Branch data Line data Source code
1 : : /*----------------------------------------------------------------------
2 : : *
3 : : * tableam.c
4 : : * Table access method routines too big to be inline functions.
5 : : *
6 : : * Portions Copyright (c) 1996-2026, PostgreSQL Global Development Group
7 : : * Portions Copyright (c) 1994, Regents of the University of California
8 : : *
9 : : *
10 : : * IDENTIFICATION
11 : : * src/backend/access/table/tableam.c
12 : : *
13 : : * NOTES
14 : : * Note that most functions in here are documented in tableam.h, rather than
15 : : * here. That's because there's a lot of inline functions in tableam.h and
16 : : * it'd be harder to understand if one constantly had to switch between files.
17 : : *
18 : : *----------------------------------------------------------------------
19 : : */
20 : : #include "postgres.h"
21 : :
22 : : #include <math.h>
23 : :
24 : : #include "access/syncscan.h"
25 : : #include "access/tableam.h"
26 : : #include "access/xact.h"
27 : : #include "optimizer/optimizer.h"
28 : : #include "optimizer/plancat.h"
29 : : #include "port/pg_bitutils.h"
30 : : #include "storage/bufmgr.h"
31 : : #include "storage/shmem.h"
32 : : #include "storage/smgr.h"
33 : :
34 : : /*
35 : : * Constants to control the behavior of block allocation to parallel workers
36 : : * during a parallel seqscan. Technically these values do not need to be
37 : : * powers of 2, but having them as powers of 2 makes the math more optimal
38 : : * and makes the ramp-down stepping more even.
39 : : */
40 : :
41 : : /* The number of I/O chunks we try to break a parallel seqscan down into */
42 : : #define PARALLEL_SEQSCAN_NCHUNKS 2048
43 : : /* Ramp down size of allocations when we've only this number of chunks left */
44 : : #define PARALLEL_SEQSCAN_RAMPDOWN_CHUNKS 64
45 : : /* Cap the size of parallel I/O chunks to this number of blocks */
46 : : #define PARALLEL_SEQSCAN_MAX_CHUNK_SIZE 8192
47 : :
48 : : /* GUC variables */
49 : : char *default_table_access_method = DEFAULT_TABLE_ACCESS_METHOD;
50 : : bool synchronize_seqscans = true;
51 : :
52 : :
53 : : /* ----------------------------------------------------------------------------
54 : : * Slot functions.
55 : : * ----------------------------------------------------------------------------
56 : : */
57 : :
58 : : const TupleTableSlotOps *
59 : 3621620 : table_slot_callbacks(Relation relation)
60 : : {
61 : 3621620 : const TupleTableSlotOps *tts_cb;
62 : :
63 [ + + ]: 3621620 : if (relation->rd_tableam)
64 : 3620701 : tts_cb = relation->rd_tableam->slot_callbacks(relation);
65 [ - + ]: 919 : else if (relation->rd_rel->relkind == RELKIND_FOREIGN_TABLE)
66 : : {
67 : : /*
68 : : * Historically FDWs expect to store heap tuples in slots. Continue
69 : : * handing them one, to make it less painful to adapt FDWs to new
70 : : * versions. The cost of a heap slot over a virtual slot is pretty
71 : : * small.
72 : : */
73 : 0 : tts_cb = &TTSOpsHeapTuple;
74 : 0 : }
75 : : else
76 : : {
77 : : /*
78 : : * These need to be supported, as some parts of the code (like COPY)
79 : : * need to create slots for such relations too. It seems better to
80 : : * centralize the knowledge that a heap slot is the right thing in
81 : : * that case here.
82 : : */
83 [ + + + - ]: 919 : Assert(relation->rd_rel->relkind == RELKIND_VIEW ||
84 : : relation->rd_rel->relkind == RELKIND_PARTITIONED_TABLE);
85 : 919 : tts_cb = &TTSOpsVirtual;
86 : : }
87 : :
88 : 7243240 : return tts_cb;
89 : 3621620 : }
90 : :
91 : : TupleTableSlot *
92 : 3172484 : table_slot_create(Relation relation, List **reglist)
93 : : {
94 : 3172484 : const TupleTableSlotOps *tts_cb;
95 : 3172484 : TupleTableSlot *slot;
96 : :
97 : 3172484 : tts_cb = table_slot_callbacks(relation);
98 : 3172484 : slot = MakeSingleTupleTableSlot(RelationGetDescr(relation), tts_cb);
99 : :
100 [ + + ]: 3172484 : if (reglist)
101 : 410533 : *reglist = lappend(*reglist, slot);
102 : :
103 : 6344968 : return slot;
104 : 3172484 : }
105 : :
106 : :
107 : : /* ----------------------------------------------------------------------------
108 : : * Table scan functions.
