Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * indexcmds.c
4 : * POSTGRES define and remove index code.
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/commands/indexcmds.c
12 : *
13 : *-------------------------------------------------------------------------
14 : */
15 :
16 : #include "postgres.h"
17 :
18 : #include "access/amapi.h"
19 : #include "access/gist.h"
20 : #include "access/heapam.h"
21 : #include "access/htup_details.h"
22 : #include "access/reloptions.h"
23 : #include "access/sysattr.h"
24 : #include "access/tableam.h"
25 : #include "access/xact.h"
26 : #include "catalog/catalog.h"
27 : #include "catalog/index.h"
28 : #include "catalog/indexing.h"
29 : #include "catalog/namespace.h"
30 : #include "catalog/pg_am.h"
31 : #include "catalog/pg_authid.h"
32 : #include "catalog/pg_collation.h"
33 : #include "catalog/pg_constraint.h"
34 : #include "catalog/pg_database.h"
35 : #include "catalog/pg_inherits.h"
36 : #include "catalog/pg_namespace.h"
37 : #include "catalog/pg_opclass.h"
38 : #include "catalog/pg_tablespace.h"
39 : #include "catalog/pg_type.h"
40 : #include "commands/comment.h"
41 : #include "commands/defrem.h"
42 : #include "commands/event_trigger.h"
43 : #include "commands/progress.h"
44 : #include "commands/tablecmds.h"
45 : #include "commands/tablespace.h"
46 : #include "mb/pg_wchar.h"
47 : #include "miscadmin.h"
48 : #include "nodes/makefuncs.h"
49 : #include "nodes/nodeFuncs.h"
50 : #include "optimizer/optimizer.h"
51 : #include "parser/parse_coerce.h"
52 : #include "parser/parse_oper.h"
53 : #include "parser/parse_utilcmd.h"
54 : #include "partitioning/partdesc.h"
55 : #include "pgstat.h"
56 : #include "rewrite/rewriteManip.h"
57 : #include "storage/lmgr.h"
58 : #include "storage/proc.h"
59 : #include "storage/procarray.h"
60 : #include "utils/acl.h"
61 : #include "utils/builtins.h"
62 : #include "utils/fmgroids.h"
63 : #include "utils/guc.h"
64 : #include "utils/injection_point.h"
65 : #include "utils/inval.h"
66 : #include "utils/lsyscache.h"
67 : #include "utils/memutils.h"
68 : #include "utils/partcache.h"
69 : #include "utils/pg_rusage.h"
70 : #include "utils/regproc.h"
71 : #include "utils/snapmgr.h"
72 : #include "utils/syscache.h"
73 :
74 :
75 : /* non-export function prototypes */
76 : static bool CompareOpclassOptions(const Datum *opts1, const Datum *opts2, int natts);
77 : static void CheckPredicate(Expr *predicate);
78 : static void ComputeIndexAttrs(ParseState *pstate,
79 : IndexInfo *indexInfo,
80 : Oid *typeOids,
81 : Oid *collationOids,
82 : Oid *opclassOids,
83 : Datum *opclassOptions,
84 : int16 *colOptions,
85 : const List *attList,
86 : const List *exclusionOpNames,
87 : Oid relId,
88 : const char *accessMethodName,
89 : Oid accessMethodId,
90 : bool amcanorder,
91 : bool isconstraint,
92 : bool iswithoutoverlaps,
93 : Oid ddl_userid,
94 : int ddl_sec_context,
95 : int *ddl_save_nestlevel);
96 : static char *ChooseIndexName(const char *tabname, Oid namespaceId,
97 : const List *colnames, const List *exclusionOpNames,
98 : bool primary, bool isconstraint);
99 : static char *ChooseIndexNameAddition(const List *colnames);
100 : static List *ChooseIndexColumnNames(const List *indexElems);
101 : static void ReindexIndex(const ReindexStmt *stmt, const ReindexParams *params,
102 : bool isTopLevel);
103 : static void RangeVarCallbackForReindexIndex(const RangeVar *relation,
104 : Oid relId, Oid oldRelId, void *arg);
105 : static Oid ReindexTable(const ReindexStmt *stmt, const ReindexParams *params,
106 : bool isTopLevel);
107 : static void ReindexMultipleTables(const ReindexStmt *stmt,
108 : const ReindexParams *params);
109 : static void reindex_error_callback(void *arg);
110 : static void ReindexPartitions(const ReindexStmt *stmt, Oid relid,
111 : const ReindexParams *params, bool isTopLevel);
112 : static void ReindexMultipleInternal(const ReindexStmt *stmt, const List *relids,
113 : const ReindexParams *params);
114 : static bool ReindexRelationConcurrently(const ReindexStmt *stmt,
115 : Oid relationOid,
116 : const ReindexParams *params);
117 : static void update_relispartition(Oid relationId, bool newval);
118 : static inline void set_indexsafe_procflags(void);
119 :
120 : /*
121 : * callback argument type for RangeVarCallbackForReindexIndex()
122 : */
123 : struct ReindexIndexCallbackState
124 : {
125 : ReindexParams params; /* options from statement */
126 : Oid locked_table_oid; /* tracks previously locked table */
127 : };
128 :
129 : /*
130 : * callback arguments for reindex_error_callback()
131 : */
132 : typedef struct ReindexErrorInfo
133 : {
134 : char *relname;
135 : char *relnamespace;
136 : char relkind;
137 : } ReindexErrorInfo;
138 :
139 : /*
140 : * CheckIndexCompatible
141 : * Determine whether an existing index definition is compatible with a
142 : * prospective index definition, such that the existing index storage
143 : * could become the storage of the new index, avoiding a rebuild.
144 : *
145 : * 'oldId': the OID of the existing index
146 : * 'accessMethodName': name of the AM to use.
147 : * 'attributeList': a list of IndexElem specifying columns and expressions
148 : * to index on.
149 : * 'exclusionOpNames': list of names of exclusion-constraint operators,
150 : * or NIL if not an exclusion constraint.
151 : * 'isWithoutOverlaps': true iff this index has a WITHOUT OVERLAPS clause.
152 : *
153 : * This is tailored to the needs of ALTER TABLE ALTER TYPE, which recreates
154 : * any indexes that depended on a changing column from their pg_get_indexdef
155 : * or pg_get_constraintdef definitions. We omit some of the sanity checks of
156 : * DefineIndex. We assume that the old and new indexes have the same number
157 : * of columns and that if one has an expression column or predicate, both do.
158 : * Errors arising from the attribute list still apply.
159 : *
160 : * Most column type changes that can skip a table rewrite do not invalidate
161 : * indexes. We acknowledge this when all operator classes, collations and
162 : * exclusion operators match. Though we could further permit intra-opfamily
163 : * changes for btree and hash indexes, that adds subtle complexity with no
164 : * concrete benefit for core types. Note, that INCLUDE columns aren't
165 : * checked by this function, for them it's enough that table rewrite is
166 : * skipped.
167 : *
168 : * When a comparison or exclusion operator has a polymorphic input type, the
169 : * actual input types must also match. This defends against the possibility
170 : * that operators could vary behavior in response to get_fn_expr_argtype().
171 : * At present, this hazard is theoretical: check_exclusion_constraint() and
172 : * all core index access methods decline to set fn_expr for such calls.
173 : *
174 : * We do not yet implement a test to verify compatibility of expression
175 : * columns or predicates, so assume any such index is incompatible.
176 : */
177 : bool
178 104 : CheckIndexCompatible(Oid oldId,
179 : const char *accessMethodName,
180 : const List *attributeList,
181 : const List *exclusionOpNames,
182 : bool isWithoutOverlaps)
183 : {
184 : bool isconstraint;
185 : Oid *typeIds;
186 : Oid *collationIds;
187 : Oid *opclassIds;
188 : Datum *opclassOptions;
189 : Oid accessMethodId;
190 : Oid relationId;
191 : HeapTuple tuple;
192 : Form_pg_index indexForm;
193 : Form_pg_am accessMethodForm;
194 : const IndexAmRoutine *amRoutine;
195 : bool amcanorder;
196 : bool amsummarizing;
197 : int16 *coloptions;
198 : IndexInfo *indexInfo;
199 : int numberOfAttributes;
200 : int old_natts;
201 104 : bool ret = true;
202 : oidvector *old_indclass;
203 : oidvector *old_indcollation;
204 : Relation irel;
205 : int i;
206 : Datum d;
207 :
208 : /* Caller should already have the relation locked in some way. */
209 104 : relationId = IndexGetRelation(oldId, false);
210 :
211 : /*
212 : * We can pretend isconstraint = false unconditionally. It only serves to
213 : * decide the text of an error message that should never happen for us.
214 : */
215 104 : isconstraint = false;
216 :
217 104 : numberOfAttributes = list_length(attributeList);
218 : Assert(numberOfAttributes > 0);
219 : Assert(numberOfAttributes <= INDEX_MAX_KEYS);
220 :
221 : /* look up the access method */
222 104 : tuple = SearchSysCache1(AMNAME, PointerGetDatum(accessMethodName));
223 104 : if (!HeapTupleIsValid(tuple))
224 0 : ereport(ERROR,
225 : (errcode(ERRCODE_UNDEFINED_OBJECT),
226 : errmsg("access method \"%s\" does not exist",
227 : accessMethodName)));
228 104 : accessMethodForm = (Form_pg_am) GETSTRUCT(tuple);
229 104 : accessMethodId = accessMethodForm->oid;
230 104 : amRoutine = GetIndexAmRoutine(accessMethodForm->amhandler);
231 104 : ReleaseSysCache(tuple);
232 :
233 104 : amcanorder = amRoutine->amcanorder;
234 104 : amsummarizing = amRoutine->amsummarizing;
235 :
236 : /*
237 : * Compute the operator classes, collations, and exclusion operators for
238 : * the new index, so we can test whether it's compatible with the existing
239 : * one. Note that ComputeIndexAttrs might fail here, but that's OK:
240 : * DefineIndex would have failed later. Our attributeList contains only
241 : * key attributes, thus we're filling ii_NumIndexAttrs and
242 : * ii_NumIndexKeyAttrs with same value.
243 : */
244 104 : indexInfo = makeIndexInfo(numberOfAttributes, numberOfAttributes,
245 : accessMethodId, NIL, NIL, false, false,
246 : false, false, amsummarizing, isWithoutOverlaps);
247 104 : typeIds = palloc_array(Oid, numberOfAttributes);
248 104 : collationIds = palloc_array(Oid, numberOfAttributes);
249 104 : opclassIds = palloc_array(Oid, numberOfAttributes);
250 104 : opclassOptions = palloc_array(Datum, numberOfAttributes);
251 104 : coloptions = palloc_array(int16, numberOfAttributes);
252 104 : ComputeIndexAttrs(NULL, indexInfo,
253 : typeIds, collationIds, opclassIds, opclassOptions,
254 : coloptions, attributeList,
255 : exclusionOpNames, relationId,
256 : accessMethodName, accessMethodId,
257 : amcanorder, isconstraint, isWithoutOverlaps, InvalidOid,
258 : 0, NULL);
259 :
260 : /* Get the soon-obsolete pg_index tuple. */
261 104 : tuple = SearchSysCache1(INDEXRELID, ObjectIdGetDatum(oldId));
262 104 : if (!HeapTupleIsValid(tuple))
263 0 : elog(ERROR, "cache lookup failed for index %u", oldId);
264 104 : indexForm = (Form_pg_index) GETSTRUCT(tuple);
265 :
266 : /*
267 : * We don't assess expressions or predicates; assume incompatibility.
268 : * Also, if the index is invalid for any reason, treat it as incompatible.
269 : */
270 208 : if (!(heap_attisnull(tuple, Anum_pg_index_indpred, NULL) &&
271 104 : heap_attisnull(tuple, Anum_pg_index_indexprs, NULL) &&
272 104 : indexForm->indisvalid))
273 : {
274 0 : ReleaseSysCache(tuple);
275 0 : return false;
276 : }
277 :
278 : /* Any change in operator class or collation breaks compatibility. */
279 104 : old_natts = indexForm->indnkeyatts;
280 : Assert(old_natts == numberOfAttributes);
281 :
282 104 : d = SysCacheGetAttrNotNull(INDEXRELID, tuple, Anum_pg_index_indcollation);
283 104 : old_indcollation = (oidvector *) DatumGetPointer(d);
284 :
285 104 : d = SysCacheGetAttrNotNull(INDEXRELID, tuple, Anum_pg_index_indclass);
286 104 : old_indclass = (oidvector *) DatumGetPointer(d);
287 :
288 208 : ret = (memcmp(old_indclass->values, opclassIds, old_natts * sizeof(Oid)) == 0 &&
289 104 : memcmp(old_indcollation->values, collationIds, old_natts * sizeof(Oid)) == 0);
290 :
291 104 : ReleaseSysCache(tuple);
292 :
293 104 : if (!ret)
294 0 : return false;
295 :
296 : /* For polymorphic opcintype, column type changes break compatibility. */
297 104 : irel = index_open(oldId, AccessShareLock); /* caller probably has a lock */
298 214 : for (i = 0; i < old_natts; i++)
299 : {
300 110 : if (IsPolymorphicType(get_opclass_input_type(opclassIds[i])) &&
301 0 : TupleDescAttr(irel->rd_att, i)->atttypid != typeIds[i])
302 : {
303 0 : ret = false;
304 0 : break;
305 : }
306 : }
307 :
308 : /* Any change in opclass options break compatibility. */
309 104 : if (ret)
310 : {
311 104 : Datum *oldOpclassOptions = palloc_array(Datum, old_natts);
312 :
313 214 : for (i = 0; i < old_natts; i++)
314 110 : oldOpclassOptions[i] = get_attoptions(oldId, i + 1);
315 :
316 104 : ret = CompareOpclassOptions(oldOpclassOptions, opclassOptions, old_natts);
317 :
318 104 : pfree(oldOpclassOptions);
319 : }
320 :
321 : /* Any change in exclusion operator selections breaks compatibility. */
322 104 : if (ret && indexInfo->ii_ExclusionOps != NULL)
323 : {
324 : Oid *old_operators,
325 : *old_procs;
326 : uint16 *old_strats;
327 :
328 0 : RelationGetExclusionInfo(irel, &old_operators, &old_procs, &old_strats);
329 0 : ret = memcmp(old_operators, indexInfo->ii_ExclusionOps,
330 : old_natts * sizeof(Oid)) == 0;
331 :
332 : /* Require an exact input type match for polymorphic operators. */
333 0 : if (ret)
334 : {
335 0 : for (i = 0; i < old_natts && ret; i++)
336 : {
337 : Oid left,
338 : right;
339 :
340 0 : op_input_types(indexInfo->ii_ExclusionOps[i], &left, &right);
341 0 : if ((IsPolymorphicType(left) || IsPolymorphicType(right)) &&
342 0 : TupleDescAttr(irel->rd_att, i)->atttypid != typeIds[i])
343 : {
344 0 : ret = false;
345 0 : break;
346 : }
347 : }
348 : }
349 : }
350 :
351 104 : index_close(irel, NoLock);
352 104 : return ret;
353 : }
354 :
355 : /*
356 : * CompareOpclassOptions
357 : *
358 : * Compare per-column opclass options which are represented by arrays of text[]
359 : * datums. Both elements of arrays and array themselves can be NULL.
360 : */
361 : static bool
362 104 : CompareOpclassOptions(const Datum *opts1, const Datum *opts2, int natts)
363 : {
364 : int i;
365 : FmgrInfo fm;
366 :
367 104 : if (!opts1 && !opts2)
368 0 : return true;
369 :
370 104 : fmgr_info(F_ARRAY_EQ, &fm);
371 214 : for (i = 0; i < natts; i++)
372 : {
373 110 : Datum opt1 = opts1 ? opts1[i] : (Datum) 0;
374 110 : Datum opt2 = opts2 ? opts2[i] : (Datum) 0;
375 :
376 110 : if (opt1 == (Datum) 0)
377 : {
378 108 : if (opt2 == (Datum) 0)
379 108 : continue;
380 : else
381 0 : return false;
382 : }
383 2 : else if (opt2 == (Datum) 0)
384 0 : return false;
385 :
386 : /*
387 : * Compare non-NULL text[] datums. Use C collation to enforce binary
388 : * equivalence of texts, because we don't know anything about the
389 : * semantics of opclass options.
390 : */
391 2 : if (!DatumGetBool(FunctionCall2Coll(&fm, C_COLLATION_OID, opt1, opt2)))
392 0 : return false;
393 : }
394 :
395 104 : return true;
396 : }
397 :
398 : /*
399 : * WaitForOlderSnapshots
400 : *
401 : * Wait for transactions that might have an older snapshot than the given xmin
402 : * limit, because it might not contain tuples deleted just before it has
403 : * been taken. Obtain a list of VXIDs of such transactions, and wait for them
404 : * individually. This is used when building an index concurrently.
405 : *
406 : * We can exclude any running transactions that have xmin > the xmin given;
407 : * their oldest snapshot must be newer than our xmin limit.
408 : * We can also exclude any transactions that have xmin = zero, since they
409 : * evidently have no live snapshot at all (and any one they might be in
410 : * process of taking is certainly newer than ours). Transactions in other
411 : * DBs can be ignored too, since they'll never even be able to see the
412 : * index being worked on.
413 : *
414 : * We can also exclude autovacuum processes and processes running manual
415 : * lazy VACUUMs, because they won't be fazed by missing index entries
416 : * either. (Manual ANALYZEs, however, can't be excluded because they
417 : * might be within transactions that are going to do arbitrary operations
418 : * later.) Processes running CREATE INDEX CONCURRENTLY or REINDEX CONCURRENTLY
419 : * on indexes that are neither expressional nor partial are also safe to
420 : * ignore, since we know that those processes won't examine any data
421 : * outside the table they're indexing.
422 : *
423 : * Also, GetCurrentVirtualXIDs never reports our own vxid, so we need not
424 : * check for that.
425 : *
426 : * If a process goes idle-in-transaction with xmin zero, we do not need to
427 : * wait for it anymore, per the above argument. We do not have the
428 : * infrastructure right now to stop waiting if that happens, but we can at
429 : * least avoid the folly of waiting when it is idle at the time we would
430 : * begin to wait. We do this by repeatedly rechecking the output of
431 : * GetCurrentVirtualXIDs. If, during any iteration, a particular vxid
432 : * doesn't show up in the output, we know we can forget about it.
