
1177 : * If btbulkdelete was _not_ called, then we have a choice to make: we
1178 : * must decide whether or not a btvacuumscan() call is needed now (i.e.
1179 : * whether the ongoing VACUUM operation can entirely avoid a physical scan
1180 : * of the index). A call to _bt_vacuum_needs_cleanup() decides it for us
1181 : * now.
1182 : */
1183 243552 : if (stats == NULL)
1184 : {
1185 : /* Check if VACUUM operation can entirely avoid btvacuumscan() call */
1186 241270 : if (!_bt_vacuum_needs_cleanup(info->index))
1187 241258 : return NULL;
1188 :
1189 : /*
1190 : * Since we aren't going to actually delete any leaf items, there's no
1191 : * need to go through all the vacuum-cycle-ID pushups here.
1192 : *
1193 : * Posting list tuples are a source of inaccuracy for cleanup-only
1194 : * scans. btvacuumscan() will assume that the number of index tuples
1195 : * from each page can be used as num_index_tuples, even though
1196 : * num_index_tuples is supposed to represent the number of TIDs in the
1197 : * index. This naive approach can underestimate the number of tuples
1198 : * in the index significantly.
1199 : *
1200 : * We handle the problem by making num_index_tuples an estimate in
1201 : * cleanup-only case.
1202 : */
1203 12 : stats = palloc0_object(IndexBulkDeleteResult);
1204 12 : btvacuumscan(info, stats, NULL, NULL, 0);
1205 12 : stats->estimated_count = true;
1206 : }
1207 :
1208 : /*
1209 : * Maintain num_delpages value in metapage for _bt_vacuum_needs_cleanup().
1210 : *
1211 : * num_delpages is the number of deleted pages now in the index that were
1212 : * not safe to place in the FSM to be recycled just yet. num_delpages is
1213 : * greater than 0 only when _bt_pagedel() actually deleted pages during
1214 : * our call to btvacuumscan(). Even then, _bt_pendingfsm_finalize() must
1215 : * have failed to place any newly deleted pages in the FSM just moments
1216 : * ago. (Actually, there are edge cases where recycling of the current
1217 : * VACUUM's newly deleted pages does not even become safe by the time the
1218 : * next VACUUM comes around. See nbtree/README.)
1219 : */
1220 : Assert(stats->pages_deleted >= stats->pages_free);
1221 2294 : num_delpages = stats->pages_deleted - stats->pages_free;
1222 2294 : _bt_set_cleanup_info(info->index, num_delpages);
1223 :
1224 : /*
1225 : * It's quite possible for us to be fooled by concurrent page splits into
1226 : * double-counting some index tuples, so disbelieve any total that exceeds
1227 : * the underlying heap's count ... if we know that accurately. Otherwise
1228 : * this might just make matters worse.
1229 : */
1230 2294 : if (!info->estimated_count)
1231 : {
1232 2234 : if (stats->num_index_tuples > info->num_heap_tuples)
1233 20 : stats->num_index_tuples = info->num_heap_tuples;
1234 : }
1235 :
1236 2294 : return stats;
1237 : }
1238 :
1239 : /*
1240 : * btvacuumscan --- scan the index for VACUUMing purposes
1241 : *
1242 : * This combines the functions of looking for leaf tuples that are deletable
1243 : * according to the vacuum callback, looking for empty pages that can be
1244 : * deleted, and looking for old deleted pages that can be recycled. Both
1245 : * btbulkdelete and btvacuumcleanup invoke this (the latter only if no
1246 : * btbulkdelete call occurred and _bt_vacuum_needs_cleanup returned true).
1247 : *
1248 : * The caller is responsible for initially allocating/zeroing a stats struct
1249 : * and for obtaining a vacuum cycle ID if necessary.
1250 : */
1251 : static void
1252 2914 : btvacuumscan(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
1253 : IndexBulkDeleteCallback callback, void *callback_state,
1254 : BTCycleId cycleid)
1255 : {
1256 2914 : Relation rel = info->index;
1257 : BTVacState vstate;
1258 : BlockNumber num_pages;
1259 : bool needLock;
1260 : BlockRangeReadStreamPrivate p;
1261 2914 : ReadStream *stream = NULL;
1262 :
1263 : /*
1264 : * Reset fields that track information about the entire index now. This
1265 : * avoids double-counting in the case where a single VACUUM command
1266 : * requires multiple scans of the index.