109 : : * ----------------------------------------------------------------------------
110 : : */
111 : :
112 : : TableScanDesc
113 : 6127 : table_beginscan_catalog(Relation relation, int nkeys, ScanKeyData *key)
114 : : {
115 : 6127 : uint32 flags = SO_TYPE_SEQSCAN |
116 : : SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE | SO_TEMP_SNAPSHOT;
117 : 6127 : Oid relid = RelationGetRelid(relation);
118 : 6127 : Snapshot snapshot = RegisterSnapshot(GetCatalogSnapshot(relid));
119 : :
120 : 18381 : return relation->rd_tableam->scan_begin(relation, snapshot, nkeys, key,
121 : 6127 : NULL, flags);
122 : 6127 : }
123 : :
124 : :
125 : : /* ----------------------------------------------------------------------------
126 : : * Parallel table scan related functions.
127 : : * ----------------------------------------------------------------------------
128 : : */
129 : :
130 : : Size
131 : 183 : table_parallelscan_estimate(Relation rel, Snapshot snapshot)
132 : : {
133 : 183 : Size sz = 0;
134 : :
135 [ + + - + ]: 183 : if (IsMVCCSnapshot(snapshot))
136 : 155 : sz = add_size(sz, EstimateSnapshotSpace(snapshot));
137 : : else
138 [ + - ]: 28 : Assert(snapshot == SnapshotAny);
139 : :
140 : 183 : sz = add_size(sz, rel->rd_tableam->parallelscan_estimate(rel));
141 : :
142 : 366 : return sz;
143 : 183 : }
144 : :
145 : : void
146 : 183 : table_parallelscan_initialize(Relation rel, ParallelTableScanDesc pscan,
147 : : Snapshot snapshot)
148 : : {
149 : 183 : Size snapshot_off = rel->rd_tableam->parallelscan_initialize(rel, pscan);
150 : :
151 : 183 : pscan->phs_snapshot_off = snapshot_off;
152 : :
153 [ + + - + ]: 183 : if (IsMVCCSnapshot(snapshot))
154 : : {
155 : 155 : SerializeSnapshot(snapshot, (char *) pscan + pscan->phs_snapshot_off);
156 : 155 : pscan->phs_snapshot_any = false;
157 : 155 : }
158 : : else
159 : : {
160 [ + - ]: 28 : Assert(snapshot == SnapshotAny);
161 : 28 : pscan->phs_snapshot_any = true;
162 : : }
163 : 183 : }
164 : :
165 : : TableScanDesc
166 : 659 : table_beginscan_parallel(Relation relation, ParallelTableScanDesc pscan)
167 : : {
168 : 659 : Snapshot snapshot;
169 : 659 : uint32 flags = SO_TYPE_SEQSCAN |
170 : : SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
171 : :
172 [ + - ]: 659 : Assert(RelFileLocatorEquals(relation->rd_locator, pscan->phs_locator));
173 : :
174 [ + + ]: 659 : if (!pscan->phs_snapshot_any)
175 : : {
176 : : /* Snapshot was serialized -- restore it */
177 : 603 : snapshot = RestoreSnapshot((char *) pscan + pscan->phs_snapshot_off);
178 : 603 : RegisterSnapshot(snapshot);
179 : 603 : flags |= SO_TEMP_SNAPSHOT;
180 : 603 : }
181 : : else
182 : : {
183 : : /* SnapshotAny passed by caller (not serialized) */
184 : 56 : snapshot = SnapshotAny;
185 : : }
186 : :
187 : 1977 : return relation->rd_tableam->scan_begin(relation, snapshot, 0, NULL,
188 : 659 : pscan, flags);
189 : 659 : }
190 : :
191 : : TableScanDesc
192 : 20 : table_beginscan_parallel_tidrange(Relation relation,
193 : : ParallelTableScanDesc pscan)
194 : : {
195 : 20 : Snapshot snapshot;
196 : 20 : uint32 flags = SO_TYPE_TIDRANGESCAN | SO_ALLOW_PAGEMODE;
197 : 20 : TableScanDesc sscan;
198 : :
199 [ + - ]: 20 : Assert(RelFileLocatorEquals(relation->rd_locator, pscan->phs_locator));
200 : :
201 : : /* disable syncscan in parallel tid range scan. */
202 : 20 : pscan->phs_syncscan = false;
203 : :
204 [ - + ]: 20 : if (!pscan->phs_snapshot_any)
205 : : {
206 : : /* Snapshot was serialized -- restore it */
207 : 20 : snapshot = RestoreSnapshot((char *) pscan + pscan->phs_snapshot_off);
208 : 20 : RegisterSnapshot(snapshot);
209 : 20 : flags |= SO_TEMP_SNAPSHOT;
210 : 20 : }
211 : : else
212 : : {
213 : : /* SnapshotAny passed by caller (not serialized) */
214 : 0 : snapshot = SnapshotAny;
215 : : }
216 : :
217 : 40 : sscan = relation->rd_tableam->scan_begin(relation, snapshot, 0, NULL,
218 : 20 : pscan, flags);
219 : 40 : return sscan;
220 : 20 : }
221 : :
222 : :
223 : : /* ----------------------------------------------------------------------------
224 : : * Index scan related functions.