433 : */
434 : void
435 706 : WaitForOlderSnapshots(TransactionId limitXmin, bool progress)
436 : {
437 : int n_old_snapshots;
438 : int i;
439 : VirtualTransactionId *old_snapshots;
440 :
441 706 : old_snapshots = GetCurrentVirtualXIDs(limitXmin, true, false,
442 : PROC_IS_AUTOVACUUM | PROC_IN_VACUUM
443 : | PROC_IN_SAFE_IC,
444 : &n_old_snapshots);
445 706 : if (progress)
446 692 : pgstat_progress_update_param(PROGRESS_WAITFOR_TOTAL, n_old_snapshots);
447 :
448 1024 : for (i = 0; i < n_old_snapshots; i++)
449 : {
450 318 : if (!VirtualTransactionIdIsValid(old_snapshots[i]))
451 58 : continue; /* found uninteresting in previous cycle */
452 :
453 260 : if (i > 0)
454 : {
455 : /* see if anything's changed ... */
456 : VirtualTransactionId *newer_snapshots;
457 : int n_newer_snapshots;
458 : int j;
459 : int k;
460 :
461 118 : newer_snapshots = GetCurrentVirtualXIDs(limitXmin,
462 : true, false,
463 : PROC_IS_AUTOVACUUM | PROC_IN_VACUUM
464 : | PROC_IN_SAFE_IC,
465 : &n_newer_snapshots);
466 432 : for (j = i; j < n_old_snapshots; j++)
467 : {
468 314 : if (!VirtualTransactionIdIsValid(old_snapshots[j]))
469 52 : continue; /* found uninteresting in previous cycle */
470 822 : for (k = 0; k < n_newer_snapshots; k++)
471 : {
472 712 : if (VirtualTransactionIdEquals(old_snapshots[j],
473 : newer_snapshots[k]))
474 152 : break;
475 : }
476 262 : if (k >= n_newer_snapshots) /* not there anymore */
477 110 : SetInvalidVirtualTransactionId(old_snapshots[j]);
478 : }
479 118 : pfree(newer_snapshots);
480 : }
481 :
482 260 : if (VirtualTransactionIdIsValid(old_snapshots[i]))
483 : {
484 : /* If requested, publish who we're going to wait for. */
485 208 : if (progress)
486 : {
487 208 : PGPROC *holder = ProcNumberGetProc(old_snapshots[i].procNumber);
488 :
489 208 : if (holder)
490 208 : pgstat_progress_update_param(PROGRESS_WAITFOR_CURRENT_PID,
491 208 : holder->pid);
492 : }
493 208 : VirtualXactLock(old_snapshots[i], true);
494 : }
495 :
496 260 : if (progress)
497 260 : pgstat_progress_update_param(PROGRESS_WAITFOR_DONE, i + 1);
498 : }
499 706 : }
500 :
501 :
502 : /*
503 : * DefineIndex
504 : * Creates a new index.
505 : *
506 : * This function manages the current userid according to the needs of pg_dump.
507 : * Recreating old-database catalog entries in new-database is fine, regardless
508 : * of which users would have permission to recreate those entries now. That's
509 : * just preservation of state. Running opaque expressions, like calling a
510 : * function named in a catalog entry or evaluating a pg_node_tree in a catalog
511 : * entry, as anyone other than the object owner, is not fine. To adhere to
512 : * those principles and to remain fail-safe, use the table owner userid for
513 : * most ACL checks. Use the original userid for ACL checks reached without
514 : * traversing opaque expressions. (pg_dump can predict such ACL checks from
515 : * catalogs.) Overall, this is a mess. Future DDL development should
516 : * consider offering one DDL command for catalog setup and a separate DDL
517 : * command for steps that run opaque expressions.
518 : *
519 : * 'pstate': ParseState struct (used only for error reports; pass NULL if
520 : * not available)
521 : * 'tableId': the OID of the table relation on which the index is to be
522 : * created
523 : * 'stmt': IndexStmt describing the properties of the new index.
524 : * 'indexRelationId': normally InvalidOid, but during bootstrap can be
525 : * nonzero to specify a preselected OID for the index.
526 : * 'parentIndexId': the OID of the parent index; InvalidOid if not the child
527 : * of a partitioned index.
528 : * 'parentConstraintId': the OID of the parent constraint; InvalidOid if not
529 : * the child of a constraint (only used when recursing)
530 : * 'total_parts': total number of direct and indirect partitions of relation;
531 : * pass -1 if not known or rel is not partitioned.
532 : * 'is_alter_table': this is due to an ALTER rather than a CREATE operation.
533 : * 'check_rights': check for CREATE rights in namespace and tablespace. (This
534 : * should be true except when ALTER is deleting/recreating an index.)
535 : * 'check_not_in_use': check for table not already in use in current session.
536 : * This should be true unless caller is holding the table open, in which
537 : * case the caller had better have checked it earlier.
538 : * 'skip_build': make the catalog entries but don't create the index files
539 : * 'quiet': suppress the NOTICE chatter ordinarily provided for constraints.
540 : *
541 : * Returns the object address of the created index.
542 : */
543 : ObjectAddress
544 31682 : DefineIndex(ParseState *pstate,
545 : Oid tableId,
546 : IndexStmt *stmt,
547 : Oid indexRelationId,
548 : Oid parentIndexId,
549 : Oid parentConstraintId,
550 : int total_parts,
551 : bool is_alter_table,
552 : bool check_rights,
553 : bool check_not_in_use,
554 : bool skip_build,
555 : bool quiet)
556 : {
557 : bool concurrent;
558 : char *indexRelationName;
559 : char *accessMethodName;
560 : Oid *typeIds;
561 : Oid *collationIds;
562 : Oid *opclassIds;
563 : Datum *opclassOptions;
564 : Oid accessMethodId;
565 : Oid namespaceId;
566 : Oid tablespaceId;
567 31682 : Oid createdConstraintId = InvalidOid;
568 : List *indexColNames;
569 : List *allIndexParams;
570 : Relation rel;
571 : HeapTuple tuple;
572 : Form_pg_am accessMethodForm;
573 : const IndexAmRoutine *amRoutine;
574 : bool amcanorder;
575 : bool amissummarizing;
576 : amoptions_function amoptions;
577 : bool exclusion;
578 : bool partitioned;
579 : bool safe_index;
580 : Datum reloptions;
581 : int16 *coloptions;
582 : IndexInfo *indexInfo;
583 : bits16 flags;
584 : bits16 constr_flags;
585 : int numberOfAttributes;
586 : int numberOfKeyAttributes;
587 : TransactionId limitXmin;
588 : ObjectAddress address;
589 : LockRelId heaprelid;
590 : LOCKTAG heaplocktag;
591 : LOCKMODE lockmode;
592 : Snapshot snapshot;
593 : Oid root_save_userid;
594 : int root_save_sec_context;
595 : int root_save_nestlevel;
596 :
597 31682 : root_save_nestlevel = NewGUCNestLevel();
598 :
599 31682 : RestrictSearchPath();
600 :
601 : /*
602 : * Some callers need us to run with an empty default_tablespace; this is a
603 : * necessary hack to be able to reproduce catalog state accurately when
604 : * recreating indexes after table-rewriting ALTER TABLE.
605 : */
606 31682 : if (stmt->reset_default_tblspc)
607 456 : (void) set_config_option("default_tablespace", "",
608 : PGC_USERSET, PGC_S_SESSION,
609 : GUC_ACTION_SAVE, true, 0, false);
610 :
611 : /*
612 : * Force non-concurrent build on temporary relations, even if CONCURRENTLY
613 : * was requested. Other backends can't access a temporary relation, so
614 : * there's no harm in grabbing a stronger lock, and a non-concurrent DROP
615 : * is more efficient. Do this before any use of the concurrent option is
616 : * done.
617 : */
618 31682 : if (stmt->concurrent && get_rel_persistence(tableId) != RELPERSISTENCE_TEMP)
619 194 : concurrent = true;
620 : else
621 31488 : concurrent = false;
622 :
623 : /*
624 : * Start progress report. If we're building a partition, this was already
625 : * done.
626 : */
627 31682 : if (!OidIsValid(parentIndexId))
628 : {
629 28534 : pgstat_progress_start_command(PROGRESS_COMMAND_CREATE_INDEX, tableId);
630 28534 : pgstat_progress_update_param(PROGRESS_CREATEIDX_COMMAND,
631 : concurrent ?
632 : PROGRESS_CREATEIDX_COMMAND_CREATE_CONCURRENTLY :
633 : PROGRESS_CREATEIDX_COMMAND_CREATE);
634 : }
635 :
636 : /*
637 : * No index OID to report yet
638 : */
639 31682 : pgstat_progress_update_param(PROGRESS_CREATEIDX_INDEX_OID,
640 : InvalidOid);
641 :
642 : /*
643 : * count key attributes in index
644 : */
645 31682 : numberOfKeyAttributes = list_length(stmt->indexParams);
646 :
647 : /*
648 : * Calculate the new list of index columns including both key columns and
649 : * INCLUDE columns. Later we can determine which of these are key
650 : * columns, and which are just part of the INCLUDE list by checking the
651 : * list position. A list item in a position less than ii_NumIndexKeyAttrs
652 : * is part of the key columns, and anything equal to and over is part of
653 : * the INCLUDE columns.
654 : */
655 31682 : allIndexParams = list_concat_copy(stmt->indexParams,
656 31682 : stmt->indexIncludingParams);
657 31682 : numberOfAttributes = list_length(allIndexParams);
658 :
659 31682 : if (numberOfKeyAttributes <= 0)
660 0 : ereport(ERROR,
661 : (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
662 : errmsg("must specify at least one column")));
663 31682 : if (numberOfAttributes > INDEX_MAX_KEYS)
664 0 : ereport(ERROR,
665 : (errcode(ERRCODE_TOO_MANY_COLUMNS),
666 : errmsg("cannot use more than %d columns in an index",
667 : INDEX_MAX_KEYS)));
668 :
669 : /*
670 : * Only SELECT ... FOR UPDATE/SHARE are allowed while doing a standard
671 : * index build; but for concurrent builds we allow INSERT/UPDATE/DELETE
672 : * (but not VACUUM).
673 : *
674 : * NB: Caller is responsible for making sure that tableId refers to the
675 : * relation on which the index should be built; except in bootstrap mode,
676 : * this will typically require the caller to have already locked the
677 : * relation. To avoid lock upgrade hazards, that lock should be at least
678 : * as strong as the one we take here.
679 : *
680 : * NB: If the lock strength here ever changes, code that is run by
681 : * parallel workers under the control of certain particular ambuild
682 : * functions will need to be updated, too.
683 : */
684 31682 : lockmode = concurrent ? ShareUpdateExclusiveLock : ShareLock;
685 31682 : rel = table_open(tableId, lockmode);
686 :
687 : /*
688 : * Switch to the table owner's userid, so that any index functions are run
689 : * as that user. Also lock down security-restricted operations. We
690 : * already arranged to make GUC variable changes local to this command.
691 : */
692 31682 : GetUserIdAndSecContext(&root_save_userid, &root_save_sec_context);
693 31682 : SetUserIdAndSecContext(rel->rd_rel->relowner,
694 : root_save_sec_context | SECURITY_RESTRICTED_OPERATION);
695 :
696 31682 : namespaceId = RelationGetNamespace(rel);
697 :
698 : /*
699 : * It has exclusion constraint behavior if it's an EXCLUDE constraint or a
700 : * temporal PRIMARY KEY/UNIQUE constraint
701 : */
702 31682 : exclusion = stmt->excludeOpNames || stmt->iswithoutoverlaps;
703 :
704 : /* Ensure that it makes sense to index this kind of relation */
705 31682 : switch (rel->rd_rel->relkind)
706 : {
707 31676 : case RELKIND_RELATION:
708 : case RELKIND_MATVIEW:
709 : case RELKIND_PARTITIONED_TABLE:
710 : /* OK */
711 31676 : break;
712 6 : default:
713 6 : ereport(ERROR,
714 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
715 : errmsg("cannot create index on relation \"%s\"",
716 : RelationGetRelationName(rel)),
717 : errdetail_relkind_not_supported(rel->rd_rel->relkind)));
718 : break;
719 : }
720 :
721 : /*
722 : * Establish behavior for partitioned tables, and verify sanity of
723 : * parameters.
724 : *
725 : * We do not build an actual index in this case; we only create a few
726 : * catalog entries. The actual indexes are built by recursing for each
727 : * partition.
728 : */
729 31676 : partitioned = rel->rd_rel->relkind == RELKIND_PARTITIONED_TABLE;
730 31676 : if (partitioned)
731 : {
732 : /*
733 : * Note: we check 'stmt->concurrent' rather than 'concurrent', so that
734 : * the error is thrown also for temporary tables. Seems better to be
735 : * consistent, even though we could do it on temporary table because
736 : * we're not actually doing it concurrently.
737 : */
738 2254 : if (stmt->concurrent)
739 6 : ereport(ERROR,
740 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
741 : errmsg("cannot create index on partitioned table \"%s\" concurrently",
742 : RelationGetRelationName(rel))));
743 : }
744 :
745 : /*
746 : * Don't try to CREATE INDEX on temp tables of other backends.
747 : */
748 31670 : if (RELATION_IS_OTHER_TEMP(rel))
749 0 : ereport(ERROR,
750 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
751 : errmsg("cannot create indexes on temporary tables of other sessions")));
752 :
753 : /*
754 : * Unless our caller vouches for having checked this already, insist that
755 : * the table not be in use by our own session, either. Otherwise we might
756 : * fail to make entries in the new index (for instance, if an INSERT or
757 : * UPDATE is in progress and has already made its list of target indexes).
758 : */
759 31670 : if (check_not_in_use)
760 15040 : CheckTableNotInUse(rel, "CREATE INDEX");
761 :
762 : /*
763 : * Verify we (still) have CREATE rights in the rel's namespace.
764 : * (Presumably we did when the rel was created, but maybe not anymore.)
765 : * Skip check if caller doesn't want it. Also skip check if
766 : * bootstrapping, since permissions machinery may not be working yet.
767 : */
768 31664 : if (check_rights && !IsBootstrapProcessingMode())
769 : {
770 : AclResult aclresult;
771 :
772 16278 : aclresult = object_aclcheck(NamespaceRelationId, namespaceId, root_save_userid,
773 : ACL_CREATE);
774 16278 : if (aclresult != ACLCHECK_OK)
775 0 : aclcheck_error(aclresult, OBJECT_SCHEMA,
776 0 : get_namespace_name(namespaceId));
777 : }
778 :
779 : /*
780 : * Select tablespace to use. If not specified, use default tablespace
781 : * (which may in turn default to database's default).
782 : */
783 31664 : if (stmt->tableSpace)
784 : {
785 248 : tablespaceId = get_tablespace_oid(stmt->tableSpace, false);
786 248 : if (partitioned && tablespaceId == MyDatabaseTableSpace)
787 6 : ereport(ERROR,
788 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
789 : errmsg("cannot specify default tablespace for partitioned relations")));
790 : }
791 : else
792 : {
793 31416 : tablespaceId = GetDefaultTablespace(rel->rd_rel->relpersistence,
794 : partitioned);
795 : /* note InvalidOid is OK in this case */
796 : }
797 :
798 : /* Check tablespace permissions */
799 31652 : if (check_rights &&
800 116 : OidIsValid(tablespaceId) && tablespaceId != MyDatabaseTableSpace)
801 : {
802 : AclResult aclresult;
803 :
804 116 : aclresult = object_aclcheck(TableSpaceRelationId, tablespaceId, root_save_userid,
805 : ACL_CREATE);
806 116 : if (aclresult != ACLCHECK_OK)
807 0 : aclcheck_error(aclresult, OBJECT_TABLESPACE,
808 0 : get_tablespace_name(tablespaceId));
809 : }
810 :
811 : /*
812 : * Force shared indexes into the pg_global tablespace. This is a bit of a
813 : * hack but seems simpler than marking them in the BKI commands. On the
814 : * other hand, if it's not shared, don't allow it to be placed there.
815 : */
816 31652 : if (rel->rd_rel->relisshared)
817 2142 : tablespaceId = GLOBALTABLESPACE_OID;
818 29510 : else if (tablespaceId == GLOBALTABLESPACE_OID)
819 0 : ereport(ERROR,
820 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
821 : errmsg("only shared relations can be placed in pg_global tablespace")));
822 :
823 : /*
824 : * Choose the index column names.
825 : */
826 31652 : indexColNames = ChooseIndexColumnNames(allIndexParams);
827 :
828 : /*
829 : * Select name for index if caller didn't specify
830 : */
831 31652 : indexRelationName = stmt->idxname;
832 31652 : if (indexRelationName == NULL)
833 12316 : indexRelationName = ChooseIndexName(RelationGetRelationName(rel),
834 : namespaceId,
835 : indexColNames,
836 12316 : stmt->excludeOpNames,
837 12316 : stmt->primary,
838 12316 : stmt->isconstraint);
839 :
840 : /*
841 : * look up the access method, verify it can handle the requested features
842 : */
843 31652 : accessMethodName = stmt->accessMethod;
844 31652 : tuple = SearchSysCache1(AMNAME, PointerGetDatum(accessMethodName));
845 31652 : if (!HeapTupleIsValid(tuple))
846 : {
847 : /*
848 : * Hack to provide more-or-less-transparent updating of old RTREE
849 : * indexes to GiST: if RTREE is requested and not found, use GIST.