1267 : *
1268 : * Avoid resetting the tuples_removed and pages_newly_deleted fields here,
1269 : * since they track information about the VACUUM command, and so must last
1270 : * across each call to btvacuumscan().
1271 : *
1272 : * (Note that pages_free is treated as state about the whole index, not
1273 : * the current VACUUM. This is appropriate because RecordFreeIndexPage()
1274 : * calls are idempotent, and get repeated for the same deleted pages in
1275 : * some scenarios. The point for us is to track the number of recyclable
1276 : * pages in the index at the end of the VACUUM command.)
1277 : */
1278 2914 : stats->num_pages = 0;
1279 2914 : stats->num_index_tuples = 0;
1280 2914 : stats->pages_deleted = 0;
1281 2914 : stats->pages_free = 0;
1282 :
1283 : /* Set up info to pass down to btvacuumpage */
1284 2914 : vstate.info = info;
1285 2914 : vstate.stats = stats;
1286 2914 : vstate.callback = callback;
1287 2914 : vstate.callback_state = callback_state;
1288 2914 : vstate.cycleid = cycleid;
1289 :
1290 : /* Create a temporary memory context to run _bt_pagedel in */
1291 2914 : vstate.pagedelcontext = AllocSetContextCreate(CurrentMemoryContext,
1292 : "_bt_pagedel",
1293 : ALLOCSET_DEFAULT_SIZES);
1294 :
1295 : /* Initialize vstate fields used by _bt_pendingfsm_finalize */
1296 2914 : vstate.bufsize = 0;
1297 2914 : vstate.maxbufsize = 0;
1298 2914 : vstate.pendingpages = NULL;
1299 2914 : vstate.npendingpages = 0;
1300 : /* Consider applying _bt_pendingfsm_finalize optimization */
1301 2914 : _bt_pendingfsm_init(rel, &vstate, (callback == NULL));
1302 :
1303 : /*
1304 : * The outer loop iterates over all index pages except the metapage, in
1305 : * physical order (we hope the kernel will cooperate in providing
1306 : * read-ahead for speed). It is critical that we visit all leaf pages,
1307 : * including ones added after we start the scan, else we might fail to
1308 : * delete some deletable tuples. Hence, we must repeatedly check the
1309 : * relation length. We must acquire the relation-extension lock while
1310 : * doing so to avoid a race condition: if someone else is extending the
1311 : * relation, there is a window where bufmgr/smgr have created a new
1312 : * all-zero page but it hasn't yet been write-locked by _bt_getbuf(). If
1313 : * we manage to scan such a page here, we'll improperly assume it can be
1314 : * recycled. Taking the lock synchronizes things enough to prevent a
1315 : * problem: either num_pages won't include the new page, or _bt_getbuf
1316 : * already has write lock on the buffer and it will be fully initialized
1317 : * before we can examine it. Also, we need not worry if a page is added
1318 : * immediately after we look; the page splitting code already has
1319 : * write-lock on the left page before it adds a right page, so we must
1320 : * already have processed any tuples due to be moved into such a page.
1321 : *
1322 : * XXX: Now that new pages are locked with RBM_ZERO_AND_LOCK, I don't
1323 : * think the use of the extension lock is still required.
1324 : *
1325 : * We can skip locking for new or temp relations, however, since no one
1326 : * else could be accessing them.
1327 : */
1328 2914 : needLock = !RELATION_IS_LOCAL(rel);
1329 :
1330 2914 : p.current_blocknum = BTREE_METAPAGE + 1;
1331 :
1332 : /*
1333 : * It is safe to use batchmode as block_range_read_stream_cb takes no
1334 : * locks.