225 : : * ----------------------------------------------------------------------------
226 : : */
227 : :
228 : : /*
229 : : * To perform that check simply start an index scan, create the necessary
230 : : * slot, do the heap lookup, and shut everything down again. This could be
231 : : * optimized, but is unlikely to matter from a performance POV. If there
232 : : * frequently are live index pointers also matching a unique index key, the
233 : : * CPU overhead of this routine is unlikely to matter.
234 : : *
235 : : * Note that *tid may be modified when we return true if the AM supports
236 : : * storing multiple row versions reachable via a single index entry (like
237 : : * heap's HOT).
238 : : */
239 : : bool
240 : 1847860 : table_index_fetch_tuple_check(Relation rel,
241 : : ItemPointer tid,
242 : : Snapshot snapshot,
243 : : bool *all_dead)
244 : : {
245 : 1847860 : IndexFetchTableData *scan;
246 : 1847860 : TupleTableSlot *slot;
247 : 1847860 : bool call_again = false;
248 : 1847860 : bool found;
249 : :
250 : 1847860 : slot = table_slot_create(rel, NULL);
251 : 1847860 : scan = table_index_fetch_begin(rel);
252 : 3695720 : found = table_index_fetch_tuple(scan, tid, snapshot, slot, &call_again,
253 : 1847860 : all_dead);
254 : 1847860 : table_index_fetch_end(scan);
255 : 1847860 : ExecDropSingleTupleTableSlot(slot);
256 : :
257 : 3695720 : return found;
258 : 1847860 : }
259 : :
260 : :
261 : : /* ------------------------------------------------------------------------
262 : : * Functions for non-modifying operations on individual tuples
263 : : * ------------------------------------------------------------------------
264 : : */
265 : :
266 : : void
267 : 51 : table_tuple_get_latest_tid(TableScanDesc scan, ItemPointer tid)
268 : : {
269 : 51 : Relation rel = scan->rs_rd;
270 : 51 : const TableAmRoutine *tableam = rel->rd_tableam;
271 : :
272 : : /*
273 : : * We don't expect direct calls to table_tuple_get_latest_tid with valid
274 : : * CheckXidAlive for catalog or regular tables. See detailed comments in
275 : : * xact.c where these variables are declared.
276 : : */
277 [ + - + - ]: 51 : if (unlikely(TransactionIdIsValid(CheckXidAlive) && !bsysscan))
278 [ # # # # ]: 0 : elog(ERROR, "unexpected table_tuple_get_latest_tid call during logical decoding");
279 : :
280 : : /*
281 : : * Since this can be called with user-supplied TID, don't trust the input
282 : : * too much.
283 : : */
284 [ + + ]: 51 : if (!tableam->tuple_tid_valid(scan, tid))
285 [ + - + - ]: 2 : ereport(ERROR,
286 : : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
287 : : errmsg("tid (%u, %u) is not valid for relation \"%s\"",
288 : : ItemPointerGetBlockNumberNoCheck(tid),
289 : : ItemPointerGetOffsetNumberNoCheck(tid),
290 : : RelationGetRelationName(rel))));
291 : :
292 : 49 : tableam->tuple_get_latest_tid(scan, tid);
293 : 49 : }
294 : :
295 : :
296 : : /* ----------------------------------------------------------------------------
297 : : * Functions to make modifications a bit simpler.
298 : : * ----------------------------------------------------------------------------
299 : : */
300 : :
301 : : /*
302 : : * simple_table_tuple_insert - insert a tuple
303 : : *
304 : : * Currently, this routine differs from table_tuple_insert only in supplying a
305 : : * default command ID and not allowing access to the speedup options.
306 : : */
307 : : void
308 : 0 : simple_table_tuple_insert(Relation rel, TupleTableSlot *slot)
309 : : {
310 : 0 : table_tuple_insert(rel, slot, GetCurrentCommandId(true), 0, NULL);
311 : 0 : }
312 : :
313 : : /*
314 : : * simple_table_tuple_delete - delete a tuple
315 : : *
316 : : * This routine may be used to delete a tuple when concurrent updates of
317 : : * the target tuple are not expected (for example, because we have a lock
318 : : * on the relation associated with the tuple). Any failure is reported
319 : : * via ereport().