850 : */
851 6 : if (strcmp(accessMethodName, "rtree") == 0)
852 : {
853 6 : ereport(NOTICE,
854 : (errmsg("substituting access method \"gist\" for obsolete method \"rtree\"")));
855 6 : accessMethodName = "gist";
856 6 : tuple = SearchSysCache1(AMNAME, PointerGetDatum(accessMethodName));
857 : }
858 :
859 6 : if (!HeapTupleIsValid(tuple))
860 0 : ereport(ERROR,
861 : (errcode(ERRCODE_UNDEFINED_OBJECT),
862 : errmsg("access method \"%s\" does not exist",
863 : accessMethodName)));
864 : }
865 31652 : accessMethodForm = (Form_pg_am) GETSTRUCT(tuple);
866 31652 : accessMethodId = accessMethodForm->oid;
867 31652 : amRoutine = GetIndexAmRoutine(accessMethodForm->amhandler);
868 :
869 31652 : pgstat_progress_update_param(PROGRESS_CREATEIDX_ACCESS_METHOD_OID,
870 : accessMethodId);
871 :
872 31652 : if (stmt->unique && !stmt->iswithoutoverlaps && !amRoutine->amcanunique)
873 0 : ereport(ERROR,
874 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
875 : errmsg("access method \"%s\" does not support unique indexes",
876 : accessMethodName)));
877 31652 : if (stmt->indexIncludingParams != NIL && !amRoutine->amcaninclude)
878 18 : ereport(ERROR,
879 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
880 : errmsg("access method \"%s\" does not support included columns",
881 : accessMethodName)));
882 31634 : if (numberOfKeyAttributes > 1 && !amRoutine->amcanmulticol)
883 0 : ereport(ERROR,
884 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
885 : errmsg("access method \"%s\" does not support multicolumn indexes",
886 : accessMethodName)));
887 31634 : if (exclusion && amRoutine->amgettuple == NULL)
888 0 : ereport(ERROR,
889 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
890 : errmsg("access method \"%s\" does not support exclusion constraints",
891 : accessMethodName)));
892 31634 : if (stmt->iswithoutoverlaps && strcmp(accessMethodName, "gist") != 0)
893 0 : ereport(ERROR,
894 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
895 : errmsg("access method \"%s\" does not support WITHOUT OVERLAPS constraints",
896 : accessMethodName)));
897 :
898 31634 : amcanorder = amRoutine->amcanorder;
899 31634 : amoptions = amRoutine->amoptions;
900 31634 : amissummarizing = amRoutine->amsummarizing;
901 :
902 31634 : ReleaseSysCache(tuple);
903 :
904 : /*
905 : * Validate predicate, if given
906 : */
907 31634 : if (stmt->whereClause)
908 448 : CheckPredicate((Expr *) stmt->whereClause);
909 :
910 : /*
911 : * Parse AM-specific options, convert to text array form, validate.
912 : */
913 31634 : reloptions = transformRelOptions((Datum) 0, stmt->options,
914 : NULL, NULL, false, false);
915 :
916 31628 : (void) index_reloptions(amoptions, reloptions, true);
917 :
918 : /*
919 : * Prepare arguments for index_create, primarily an IndexInfo structure.
920 : * Note that predicates must be in implicit-AND format. In a concurrent
921 : * build, mark it not-ready-for-inserts.
922 : */
923 31564 : indexInfo = makeIndexInfo(numberOfAttributes,
924 : numberOfKeyAttributes,
925 : accessMethodId,
926 : NIL, /* expressions, NIL for now */
927 31564 : make_ands_implicit((Expr *) stmt->whereClause),
928 31564 : stmt->unique,
929 31564 : stmt->nulls_not_distinct,
930 31564 : !concurrent,
931 : concurrent,
932 : amissummarizing,
933 31564 : stmt->iswithoutoverlaps);
934 :
935 31564 : typeIds = palloc_array(Oid, numberOfAttributes);
936 31564 : collationIds = palloc_array(Oid, numberOfAttributes);
937 31564 : opclassIds = palloc_array(Oid, numberOfAttributes);
938 31564 : opclassOptions = palloc_array(Datum, numberOfAttributes);
939 31564 : coloptions = palloc_array(int16, numberOfAttributes);
940 31564 : ComputeIndexAttrs(pstate,
941 : indexInfo,
942 : typeIds, collationIds, opclassIds, opclassOptions,
943 : coloptions, allIndexParams,
944 31564 : stmt->excludeOpNames, tableId,
945 : accessMethodName, accessMethodId,
946 31564 : amcanorder, stmt->isconstraint, stmt->iswithoutoverlaps,
947 : root_save_userid, root_save_sec_context,
948 : &root_save_nestlevel);
949 :
950 : /*
951 : * Extra checks when creating a PRIMARY KEY index.
952 : */
953 31346 : if (stmt->primary)
954 9354 : index_check_primary_key(rel, indexInfo, is_alter_table, stmt);
955 :
956 : /*
957 : * If this table is partitioned and we're creating a unique index, primary
958 : * key, or exclusion constraint, make sure that the partition key is a
959 : * subset of the index's columns. Otherwise it would be possible to
960 : * violate uniqueness by putting values that ought to be unique in
961 : * different partitions.
962 : *
963 : * We could lift this limitation if we had global indexes, but those have
964 : * their own problems, so this is a useful feature combination.
965 : */
966 31310 : if (partitioned && (stmt->unique || exclusion))
967 : {
968 1328 : PartitionKey key = RelationGetPartitionKey(rel);
969 : const char *constraint_type;
970 : int i;
971 :
972 1328 : if (stmt->primary)
973 982 : constraint_type = "PRIMARY KEY";
974 346 : else if (stmt->unique)
975 246 : constraint_type = "UNIQUE";
976 100 : else if (stmt->excludeOpNames)
977 100 : constraint_type = "EXCLUDE";
978 : else
979 : {
980 0 : elog(ERROR, "unknown constraint type");
981 : constraint_type = NULL; /* keep compiler quiet */
982 : }
983 :
984 : /*
985 : * Verify that all the columns in the partition key appear in the
986 : * unique key definition, with the same notion of equality.
987 : */
988 2672 : for (i = 0; i < key->partnatts; i++)
989 : {
990 1448 : bool found = false;
991 : int eq_strategy;
992 : Oid ptkey_eqop;
993 : int j;
994 :
995 : /*
996 : * Identify the equality operator associated with this partkey
997 : * column. For list and range partitioning, partkeys use btree
998 : * operator classes; hash partitioning uses hash operator classes.
999 : * (Keep this in sync with ComputePartitionAttrs!)
1000 : */
1001 1448 : if (key->strategy == PARTITION_STRATEGY_HASH)
1002 66 : eq_strategy = HTEqualStrategyNumber;
1003 : else
1004 1382 : eq_strategy = BTEqualStrategyNumber;
1005 :
1006 1448 : ptkey_eqop = get_opfamily_member(key->partopfamily[i],
1007 1448 : key->partopcintype[i],
1008 1448 : key->partopcintype[i],
1009 : eq_strategy);
1010 1448 : if (!OidIsValid(ptkey_eqop))
1011 0 : elog(ERROR, "missing operator %d(%u,%u) in partition opfamily %u",
1012 : eq_strategy, key->partopcintype[i], key->partopcintype[i],
1013 : key->partopfamily[i]);
1014 :
1015 : /*
1016 : * It may be possible to support UNIQUE constraints when partition
1017 : * keys are expressions, but is it worth it? Give up for now.
1018 : */
1019 1448 : if (key->partattrs[i] == 0)
1020 12 : ereport(ERROR,
1021 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1022 : errmsg("unsupported %s constraint with partition key definition",
1023 : constraint_type),
1024 : errdetail("%s constraints cannot be used when partition keys include expressions.",
1025 : constraint_type)));
1026 :
1027 : /* Search the index column(s) for a match */
1028 1636 : for (j = 0; j < indexInfo->ii_NumIndexKeyAttrs; j++)
1029 : {
1030 1558 : if (key->partattrs[i] == indexInfo->ii_IndexAttrNumbers[j])
1031 : {
1032 : /*
1033 : * Matched the column, now what about the collation and
1034 : * equality op?
1035 : */
1036 : Oid idx_opfamily;
1037 : Oid idx_opcintype;
1038 :
1039 1358 : if (key->partcollation[i] != collationIds[j])
1040 0 : continue;
1041 :
1042 1358 : if (get_opclass_opfamily_and_input_type(opclassIds[j],
1043 : &idx_opfamily,
1044 : &idx_opcintype))
1045 : {
1046 1358 : Oid idx_eqop = InvalidOid;
1047 :
1048 1358 : if (stmt->unique && !stmt->iswithoutoverlaps)
1049 1210 : idx_eqop = get_opfamily_member_for_cmptype(idx_opfamily,
1050 : idx_opcintype,
1051 : idx_opcintype,
1052 : COMPARE_EQ);
1053 148 : else if (exclusion)
1054 148 : idx_eqop = indexInfo->ii_ExclusionOps[j];
1055 :
1056 1358 : if (!idx_eqop)
1057 0 : ereport(ERROR,
1058 : errcode(ERRCODE_UNDEFINED_OBJECT),
1059 : errmsg("could not identify an equality operator for type %s", format_type_be(idx_opcintype)),
1060 : errdetail("There is no suitable operator in operator family \"%s\" for access method \"%s\".",
1061 : get_opfamily_name(idx_opfamily, false), get_am_name(get_opfamily_method(idx_opfamily))));
1062 :
1063 1358 : if (ptkey_eqop == idx_eqop)
1064 : {
1065 1344 : found = true;
1066 1344 : break;
1067 : }
1068 14 : else if (exclusion)
1069 : {
1070 : /*
1071 : * We found a match, but it's not an equality
1072 : * operator. Instead of failing below with an
1073 : * error message about a missing column, fail now
1074 : * and explain that the operator is wrong.
1075 : */
1076 14 : Form_pg_attribute att = TupleDescAttr(RelationGetDescr(rel), key->partattrs[i] - 1);
1077 :
1078 14 : ereport(ERROR,
1079 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1080 : errmsg("cannot match partition key to index on column \"%s\" using non-equal operator \"%s\"",
1081 : NameStr(att->attname),
1082 : get_opname(indexInfo->ii_ExclusionOps[j]))));
1083 : }
1084 : }
1085 : }
1086 : }
1087 :
1088 1422 : if (!found)
1089 : {
1090 : Form_pg_attribute att;
1091 :
1092 78 : att = TupleDescAttr(RelationGetDescr(rel),
1093 78 : key->partattrs[i] - 1);
1094 78 : ereport(ERROR,
1095 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1096 : /* translator: %s is UNIQUE, PRIMARY KEY, etc */
1097 : errmsg("%s constraint on partitioned table must include all partitioning columns",
1098 : constraint_type),
1099 : /* translator: first %s is UNIQUE, PRIMARY KEY, etc */
1100 : errdetail("%s constraint on table \"%s\" lacks column \"%s\" which is part of the partition key.",
1101 : constraint_type, RelationGetRelationName(rel),
1102 : NameStr(att->attname))));
1103 : }
1104 : }
1105 : }
1106 :
1107 :
1108 : /*
1109 : * We disallow indexes on system columns. They would not necessarily get
1110 : * updated correctly, and they don't seem useful anyway.
1111 : *
1112 : * Also disallow virtual generated columns in indexes (use expression
1113 : * index instead).
1114 : */
1115 75080 : for (int i = 0; i < indexInfo->ii_NumIndexAttrs; i++)
1116 : {
1117 43898 : AttrNumber attno = indexInfo->ii_IndexAttrNumbers[i];
1118 :
1119 43898 : if (attno < 0)
1120 6 : ereport(ERROR,
1121 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1122 : errmsg("index creation on system columns is not supported")));
1123 :
1124 :
1125 43892 : if (TupleDescAttr(RelationGetDescr(rel), attno - 1)->attgenerated == ATTRIBUTE_GENERATED_VIRTUAL)
1126 18 : ereport(ERROR,
1127 : errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1128 : stmt->primary ?
1129 : errmsg("primary keys on virtual generated columns are not supported") :
1130 : stmt->isconstraint ?
1131 : errmsg("unique constraints on virtual generated columns are not supported") :
1132 : errmsg("indexes on virtual generated columns are not supported"));
1133 : }
1134 :
1135 : /*
1136 : * Also check for system and generated columns used in expressions or
1137 : * predicates.
1138 : */
1139 31182 : if (indexInfo->ii_Expressions || indexInfo->ii_Predicate)
1140 : {
1141 1276 : Bitmapset *indexattrs = NULL;
1142 : int j;
1143 :
1144 1276 : pull_varattnos((Node *) indexInfo->ii_Expressions, 1, &indexattrs);
1145 1276 : pull_varattnos((Node *) indexInfo->ii_Predicate, 1, &indexattrs);
1146 :
1147 8920 : for (int i = FirstLowInvalidHeapAttributeNumber + 1; i < 0; i++)
1148 : {
1149 7656 : if (bms_is_member(i - FirstLowInvalidHeapAttributeNumber,
1150 : indexattrs))
1151 12 : ereport(ERROR,
1152 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1153 : errmsg("index creation on system columns is not supported")));
1154 : }
1155 :
1156 : /*
1157 : * XXX Virtual generated columns in index expressions or predicates
1158 : * could be supported, but it needs support in
1159 : * RelationGetIndexExpressions() and RelationGetIndexPredicate().
1160 : */
1161 1264 : j = -1;
1162 2742 : while ((j = bms_next_member(indexattrs, j)) >= 0)
1163 : {
1164 1478 : AttrNumber attno = j + FirstLowInvalidHeapAttributeNumber;
1165 :
1166 1478 : if (TupleDescAttr(RelationGetDescr(rel), attno - 1)->attgenerated == ATTRIBUTE_GENERATED_VIRTUAL)
1167 0 : ereport(ERROR,
1168 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1169 : stmt->isconstraint ?
1170 : errmsg("unique constraints on virtual generated columns are not supported") :
1171 : errmsg("indexes on virtual generated columns are not supported")));
1172 : }
1173 : }
1174 :
1175 : /* Is index safe for others to ignore? See set_indexsafe_procflags() */
1176 61414 : safe_index = indexInfo->ii_Expressions == NIL &&
1177 30244 : indexInfo->ii_Predicate == NIL;
1178 :
1179 : /*
1180 : * Report index creation if appropriate (delay this till after most of the
1181 : * error checks)
1182 : */
1183 31170 : if (stmt->isconstraint && !quiet)
1184 : {
1185 : const char *constraint_type;
1186 :
1187 10276 : if (stmt->primary)
1188 9150 : constraint_type = "PRIMARY KEY";
1189 1126 : else if (stmt->unique)
1190 950 : constraint_type = "UNIQUE";
1191 176 : else if (stmt->excludeOpNames)
1192 176 : constraint_type = "EXCLUDE";
1193 : else
1194 : {
1195 0 : elog(ERROR, "unknown constraint type");
1196 : constraint_type = NULL; /* keep compiler quiet */
1197 : }
1198 :
1199 10276 : ereport(DEBUG1,
1200 : (errmsg_internal("%s %s will create implicit index \"%s\" for table \"%s\"",
1201 : is_alter_table ? "ALTER TABLE / ADD" : "CREATE TABLE /",
1202 : constraint_type,
1203 : indexRelationName, RelationGetRelationName(rel))));
1204 : }
1205 :
1206 : /*
1207 : * A valid stmt->oldNumber implies that we already have a built form of
1208 : * the index. The caller should also decline any index build.
1209 : */
1210 : Assert(!RelFileNumberIsValid(stmt->oldNumber) || (skip_build && !concurrent));
1211 :
1212 : /*
1213 : * Make the catalog entries for the index, including constraints. This
1214 : * step also actually builds the index, except if caller requested not to
1215 : * or in concurrent mode, in which case it'll be done later, or doing a
1216 : * partitioned index (because those don't have storage).
1217 : */
1218 31170 : flags = constr_flags = 0;
1219 31170 : if (stmt->isconstraint)
1220 10516 : flags |= INDEX_CREATE_ADD_CONSTRAINT;
1221 31170 : if (skip_build || concurrent || partitioned)
1222 15406 : flags |= INDEX_CREATE_SKIP_BUILD;
1223 31170 : if (stmt->if_not_exists)
1224 18 : flags |= INDEX_CREATE_IF_NOT_EXISTS;
1225 31170 : if (concurrent)
1226 188 : flags |= INDEX_CREATE_CONCURRENT;
1227 31170 : if (partitioned)
1228 2132 : flags |= INDEX_CREATE_PARTITIONED;
1229 31170 : if (stmt->primary)
1230 9276 : flags |= INDEX_CREATE_IS_PRIMARY;
1231 :
1232 : /*
1233 : * If the table is partitioned, and recursion was declined but partitions
1234 : * exist, mark the index as invalid.
1235 : */
1236 31170 : if (partitioned && stmt->relation && !stmt->relation->inh)
1237 : {
1238 250 : PartitionDesc pd = RelationGetPartitionDesc(rel, true);
1239 :
1240 250 : if (pd->nparts != 0)
1241 230 : flags |= INDEX_CREATE_INVALID;
1242 : }
1243 :
1244 31170 : if (stmt->deferrable)
1245 132 : constr_flags |= INDEX_CONSTR_CREATE_DEFERRABLE;
1246 31170 : if (stmt->initdeferred)
1247 38 : constr_flags |= INDEX_CONSTR_CREATE_INIT_DEFERRED;
1248 31170 : if (stmt->iswithoutoverlaps)
1249 632 : constr_flags |= INDEX_CONSTR_CREATE_WITHOUT_OVERLAPS;
1250 :
1251 : indexRelationId =
1252 31170 : index_create(rel, indexRelationName, indexRelationId, parentIndexId,
1253 : parentConstraintId,
1254 : stmt->oldNumber, indexInfo, indexColNames,
1255 : accessMethodId, tablespaceId,
1256 : collationIds, opclassIds, opclassOptions,
1257 : coloptions, NULL, reloptions,
1258 : flags, constr_flags,
1259 31170 : allowSystemTableMods, !check_rights,
1260 31170 : &createdConstraintId);
1261 :
1262 30938 : ObjectAddressSet(address, RelationRelationId, indexRelationId);
1263 :
1264 30938 : if (!OidIsValid(indexRelationId))
1265 : {
1266 : /*
1267 : * Roll back any GUC changes executed by index functions. Also revert
1268 : * to original default_tablespace if we changed it above.
1269 : */
1270 18 : AtEOXact_GUC(false, root_save_nestlevel);
1271 :
1272 : /* Restore userid and security context */
1273 18 : SetUserIdAndSecContext(root_save_userid, root_save_sec_context);
1274 :
1275 18 : table_close(rel, NoLock);
1276 :
1277 : /* If this is the top-level index, we're done */
1278 18 : if (!OidIsValid(parentIndexId))
1279 18 : pgstat_progress_end_command();
1280 :
1281 18 : return address;
1282 : }
1283 :
1284 : /*
1285 : * Roll back any GUC changes executed by index functions, and keep
1286 : * subsequent changes local to this command. This is essential if some
1287 : * index function changed a behavior-affecting GUC, e.g. search_path.