1335 : */
1336 2914 : stream = read_stream_begin_relation(READ_STREAM_MAINTENANCE |
1337 : READ_STREAM_FULL |
1338 : READ_STREAM_USE_BATCHING,
1339 : info->strategy,
1340 : rel,
1341 : MAIN_FORKNUM,
1342 : block_range_read_stream_cb,
1343 : &p,
1344 : 0);
1345 : for (;;)
1346 : {
1347 : /* Get the current relation length */
1348 5554 : if (needLock)
1349 5550 : LockRelationForExtension(rel, ExclusiveLock);
1350 5554 : num_pages = RelationGetNumberOfBlocks(rel);
1351 5554 : if (needLock)
1352 5550 : UnlockRelationForExtension(rel, ExclusiveLock);
1353 :
1354 5554 : if (info->report_progress)
1355 958 : pgstat_progress_update_param(PROGRESS_SCAN_BLOCKS_TOTAL,
1356 : num_pages);
1357 :
1358 : /* Quit if we've scanned the whole relation */
1359 5554 : if (p.current_blocknum >= num_pages)
1360 2914 : break;
1361 :
1362 2640 : p.last_exclusive = num_pages;
1363 :
1364 : /* Iterate over pages, then loop back to recheck relation length */
1365 : while (true)
1366 25508 : {
1367 : BlockNumber current_block;
1368 : Buffer buf;
1369 :
1370 : /* call vacuum_delay_point while not holding any buffer lock */
1371 28148 : vacuum_delay_point(false);
1372 :
1373 28148 : buf = read_stream_next_buffer(stream, NULL);
1374 :
1375 28148 : if (!BufferIsValid(buf))
1376 2640 : break;
1377 :
1378 25508 : current_block = btvacuumpage(&vstate, buf);
1379 :
1380 25508 : if (info->report_progress)
1381 984 : pgstat_progress_update_param(PROGRESS_SCAN_BLOCKS_DONE,
1382 : current_block);
1383 : }
1384 :
1385 : /*
1386 : * We have to reset the read stream to use it again. After returning
1387 : * InvalidBuffer, the read stream API won't invoke our callback again
1388 : * until the stream has been reset.
1389 : */
1390 2640 : read_stream_reset(stream);
1391 : }
1392 :
1393 2914 : read_stream_end(stream);
1394 :
1395 : /* Set statistics num_pages field to final size of index */
1396 2914 : stats->num_pages = num_pages;
1397 :
1398 2914 : MemoryContextDelete(vstate.pagedelcontext);
1399 :
1400 : /*
1401 : * If there were any calls to _bt_pagedel() during scan of the index then
1402 : * see if any of the resulting pages can be placed in the FSM now. When
1403 : * it's not safe we'll have to leave it up to a future VACUUM operation.
1404 : *
1405 : * Finally, if we placed any pages in the FSM (either just now or during
1406 : * the scan), forcibly update the upper-level FSM pages to ensure that
1407 : * searchers can find them.
1408 : */
1409 2914 : _bt_pendingfsm_finalize(rel, &vstate);
1410 2914 : if (stats->pages_free > 0)
1411 50 : IndexFreeSpaceMapVacuum(rel);
1412 2914 : }
1413 :
1414 : /*
1415 : * btvacuumpage --- VACUUM one page
1416 : *
1417 : * This processes a single page for btvacuumscan(). In some cases we must
1418 : * backtrack to re-examine and VACUUM pages that were on buf's page during
1419 : * a previous call here. This is how we handle page splits (that happened
1420 : * after our cycleid was acquired) whose right half page happened to reuse
1421 : * a block that we might have processed at some point before it was
1422 : * recycled (i.e. before the page split).
1423 : *
1424 : * Returns BlockNumber of a scanned page (not backtracked).
1425 : */
1426 : static BlockNumber
1427 25508 : btvacuumpage(BTVacState *vstate, Buffer buf)
1428 : {
1429 25508 : IndexVacuumInfo *info = vstate->info;
1430 25508 : IndexBulkDeleteResult *stats = vstate->stats;
1431 25508 : IndexBulkDeleteCallback callback = vstate->callback;
1432 25508 : void *callback_state = vstate->callback_state;
1433 25508 : Relation rel = info->index;
1434 25508 : Relation heaprel = info->heaprel;
1435 : bool attempt_pagedel;
1436 : BlockNumber blkno,
1437 : backtrack_to;
1438 25508 : BlockNumber scanblkno = BufferGetBlockNumber(buf);
1439 : Page page;
1440 : BTPageOpaque opaque;
1441 :
1442 25508 : blkno = scanblkno;
1443 :
1444 25508 : backtrack:
1445 :
1446 25508 : attempt_pagedel = false;
1447 25508 : backtrack_to = P_NONE;
1448 :
1449 25508 : _bt_lockbuf(rel, buf, BT_READ);
1450 25508 : page = BufferGetPage(buf);
1451 25508 : opaque = NULL;
1452 25508 : if (!PageIsNew(page))
1453 : {
1454 25508 : _bt_checkpage(rel, buf);
1455 25508 : opaque = BTPageGetOpaque(page);
1456 : }
1457 :
1458 : Assert(blkno <= scanblkno);
1459 25508 : if (blkno != scanblkno)
1460 : {
1461 : /*
1462 : * We're backtracking.