320 : : */
321 : : void
322 : 0 : simple_table_tuple_delete(Relation rel, ItemPointer tid, Snapshot snapshot)
323 : : {
324 : 0 : TM_Result result;
325 : 0 : TM_FailureData tmfd;
326 : :
327 : 0 : result = table_tuple_delete(rel, tid,
328 : 0 : GetCurrentCommandId(true),
329 : 0 : snapshot, InvalidSnapshot,
330 : : true /* wait for commit */ ,
331 : : &tmfd, false /* changingPart */ );
332 : :
333 [ # # # # : 0 : switch (result)
# ]
334 : : {
335 : : case TM_SelfModified:
336 : : /* Tuple was already updated in current command? */
337 [ # # # # ]: 0 : elog(ERROR, "tuple already updated by self");
338 : 0 : break;
339 : :
340 : : case TM_Ok:
341 : : /* done successfully */
342 : : break;
343 : :
344 : : case TM_Updated:
345 [ # # # # ]: 0 : elog(ERROR, "tuple concurrently updated");
346 : 0 : break;
347 : :
348 : : case TM_Deleted:
349 [ # # # # ]: 0 : elog(ERROR, "tuple concurrently deleted");
350 : 0 : break;
351 : :
352 : : default:
353 [ # # # # ]: 0 : elog(ERROR, "unrecognized table_tuple_delete status: %u", result);
354 : 0 : break;
355 : : }
356 : 0 : }
357 : :
358 : : /*
359 : : * simple_table_tuple_update - replace a tuple
360 : : *
361 : : * This routine may be used to update a tuple when concurrent updates of
362 : : * the target tuple are not expected (for example, because we have a lock
363 : : * on the relation associated with the tuple). Any failure is reported
364 : : * via ereport().
365 : : */
366 : : void
367 : 0 : simple_table_tuple_update(Relation rel, ItemPointer otid,
368 : : TupleTableSlot *slot,
369 : : Snapshot snapshot,
370 : : TU_UpdateIndexes *update_indexes)
371 : : {
372 : 0 : TM_Result result;
373 : 0 : TM_FailureData tmfd;
374 : 0 : LockTupleMode lockmode;
375 : :
376 : 0 : result = table_tuple_update(rel, otid, slot,
377 : 0 : GetCurrentCommandId(true),
378 : 0 : snapshot, InvalidSnapshot,
379 : : true /* wait for commit */ ,
380 : 0 : &tmfd, &lockmode, update_indexes);
381 : :
382 [ # # # # : 0 : switch (result)
# ]
383 : : {
384 : : case TM_SelfModified:
385 : : /* Tuple was already updated in current command? */
386 [ # # # # ]: 0 : elog(ERROR, "tuple already updated by self");
387 : 0 : break;
388 : :
389 : : case TM_Ok:
390 : : /* done successfully */
391 : : break;
392 : :
393 : : case TM_Updated:
394 [ # # # # ]: 0 : elog(ERROR, "tuple concurrently updated");
395 : 0 : break;
396 : :
397 : : case TM_Deleted:
398 [ # # # # ]: 0 : elog(ERROR, "tuple concurrently deleted");
399 : 0 : break;
400 : :
401 : : default:
402 [ # # # # ]: 0 : elog(ERROR, "unrecognized table_tuple_update status: %u", result);
403 : 0 : break;
404 : : }
405 : 0 : }
406 : :
407 : :
408 : : /* ----------------------------------------------------------------------------
409 : : * Helper functions to implement parallel scans for block oriented AMs.
410 : : * ----------------------------------------------------------------------------
411 : : */
412 : :
413 : : Size
414 : 183 : table_block_parallelscan_estimate(Relation rel)
415 : : {
416 : 183 : return sizeof(ParallelBlockTableScanDescData);
417 : : }
418 : :
419 : : Size
420 : 183 : table_block_parallelscan_initialize(Relation rel, ParallelTableScanDesc pscan)
421 : : {
422 : 183 : ParallelBlockTableScanDesc bpscan = (ParallelBlockTableScanDesc) pscan;
423 : :
424 : 183 : bpscan->base.phs_locator = rel->rd_locator;
425 : 183 : bpscan->phs_nblocks = RelationGetNumberOfBlocks(rel);
426 : : /* compare phs_syncscan initialization to similar logic in initscan */
427 [ + - ]: 366 : bpscan->base.phs_syncscan = synchronize_seqscans &&
428 [ - + ]: 183 : !RelationUsesLocalBuffers(rel) &&
429 : 183 : bpscan->phs_nblocks > NBuffers / 4;
430 : 183 : SpinLockInit(&bpscan->phs_mutex);
431 : 183 : bpscan->phs_startblock = InvalidBlockNumber;
432 : 183 : bpscan->phs_numblock = InvalidBlockNumber;
433 : 183 : pg_atomic_init_u64(&bpscan->phs_nallocated, 0);
434 : :
435 : 183 : return sizeof(ParallelBlockTableScanDescData);
436 : 183 : }
437 : :
438 : : void
439 : 38 : table_block_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
440 : : {
441 : 38 : ParallelBlockTableScanDesc bpscan = (ParallelBlockTableScanDesc) pscan;
442 : :
443 : 38 : pg_atomic_write_u64(&bpscan->phs_nallocated, 0);
444 : 38 : }
445 : :
446 : : /*
447 : : * find and set the scan's startblock
448 : : *
449 : : * Determine where the parallel seq scan should start. This function may be
450 : : * called many times, once by each parallel worker. We must be careful only
451 : : * to set the phs_startblock and phs_numblock fields once.