1288 : */
1289 30920 : AtEOXact_GUC(false, root_save_nestlevel);
1290 30920 : root_save_nestlevel = NewGUCNestLevel();
1291 30920 : RestrictSearchPath();
1292 :
1293 : /* Add any requested comment */
1294 30920 : if (stmt->idxcomment != NULL)
1295 78 : CreateComments(indexRelationId, RelationRelationId, 0,
1296 78 : stmt->idxcomment);
1297 :
1298 30920 : if (partitioned)
1299 : {
1300 : PartitionDesc partdesc;
1301 :
1302 : /*
1303 : * Unless caller specified to skip this step (via ONLY), process each
1304 : * partition to make sure they all contain a corresponding index.
1305 : *
1306 : * If we're called internally (no stmt->relation), recurse always.
1307 : */
1308 2132 : partdesc = RelationGetPartitionDesc(rel, true);
1309 2132 : if ((!stmt->relation || stmt->relation->inh) && partdesc->nparts > 0)
1310 : {
1311 624 : int nparts = partdesc->nparts;
1312 624 : Oid *part_oids = palloc_array(Oid, nparts);
1313 624 : bool invalidate_parent = false;
1314 : Relation parentIndex;
1315 : TupleDesc parentDesc;
1316 :
1317 : /*
1318 : * Report the total number of partitions at the start of the
1319 : * command; don't update it when being called recursively.
1320 : */
1321 624 : if (!OidIsValid(parentIndexId))
1322 : {
1323 : /*
1324 : * When called by ProcessUtilitySlow, the number of partitions
1325 : * is passed in as an optimization; but other callers pass -1
1326 : * since they don't have the value handy. This should count
1327 : * partitions the same way, ie one less than the number of
1328 : * relations find_all_inheritors reports.
1329 : *
1330 : * We assume we needn't ask find_all_inheritors to take locks,
1331 : * because that should have happened already for all callers.
1332 : * Even if it did not, this is safe as long as we don't try to
1333 : * touch the partitions here; the worst consequence would be a
1334 : * bogus progress-reporting total.
1335 : */
1336 512 : if (total_parts < 0)
1337 : {
1338 128 : List *children = find_all_inheritors(tableId, NoLock, NULL);
1339 :
1340 128 : total_parts = list_length(children) - 1;
1341 128 : list_free(children);
1342 : }
1343 :
1344 512 : pgstat_progress_update_param(PROGRESS_CREATEIDX_PARTITIONS_TOTAL,
1345 : total_parts);
1346 : }
1347 :
1348 : /* Make a local copy of partdesc->oids[], just for safety */
1349 624 : memcpy(part_oids, partdesc->oids, sizeof(Oid) * nparts);
1350 :
1351 : /*
1352 : * We'll need an IndexInfo describing the parent index. The one
1353 : * built above is almost good enough, but not quite, because (for
1354 : * example) its predicate expression if any hasn't been through
1355 : * expression preprocessing. The most reliable way to get an
1356 : * IndexInfo that will match those for child indexes is to build
1357 : * it the same way, using BuildIndexInfo().
1358 : */
1359 624 : parentIndex = index_open(indexRelationId, lockmode);
1360 624 : indexInfo = BuildIndexInfo(parentIndex);
1361 :
1362 624 : parentDesc = RelationGetDescr(rel);
1363 :
1364 : /*
1365 : * For each partition, scan all existing indexes; if one matches
1366 : * our index definition and is not already attached to some other
1367 : * parent index, attach it to the one we just created.
1368 : *
1369 : * If none matches, build a new index by calling ourselves
1370 : * recursively with the same options (except for the index name).
1371 : */
1372 1706 : for (int i = 0; i < nparts; i++)
1373 : {
1374 1106 : Oid childRelid = part_oids[i];
1375 : Relation childrel;
1376 : Oid child_save_userid;
1377 : int child_save_sec_context;
1378 : int child_save_nestlevel;
1379 : List *childidxs;
1380 : ListCell *cell;
1381 : AttrMap *attmap;
1382 1106 : bool found = false;
1383 :
1384 1106 : childrel = table_open(childRelid, lockmode);
1385 :
1386 1106 : GetUserIdAndSecContext(&child_save_userid,
1387 : &child_save_sec_context);
1388 1106 : SetUserIdAndSecContext(childrel->rd_rel->relowner,
1389 : child_save_sec_context | SECURITY_RESTRICTED_OPERATION);
1390 1106 : child_save_nestlevel = NewGUCNestLevel();
1391 1106 : RestrictSearchPath();
1392 :
1393 : /*
1394 : * Don't try to create indexes on foreign tables, though. Skip
1395 : * those if a regular index, or fail if trying to create a
1396 : * constraint index.
1397 : */
1398 1106 : if (childrel->rd_rel->relkind == RELKIND_FOREIGN_TABLE)
1399 : {
1400 18 : if (stmt->unique || stmt->primary)
1401 12 : ereport(ERROR,
1402 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
1403 : errmsg("cannot create unique index on partitioned table \"%s\"",
1404 : RelationGetRelationName(rel)),
1405 : errdetail("Table \"%s\" contains partitions that are foreign tables.",
1406 : RelationGetRelationName(rel))));
1407 :
1408 6 : AtEOXact_GUC(false, child_save_nestlevel);
1409 6 : SetUserIdAndSecContext(child_save_userid,
1410 : child_save_sec_context);
1411 6 : table_close(childrel, lockmode);
1412 6 : continue;
1413 : }
1414 :
1415 1088 : childidxs = RelationGetIndexList(childrel);
1416 : attmap =
1417 1088 : build_attrmap_by_name(RelationGetDescr(childrel),
1418 : parentDesc,
1419 : false);
1420 :
1421 1436 : foreach(cell, childidxs)
1422 : {
1423 426 : Oid cldidxid = lfirst_oid(cell);
1424 : Relation cldidx;
1425 : IndexInfo *cldIdxInfo;
1426 :
1427 : /* this index is already partition of another one */
1428 426 : if (has_superclass(cldidxid))
1429 324 : continue;
1430 :
1431 102 : cldidx = index_open(cldidxid, lockmode);
1432 102 : cldIdxInfo = BuildIndexInfo(cldidx);
1433 102 : if (CompareIndexInfo(cldIdxInfo, indexInfo,
1434 102 : cldidx->rd_indcollation,
1435 102 : parentIndex->rd_indcollation,
1436 102 : cldidx->rd_opfamily,
1437 102 : parentIndex->rd_opfamily,
1438 : attmap))
1439 : {
1440 78 : Oid cldConstrOid = InvalidOid;
1441 :
1442 : /*
1443 : * Found a match.
1444 : *
1445 : * If this index is being created in the parent
1446 : * because of a constraint, then the child needs to
1447 : * have a constraint also, so look for one. If there
1448 : * is no such constraint, this index is no good, so
1449 : * keep looking.
1450 : */
1451 78 : if (createdConstraintId != InvalidOid)
1452 : {
1453 : cldConstrOid =
1454 12 : get_relation_idx_constraint_oid(childRelid,
1455 : cldidxid);
1456 12 : if (cldConstrOid == InvalidOid)
1457 : {
1458 0 : index_close(cldidx, lockmode);
1459 0 : continue;
1460 : }
1461 : }
1462 :
1463 : /* Attach index to parent and we're done. */
1464 78 : IndexSetParentIndex(cldidx, indexRelationId);
1465 78 : if (createdConstraintId != InvalidOid)
1466 12 : ConstraintSetParentConstraint(cldConstrOid,
1467 : createdConstraintId,
1468 : childRelid);
1469 :
1470 78 : if (!cldidx->rd_index->indisvalid)
1471 18 : invalidate_parent = true;
1472 :
1473 78 : found = true;
1474 :
1475 : /*
1476 : * Report this partition as processed. Note that if
1477 : * the partition has children itself, we'd ideally
1478 : * count the children and update the progress report
1479 : * for all of them; but that seems unduly expensive.
1480 : * Instead, the progress report will act like all such
1481 : * indirect children were processed in zero time at
1482 : * the end of the command.
1483 : */
1484 78 : pgstat_progress_incr_param(PROGRESS_CREATEIDX_PARTITIONS_DONE, 1);
1485 :
1486 : /* keep lock till commit */
1487 78 : index_close(cldidx, NoLock);
1488 78 : break;
1489 : }
1490 :
1491 24 : index_close(cldidx, lockmode);
1492 : }
1493 :
1494 1088 : list_free(childidxs);
1495 1088 : AtEOXact_GUC(false, child_save_nestlevel);
1496 1088 : SetUserIdAndSecContext(child_save_userid,
1497 : child_save_sec_context);
1498 1088 : table_close(childrel, NoLock);
1499 :
1500 : /*
1501 : * If no matching index was found, create our own.
1502 : */
1503 1088 : if (!found)
1504 : {
1505 : IndexStmt *childStmt;
1506 : ObjectAddress childAddr;
1507 :
1508 : /*
1509 : * Build an IndexStmt describing the desired child index
1510 : * in the same way that we do during ATTACH PARTITION.
1511 : * Notably, we rely on generateClonedIndexStmt to produce
1512 : * a search-path-independent representation, which the
1513 : * original IndexStmt might not be.
1514 : */
1515 1010 : childStmt = generateClonedIndexStmt(NULL,
1516 : parentIndex,
1517 : attmap,
1518 : NULL);
1519 :
1520 : /*
1521 : * Recurse as the starting user ID. Callee will use that
1522 : * for permission checks, then switch again.
1523 : */
1524 : Assert(GetUserId() == child_save_userid);
1525 1010 : SetUserIdAndSecContext(root_save_userid,
1526 : root_save_sec_context);
1527 : childAddr =
1528 1010 : DefineIndex(NULL, /* original pstate not applicable */
1529 : childRelid, childStmt,
1530 : InvalidOid, /* no predefined OID */
1531 : indexRelationId, /* this is our child */
1532 : createdConstraintId,
1533 : -1,
1534 : is_alter_table, check_rights,
1535 : check_not_in_use,
1536 : skip_build, quiet);
1537 998 : SetUserIdAndSecContext(child_save_userid,
1538 : child_save_sec_context);
1539 :
1540 : /*
1541 : * Check if the index just created is valid or not, as it
1542 : * could be possible that it has been switched as invalid
1543 : * when recursing across multiple partition levels.
1544 : */
1545 998 : if (!get_index_isvalid(childAddr.objectId))
1546 6 : invalidate_parent = true;
1547 : }
1548 :
1549 1076 : free_attrmap(attmap);
1550 : }
1551 :
1552 600 : index_close(parentIndex, lockmode);
1553 :
1554 : /*
1555 : * The pg_index row we inserted for this index was marked
1556 : * indisvalid=true. But if we attached an existing index that is
1557 : * invalid, this is incorrect, so update our row to invalid too.
1558 : */
1559 600 : if (invalidate_parent)
1560 : {
1561 24 : Relation pg_index = table_open(IndexRelationId, RowExclusiveLock);
1562 : HeapTuple tup,
1563 : newtup;
1564 :
1565 24 : tup = SearchSysCache1(INDEXRELID,
1566 : ObjectIdGetDatum(indexRelationId));
1567 24 : if (!HeapTupleIsValid(tup))
1568 0 : elog(ERROR, "cache lookup failed for index %u",
1569 : indexRelationId);
1570 24 : newtup = heap_copytuple(tup);
1571 24 : ((Form_pg_index) GETSTRUCT(newtup))->indisvalid = false;
1572 24 : CatalogTupleUpdate(pg_index, &tup->t_self, newtup);
1573 24 : ReleaseSysCache(tup);
1574 24 : table_close(pg_index, RowExclusiveLock);
1575 24 : heap_freetuple(newtup);
1576 :
1577 : /*
1578 : * CCI here to make this update visible, in case this recurses
1579 : * across multiple partition levels.
1580 : */
1581 24 : CommandCounterIncrement();
1582 : }
1583 : }
1584 :
1585 : /*
1586 : * Indexes on partitioned tables are not themselves built, so we're
1587 : * done here.
1588 : */
1589 2108 : AtEOXact_GUC(false, root_save_nestlevel);
1590 2108 : SetUserIdAndSecContext(root_save_userid, root_save_sec_context);
1591 2108 : table_close(rel, NoLock);
1592 2108 : if (!OidIsValid(parentIndexId))
1593 1806 : pgstat_progress_end_command();
1594 : else
1595 : {
1596 : /* Update progress for an intermediate partitioned index itself */
1597 302 : pgstat_progress_incr_param(PROGRESS_CREATEIDX_PARTITIONS_DONE, 1);
1598 : }
1599 :
1600 2108 : return address;
1601 : }
1602 :
1603 28788 : AtEOXact_GUC(false, root_save_nestlevel);
1604 28788 : SetUserIdAndSecContext(root_save_userid, root_save_sec_context);
1605 :
1606 28788 : if (!concurrent)
1607 : {
1608 : /* Close the heap and we're done, in the non-concurrent case */
1609 28612 : table_close(rel, NoLock);
1610 :
1611 : /*
1612 : * If this is the top-level index, the command is done overall;
1613 : * otherwise, increment progress to report one child index is done.
1614 : */
1615 28612 : if (!OidIsValid(parentIndexId))
1616 25802 : pgstat_progress_end_command();
1617 : else
1618 2810 : pgstat_progress_incr_param(PROGRESS_CREATEIDX_PARTITIONS_DONE, 1);
1619 :
1620 28612 : return address;
1621 : }
1622 :
1623 : /* save lockrelid and locktag for below, then close rel */
1624 176 : heaprelid = rel->rd_lockInfo.lockRelId;
1625 176 : SET_LOCKTAG_RELATION(heaplocktag, heaprelid.dbId, heaprelid.relId);
1626 176 : table_close(rel, NoLock);
1627 :
1628 : /*
1629 : * For a concurrent build, it's important to make the catalog entries
1630 : * visible to other transactions before we start to build the index. That
1631 : * will prevent them from making incompatible HOT updates. The new index
1632 : * will be marked not indisready and not indisvalid, so that no one else
1633 : * tries to either insert into it or use it for queries.
1634 : *
1635 : * We must commit our current transaction so that the index becomes
1636 : * visible; then start another. Note that all the data structures we just
1637 : * built are lost in the commit. The only data we keep past here are the
1638 : * relation IDs.
1639 : *
1640 : * Before committing, get a session-level lock on the table, to ensure
1641 : * that neither it nor the index can be dropped before we finish. This
1642 : * cannot block, even if someone else is waiting for access, because we
1643 : * already have the same lock within our transaction.
1644 : *
1645 : * Note: we don't currently bother with a session lock on the index,
1646 : * because there are no operations that could change its state while we
1647 : * hold lock on the parent table. This might need to change later.
1648 : */
1649 176 : LockRelationIdForSession(&heaprelid, ShareUpdateExclusiveLock);
1650 :
1651 176 : PopActiveSnapshot();
1652 176 : CommitTransactionCommand();
1653 176 : StartTransactionCommand();
1654 :
1655 : /* Tell concurrent index builds to ignore us, if index qualifies */
1656 176 : if (safe_index)
1657 130 : set_indexsafe_procflags();
1658 :
1659 : /*
1660 : * The index is now visible, so we can report the OID. While on it,
1661 : * include the report for the beginning of phase 2.
1662 : */
1663 : {
1664 176 : const int progress_cols[] = {
1665 : PROGRESS_CREATEIDX_INDEX_OID,
1666 : PROGRESS_CREATEIDX_PHASE
1667 : };
1668 176 : const int64 progress_vals[] = {
1669 : indexRelationId,
1670 : PROGRESS_CREATEIDX_PHASE_WAIT_1
1671 : };
1672 :
1673 176 : pgstat_progress_update_multi_param(2, progress_cols, progress_vals);
1674 : }
1675 :
1676 : /*
1677 : * Phase 2 of concurrent index build (see comments for validate_index()
1678 : * for an overview of how this works)
1679 : *
1680 : * Now we must wait until no running transaction could have the table open
1681 : * with the old list of indexes. Use ShareLock to consider running
1682 : * transactions that hold locks that permit writing to the table. Note we
1683 : * do not need to worry about xacts that open the table for writing after
1684 : * this point; they will see the new index when they open it.
1685 : *
1686 : * Note: the reason we use actual lock acquisition here, rather than just
1687 : * checking the ProcArray and sleeping, is that deadlock is possible if
1688 : * one of the transactions in question is blocked trying to acquire an
1689 : * exclusive lock on our table. The lock code will detect deadlock and
1690 : * error out properly.
1691 : */
1692 176 : WaitForLockers(heaplocktag, ShareLock, true);
1693 :
1694 : /*
1695 : * At this moment we are sure that there are no transactions with the
1696 : * table open for write that don't have this new index in their list of
1697 : * indexes. We have waited out all the existing transactions and any new
1698 : * transaction will have the new index in its list, but the index is still
1699 : * marked as "not-ready-for-inserts". The index is consulted while
1700 : * deciding HOT-safety though. This arrangement ensures that no new HOT
1701 : * chains can be created where the new tuple and the old tuple in the
1702 : * chain have different index keys.
1703 : *
1704 : * We now take a new snapshot, and build the index using all tuples that
1705 : * are visible in this snapshot. We can be sure that any HOT updates to
1706 : * these tuples will be compatible with the index, since any updates made
1707 : * by transactions that didn't know about the index are now committed or
1708 : * rolled back. Thus, each visible tuple is either the end of its
1709 : * HOT-chain or the extension of the chain is HOT-safe for this index.
1710 : */
1711 :
1712 : /* Set ActiveSnapshot since functions in the indexes may need it */
1713 176 : PushActiveSnapshot(GetTransactionSnapshot());
1714 :
1715 : /* Perform concurrent build of index */
1716 176 : index_concurrently_build(tableId, indexRelationId);
1717 :
1718 : /* we can do away with our snapshot */
1719 158 : PopActiveSnapshot();
1720 :
1721 : /*
1722 : * Commit this transaction to make the indisready update visible.