1463 : *
1464 : * We followed a right link to a sibling leaf page (a page that
1465 : * happens to be from a block located before scanblkno). The only
1466 : * case we want to do anything with is a live leaf page having the
1467 : * current vacuum cycle ID.
1468 : *
1469 : * The page had better be in a state that's consistent with what we
1470 : * expect. Check for conditions that imply corruption in passing. It
1471 : * can't be half-dead because only an interrupted VACUUM process can
1472 : * leave pages in that state, so we'd definitely have dealt with it
1473 : * back when the page was the scanblkno page (half-dead pages are
1474 : * always marked fully deleted by _bt_pagedel(), barring corruption).
1475 : */
1476 0 : if (!opaque || !P_ISLEAF(opaque) || P_ISHALFDEAD(opaque))
1477 : {
1478 : Assert(false);
1479 0 : ereport(LOG,
1480 : (errcode(ERRCODE_INDEX_CORRUPTED),
1481 : errmsg_internal("right sibling %u of scanblkno %u unexpectedly in an inconsistent state in index \"%s\"",
1482 : blkno, scanblkno, RelationGetRelationName(rel))));
1483 0 : _bt_relbuf(rel, buf);
1484 0 : return scanblkno;
1485 : }
1486 :
1487 : /*
1488 : * We may have already processed the page in an earlier call, when the
1489 : * page was scanblkno. This happens when the leaf page split occurred
1490 : * after the scan began, but before the right sibling page became the
1491 : * scanblkno.
1492 : *
1493 : * Page may also have been deleted by current btvacuumpage() call,
1494 : * since _bt_pagedel() sometimes deletes the right sibling page of
1495 : * scanblkno in passing (it does so after we decided where to
1496 : * backtrack to). We don't need to process this page as a deleted
1497 : * page a second time now (in fact, it would be wrong to count it as a
1498 : * deleted page in the bulk delete statistics a second time).
1499 : */
1500 0 : if (opaque->btpo_cycleid != vstate->cycleid || P_ISDELETED(opaque))
1501 : {
1502 : /* Done with current scanblkno (and all lower split pages) */
1503 0 : _bt_relbuf(rel, buf);
1504 0 : return scanblkno;
1505 : }
1506 : }
1507 :
1508 25508 : if (!opaque || BTPageIsRecyclable(page, heaprel))
1509 : {
1510 : /* Okay to recycle this page (which could be leaf or internal) */
1511 278 : RecordFreeIndexPage(rel, blkno);
1512 278 : stats->pages_deleted++;
1513 278 : stats->pages_free++;
1514 : }
1515 25230 : else if (P_ISDELETED(opaque))
1516 : {
1517 : /*
1518 : * Already deleted page (which could be leaf or internal). Can't
1519 : * recycle yet.
1520 : */
1521 138 : stats->pages_deleted++;
1522 : }
1523 25092 : else if (P_ISHALFDEAD(opaque))
1524 : {
1525 : /* Half-dead leaf page (from interrupted VACUUM) -- finish deleting */
1526 0 : attempt_pagedel = true;
1527 :
1528 : /*
1529 : * _bt_pagedel() will increment both pages_newly_deleted and
1530 : * pages_deleted stats in all cases (barring corruption)
1531 : */
1532 : }
1533 25092 : else if (P_ISLEAF(opaque))
1534 : {
1535 : OffsetNumber deletable[MaxIndexTuplesPerPage];
1536 : int ndeletable;
1537 : BTVacuumPosting updatable[MaxIndexTuplesPerPage];
1538 : int nupdatable;
1539 : OffsetNumber offnum,
1540 : minoff,
1541 : maxoff;
1542 : int nhtidsdead,
1543 : nhtidslive;
1544 :
1545 : /*
1546 : * Trade in the initial read lock for a full cleanup lock on this
1547 : * page. We must get such a lock on every leaf page over the course
1548 : * of the vacuum scan, whether or not it actually contains any
1549 : * deletable tuples --- see nbtree/README.