452 : : *
453 : : * Callers may optionally specify a non-InvalidBlockNumber value for
454 : : * 'startblock' to force the scan to start at the given page. Likewise,
455 : : * 'numblocks' can be specified as a non-InvalidBlockNumber to limit the
456 : : * number of blocks to scan to that many blocks.
457 : : */
458 : : void
459 : 519 : table_block_parallelscan_startblock_init(Relation rel,
460 : : ParallelBlockTableScanWorker pbscanwork,
461 : : ParallelBlockTableScanDesc pbscan,
462 : : BlockNumber startblock,
463 : : BlockNumber numblocks)
464 : : {
465 : : StaticAssertDecl(MaxBlockNumber <= 0xFFFFFFFE,
466 : : "pg_nextpower2_32 may be too small for non-standard BlockNumber width");
467 : :
468 : 519 : BlockNumber sync_startpage = InvalidBlockNumber;
469 : 519 : BlockNumber scan_nblocks;
470 : :
471 : : /* Reset the state we use for controlling allocation size. */
472 : 519 : memset(pbscanwork, 0, sizeof(*pbscanwork));
473 : :
474 : : retry:
475 : : /* Grab the spinlock. */
476 [ - + ]: 519 : SpinLockAcquire(&pbscan->phs_mutex);
477 : :
478 : : /*
479 : : * When the caller specified a limit on the number of blocks to scan, set
480 : : * that in the ParallelBlockTableScanDesc, if it's not been done by
481 : : * another worker already.
482 : : */
483 [ + + + + ]: 519 : if (numblocks != InvalidBlockNumber &&
484 : 20 : pbscan->phs_numblock == InvalidBlockNumber)
485 : : {
486 : 4 : pbscan->phs_numblock = numblocks;
487 : 4 : }
488 : :
489 : : /*
490 : : * If the scan's phs_startblock has not yet been initialized, we must do
491 : : * so now. If a startblock was specified, start there, otherwise if this
492 : : * is not a synchronized scan, we just start at block 0, but if it is a
493 : : * synchronized scan, we must get the starting position from the
494 : : * synchronized scan machinery. We can't hold the spinlock while doing
495 : : * that, though, so release the spinlock, get the information we need, and
496 : : * retry. If nobody else has initialized the scan in the meantime, we'll
497 : : * fill in the value we fetched on the second time through.
498 : : */
499 [ + + ]: 519 : if (pbscan->phs_startblock == InvalidBlockNumber)
500 : : {
501 [ + + ]: 179 : if (startblock != InvalidBlockNumber)
502 : 4 : pbscan->phs_startblock = startblock;
503 [ + - ]: 175 : else if (!pbscan->base.phs_syncscan)
504 : 175 : pbscan->phs_startblock = 0;
505 [ # # ]: 0 : else if (sync_startpage != InvalidBlockNumber)
506 : 0 : pbscan->phs_startblock = sync_startpage;
507 : : else
508 : : {
509 : 0 : SpinLockRelease(&pbscan->phs_mutex);
510 : 0 : sync_startpage = ss_get_location(rel, pbscan->phs_nblocks);
511 : 0 : goto retry;
512 : : }
513 : 179 : }
514 : 519 : SpinLockRelease(&pbscan->phs_mutex);
515 : :
516 : : /*
517 : : * Figure out how many blocks we're going to scan; either all of them, or
518 : : * just phs_numblock's worth, if a limit has been imposed.
519 : : */
520 [ + + ]: 519 : if (pbscan->phs_numblock == InvalidBlockNumber)
521 : 499 : scan_nblocks = pbscan->phs_nblocks;
522 : : else
523 : 20 : scan_nblocks = pbscan->phs_numblock;
524 : :
525 : : /*
526 : : * We determine the chunk size based on scan_nblocks. First we split
527 : : * scan_nblocks into PARALLEL_SEQSCAN_NCHUNKS chunks then we calculate the
528 : : * next highest power of 2 number of the result. This means we split the
529 : : * blocks we're scanning into somewhere between PARALLEL_SEQSCAN_NCHUNKS
530 : : * and PARALLEL_SEQSCAN_NCHUNKS / 2 chunks.
531 : : */
532 [ - + ]: 519 : pbscanwork->phsw_chunk_size = pg_nextpower2_32(Max(scan_nblocks /
533 : : PARALLEL_SEQSCAN_NCHUNKS, 1));
534 : :
535 : : /*
536 : : * Ensure we don't go over the maximum chunk size with larger tables. This
537 : : * means we may get much more than PARALLEL_SEQSCAN_NCHUNKS for larger
538 : : * tables. Too large a chunk size has been shown to be detrimental to
539 : : * sequential scan performance.