1723 : */
1724 158 : CommitTransactionCommand();
1725 158 : StartTransactionCommand();
1726 :
1727 : /* Tell concurrent index builds to ignore us, if index qualifies */
1728 158 : if (safe_index)
1729 118 : set_indexsafe_procflags();
1730 :
1731 : /*
1732 : * Phase 3 of concurrent index build
1733 : *
1734 : * We once again wait until no transaction can have the table open with
1735 : * the index marked as read-only for updates.
1736 : */
1737 158 : pgstat_progress_update_param(PROGRESS_CREATEIDX_PHASE,
1738 : PROGRESS_CREATEIDX_PHASE_WAIT_2);
1739 158 : WaitForLockers(heaplocktag, ShareLock, true);
1740 :
1741 : /*
1742 : * Now take the "reference snapshot" that will be used by validate_index()
1743 : * to filter candidate tuples. Beware! There might still be snapshots in
1744 : * use that treat some transaction as in-progress that our reference
1745 : * snapshot treats as committed. If such a recently-committed transaction
1746 : * deleted tuples in the table, we will not include them in the index; yet
1747 : * those transactions which see the deleting one as still-in-progress will
1748 : * expect such tuples to be there once we mark the index as valid.
1749 : *
1750 : * We solve this by waiting for all endangered transactions to exit before
1751 : * we mark the index as valid.
1752 : *
1753 : * We also set ActiveSnapshot to this snap, since functions in indexes may
1754 : * need a snapshot.
1755 : */
1756 158 : snapshot = RegisterSnapshot(GetTransactionSnapshot());
1757 158 : PushActiveSnapshot(snapshot);
1758 :
1759 : /*
1760 : * Scan the index and the heap, insert any missing index entries.
1761 : */
1762 158 : validate_index(tableId, indexRelationId, snapshot);
1763 :
1764 : /*
1765 : * Drop the reference snapshot. We must do this before waiting out other
1766 : * snapshot holders, else we will deadlock against other processes also
1767 : * doing CREATE INDEX CONCURRENTLY, which would see our snapshot as one
1768 : * they must wait for. But first, save the snapshot's xmin to use as
1769 : * limitXmin for GetCurrentVirtualXIDs().
1770 : */
1771 158 : limitXmin = snapshot->xmin;
1772 :
1773 158 : PopActiveSnapshot();
1774 158 : UnregisterSnapshot(snapshot);
1775 :
1776 : /*
1777 : * The snapshot subsystem could still contain registered snapshots that
1778 : * are holding back our process's advertised xmin; in particular, if
1779 : * default_transaction_isolation = serializable, there is a transaction
1780 : * snapshot that is still active. The CatalogSnapshot is likewise a
1781 : * hazard. To ensure no deadlocks, we must commit and start yet another
1782 : * transaction, and do our wait before any snapshot has been taken in it.
1783 : */
1784 158 : CommitTransactionCommand();
1785 158 : StartTransactionCommand();
1786 :
1787 : /* Tell concurrent index builds to ignore us, if index qualifies */
1788 158 : if (safe_index)
1789 118 : set_indexsafe_procflags();
1790 :
1791 : /* We should now definitely not be advertising any xmin. */
1792 : Assert(MyProc->xmin == InvalidTransactionId);
1793 :
1794 : /*
1795 : * The index is now valid in the sense that it contains all currently
1796 : * interesting tuples. But since it might not contain tuples deleted just
1797 : * before the reference snap was taken, we have to wait out any
1798 : * transactions that might have older snapshots.
1799 : */
1800 158 : INJECTION_POINT("define-index-before-set-valid", NULL);
1801 158 : pgstat_progress_update_param(PROGRESS_CREATEIDX_PHASE,
1802 : PROGRESS_CREATEIDX_PHASE_WAIT_3);
1803 158 : WaitForOlderSnapshots(limitXmin, true);
1804 :
1805 : /*
1806 : * Updating pg_index might involve TOAST table access, so ensure we have a
1807 : * valid snapshot.
1808 : */
1809 158 : PushActiveSnapshot(GetTransactionSnapshot());
1810 :
1811 : /*
1812 : * Index can now be marked valid -- update its pg_index entry
1813 : */
1814 158 : index_set_state_flags(indexRelationId, INDEX_CREATE_SET_VALID);
1815 :
1816 158 : PopActiveSnapshot();
1817 :
1818 : /*
1819 : * The pg_index update will cause backends (including this one) to update
1820 : * relcache entries for the index itself, but we should also send a
1821 : * relcache inval on the parent table to force replanning of cached plans.
1822 : * Otherwise existing sessions might fail to use the new index where it
1823 : * would be useful. (Note that our earlier commits did not create reasons
1824 : * to replan; so relcache flush on the index itself was sufficient.)
1825 : */
1826 158 : CacheInvalidateRelcacheByRelid(heaprelid.relId);
1827 :
1828 : /*
1829 : * Last thing to do is release the session-level lock on the parent table.
1830 : */
1831 158 : UnlockRelationIdForSession(&heaprelid, ShareUpdateExclusiveLock);
1832 :
1833 158 : pgstat_progress_end_command();
1834 :
1835 158 : return address;
1836 : }
1837 :
1838 :
1839 : /*
1840 : * CheckPredicate
1841 : * Checks that the given partial-index predicate is valid.
1842 : *
1843 : * This used to also constrain the form of the predicate to forms that
1844 : * indxpath.c could do something with. However, that seems overly
1845 : * restrictive. One useful application of partial indexes is to apply
1846 : * a UNIQUE constraint across a subset of a table, and in that scenario
1847 : * any evaluable predicate will work. So accept any predicate here
1848 : * (except ones requiring a plan), and let indxpath.c fend for itself.
1849 : */
1850 : static void
1851 448 : CheckPredicate(Expr *predicate)
1852 : {
1853 : /*
1854 : * transformExpr() should have already rejected subqueries, aggregates,
1855 : * and window functions, based on the EXPR_KIND_ for a predicate.
1856 : */
1857 :
1858 : /*
1859 : * A predicate using mutable functions is probably wrong, for the same
1860 : * reasons that we don't allow an index expression to use one.
1861 : */
1862 448 : if (contain_mutable_functions_after_planning(predicate))
1863 0 : ereport(ERROR,
1864 : (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
1865 : errmsg("functions in index predicate must be marked IMMUTABLE")));
1866 448 : }
1867 :
1868 : /*
1869 : * Compute per-index-column information, including indexed column numbers
1870 : * or index expressions, opclasses and their options. Note, all output vectors
1871 : * should be allocated for all columns, including "including" ones.
1872 : *
1873 : * If the caller switched to the table owner, ddl_userid is the role for ACL
1874 : * checks reached without traversing opaque expressions. Otherwise, it's
1875 : * InvalidOid, and other ddl_* arguments are undefined.
1876 : */
1877 : static void
1878 31668 : ComputeIndexAttrs(ParseState *pstate,
1879 : IndexInfo *indexInfo,
1880 : Oid *typeOids,
1881 : Oid *collationOids,
1882 : Oid *opclassOids,
1883 : Datum *opclassOptions,
1884 : int16 *colOptions,
1885 : const List *attList, /* list of IndexElem's */
1886 : const List *exclusionOpNames,
1887 : Oid relId,
1888 : const char *accessMethodName,
1889 : Oid accessMethodId,
1890 : bool amcanorder,
1891 : bool isconstraint,
1892 : bool iswithoutoverlaps,
1893 : Oid ddl_userid,
1894 : int ddl_sec_context,
1895 : int *ddl_save_nestlevel)
1896 : {
1897 : ListCell *nextExclOp;
1898 : ListCell *lc;
1899 : int attn;
1900 31668 : int nkeycols = indexInfo->ii_NumIndexKeyAttrs;
1901 : Oid save_userid;
1902 : int save_sec_context;
1903 :
1904 : /* Allocate space for exclusion operator info, if needed */
1905 31668 : if (exclusionOpNames)
1906 : {
1907 : Assert(list_length(exclusionOpNames) == nkeycols);
1908 318 : indexInfo->ii_ExclusionOps = palloc_array(Oid, nkeycols);
1909 318 : indexInfo->ii_ExclusionProcs = palloc_array(Oid, nkeycols);
1910 318 : indexInfo->ii_ExclusionStrats = palloc_array(uint16, nkeycols);
1911 318 : nextExclOp = list_head(exclusionOpNames);
1912 : }
1913 : else
1914 31350 : nextExclOp = NULL;
1915 :
1916 : /*
1917 : * If this is a WITHOUT OVERLAPS constraint, we need space for exclusion
1918 : * ops, but we don't need to parse anything, so we can let nextExclOp be
1919 : * NULL. Note that for partitions/inheriting/LIKE, exclusionOpNames will
1920 : * be set, so we already allocated above.
1921 : */
1922 31668 : if (iswithoutoverlaps)
1923 : {
1924 632 : if (exclusionOpNames == NIL)
1925 : {
1926 554 : indexInfo->ii_ExclusionOps = palloc_array(Oid, nkeycols);
1927 554 : indexInfo->ii_ExclusionProcs = palloc_array(Oid, nkeycols);
1928 554 : indexInfo->ii_ExclusionStrats = palloc_array(uint16, nkeycols);
1929 : }
1930 632 : nextExclOp = NULL;
1931 : }
1932 :
1933 31668 : if (OidIsValid(ddl_userid))
1934 31564 : GetUserIdAndSecContext(&save_userid, &save_sec_context);
1935 :
1936 : /*
1937 : * process attributeList
1938 : */
1939 31668 : attn = 0;
1940 75836 : foreach(lc, attList)
1941 : {
1942 44386 : IndexElem *attribute = (IndexElem *) lfirst(lc);
1943 : Oid atttype;
1944 : Oid attcollation;
1945 :
1946 : /*
1947 : * Process the column-or-expression to be indexed.
1948 : */
1949 44386 : if (attribute->name != NULL)
1950 : {
1951 : /* Simple index attribute */
1952 : HeapTuple atttuple;
1953 : Form_pg_attribute attform;
1954 :
1955 : Assert(attribute->expr == NULL);
1956 43246 : atttuple = SearchSysCacheAttName(relId, attribute->name);
1957 43246 : if (!HeapTupleIsValid(atttuple))
1958 : {
1959 : /* difference in error message spellings is historical */
1960 30 : if (isconstraint)
1961 18 : ereport(ERROR,
1962 : (errcode(ERRCODE_UNDEFINED_COLUMN),
1963 : errmsg("column \"%s\" named in key does not exist",
1964 : attribute->name),
1965 : parser_errposition(pstate, attribute->location)));
1966 : else
1967 12 : ereport(ERROR,
1968 : (errcode(ERRCODE_UNDEFINED_COLUMN),
1969 : errmsg("column \"%s\" does not exist",
1970 : attribute->name),
1971 : parser_errposition(pstate, attribute->location)));
1972 : }
1973 43216 : attform = (Form_pg_attribute) GETSTRUCT(atttuple);
1974 43216 : indexInfo->ii_IndexAttrNumbers[attn] = attform->attnum;
1975 43216 : atttype = attform->atttypid;
1976 43216 : attcollation = attform->attcollation;
1977 43216 : ReleaseSysCache(atttuple);
1978 : }
1979 : else
1980 : {
1981 : /* Index expression */
1982 1140 : Node *expr = attribute->expr;
1983 :
1984 : Assert(expr != NULL);
1985 :
1986 1140 : if (attn >= nkeycols)
1987 0 : ereport(ERROR,
1988 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1989 : errmsg("expressions are not supported in included columns"),
1990 : parser_errposition(pstate, attribute->location)));
1991 1140 : atttype = exprType(expr);
1992 1140 : attcollation = exprCollation(expr);
1993 :
1994 : /*
1995 : * Strip any top-level COLLATE clause. This ensures that we treat
1996 : * "x COLLATE y" and "(x COLLATE y)" alike.
1997 : */
1998 1170 : while (IsA(expr, CollateExpr))
1999 30 : expr = (Node *) ((CollateExpr *) expr)->arg;
2000 :
2001 1140 : if (IsA(expr, Var) &&
2002 12 : ((Var *) expr)->varattno != InvalidAttrNumber)
2003 : {
2004 : /*
2005 : * User wrote "(column)" or "(column COLLATE something)".
2006 : * Treat it like simple attribute anyway.
2007 : */
2008 12 : indexInfo->ii_IndexAttrNumbers[attn] = ((Var *) expr)->varattno;
2009 : }
2010 : else
2011 : {
2012 1128 : indexInfo->ii_IndexAttrNumbers[attn] = 0; /* marks expression */
2013 1128 : indexInfo->ii_Expressions = lappend(indexInfo->ii_Expressions,
2014 : expr);
2015 :
2016 : /*
2017 : * transformExpr() should have already rejected subqueries,
2018 : * aggregates, and window functions, based on the EXPR_KIND_
2019 : * for an index expression.
2020 : */
2021 :
2022 : /*
2023 : * An expression using mutable functions is probably wrong,
2024 : * since if you aren't going to get the same result for the
2025 : * same data every time, it's not clear what the index entries
2026 : * mean at all.
2027 : */
2028 1128 : if (contain_mutable_functions_after_planning((Expr *) expr))
2029 168 : ereport(ERROR,
2030 : (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
2031 : errmsg("functions in index expression must be marked IMMUTABLE"),
2032 : parser_errposition(pstate, attribute->location)));
2033 : }
2034 : }
2035 :
2036 44188 : typeOids[attn] = atttype;
2037 :
2038 : /*
2039 : * Included columns have no collation, no opclass and no ordering
2040 : * options.
2041 : */
2042 44188 : if (attn >= nkeycols)
2043 : {
2044 644 : if (attribute->collation)
2045 0 : ereport(ERROR,
2046 : (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
2047 : errmsg("including column does not support a collation"),
2048 : parser_errposition(pstate, attribute->location)));
2049 644 : if (attribute->opclass)
2050 0 : ereport(ERROR,
2051 : (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
2052 : errmsg("including column does not support an operator class"),
2053 : parser_errposition(pstate, attribute->location)));
2054 644 : if (attribute->ordering != SORTBY_DEFAULT)
2055 0 : ereport(ERROR,
2056 : (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
2057 : errmsg("including column does not support ASC/DESC options"),
2058 : parser_errposition(pstate, attribute->location)));
2059 644 : if (attribute->nulls_ordering != SORTBY_NULLS_DEFAULT)
2060 0 : ereport(ERROR,
2061 : (errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
2062 : errmsg("including column does not support NULLS FIRST/LAST options"),
2063 : parser_errposition(pstate, attribute->location)));
2064 :
2065 644 : opclassOids[attn] = InvalidOid;
2066 644 : opclassOptions[attn] = (Datum) 0;
2067 644 : colOptions[attn] = 0;
2068 644 : collationOids[attn] = InvalidOid;
2069 644 : attn++;
2070 :
2071 644 : continue;
2072 : }
2073 :
2074 : /*
2075 : * Apply collation override if any. Use of ddl_userid is necessary
2076 : * due to ACL checks therein, and it's safe because collations don't
2077 : * contain opaque expressions (or non-opaque expressions).
2078 : */
2079 43544 : if (attribute->collation)
2080 : {
2081 118 : if (OidIsValid(ddl_userid))
2082 : {
2083 118 : AtEOXact_GUC(false, *ddl_save_nestlevel);
2084 118 : SetUserIdAndSecContext(ddl_userid, ddl_sec_context);
2085 : }
2086 118 : attcollation = get_collation_oid(attribute->collation, false);
2087 116 : if (OidIsValid(ddl_userid))
2088 : {
2089 116 : SetUserIdAndSecContext(save_userid, save_sec_context);
2090 116 : *ddl_save_nestlevel = NewGUCNestLevel();
2091 116 : RestrictSearchPath();
2092 : }
2093 : }
2094 :
2095 : /*
2096 : * Check we have a collation iff it's a collatable type. The only
2097 : * expected failures here are (1) COLLATE applied to a noncollatable
2098 : * type, or (2) index expression had an unresolved collation. But we
2099 : * might as well code this to be a complete consistency check.
2100 : */
2101 43542 : if (type_is_collatable(atttype))
2102 : {
2103 6730 : if (!OidIsValid(attcollation))
2104 0 : ereport(ERROR,
2105 : (errcode(ERRCODE_INDETERMINATE_COLLATION),
2106 : errmsg("could not determine which collation to use for index expression"),
2107 : errhint("Use the COLLATE clause to set the collation explicitly."),
2108 : parser_errposition(pstate, attribute->location)));
2109 : }
2110 : else
2111 : {
2112 36812 : if (OidIsValid(attcollation))
2113 12 : ereport(ERROR,
2114 : (errcode(ERRCODE_DATATYPE_MISMATCH),
2115 : errmsg("collations are not supported by type %s",
2116 : format_type_be(atttype)),
2117 : parser_errposition(pstate, attribute->location)));
2118 : }
2119 :
2120 43530 : collationOids[attn] = attcollation;
2121 :
2122 : /*
2123 : * Identify the opclass to use. Use of ddl_userid is necessary due to
2124 : * ACL checks therein. This is safe despite opclasses containing
2125 : * opaque expressions (specifically, functions), because only
2126 : * superusers can define opclasses.
2127 : */
2128 43530 : if (OidIsValid(ddl_userid))
2129 : {
2130 43420 : AtEOXact_GUC(false, *ddl_save_nestlevel);
2131 43420 : SetUserIdAndSecContext(ddl_userid, ddl_sec_context);
2132 : }
2133 43530 : opclassOids[attn] = ResolveOpClass(attribute->opclass,
2134 : atttype,
2135 : accessMethodName,
2136 : accessMethodId);
2137 43524 : if (OidIsValid(ddl_userid))
2138 : {
2139 43414 : SetUserIdAndSecContext(save_userid, save_sec_context);
2140 43414 : *ddl_save_nestlevel = NewGUCNestLevel();
2141 43414 : RestrictSearchPath();
2142 : }
2143 :
2144 : /*
2145 : * Identify the exclusion operator, if any.