1550 : */
1551 23546 : _bt_upgradelockbufcleanup(rel, buf);
1552 :
1553 : /*
1554 : * Check whether we need to backtrack to earlier pages. What we are
1555 : * concerned about is a page split that happened since we started the
1556 : * vacuum scan. If the split moved tuples on the right half of the
1557 : * split (i.e. the tuples that sort high) to a block that we already
1558 : * passed over, then we might have missed the tuples. We need to
1559 : * backtrack now. (Must do this before possibly clearing btpo_cycleid
1560 : * or deleting scanblkno page below!)
1561 : */
1562 23546 : if (vstate->cycleid != 0 &&
1563 23414 : opaque->btpo_cycleid == vstate->cycleid &&
1564 0 : !(opaque->btpo_flags & BTP_SPLIT_END) &&
1565 0 : !P_RIGHTMOST(opaque) &&
1566 0 : opaque->btpo_next < scanblkno)
1567 0 : backtrack_to = opaque->btpo_next;
1568 :
1569 23546 : ndeletable = 0;
1570 23546 : nupdatable = 0;
1571 23546 : minoff = P_FIRSTDATAKEY(opaque);
1572 23546 : maxoff = PageGetMaxOffsetNumber(page);
1573 23546 : nhtidsdead = 0;
1574 23546 : nhtidslive = 0;
1575 23546 : if (callback)
1576 : {
1577 : /* btbulkdelete callback tells us what to delete (or update) */
1578 23414 : for (offnum = minoff;
1579 4581652 : offnum <= maxoff;
1580 4558238 : offnum = OffsetNumberNext(offnum))
1581 : {
1582 : IndexTuple itup;
1583 :
1584 4558238 : itup = (IndexTuple) PageGetItem(page,
1585 4558238 : PageGetItemId(page, offnum));
1586 :
1587 : Assert(!BTreeTupleIsPivot(itup));
1588 4558238 : if (!BTreeTupleIsPosting(itup))
1589 : {
1590 : /* Regular tuple, standard table TID representation */
1591 4412242 : if (callback(&itup->t_tid, callback_state))
1592 : {
1593 1679956 : deletable[ndeletable++] = offnum;
1594 1679956 : nhtidsdead++;
1595 : }
1596 : else
1597 2732286 : nhtidslive++;
1598 : }
1599 : else
1600 : {
1601 : BTVacuumPosting vacposting;
1602 : int nremaining;
1603 :
1604 : /* Posting list tuple */
1605 145996 : vacposting = btreevacuumposting(vstate, itup, offnum,
1606 : &nremaining);
1607 145996 : if (vacposting == NULL)
1608 : {
1609 : /*
1610 : * All table TIDs from the posting tuple remain, so no
1611 : * delete or update required
1612 : */
1613 : Assert(nremaining == BTreeTupleGetNPosting(itup));
1614 : }
1615 91132 : else if (nremaining > 0)
1616 : {
1617 :
1618 : /*
1619 : * Store metadata about posting list tuple in
1620 : * updatable array for entire page. Existing tuple
1621 : * will be updated during the later call to
1622 : * _bt_delitems_vacuum().
1623 : */
1624 : Assert(nremaining < BTreeTupleGetNPosting(itup));
1625 37688 : updatable[nupdatable++] = vacposting;
1626 37688 : nhtidsdead += BTreeTupleGetNPosting(itup) - nremaining;
1627 : }
1628 : else
1629 : {
1630 : /*
1631 : * All table TIDs from the posting list must be
1632 : * deleted. We'll delete the index tuple completely
1633 : * (no update required).
1634 : */
1635 : Assert(nremaining == 0);
1636 53444 : deletable[ndeletable++] = offnum;
1637 53444 : nhtidsdead += BTreeTupleGetNPosting(itup);
1638 53444 : pfree(vacposting);
1639 : }
1640 :
1641 145996 : nhtidslive += nremaining;
1642 : }
1643 : }
1644 : }
1645 :
1646 : /*
1647 : * Apply any needed deletes or updates. We issue just one
1648 : * _bt_delitems_vacuum() call per page, so as to minimize WAL traffic.