540 : : */
541 [ + - ]: 519 : pbscanwork->phsw_chunk_size = Min(pbscanwork->phsw_chunk_size,
542 : : PARALLEL_SEQSCAN_MAX_CHUNK_SIZE);
543 : 519 : }
544 : :
545 : : /*
546 : : * get the next page to scan
547 : : *
548 : : * Get the next page to scan. Even if there are no pages left to scan,
549 : : * another backend could have grabbed a page to scan and not yet finished
550 : : * looking at it, so it doesn't follow that the scan is done when the first
551 : : * backend gets an InvalidBlockNumber return.
552 : : */
553 : : BlockNumber
554 : 30515 : table_block_parallelscan_nextpage(Relation rel,
555 : : ParallelBlockTableScanWorker pbscanwork,
556 : : ParallelBlockTableScanDesc pbscan)
557 : : {
558 : 30515 : BlockNumber scan_nblocks;
559 : 30515 : BlockNumber page;
560 : 30515 : uint64 nallocated;
561 : :
562 : : /*
563 : : * The logic below allocates block numbers out to parallel workers in a
564 : : * way that each worker will receive a set of consecutive block numbers to
565 : : * scan. Earlier versions of this would allocate the next highest block
566 : : * number to the next worker to call this function. This would generally
567 : : * result in workers never receiving consecutive block numbers. Some
568 : : * operating systems would not detect the sequential I/O pattern due to
569 : : * each backend being a different process which could result in poor
570 : : * performance due to inefficient or no readahead. To work around this
571 : : * issue, we now allocate a range of block numbers for each worker and
572 : : * when they come back for another block, we give them the next one in
573 : : * that range until the range is complete. When the worker completes the
574 : : * range of blocks we then allocate another range for it and return the
575 : : * first block number from that range.
576 : : *
577 : : * Here we name these ranges of blocks "chunks". The initial size of
578 : : * these chunks is determined in table_block_parallelscan_startblock_init
579 : : * based on the number of blocks to scan. Towards the end of the scan, we
580 : : * start making reductions in the size of the chunks in order to attempt
581 : : * to divide the remaining work over all the workers as evenly as
582 : : * possible.
583 : : *
584 : : * Here pbscanwork is local worker memory. phsw_chunk_remaining tracks
585 : : * the number of blocks remaining in the chunk. When that reaches 0 then
586 : : * we must allocate a new chunk for the worker.
587 : : *
588 : : * phs_nallocated tracks how many blocks have been allocated to workers
589 : : * already. When phs_nallocated >= rs_nblocks, all blocks have been
590 : : * allocated.
591 : : *
592 : : * Because we use an atomic fetch-and-add to fetch the current value, the
593 : : * phs_nallocated counter will exceed rs_nblocks, because workers will
594 : : * still increment the value, when they try to allocate the next block but
595 : : * all blocks have been allocated already. The counter must be 64 bits
596 : : * wide because of that, to avoid wrapping around when scan_nblocks is
597 : : * close to 2^32.
598 : : *
599 : : * The actual block to return is calculated by adding the counter to the
600 : : * starting block number, modulo phs_nblocks.
601 : : */
602 : :
603 : : /* First, figure out how many blocks we're planning on scanning */
604 [ + + ]: 30515 : if (pbscan->phs_numblock == InvalidBlockNumber)
605 : 30412 : scan_nblocks = pbscan->phs_nblocks;
606 : : else
607 : 103 : scan_nblocks = pbscan->phs_numblock;
608 : :
609 : : /*
610 : : * Now check if we have any remaining blocks in a previous chunk for this
611 : : * worker. We must consume all of the blocks from that before we allocate
612 : : * a new chunk to the worker.
613 : : */
614 [ - + ]: 30515 : if (pbscanwork->phsw_chunk_remaining > 0)
615 : : {
616 : : /*
617 : : * Give them the next block in the range and update the remaining
618 : : * number of blocks.
619 : : */
620 : 0 : nallocated = ++pbscanwork->phsw_nallocated;
621 : 0 : pbscanwork->phsw_chunk_remaining--;
622 : 0 : }
623 : : else
624 : : {
625 : : /*
626 : : * When we've only got PARALLEL_SEQSCAN_RAMPDOWN_CHUNKS chunks
627 : : * remaining in the scan, we half the chunk size. Since we reduce the
628 : : * chunk size here, we'll hit this again after doing
629 : : * PARALLEL_SEQSCAN_RAMPDOWN_CHUNKS at the new size. After a few
630 : : * iterations of this, we'll end up doing the last few blocks with the
631 : : * chunk size set to 1.