2146 : */
2147 43524 : if (nextExclOp)
2148 : {
2149 348 : List *opname = (List *) lfirst(nextExclOp);
2150 : Oid opid;
2151 : Oid opfamily;
2152 : int strat;
2153 :
2154 : /*
2155 : * Find the operator --- it must accept the column datatype
2156 : * without runtime coercion (but binary compatibility is OK).
2157 : * Operators contain opaque expressions (specifically, functions).
2158 : * compatible_oper_opid() boils down to oper() and
2159 : * IsBinaryCoercible(). PostgreSQL would have security problems
2160 : * elsewhere if oper() started calling opaque expressions.
2161 : */
2162 348 : if (OidIsValid(ddl_userid))
2163 : {
2164 348 : AtEOXact_GUC(false, *ddl_save_nestlevel);
2165 348 : SetUserIdAndSecContext(ddl_userid, ddl_sec_context);
2166 : }
2167 348 : opid = compatible_oper_opid(opname, atttype, atttype, false);
2168 348 : if (OidIsValid(ddl_userid))
2169 : {
2170 348 : SetUserIdAndSecContext(save_userid, save_sec_context);
2171 348 : *ddl_save_nestlevel = NewGUCNestLevel();
2172 348 : RestrictSearchPath();
2173 : }
2174 :
2175 : /*
2176 : * Only allow commutative operators to be used in exclusion
2177 : * constraints. If X conflicts with Y, but Y does not conflict
2178 : * with X, bad things will happen.
2179 : */
2180 348 : if (get_commutator(opid) != opid)
2181 0 : ereport(ERROR,
2182 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
2183 : errmsg("operator %s is not commutative",
2184 : format_operator(opid)),
2185 : errdetail("Only commutative operators can be used in exclusion constraints."),
2186 : parser_errposition(pstate, attribute->location)));
2187 :
2188 : /*
2189 : * Operator must be a member of the right opfamily, too
2190 : */
2191 348 : opfamily = get_opclass_family(opclassOids[attn]);
2192 348 : strat = get_op_opfamily_strategy(opid, opfamily);
2193 348 : if (strat == 0)
2194 0 : ereport(ERROR,
2195 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
2196 : errmsg("operator %s is not a member of operator family \"%s\"",
2197 : format_operator(opid),
2198 : get_opfamily_name(opfamily, false)),
2199 : errdetail("The exclusion operator must be related to the index operator class for the constraint."),
2200 : parser_errposition(pstate, attribute->location)));
2201 :
2202 348 : indexInfo->ii_ExclusionOps[attn] = opid;
2203 348 : indexInfo->ii_ExclusionProcs[attn] = get_opcode(opid);
2204 348 : indexInfo->ii_ExclusionStrats[attn] = strat;
2205 348 : nextExclOp = lnext(exclusionOpNames, nextExclOp);
2206 : }
2207 43176 : else if (iswithoutoverlaps)
2208 : {
2209 : CompareType cmptype;
2210 : StrategyNumber strat;
2211 : Oid opid;
2212 :
2213 1298 : if (attn == nkeycols - 1)
2214 632 : cmptype = COMPARE_OVERLAP;
2215 : else
2216 666 : cmptype = COMPARE_EQ;
2217 1298 : GetOperatorFromCompareType(opclassOids[attn], InvalidOid, cmptype, &opid, &strat);
2218 1298 : indexInfo->ii_ExclusionOps[attn] = opid;
2219 1298 : indexInfo->ii_ExclusionProcs[attn] = get_opcode(opid);
2220 1298 : indexInfo->ii_ExclusionStrats[attn] = strat;
2221 : }
2222 :
2223 : /*
2224 : * Set up the per-column options (indoption field). For now, this is
2225 : * zero for any un-ordered index, while ordered indexes have DESC and
2226 : * NULLS FIRST/LAST options.
2227 : */
2228 43524 : colOptions[attn] = 0;
2229 43524 : if (amcanorder)
2230 : {
2231 : /* default ordering is ASC */
2232 39320 : if (attribute->ordering == SORTBY_DESC)
2233 42 : colOptions[attn] |= INDOPTION_DESC;
2234 : /* default null ordering is LAST for ASC, FIRST for DESC */
2235 39320 : if (attribute->nulls_ordering == SORTBY_NULLS_DEFAULT)
2236 : {
2237 39290 : if (attribute->ordering == SORTBY_DESC)
2238 30 : colOptions[attn] |= INDOPTION_NULLS_FIRST;
2239 : }
2240 30 : else if (attribute->nulls_ordering == SORTBY_NULLS_FIRST)
2241 12 : colOptions[attn] |= INDOPTION_NULLS_FIRST;
2242 : }
2243 : else
2244 : {
2245 : /* index AM does not support ordering */
2246 4204 : if (attribute->ordering != SORTBY_DEFAULT)
2247 0 : ereport(ERROR,
2248 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
2249 : errmsg("access method \"%s\" does not support ASC/DESC options",
2250 : accessMethodName),
2251 : parser_errposition(pstate, attribute->location)));
2252 4204 : if (attribute->nulls_ordering != SORTBY_NULLS_DEFAULT)
2253 0 : ereport(ERROR,
2254 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
2255 : errmsg("access method \"%s\" does not support NULLS FIRST/LAST options",
2256 : accessMethodName),
2257 : parser_errposition(pstate, attribute->location)));
2258 : }
2259 :
2260 : /* Set up the per-column opclass options (attoptions field). */
2261 43524 : if (attribute->opclassopts)
2262 : {
2263 : Assert(attn < nkeycols);
2264 :
2265 144 : opclassOptions[attn] =
2266 144 : transformRelOptions((Datum) 0, attribute->opclassopts,
2267 : NULL, NULL, false, false);
2268 : }
2269 : else
2270 43380 : opclassOptions[attn] = (Datum) 0;
2271 :
2272 43524 : attn++;
2273 : }
2274 31450 : }
2275 :
2276 : /*
2277 : * Resolve possibly-defaulted operator class specification
2278 : *
2279 : * Note: This is used to resolve operator class specifications in index and
2280 : * partition key definitions.
2281 : */
2282 : Oid
2283 43668 : ResolveOpClass(const List *opclass, Oid attrType,
2284 : const char *accessMethodName, Oid accessMethodId)
2285 : {
2286 : char *schemaname;
2287 : char *opcname;
2288 : HeapTuple tuple;
2289 : Form_pg_opclass opform;
2290 : Oid opClassId,
2291 : opInputType;
2292 :
2293 43668 : if (opclass == NIL)
2294 : {
2295 : /* no operator class specified, so find the default */
2296 21476 : opClassId = GetDefaultOpClass(attrType, accessMethodId);
2297 21476 : if (!OidIsValid(opClassId))
2298 6 : ereport(ERROR,
2299 : (errcode(ERRCODE_UNDEFINED_OBJECT),
2300 : errmsg("data type %s has no default operator class for access method \"%s\"",
2301 : format_type_be(attrType), accessMethodName),
2302 : errhint("You must specify an operator class for the index or define a default operator class for the data type.")));
2303 21470 : return opClassId;
2304 : }
2305 :
2306 : /*
2307 : * Specific opclass name given, so look up the opclass.
2308 : */
2309 :
2310 : /* deconstruct the name list */
2311 22192 : DeconstructQualifiedName(opclass, &schemaname, &opcname);
2312 :
2313 22192 : if (schemaname)
2314 : {
2315 : /* Look in specific schema only */
2316 : Oid namespaceId;
2317 :
2318 28 : namespaceId = LookupExplicitNamespace(schemaname, false);
2319 28 : tuple = SearchSysCache3(CLAAMNAMENSP,
2320 : ObjectIdGetDatum(accessMethodId),
2321 : PointerGetDatum(opcname),
2322 : ObjectIdGetDatum(namespaceId));
2323 : }
2324 : else
2325 : {
2326 : /* Unqualified opclass name, so search the search path */
2327 22164 : opClassId = OpclassnameGetOpcid(accessMethodId, opcname);
2328 22164 : if (!OidIsValid(opClassId))
2329 12 : ereport(ERROR,
2330 : (errcode(ERRCODE_UNDEFINED_OBJECT),
2331 : errmsg("operator class \"%s\" does not exist for access method \"%s\"",
2332 : opcname, accessMethodName)));
2333 22152 : tuple = SearchSysCache1(CLAOID, ObjectIdGetDatum(opClassId));
2334 : }
2335 :
2336 22180 : if (!HeapTupleIsValid(tuple))
2337 0 : ereport(ERROR,
2338 : (errcode(ERRCODE_UNDEFINED_OBJECT),
2339 : errmsg("operator class \"%s\" does not exist for access method \"%s\"",
2340 : NameListToString(opclass), accessMethodName)));
2341 :
2342 : /*
2343 : * Verify that the index operator class accepts this datatype. Note we
2344 : * will accept binary compatibility.
2345 : */
2346 22180 : opform = (Form_pg_opclass) GETSTRUCT(tuple);
2347 22180 : opClassId = opform->oid;
2348 22180 : opInputType = opform->opcintype;
2349 :
2350 22180 : if (!IsBinaryCoercible(attrType, opInputType))
2351 0 : ereport(ERROR,
2352 : (errcode(ERRCODE_DATATYPE_MISMATCH),
2353 : errmsg("operator class \"%s\" does not accept data type %s",
2354 : NameListToString(opclass), format_type_be(attrType))));
2355 :
2356 22180 : ReleaseSysCache(tuple);
2357 :
2358 22180 : return opClassId;
2359 : }
2360 :
2361 : /*
2362 : * GetDefaultOpClass
2363 : *
2364 : * Given the OIDs of a datatype and an access method, find the default
2365 : * operator class, if any. Returns InvalidOid if there is none.
2366 : */
2367 : Oid
2368 117850 : GetDefaultOpClass(Oid type_id, Oid am_id)
2369 : {
2370 117850 : Oid result = InvalidOid;
2371 117850 : int nexact = 0;
2372 117850 : int ncompatible = 0;
2373 117850 : int ncompatiblepreferred = 0;
2374 : Relation rel;
2375 : ScanKeyData skey[1];
2376 : SysScanDesc scan;
2377 : HeapTuple tup;
2378 : TYPCATEGORY tcategory;
2379 :
2380 : /* If it's a domain, look at the base type instead */
2381 117850 : type_id = getBaseType(type_id);
2382 :
2383 117850 : tcategory = TypeCategory(type_id);
2384 :
2385 : /*
2386 : * We scan through all the opclasses available for the access method,
2387 : * looking for one that is marked default and matches the target type
2388 : * (either exactly or binary-compatibly, but prefer an exact match).
2389 : *
2390 : * We could find more than one binary-compatible match. If just one is
2391 : * for a preferred type, use that one; otherwise we fail, forcing the user
2392 : * to specify which one he wants. (The preferred-type special case is a
2393 : * kluge for varchar: it's binary-compatible to both text and bpchar, so
2394 : * we need a tiebreaker.) If we find more than one exact match, then
2395 : * someone put bogus entries in pg_opclass.
2396 : */
2397 117850 : rel = table_open(OperatorClassRelationId, AccessShareLock);
2398 :
2399 117850 : ScanKeyInit(&skey[0],
2400 : Anum_pg_opclass_opcmethod,
2401 : BTEqualStrategyNumber, F_OIDEQ,
2402 : ObjectIdGetDatum(am_id));
2403 :
2404 117850 : scan = systable_beginscan(rel, OpclassAmNameNspIndexId, true,
2405 : NULL, 1, skey);
2406 :
2407 5275588 : while (HeapTupleIsValid(tup = systable_getnext(scan)))
2408 : {
2409 5157738 : Form_pg_opclass opclass = (Form_pg_opclass) GETSTRUCT(tup);
2410 :
2411 : /* ignore altogether if not a default opclass */
2412 5157738 : if (!opclass->opcdefault)
2413 743230 : continue;
2414 4414508 : if (opclass->opcintype == type_id)
2415 : {
2416 104656 : nexact++;
2417 104656 : result = opclass->oid;
2418 : }
2419 6462642 : else if (nexact == 0 &&
2420 2152790 : IsBinaryCoercible(type_id, opclass->opcintype))
2421 : {
2422 25274 : if (IsPreferredType(tcategory, opclass->opcintype))
2423 : {
2424 2110 : ncompatiblepreferred++;
2425 2110 : result = opclass->oid;
2426 : }
2427 23164 : else if (ncompatiblepreferred == 0)
2428 : {
2429 23164 : ncompatible++;
2430 23164 : result = opclass->oid;
2431 : }
2432 : }
2433 : }
2434 :
2435 117850 : systable_endscan(scan);
2436 :
2437 117850 : table_close(rel, AccessShareLock);
2438 :
2439 : /* raise error if pg_opclass contains inconsistent data */
2440 117850 : if (nexact > 1)
2441 0 : ereport(ERROR,
2442 : (errcode(ERRCODE_DUPLICATE_OBJECT),
2443 : errmsg("there are multiple default operator classes for data type %s",
2444 : format_type_be(type_id))));
2445 :
2446 117850 : if (nexact == 1 ||
2447 11086 : ncompatiblepreferred == 1 ||
2448 11086 : (ncompatiblepreferred == 0 && ncompatible == 1))
2449 116582 : return result;
2450 :
2451 1268 : return InvalidOid;
2452 : }
2453 :
2454 : /*
2455 : * GetOperatorFromCompareType
2456 : *
2457 : * opclass - the opclass to use
2458 : * rhstype - the type for the right-hand side, or InvalidOid to use the type of the given opclass.
2459 : * cmptype - kind of operator to find
2460 : * opid - holds the operator we found
2461 : * strat - holds the output strategy number
2462 : *
2463 : * Finds an operator from a CompareType. This is used for temporal index
2464 : * constraints (and other temporal features) to look up equality and overlaps
2465 : * operators. We ask an opclass support function to translate from the
2466 : * compare type to the internal strategy numbers. Raises ERROR on search
2467 : * failure.
2468 : */
2469 : void
2470 2122 : GetOperatorFromCompareType(Oid opclass, Oid rhstype, CompareType cmptype,
2471 : Oid *opid, StrategyNumber *strat)
2472 : {
2473 : Oid amid;
2474 : Oid opfamily;
2475 : Oid opcintype;
2476 :
2477 : Assert(cmptype == COMPARE_EQ || cmptype == COMPARE_OVERLAP || cmptype == COMPARE_CONTAINED_BY);
2478 :
2479 : /*
2480 : * Use the opclass to get the opfamily, opcintype, and access method. If
2481 : * any of this fails, quit early.
2482 : */
2483 2122 : if (!get_opclass_opfamily_and_input_type(opclass, &opfamily, &opcintype))
2484 0 : elog(ERROR, "cache lookup failed for opclass %u", opclass);
2485 :
2486 2122 : amid = get_opclass_method(opclass);
2487 :
2488 : /*
2489 : * Ask the index AM to translate to its internal stratnum
2490 : */
2491 2122 : *strat = IndexAmTranslateCompareType(cmptype, amid, opfamily, true);
2492 2122 : if (*strat == InvalidStrategy)
2493 0 : ereport(ERROR,
2494 : errcode(ERRCODE_UNDEFINED_OBJECT),
2495 : cmptype == COMPARE_EQ ? errmsg("could not identify an equality operator for type %s", format_type_be(opcintype)) :
2496 : cmptype == COMPARE_OVERLAP ? errmsg("could not identify an overlaps operator for type %s", format_type_be(opcintype)) :
2497 : cmptype == COMPARE_CONTAINED_BY ? errmsg("could not identify a contained-by operator for type %s", format_type_be(opcintype)) : 0,
2498 : errdetail("Could not translate compare type %d for operator family \"%s\" of access method \"%s\".",
2499 : cmptype, get_opfamily_name(opfamily, false), get_am_name(amid)));
2500 :
2501 : /*
2502 : * We parameterize rhstype so foreign keys can ask for a <@ operator whose
2503 : * rhs matches the aggregate function. For example range_agg returns
2504 : * anymultirange.
2505 : */
2506 2122 : if (!OidIsValid(rhstype))
2507 1710 : rhstype = opcintype;
2508 2122 : *opid = get_opfamily_member(opfamily, opcintype, rhstype, *strat);
2509 :
2510 2122 : if (!OidIsValid(*opid))
2511 0 : ereport(ERROR,
2512 : errcode(ERRCODE_UNDEFINED_OBJECT),
2513 : cmptype == COMPARE_EQ ? errmsg("could not identify an equality operator for type %s", format_type_be(opcintype)) :
2514 : cmptype == COMPARE_OVERLAP ? errmsg("could not identify an overlaps operator for type %s", format_type_be(opcintype)) :
2515 : cmptype == COMPARE_CONTAINED_BY ? errmsg("could not identify a contained-by operator for type %s", format_type_be(opcintype)) : 0,
2516 : errdetail("There is no suitable operator in operator family \"%s\" for access method \"%s\".",
2517 : get_opfamily_name(opfamily, false), get_am_name(amid)));
2518 2122 : }
2519 :
2520 : /*
2521 : * makeObjectName()
2522 : *
2523 : * Create a name for an implicitly created index, sequence, constraint,
2524 : * extended statistics, etc.
2525 : *
2526 : * The parameters are typically: the original table name, the original field
2527 : * name, and a "type" string (such as "seq" or "pkey"). The field name
2528 : * and/or type can be NULL if not relevant.
2529 : *
2530 : * The result is a palloc'd string.
2531 : *
2532 : * The basic result we want is "name1_name2_label", omitting "_name2" or
2533 : * "_label" when those parameters are NULL. However, we must generate
2534 : * a name with less than NAMEDATALEN characters! So, we truncate one or
2535 : * both names if necessary to make a short-enough string. The label part
2536 : * is never truncated (so it had better be reasonably short).
2537 : *
2538 : * The caller is responsible for checking uniqueness of the generated
2539 : * name and retrying as needed; retrying will be done by altering the
2540 : * "label" string (which is why we never truncate that part).