1649 : */
1650 23546 : if (ndeletable > 0 || nupdatable > 0)
1651 : {
1652 : Assert(nhtidsdead >= ndeletable + nupdatable);
1653 15016 : _bt_delitems_vacuum(rel, buf, deletable, ndeletable, updatable,
1654 : nupdatable);
1655 :
1656 15016 : stats->tuples_removed += nhtidsdead;
1657 : /* must recompute maxoff */
1658 15016 : maxoff = PageGetMaxOffsetNumber(page);
1659 :
1660 : /* can't leak memory here */
1661 52704 : for (int i = 0; i < nupdatable; i++)
1662 37688 : pfree(updatable[i]);
1663 : }
1664 : else
1665 : {
1666 : /*
1667 : * If the leaf page has been split during this vacuum cycle, it
1668 : * seems worth expending a write to clear btpo_cycleid even if we
1669 : * don't have any deletions to do. (If we do, _bt_delitems_vacuum
1670 : * takes care of this.) This ensures we won't process the page
1671 : * again.
1672 : *
1673 : * We treat this like a hint-bit update because there's no need to
1674 : * WAL-log it.
1675 : */
1676 : Assert(nhtidsdead == 0);
1677 8530 : if (vstate->cycleid != 0 &&
1678 8398 : opaque->btpo_cycleid == vstate->cycleid)
1679 : {
1680 0 : opaque->btpo_cycleid = 0;
1681 0 : MarkBufferDirtyHint(buf, true);
1682 : }
1683 : }
1684 :
1685 : /*
1686 : * If the leaf page is now empty, try to delete it; else count the
1687 : * live tuples (live table TIDs in posting lists are counted as
1688 : * separate live tuples). We don't delete when backtracking, though,
1689 : * since that would require teaching _bt_pagedel() about backtracking
1690 : * (doesn't seem worth adding more complexity to deal with that).
1691 : *
1692 : * We don't count the number of live TIDs during cleanup-only calls to
1693 : * btvacuumscan (i.e. when callback is not set). We count the number
1694 : * of index tuples directly instead. This avoids the expense of
1695 : * directly examining all of the tuples on each page. VACUUM will
1696 : * treat num_index_tuples as an estimate in cleanup-only case, so it
1697 : * doesn't matter that this underestimates num_index_tuples
1698 : * significantly in some cases.
1699 : */
1700 23546 : if (minoff > maxoff)
1701 5786 : attempt_pagedel = (blkno == scanblkno);
1702 17760 : else if (callback)
1703 17636 : stats->num_index_tuples += nhtidslive;
1704 : else
1705 124 : stats->num_index_tuples += maxoff - minoff + 1;
1706 :
1707 : Assert(!attempt_pagedel || nhtidslive == 0);
1708 : }
1709 :
1710 25508 : if (attempt_pagedel)
1711 : {
1712 : MemoryContext oldcontext;
1713 :
1714 : /* Run pagedel in a temp context to avoid memory leakage */
1715 5786 : MemoryContextReset(vstate->pagedelcontext);
1716 5786 : oldcontext = MemoryContextSwitchTo(vstate->pagedelcontext);
1717 :
1718 : /*
1719 : * _bt_pagedel maintains the bulk delete stats on our behalf;
1720 : * pages_newly_deleted and pages_deleted are likely to be incremented
1721 : * during call
1722 : */
1723 : Assert(blkno == scanblkno);
1724 5786 : _bt_pagedel(rel, buf, vstate);
1725 :
1726 5786 : MemoryContextSwitchTo(oldcontext);
1727 : /* pagedel released buffer, so we shouldn't */
1728 : }
1729 : else
1730 19722 : _bt_relbuf(rel, buf);
1731 :
1732 25508 : if (backtrack_to != P_NONE)
1733 : {
1734 0 : blkno = backtrack_to;
1735 :
1736 : /* check for vacuum delay while not holding any buffer lock */
1737 0 : vacuum_delay_point(false);
1738 :
1739 : /*
1740 : * We can't use _bt_getbuf() here because it always applies
1741 : * _bt_checkpage(), which will barf on an all-zero page. We want to
1742 : * recycle all-zero pages, not fail. Also, we want to use a
1743 : * nondefault buffer access strategy.
1744 : */
1745 0 : buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL,
1746 : info->strategy);
1747 0 : goto backtrack;
1748 : }
1749 :
1750 25508 : return scanblkno;
1751 : }
1752 :
1753 : /*
1754 : * btreevacuumposting --- determine TIDs still needed in posting list
1755 : *
1756 : * Returns metadata describing how to build replacement tuple without the TIDs
1757 : * that VACUUM needs to delete. Returned value is NULL in the common case
1758 : * where no changes are needed to caller's posting list tuple (we avoid
1759 : * allocating memory here as an optimization).