632 : : */
633 [ - + # # ]: 30515 : if (pbscanwork->phsw_chunk_size > 1 &&
634 : 0 : pbscanwork->phsw_nallocated > scan_nblocks -
635 : 0 : (pbscanwork->phsw_chunk_size * PARALLEL_SEQSCAN_RAMPDOWN_CHUNKS))
636 : 0 : pbscanwork->phsw_chunk_size >>= 1;
637 : :
638 : 30515 : nallocated = pbscanwork->phsw_nallocated =
639 : 61030 : pg_atomic_fetch_add_u64(&pbscan->phs_nallocated,
640 : 30515 : pbscanwork->phsw_chunk_size);
641 : :
642 : : /*
643 : : * Set the remaining number of blocks in this chunk so that subsequent
644 : : * calls from this worker continue on with this chunk until it's done.
645 : : */
646 : 30515 : pbscanwork->phsw_chunk_remaining = pbscanwork->phsw_chunk_size - 1;
647 : : }
648 : :
649 : : /* Check if we've run out of blocks to scan */
650 [ + + ]: 30515 : if (nallocated >= scan_nblocks)
651 : 519 : page = InvalidBlockNumber; /* all blocks have been allocated */
652 : : else
653 : 29996 : page = (nallocated + pbscan->phs_startblock) % pbscan->phs_nblocks;
654 : :
655 : : /*
656 : : * Report scan location. Normally, we report the current page number.
657 : : * When we reach the end of the scan, though, we report the starting page,
658 : : * not the ending page, just so the starting positions for later scans
659 : : * doesn't slew backwards. We only report the position at the end of the
660 : : * scan once, though: subsequent callers will report nothing.
661 : : */
662 [ + - ]: 30515 : if (pbscan->base.phs_syncscan)
663 : : {
664 [ # # ]: 0 : if (page != InvalidBlockNumber)
665 : 0 : ss_report_location(rel, page);
666 [ # # ]: 0 : else if (nallocated == pbscan->phs_nblocks)
667 : 0 : ss_report_location(rel, pbscan->phs_startblock);
668 : 0 : }
669 : :
670 : 61030 : return page;
671 : 30515 : }
672 : :
673 : : /* ----------------------------------------------------------------------------
674 : : * Helper functions to implement relation sizing for block oriented AMs.
675 : : * ----------------------------------------------------------------------------
676 : : */
677 : :
678 : : /*
679 : : * table_block_relation_size
680 : : *
681 : : * If a table AM uses the various relation forks as the sole place where data
682 : : * is stored, and if it uses them in the expected manner (e.g. the actual data
683 : : * is in the main fork rather than some other), it can use this implementation
684 : : * of the relation_size callback rather than implementing its own.
685 : : */
686 : : uint64
687 : 405815 : table_block_relation_size(Relation rel, ForkNumber forkNumber)
688 : : {
689 : 405815 : uint64 nblocks = 0;
690 : :
691 : : /* InvalidForkNumber indicates returning the size for all forks */
692 [ - + ]: 405815 : if (forkNumber == InvalidForkNumber)
693 : : {
694 [ # # ]: 0 : for (int i = 0; i < MAX_FORKNUM; i++)
695 : 0 : nblocks += smgrnblocks(RelationGetSmgr(rel), i);
696 : 0 : }
697 : : else
698 : 405815 : nblocks = smgrnblocks(RelationGetSmgr(rel), forkNumber);
699 : :
700 : 811630 : return nblocks * BLCKSZ;
701 : 405815 : }
702 : :
703 : : /*
704 : : * table_block_relation_estimate_size
705 : : *
706 : : * This function can't be directly used as the implementation of the
707 : : * relation_estimate_size callback, because it has a few additional parameters.
708 : : * Instead, it is intended to be used as a helper function; the caller can
709 : : * pass through the arguments to its relation_estimate_size function plus the
710 : : * additional values required here.
711 : : *
712 : : * overhead_bytes_per_tuple should contain the approximate number of bytes
713 : : * of storage required to store a tuple above and beyond what is required for
714 : : * the tuple data proper. Typically, this would include things like the
715 : : * size of the tuple header and item pointer. This is only used for query
716 : : * planning, so a table AM where the value is not constant could choose to
717 : : * pass a "best guess".
718 : : *
719 : : * usable_bytes_per_page should contain the approximate number of bytes per
720 : : * page usable for tuple data, excluding the page header and any anticipated
721 : : * special space.