2541 : */
2542 : char *
2543 119788 : makeObjectName(const char *name1, const char *name2, const char *label)
2544 : {
2545 : char *name;
2546 119788 : int overhead = 0; /* chars needed for label and underscores */
2547 : int availchars; /* chars available for name(s) */
2548 : int name1chars; /* chars allocated to name1 */
2549 : int name2chars; /* chars allocated to name2 */
2550 : int ndx;
2551 :
2552 119788 : name1chars = strlen(name1);
2553 119788 : if (name2)
2554 : {
2555 105408 : name2chars = strlen(name2);
2556 105408 : overhead++; /* allow for separating underscore */
2557 : }
2558 : else
2559 14380 : name2chars = 0;
2560 119788 : if (label)
2561 45126 : overhead += strlen(label) + 1;
2562 :
2563 119788 : availchars = NAMEDATALEN - 1 - overhead;
2564 : Assert(availchars > 0); /* else caller chose a bad label */
2565 :
2566 : /*
2567 : * If we must truncate, preferentially truncate the longer name. This
2568 : * logic could be expressed without a loop, but it's simple and obvious as
2569 : * a loop.
2570 : */
2571 119854 : while (name1chars + name2chars > availchars)
2572 : {
2573 66 : if (name1chars > name2chars)
2574 0 : name1chars--;
2575 : else
2576 66 : name2chars--;
2577 : }
2578 :
2579 119788 : name1chars = pg_mbcliplen(name1, name1chars, name1chars);
2580 119788 : if (name2)
2581 105408 : name2chars = pg_mbcliplen(name2, name2chars, name2chars);
2582 :
2583 : /* Now construct the string using the chosen lengths */
2584 119788 : name = palloc(name1chars + name2chars + overhead + 1);
2585 119788 : memcpy(name, name1, name1chars);
2586 119788 : ndx = name1chars;
2587 119788 : if (name2)
2588 : {
2589 105408 : name[ndx++] = '_';
2590 105408 : memcpy(name + ndx, name2, name2chars);
2591 105408 : ndx += name2chars;
2592 : }
2593 119788 : if (label)
2594 : {
2595 45126 : name[ndx++] = '_';
2596 45126 : strcpy(name + ndx, label);
2597 : }
2598 : else
2599 74662 : name[ndx] = '\0';
2600 :
2601 119788 : return name;
2602 : }
2603 :
2604 : /*
2605 : * Select a nonconflicting name for a new relation. This is ordinarily
2606 : * used to choose index names (which is why it's here) but it can also
2607 : * be used for sequences, or any autogenerated relation kind.
2608 : *
2609 : * name1, name2, and label are used the same way as for makeObjectName(),
2610 : * except that the label can't be NULL; digits will be appended to the label
2611 : * if needed to create a name that is unique within the specified namespace.
2612 : *
2613 : * If isconstraint is true, we also avoid choosing a name matching any
2614 : * existing constraint in the same namespace. (This is stricter than what
2615 : * Postgres itself requires, but the SQL standard says that constraint names
2616 : * should be unique within schemas, so we follow that for autogenerated
2617 : * constraint names.)
2618 : *
2619 : * Note: it is theoretically possible to get a collision anyway, if someone
2620 : * else chooses the same name concurrently. We shorten the race condition
2621 : * window by checking for conflicting relations using SnapshotDirty, but
2622 : * that doesn't close the window entirely. This is fairly unlikely to be
2623 : * a problem in practice, especially if one is holding an exclusive lock on
2624 : * the relation identified by name1. However, if choosing multiple names
2625 : * within a single command, you'd better create the new object and do
2626 : * CommandCounterIncrement before choosing the next one!
2627 : *
2628 : * Returns a palloc'd string.
2629 : */
2630 : char *
2631 14726 : ChooseRelationName(const char *name1, const char *name2,
2632 : const char *label, Oid namespaceid,
2633 : bool isconstraint)
2634 : {
2635 14726 : int pass = 0;
2636 14726 : char *relname = NULL;
2637 : char modlabel[NAMEDATALEN];
2638 : SnapshotData SnapshotDirty;
2639 : Relation pgclassrel;
2640 :
2641 : /* prepare to search pg_class with a dirty snapshot */
2642 14726 : InitDirtySnapshot(SnapshotDirty);
2643 14726 : pgclassrel = table_open(RelationRelationId, AccessShareLock);
2644 :
2645 : /* try the unmodified label first */
2646 14726 : strlcpy(modlabel, label, sizeof(modlabel));
2647 :
2648 : for (;;)
2649 1098 : {
2650 : ScanKeyData key[2];
2651 : SysScanDesc scan;
2652 : bool collides;
2653 :
2654 15824 : relname = makeObjectName(name1, name2, modlabel);
2655 :
2656 : /* is there any conflicting relation name? */
2657 15824 : ScanKeyInit(&key[0],
2658 : Anum_pg_class_relname,
2659 : BTEqualStrategyNumber, F_NAMEEQ,
2660 : CStringGetDatum(relname));
2661 15824 : ScanKeyInit(&key[1],
2662 : Anum_pg_class_relnamespace,
2663 : BTEqualStrategyNumber, F_OIDEQ,
2664 : ObjectIdGetDatum(namespaceid));
2665 :
2666 15824 : scan = systable_beginscan(pgclassrel, ClassNameNspIndexId,
2667 : true /* indexOK */ ,
2668 : &SnapshotDirty,
2669 : 2, key);
2670 :
2671 15824 : collides = HeapTupleIsValid(systable_getnext(scan));
2672 :
2673 15824 : systable_endscan(scan);
2674 :
2675 : /* break out of loop if no conflict */
2676 15824 : if (!collides)
2677 : {
2678 14732 : if (!isconstraint ||
2679 9294 : !ConstraintNameExists(relname, namespaceid))
2680 : break;
2681 : }
2682 :
2683 : /* found a conflict, so try a new name component */
2684 1098 : pfree(relname);
2685 1098 : snprintf(modlabel, sizeof(modlabel), "%s%d", label, ++pass);
2686 : }
2687 :
2688 14726 : table_close(pgclassrel, AccessShareLock);
2689 :
2690 14726 : return relname;
2691 : }
2692 :
2693 : /*
2694 : * Select the name to be used for an index.
2695 : *
2696 : * The argument list is pretty ad-hoc :-(
2697 : */
2698 : static char *
2699 12316 : ChooseIndexName(const char *tabname, Oid namespaceId,
2700 : const List *colnames, const List *exclusionOpNames,
2701 : bool primary, bool isconstraint)
2702 : {
2703 : char *indexname;
2704 :
2705 12316 : if (primary)
2706 : {
2707 : /* the primary key's name does not depend on the specific column(s) */
2708 8316 : indexname = ChooseRelationName(tabname,
2709 : NULL,
2710 : "pkey",
2711 : namespaceId,
2712 : true);
2713 : }
2714 4000 : else if (exclusionOpNames != NIL)
2715 : {
2716 206 : indexname = ChooseRelationName(tabname,
2717 206 : ChooseIndexNameAddition(colnames),
2718 : "excl",
2719 : namespaceId,
2720 : true);
2721 : }
2722 3794 : else if (isconstraint)
2723 : {
2724 766 : indexname = ChooseRelationName(tabname,
2725 766 : ChooseIndexNameAddition(colnames),
2726 : "key",
2727 : namespaceId,
2728 : true);
2729 : }
2730 : else
2731 : {
2732 3028 : indexname = ChooseRelationName(tabname,
2733 3028 : ChooseIndexNameAddition(colnames),
2734 : "idx",
2735 : namespaceId,
2736 : false);
2737 : }
2738 :
2739 12316 : return indexname;
2740 : }
2741 :
2742 : /*
2743 : * Generate "name2" for a new index given the list of column names for it
2744 : * (as produced by ChooseIndexColumnNames). This will be passed to
2745 : * ChooseRelationName along with the parent table name and a suitable label.
2746 : *
2747 : * We know that less than NAMEDATALEN characters will actually be used,
2748 : * so we can truncate the result once we've generated that many.
2749 : *
2750 : * XXX See also ChooseForeignKeyConstraintNameAddition and
2751 : * ChooseExtendedStatisticNameAddition.
2752 : */
2753 : static char *
2754 4000 : ChooseIndexNameAddition(const List *colnames)
2755 : {
2756 : char buf[NAMEDATALEN * 2];
2757 4000 : int buflen = 0;
2758 : ListCell *lc;
2759 :
2760 4000 : buf[0] = '\0';
2761 9070 : foreach(lc, colnames)
2762 : {
2763 5070 : const char *name = (const char *) lfirst(lc);
2764 :
2765 5070 : if (buflen > 0)
2766 1070 : buf[buflen++] = '_'; /* insert _ between names */
2767 :
2768 : /*
2769 : * At this point we have buflen <= NAMEDATALEN. name should be less
2770 : * than NAMEDATALEN already, but use strlcpy for paranoia.
2771 : */
2772 5070 : strlcpy(buf + buflen, name, NAMEDATALEN);
2773 5070 : buflen += strlen(buf + buflen);
2774 5070 : if (buflen >= NAMEDATALEN)
2775 0 : break;
2776 : }
2777 4000 : return pstrdup(buf);
2778 : }
2779 :
2780 : /*
2781 : * Select the actual names to be used for the columns of an index, given the
2782 : * list of IndexElems for the columns. This is mostly about ensuring the
2783 : * names are unique so we don't get a conflicting-attribute-names error.
2784 : *
2785 : * Returns a List of plain strings (char *, not String nodes).
2786 : */
2787 : static List *
2788 31652 : ChooseIndexColumnNames(const List *indexElems)
2789 : {
2790 31652 : List *result = NIL;
2791 : ListCell *lc;
2792 :
2793 76074 : foreach(lc, indexElems)
2794 : {
2795 44422 : IndexElem *ielem = (IndexElem *) lfirst(lc);
2796 : const char *origname;
2797 : const char *curname;
2798 : int i;
2799 : char buf[NAMEDATALEN];
2800 :
2801 : /* Get the preliminary name from the IndexElem */
2802 44422 : if (ielem->indexcolname)
2803 4384 : origname = ielem->indexcolname; /* caller-specified name */
2804 40038 : else if (ielem->name)
2805 39628 : origname = ielem->name; /* simple column reference */
2806 : else
2807 410 : origname = "expr"; /* default name for expression */
2808 :
2809 : /* If it conflicts with any previous column, tweak it */
2810 44422 : curname = origname;
2811 44422 : for (i = 1;; i++)
2812 62 : {
2813 : ListCell *lc2;
2814 : char nbuf[32];
2815 : int nlen;
2816 :
2817 69486 : foreach(lc2, result)
2818 : {
2819 25064 : if (strcmp(curname, (char *) lfirst(lc2)) == 0)
2820 62 : break;
2821 : }
2822 44484 : if (lc2 == NULL)
2823 44422 : break; /* found nonconflicting name */
2824 :
2825 62 : sprintf(nbuf, "%d", i);
2826 :
2827 : /* Ensure generated names are shorter than NAMEDATALEN */
2828 62 : nlen = pg_mbcliplen(origname, strlen(origname),
2829 62 : NAMEDATALEN - 1 - strlen(nbuf));
2830 62 : memcpy(buf, origname, nlen);
2831 62 : strcpy(buf + nlen, nbuf);
2832 62 : curname = buf;
2833 : }
2834 :
2835 : /* And attach to the result list */
2836 44422 : result = lappend(result, pstrdup(curname));
2837 : }
2838 31652 : return result;
2839 : }
2840 :
2841 : /*
2842 : * ExecReindex
2843 : *
2844 : * Primary entry point for manual REINDEX commands. This is mainly a
2845 : * preparation wrapper for the real operations that will happen in
2846 : * each subroutine of REINDEX.
2847 : */
2848 : void
2849 1130 : ExecReindex(ParseState *pstate, const ReindexStmt *stmt, bool isTopLevel)
2850 : {
2851 1130 : ReindexParams params = {0};
2852 : ListCell *lc;
2853 1130 : bool concurrently = false;
2854 1130 : bool verbose = false;
2855 1130 : char *tablespacename = NULL;
2856 :
2857 : /* Parse option list */
2858 1906 : foreach(lc, stmt->params)
2859 : {
2860 776 : DefElem *opt = (DefElem *) lfirst(lc);
2861 :
2862 776 : if (strcmp(opt->defname, "verbose") == 0)
2863 14 : verbose = defGetBoolean(opt);
2864 762 : else if (strcmp(opt->defname, "concurrently") == 0)
2865 634 : concurrently = defGetBoolean(opt);
2866 128 : else if (strcmp(opt->defname, "tablespace") == 0)
2867 128 : tablespacename = defGetString(opt);
2868 : else
2869 0 : ereport(ERROR,
2870 : (errcode(ERRCODE_SYNTAX_ERROR),
2871 : errmsg("unrecognized %s option \"%s\"",
2872 : "REINDEX", opt->defname),
2873 : parser_errposition(pstate, opt->location)));
2874 : }
2875 :
2876 1130 : if (concurrently)
2877 634 : PreventInTransactionBlock(isTopLevel,
2878 : "REINDEX CONCURRENTLY");
2879 :
2880 1104 : params.options =
2881 2208 : (verbose ? REINDEXOPT_VERBOSE : 0) |
2882 1104 : (concurrently ? REINDEXOPT_CONCURRENTLY : 0);
2883 :
2884 : /*
2885 : * Assign the tablespace OID to move indexes to, with InvalidOid to do
2886 : * nothing.
2887 : */
2888 1104 : if (tablespacename != NULL)
2889 : {
2890 128 : params.tablespaceOid = get_tablespace_oid(tablespacename, false);
2891 :
2892 : /* Check permissions except when moving to database's default */
2893 128 : if (OidIsValid(params.tablespaceOid) &&
2894 128 : params.tablespaceOid != MyDatabaseTableSpace)
2895 : {
2896 : AclResult aclresult;
2897 :
2898 128 : aclresult = object_aclcheck(TableSpaceRelationId, params.tablespaceOid,
2899 : GetUserId(), ACL_CREATE);
2900 128 : if (aclresult != ACLCHECK_OK)
2901 12 : aclcheck_error(aclresult, OBJECT_TABLESPACE,
2902 12 : get_tablespace_name(params.tablespaceOid));
2903 : }
2904 : }
2905 : else
2906 976 : params.tablespaceOid = InvalidOid;
2907 :
2908 1092 : switch (stmt->kind)
2909 : {
2910 412 : case REINDEX_OBJECT_INDEX:
2911 412 : ReindexIndex(stmt, ¶ms, isTopLevel);
2912 306 : break;
2913 502 : case REINDEX_OBJECT_TABLE:
2914 502 : ReindexTable(stmt, ¶ms, isTopLevel);
2915 380 : break;
2916 178 : case REINDEX_OBJECT_SCHEMA:
2917 : case REINDEX_OBJECT_SYSTEM:
2918 : case REINDEX_OBJECT_DATABASE:
2919 :
2920 : /*
2921 : * This cannot run inside a user transaction block; if we were
2922 : * inside a transaction, then its commit- and
2923 : * start-transaction-command calls would not have the intended
2924 : * effect!
2925 : */
2926 178 : PreventInTransactionBlock(isTopLevel,
2927 242 : (stmt->kind == REINDEX_OBJECT_SCHEMA) ? "REINDEX SCHEMA" :
2928 64 : (stmt->kind == REINDEX_OBJECT_SYSTEM) ? "REINDEX SYSTEM" :
2929 : "REINDEX DATABASE");
2930 172 : ReindexMultipleTables(stmt, ¶ms);
2931 122 : break;
2932 0 : default:
2933 0 : elog(ERROR, "unrecognized object type: %d",
2934 : (int) stmt->kind);
2935 : break;
2936 : }
2937 808 : }
2938 :
2939 : /*
2940 : * ReindexIndex
2941 : * Recreate a specific index.
2942 : */
2943 : static void
2944 412 : ReindexIndex(const ReindexStmt *stmt, const ReindexParams *params, bool isTopLevel)
2945 : {
2946 412 : const RangeVar *indexRelation = stmt->relation;
2947 : struct ReindexIndexCallbackState state;
2948 : Oid indOid;
2949 : char persistence;
2950 : char relkind;
2951 :
2952 : /*
2953 : * Find and lock index, and check permissions on table; use callback to
2954 : * obtain lock on table first, to avoid deadlock hazard. The lock level
2955 : * used here must match the index lock obtained in reindex_index().
2956 : *
2957 : * If it's a temporary index, we will perform a non-concurrent reindex,
2958 : * even if CONCURRENTLY was requested. In that case, reindex_index() will
2959 : * upgrade the lock, but that's OK, because other sessions can't hold
2960 : * locks on our temporary table.
2961 : */
2962 412 : state.params = *params;
2963 412 : state.locked_table_oid = InvalidOid;
2964 412 : indOid = RangeVarGetRelidExtended(indexRelation,
2965 412 : (params->options & REINDEXOPT_CONCURRENTLY) != 0 ?
2966 : ShareUpdateExclusiveLock : AccessExclusiveLock,
2967 : 0,
2968 : RangeVarCallbackForReindexIndex,
2969 : &state);
2970 :
2971 : /*
2972 : * Obtain the current persistence and kind of the existing index. We
2973 : * already hold a lock on the index.
2974 : */
2975 364 : persistence = get_rel_persistence(indOid);
2976 364 : relkind = get_rel_relkind(indOid);
2977 :
2978 364 : if (relkind == RELKIND_PARTITIONED_INDEX)
2979 36 : ReindexPartitions(stmt, indOid, params, isTopLevel);
2980 328 : else if ((params->options & REINDEXOPT_CONCURRENTLY) != 0 &&
2981 : persistence != RELPERSISTENCE_TEMP)
2982 198 : ReindexRelationConcurrently(stmt, indOid, params);
2983 : else
2984 : {
2985 130 : ReindexParams newparams = *params;
2986 :
2987 130 : newparams.options |= REINDEXOPT_REPORT_PROGRESS;
2988 130 : reindex_index(stmt, indOid, false, persistence, &newparams);
2989 : }
2990 306 : }
2991 :
2992 : /*
2993 : * Check permissions on table before acquiring relation lock; also lock
2994 : * the heap before the RangeVarGetRelidExtended takes the index lock, to avoid
2995 : * deadlocks.
2996 : */
2997 : static void
2998 430 : RangeVarCallbackForReindexIndex(const RangeVar *relation,
2999 : Oid relId, Oid oldRelId, void *arg)
3000 : {
3001 : char relkind;
3002 430 : struct ReindexIndexCallbackState *state = arg;
3003 : LOCKMODE table_lockmode;
3004 : Oid table_oid;
3005 : AclResult aclresult;
3006 :
3007 : /*
3008 : * Lock level here should match table lock in reindex_index() for
3009 : * non-concurrent case and table locks used by index_concurrently_*() for
3010 : * concurrent case.