1760 : *
1761 : * The number of TIDs that should remain in the posting list tuple is set for
1762 : * caller in *nremaining.
1763 : */
1764 : static BTVacuumPosting
1765 145996 : btreevacuumposting(BTVacState *vstate, IndexTuple posting,
1766 : OffsetNumber updatedoffset, int *nremaining)
1767 : {
1768 145996 : int live = 0;
1769 145996 : int nitem = BTreeTupleGetNPosting(posting);
1770 145996 : ItemPointer items = BTreeTupleGetPosting(posting);
1771 145996 : BTVacuumPosting vacposting = NULL;
1772 :
1773 824210 : for (int i = 0; i < nitem; i++)
1774 : {
1775 678214 : if (!vstate->callback(items + i, vstate->callback_state))
1776 : {
1777 : /* Live table TID */
1778 362094 : live++;
1779 : }
1780 316120 : else if (vacposting == NULL)
1781 : {
1782 : /*
1783 : * First dead table TID encountered.
1784 : *
1785 : * It's now clear that we need to delete one or more dead table
1786 : * TIDs, so start maintaining metadata describing how to update
1787 : * existing posting list tuple.
1788 : */
1789 91132 : vacposting = palloc(offsetof(BTVacuumPostingData, deletetids) +
1790 : nitem * sizeof(uint16));
1791 :
1792 91132 : vacposting->itup = posting;
1793 91132 : vacposting->updatedoffset = updatedoffset;
1794 91132 : vacposting->ndeletedtids = 0;
1795 91132 : vacposting->deletetids[vacposting->ndeletedtids++] = i;
1796 : }
1797 : else
1798 : {
1799 : /* Second or subsequent dead table TID */
1800 224988 : vacposting->deletetids[vacposting->ndeletedtids++] = i;
1801 : }
1802 : }
1803 :
1804 145996 : *nremaining = live;
1805 145996 : return vacposting;
1806 : }
1807 :
1808 : /*
1809 : * btcanreturn() -- Check whether btree indexes support index-only scans.
1810 : *
1811 : * btrees always do, so this is trivial.
1812 : */
1813 : bool
1814 1141680 : btcanreturn(Relation index, int attno)
1815 : {
1816 1141680 : return true;
1817 : }
1818 :
1819 : /*
1820 : * btgettreeheight() -- Compute tree height for use by btcostestimate().
1821 : */
1822 : int
1823 731924 : btgettreeheight(Relation rel)
1824 : {
1825 731924 : return _bt_getrootheight(rel);
1826 : }
1827 :
1828 : CompareType
1829 0 : bttranslatestrategy(StrategyNumber strategy, Oid opfamily)
1830 : {
1831 0 : switch (strategy)
1832 : {
1833 0 : case BTLessStrategyNumber:
1834 0 : return COMPARE_LT;
1835 0 : case BTLessEqualStrategyNumber:
1836 0 : return COMPARE_LE;
1837 0 : case BTEqualStrategyNumber:
1838 0 : return COMPARE_EQ;
1839 0 : case BTGreaterEqualStrategyNumber:
1840 0 : return COMPARE_GE;
1841 0 : case BTGreaterStrategyNumber:
1842 0 : return COMPARE_GT;
1843 0 : default:
1844 0 : return COMPARE_INVALID;
1845 : }
1846 : }
1847 :
1848 : StrategyNumber
1849 0 : bttranslatecmptype(CompareType cmptype, Oid opfamily)
1850 : {
1851 0 : switch (cmptype)
1852 : {
1853 0 : case COMPARE_LT:
1854 0 : return BTLessStrategyNumber;
1855 0 : case COMPARE_LE:
1856 0 : return BTLessEqualStrategyNumber;
1857 0 : case COMPARE_EQ:
1858 0 : return BTEqualStrategyNumber;
1859 0 : case COMPARE_GE:
1860 0 : return BTGreaterEqualStrategyNumber;
1861 0 : case COMPARE_GT:
1862 0 : return BTGreaterStrategyNumber;
1863 0 : default:
1864 0 : return InvalidStrategy;
1865 : }
1866 : }
|