722 : : */
723 : : void
724 : 47801 : table_block_relation_estimate_size(Relation rel, int32 *attr_widths,
725 : : BlockNumber *pages, double *tuples,
726 : : double *allvisfrac,
727 : : Size overhead_bytes_per_tuple,
728 : : Size usable_bytes_per_page)
729 : : {
730 : 47801 : BlockNumber curpages;
731 : 47801 : BlockNumber relpages;
732 : 47801 : double reltuples;
733 : 47801 : BlockNumber relallvisible;
734 : 47801 : double density;
735 : :
736 : : /* it should have storage, so we can call the smgr */
737 : 47801 : curpages = RelationGetNumberOfBlocks(rel);
738 : :
739 : : /* coerce values in pg_class to more desirable types */
740 : 47801 : relpages = (BlockNumber) rel->rd_rel->relpages;
741 : 47801 : reltuples = (double) rel->rd_rel->reltuples;
742 : 47801 : relallvisible = (BlockNumber) rel->rd_rel->relallvisible;
743 : :
744 : : /*
745 : : * HACK: if the relation has never yet been vacuumed, use a minimum size
746 : : * estimate of 10 pages. The idea here is to avoid assuming a
747 : : * newly-created table is really small, even if it currently is, because
748 : : * that may not be true once some data gets loaded into it. Once a vacuum
749 : : * or analyze cycle has been done on it, it's more reasonable to believe
750 : : * the size is somewhat stable.
751 : : *
752 : : * (Note that this is only an issue if the plan gets cached and used again
753 : : * after the table has been filled. What we're trying to avoid is using a
754 : : * nestloop-type plan on a table that has grown substantially since the
755 : : * plan was made. Normally, autovacuum/autoanalyze will occur once enough
756 : : * inserts have happened and cause cached-plan invalidation; but that
757 : : * doesn't happen instantaneously, and it won't happen at all for cases
758 : : * such as temporary tables.)
759 : : *
760 : : * We test "never vacuumed" by seeing whether reltuples < 0.
761 : : *
762 : : * If the table has inheritance children, we don't apply this heuristic.
763 : : * Totally empty parent tables are quite common, so we should be willing
764 : : * to believe that they are empty.
765 : : */
766 [ + + ]: 47801 : if (curpages < 10 &&
767 [ + + + + ]: 30258 : reltuples < 0 &&
768 : 15652 : !rel->rd_rel->relhassubclass)
769 : 15253 : curpages = 10;
770 : :
771 : : /* report estimated # pages */
772 : 47801 : *pages = curpages;
773 : : /* quick exit if rel is clearly empty */
774 [ + + ]: 47801 : if (curpages == 0)
775 : : {
776 : 1135 : *tuples = 0;
777 : 1135 : *allvisfrac = 0;
778 : 1135 : return;
779 : : }
780 : :
781 : : /* estimate number of tuples from previous tuple density */
782 [ + + + + ]: 46666 : if (reltuples >= 0 && relpages > 0)
783 : 25669 : density = reltuples / (double) relpages;
784 : : else
785 : : {
786 : : /*
787 : : * When we have no data because the relation was never yet vacuumed,
788 : : * estimate tuple width from attribute datatypes. We assume here that
789 : : * the pages are completely full, which is OK for tables but is
790 : : * probably an overestimate for indexes. Fortunately
791 : : * get_relation_info() can clamp the overestimate to the parent
792 : : * table's size.
793 : : *
794 : : * Note: this code intentionally disregards alignment considerations,
795 : : * because (a) that would be gilding the lily considering how crude
796 : : * the estimate is, (b) it creates platform dependencies in the
797 : : * default plans which are kind of a headache for regression testing,
798 : : * and (c) different table AMs might use different padding schemes.
799 : : */
800 : 20997 : int32 tuple_width;
801 : 20997 : int fillfactor;
802 : :
803 : : /*
804 : : * Without reltuples/relpages, we also need to consider fillfactor.
805 : : * The other branch considers it implicitly by calculating density
806 : : * from actual relpages/reltuples statistics.
807 : : */
808 [ + + ]: 20997 : fillfactor = RelationGetFillFactor(rel, HEAP_DEFAULT_FILLFACTOR);
809 : :
810 : 20997 : tuple_width = get_rel_data_width(rel, attr_widths);
811 : 20997 : tuple_width += overhead_bytes_per_tuple;
812 : : /* note: integer division is intentional here */
813 : 20997 : density = (usable_bytes_per_page * fillfactor / 100) / tuple_width;
814 : : /* There's at least one row on the page, even with low fillfactor. */
815 : 20997 : density = clamp_row_est(density);
816 : 20997 : }
817 : 46666 : *tuples = rint(density * (double) curpages);
818 : :
819 : : /*
820 : : * We use relallvisible as-is, rather than scaling it up like we do for
821 : : * the pages and tuples counts, on the theory that any pages added since
822 : : * the last VACUUM are most likely not marked all-visible. But costsize.c
823 : : * wants it converted to a fraction.
824 : : */
825 [ + + - + ]: 46666 : if (relallvisible == 0 || curpages <= 0)
826 : 27861 : *allvisfrac = 0;
827 [ + + ]: 18805 : else if ((double) relallvisible >= curpages)
828 : 6891 : *allvisfrac = 1;
829 : : else
830 : 11914 : *allvisfrac = (double) relallvisible / curpages;
831 [ - + ]: 47801 : }
|