3011 : */
3012 860 : table_lockmode = (state->params.options & REINDEXOPT_CONCURRENTLY) != 0 ?
3013 430 : ShareUpdateExclusiveLock : ShareLock;
3014 :
3015 : /*
3016 : * If we previously locked some other index's heap, and the name we're
3017 : * looking up no longer refers to that relation, release the now-useless
3018 : * lock.
3019 : */
3020 430 : if (relId != oldRelId && OidIsValid(oldRelId))
3021 : {
3022 6 : UnlockRelationOid(state->locked_table_oid, table_lockmode);
3023 6 : state->locked_table_oid = InvalidOid;
3024 : }
3025 :
3026 : /* If the relation does not exist, there's nothing more to do. */
3027 430 : if (!OidIsValid(relId))
3028 12 : return;
3029 :
3030 : /* If the relation does exist, check whether it's an index. */
3031 418 : relkind = get_rel_relkind(relId);
3032 418 : if (relkind != RELKIND_INDEX &&
3033 : relkind != RELKIND_PARTITIONED_INDEX)
3034 24 : ereport(ERROR,
3035 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
3036 : errmsg("\"%s\" is not an index", relation->relname)));
3037 :
3038 : /* Look up the index's table. */
3039 394 : table_oid = IndexGetRelation(relId, false);
3040 :
3041 : /*
3042 : * In the unlikely event that, upon retry, we get the same index OID with
3043 : * a different table OID, fail. RangeVarGetRelidExtended() will have
3044 : * already locked the index in this case, and it won't retry again, so we
3045 : * can't lock the newly discovered table OID without risking deadlock.
3046 : * Also, while this corner case is indeed possible, it is extremely
3047 : * unlikely to happen in practice, so it's probably not worth any more
3048 : * effort than this.
3049 : */
3050 394 : if (relId == oldRelId && table_oid != state->locked_table_oid)
3051 0 : ereport(ERROR,
3052 : (errcode(ERRCODE_UNDEFINED_OBJECT),
3053 : errmsg("index \"%s\" was concurrently dropped",
3054 : relation->relname)));
3055 :
3056 : /* Check permissions. */
3057 394 : aclresult = pg_class_aclcheck(table_oid, GetUserId(), ACL_MAINTAIN);
3058 394 : if (aclresult != ACLCHECK_OK)
3059 12 : aclcheck_error(aclresult, OBJECT_INDEX, relation->relname);
3060 :
3061 : /* Lock heap before index to avoid deadlock. */
3062 382 : if (relId != oldRelId)
3063 : {
3064 370 : LockRelationOid(table_oid, table_lockmode);
3065 370 : state->locked_table_oid = table_oid;
3066 : }
3067 : }
3068 :
3069 : /*
3070 : * ReindexTable
3071 : * Recreate all indexes of a table (and of its toast table, if any)
3072 : */
3073 : static Oid
3074 502 : ReindexTable(const ReindexStmt *stmt, const ReindexParams *params, bool isTopLevel)
3075 : {
3076 : Oid heapOid;
3077 : bool result;
3078 502 : const RangeVar *relation = stmt->relation;
3079 :
3080 : /*
3081 : * The lock level used here should match reindex_relation().
3082 : *
3083 : * If it's a temporary table, we will perform a non-concurrent reindex,
3084 : * even if CONCURRENTLY was requested. In that case, reindex_relation()
3085 : * will upgrade the lock, but that's OK, because other sessions can't hold
3086 : * locks on our temporary table.
3087 : */
3088 502 : heapOid = RangeVarGetRelidExtended(relation,
3089 502 : (params->options & REINDEXOPT_CONCURRENTLY) != 0 ?
3090 : ShareUpdateExclusiveLock : ShareLock,
3091 : 0,
3092 : RangeVarCallbackMaintainsTable, NULL);
3093 :
3094 456 : if (get_rel_relkind(heapOid) == RELKIND_PARTITIONED_TABLE)
3095 72 : ReindexPartitions(stmt, heapOid, params, isTopLevel);
3096 624 : else if ((params->options & REINDEXOPT_CONCURRENTLY) != 0 &&
3097 240 : get_rel_persistence(heapOid) != RELPERSISTENCE_TEMP)
3098 : {
3099 228 : result = ReindexRelationConcurrently(stmt, heapOid, params);
3100 :
3101 190 : if (!result)
3102 18 : ereport(NOTICE,
3103 : (errmsg("table \"%s\" has no indexes that can be reindexed concurrently",
3104 : relation->relname)));
3105 : }
3106 : else
3107 : {
3108 156 : ReindexParams newparams = *params;
3109 :
3110 156 : newparams.options |= REINDEXOPT_REPORT_PROGRESS;
3111 156 : result = reindex_relation(stmt, heapOid,
3112 : REINDEX_REL_PROCESS_TOAST |
3113 : REINDEX_REL_CHECK_CONSTRAINTS,
3114 : &newparams);
3115 124 : if (!result)
3116 12 : ereport(NOTICE,
3117 : (errmsg("table \"%s\" has no indexes to reindex",
3118 : relation->relname)));
3119 : }
3120 :
3121 380 : return heapOid;
3122 : }
3123 :
3124 : /*
3125 : * ReindexMultipleTables
3126 : * Recreate indexes of tables selected by objectName/objectKind.
3127 : *
3128 : * To reduce the probability of deadlocks, each table is reindexed in a
3129 : * separate transaction, so we can release the lock on it right away.
3130 : * That means this must not be called within a user transaction block!
3131 : */
3132 : static void
3133 172 : ReindexMultipleTables(const ReindexStmt *stmt, const ReindexParams *params)
3134 : {
3135 :
3136 : Oid objectOid;
3137 : Relation relationRelation;
3138 : TableScanDesc scan;
3139 : ScanKeyData scan_keys[1];
3140 : HeapTuple tuple;
3141 : MemoryContext private_context;
3142 : MemoryContext old;
3143 172 : List *relids = NIL;
3144 : int num_keys;
3145 172 : bool concurrent_warning = false;
3146 172 : bool tablespace_warning = false;
3147 172 : const char *objectName = stmt->name;
3148 172 : const ReindexObjectType objectKind = stmt->kind;
3149 :
3150 : Assert(objectKind == REINDEX_OBJECT_SCHEMA ||
3151 : objectKind == REINDEX_OBJECT_SYSTEM ||
3152 : objectKind == REINDEX_OBJECT_DATABASE);
3153 :
3154 : /*
3155 : * This matches the options enforced by the grammar, where the object name
3156 : * is optional for DATABASE and SYSTEM.
3157 : */
3158 : Assert(objectName || objectKind != REINDEX_OBJECT_SCHEMA);
3159 :
3160 172 : if (objectKind == REINDEX_OBJECT_SYSTEM &&
3161 34 : (params->options & REINDEXOPT_CONCURRENTLY) != 0)
3162 20 : ereport(ERROR,
3163 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3164 : errmsg("cannot reindex system catalogs concurrently")));
3165 :
3166 : /*
3167 : * Get OID of object to reindex, being the database currently being used
3168 : * by session for a database or for system catalogs, or the schema defined
3169 : * by caller. At the same time do permission checks that need different
3170 : * processing depending on the object type.
3171 : */
3172 152 : if (objectKind == REINDEX_OBJECT_SCHEMA)
3173 : {
3174 108 : objectOid = get_namespace_oid(objectName, false);
3175 :
3176 102 : if (!object_ownercheck(NamespaceRelationId, objectOid, GetUserId()) &&
3177 24 : !has_privs_of_role(GetUserId(), ROLE_PG_MAINTAIN))
3178 18 : aclcheck_error(ACLCHECK_NOT_OWNER, OBJECT_SCHEMA,
3179 : objectName);
3180 : }
3181 : else
3182 : {
3183 44 : objectOid = MyDatabaseId;
3184 :
3185 44 : if (objectName && strcmp(objectName, get_database_name(objectOid)) != 0)
3186 6 : ereport(ERROR,
3187 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3188 : errmsg("can only reindex the currently open database")));
3189 38 : if (!object_ownercheck(DatabaseRelationId, objectOid, GetUserId()) &&
3190 0 : !has_privs_of_role(GetUserId(), ROLE_PG_MAINTAIN))
3191 0 : aclcheck_error(ACLCHECK_NOT_OWNER, OBJECT_DATABASE,
3192 0 : get_database_name(objectOid));
3193 : }
3194 :
3195 : /*
3196 : * Create a memory context that will survive forced transaction commits we
3197 : * do below. Since it is a child of PortalContext, it will go away
3198 : * eventually even if we suffer an error; there's no need for special
3199 : * abort cleanup logic.
3200 : */
3201 122 : private_context = AllocSetContextCreate(PortalContext,
3202 : "ReindexMultipleTables",
3203 : ALLOCSET_SMALL_SIZES);
3204 :
3205 : /*
3206 : * Define the search keys to find the objects to reindex. For a schema, we
3207 : * select target relations using relnamespace, something not necessary for
3208 : * a database-wide operation.
3209 : */
3210 122 : if (objectKind == REINDEX_OBJECT_SCHEMA)
3211 : {
3212 84 : num_keys = 1;
3213 84 : ScanKeyInit(&scan_keys[0],
3214 : Anum_pg_class_relnamespace,
3215 : BTEqualStrategyNumber, F_OIDEQ,
3216 : ObjectIdGetDatum(objectOid));
3217 : }
3218 : else
3219 38 : num_keys = 0;
3220 :
3221 : /*
3222 : * Scan pg_class to build a list of the relations we need to reindex.
3223 : *
3224 : * We only consider plain relations and materialized views here (toast
3225 : * rels will be processed indirectly by reindex_relation).
3226 : */
3227 122 : relationRelation = table_open(RelationRelationId, AccessShareLock);
3228 122 : scan = table_beginscan_catalog(relationRelation, num_keys, scan_keys);
3229 18612 : while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
3230 : {
3231 18490 : Form_pg_class classtuple = (Form_pg_class) GETSTRUCT(tuple);
3232 18490 : Oid relid = classtuple->oid;
3233 :
3234 : /*
3235 : * Only regular tables and matviews can have indexes, so ignore any
3236 : * other kind of relation.
3237 : *
3238 : * Partitioned tables/indexes are skipped but matching leaf partitions
3239 : * are processed.
3240 : */
3241 18490 : if (classtuple->relkind != RELKIND_RELATION &&
3242 15208 : classtuple->relkind != RELKIND_MATVIEW)
3243 15190 : continue;
3244 :
3245 : /* Skip temp tables of other backends; we can't reindex them at all */
3246 3300 : if (classtuple->relpersistence == RELPERSISTENCE_TEMP &&
3247 36 : !isTempNamespace(classtuple->relnamespace))
3248 0 : continue;
3249 :
3250 : /*
3251 : * Check user/system classification. SYSTEM processes all the
3252 : * catalogs, and DATABASE processes everything that's not a catalog.
3253 : */
3254 3300 : if (objectKind == REINDEX_OBJECT_SYSTEM &&
3255 984 : !IsCatalogRelationOid(relid))
3256 88 : continue;
3257 4856 : else if (objectKind == REINDEX_OBJECT_DATABASE &&
3258 1644 : IsCatalogRelationOid(relid))
3259 1536 : continue;
3260 :
3261 : /*
3262 : * We already checked privileges on the database or schema, but we
3263 : * further restrict reindexing shared catalogs to roles with the
3264 : * MAINTAIN privilege on the relation.
3265 : */
3266 1918 : if (classtuple->relisshared &&
3267 242 : pg_class_aclcheck(relid, GetUserId(), ACL_MAINTAIN) != ACLCHECK_OK)
3268 0 : continue;
3269 :
3270 : /*
3271 : * Skip system tables, since index_create() would reject indexing them
3272 : * concurrently (and it would likely fail if we tried).
3273 : */
3274 2158 : if ((params->options & REINDEXOPT_CONCURRENTLY) != 0 &&
3275 482 : IsCatalogRelationOid(relid))
3276 : {
3277 384 : if (!concurrent_warning)
3278 6 : ereport(WARNING,
3279 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3280 : errmsg("cannot reindex system catalogs concurrently, skipping all")));
3281 384 : concurrent_warning = true;
3282 384 : continue;
3283 : }
3284 :
3285 : /*
3286 : * If a new tablespace is set, check if this relation has to be
3287 : * skipped.
3288 : */
3289 1292 : if (OidIsValid(params->tablespaceOid))
3290 : {
3291 0 : bool skip_rel = false;
3292 :
3293 : /*
3294 : * Mapped relations cannot be moved to different tablespaces (in
3295 : * particular this eliminates all shared catalogs.).
3296 : */
3297 0 : if (RELKIND_HAS_STORAGE(classtuple->relkind) &&
3298 0 : !RelFileNumberIsValid(classtuple->relfilenode))
3299 0 : skip_rel = true;
3300 :
3301 : /*
3302 : * A system relation is always skipped, even with
3303 : * allow_system_table_mods enabled.
3304 : */
3305 0 : if (IsSystemClass(relid, classtuple))
3306 0 : skip_rel = true;
3307 :
3308 0 : if (skip_rel)
3309 : {
3310 0 : if (!tablespace_warning)
3311 0 : ereport(WARNING,
3312 : (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
3313 : errmsg("cannot move system relations, skipping all")));
3314 0 : tablespace_warning = true;
3315 0 : continue;
3316 : }
3317 : }
3318 :
3319 : /* Save the list of relation OIDs in private context */
3320 1292 : old = MemoryContextSwitchTo(private_context);
3321 :
3322 : /*
3323 : * We always want to reindex pg_class first if it's selected to be
3324 : * reindexed. This ensures that if there is any corruption in
3325 : * pg_class' indexes, they will be fixed before we process any other
3326 : * tables. This is critical because reindexing itself will try to
3327 : * update pg_class.
3328 : */
3329 1292 : if (relid == RelationRelationId)
3330 16 : relids = lcons_oid(relid, relids);
3331 : else
3332 1276 : relids = lappend_oid(relids, relid);
3333 :
3334 1292 : MemoryContextSwitchTo(old);
3335 : }
3336 122 : table_endscan(scan);
3337 122 : table_close(relationRelation, AccessShareLock);
3338 :
3339 : /*
3340 : * Process each relation listed in a separate transaction. Note that this
3341 : * commits and then starts a new transaction immediately.
3342 : */
3343 122 : ReindexMultipleInternal(stmt, relids, params);
3344 :
3345 122 : MemoryContextDelete(private_context);
3346 122 : }
3347 :
3348 : /*
3349 : * Error callback specific to ReindexPartitions().
3350 : */
3351 : static void
3352 12 : reindex_error_callback(void *arg)
3353 : {
3354 12 : ReindexErrorInfo *errinfo = (ReindexErrorInfo *) arg;
3355 :
3356 : Assert(RELKIND_HAS_PARTITIONS(errinfo->relkind));
3357 :
3358 12 : if (errinfo->relkind == RELKIND_PARTITIONED_TABLE)
3359 6 : errcontext("while reindexing partitioned table \"%s.%s\"",
3360 : errinfo->relnamespace, errinfo->relname);
3361 6 : else if (errinfo->relkind == RELKIND_PARTITIONED_INDEX)
3362 6 : errcontext("while reindexing partitioned index \"%s.%s\"",
3363 : errinfo->relnamespace, errinfo->relname);
3364 12 : }
3365 :
3366 : /*
3367 : * ReindexPartitions
3368 : *
3369 : * Reindex a set of partitions, per the partitioned index or table given
3370 : * by the caller.
3371 : */
3372 : static void
3373 108 : ReindexPartitions(const ReindexStmt *stmt, Oid relid, const ReindexParams *params, bool isTopLevel)
3374 : {
3375 108 : List *partitions = NIL;
3376 108 : char relkind = get_rel_relkind(relid);
3377 108 : char *relname = get_rel_name(relid);
3378 108 : char *relnamespace = get_namespace_name(get_rel_namespace(relid));
3379 : MemoryContext reindex_context;
3380 : List *inhoids;
3381 : ListCell *lc;
3382 : ErrorContextCallback errcallback;
3383 : ReindexErrorInfo errinfo;
3384 :
3385 : Assert(RELKIND_HAS_PARTITIONS(relkind));
3386 :
3387 : /*
3388 : * Check if this runs in a transaction block, with an error callback to
3389 : * provide more context under which a problem happens.
3390 : */
3391 108 : errinfo.relname = pstrdup(relname);
3392 108 : errinfo.relnamespace = pstrdup(relnamespace);
3393 108 : errinfo.relkind = relkind;
3394 108 : errcallback.callback = reindex_error_callback;
3395 108 : errcallback.arg = &errinfo;
3396 108 : errcallback.previous = error_context_stack;
3397 108 : error_context_stack = &errcallback;
3398 :
3399 108 : PreventInTransactionBlock(isTopLevel,
3400 : relkind == RELKIND_PARTITIONED_TABLE ?
3401 : "REINDEX TABLE" : "REINDEX INDEX");
3402 :
3403 : /* Pop the error context stack */
3404 96 : error_context_stack = errcallback.previous;
3405 :
3406 : /*
3407 : * Create special memory context for cross-transaction storage.
3408 : *
3409 : * Since it is a child of PortalContext, it will go away eventually even
3410 : * if we suffer an error so there is no need for special abort cleanup
3411 : * logic.
3412 : */
3413 96 : reindex_context = AllocSetContextCreate(PortalContext, "Reindex",
3414 : ALLOCSET_DEFAULT_SIZES);
3415 :
3416 : /* ShareLock is enough to prevent schema modifications */
3417 96 : inhoids = find_all_inheritors(relid, ShareLock, NULL);
3418 :
3419 : /*
3420 : * The list of relations to reindex are the physical partitions of the
3421 : * tree so discard any partitioned table or index.
3422 : */
3423 374 : foreach(lc, inhoids)
3424 : {
3425 278 : Oid partoid = lfirst_oid(lc);
3426 278 : char partkind = get_rel_relkind(partoid);
3427 : MemoryContext old_context;
3428 :
3429 : /*
3430 : * This discards partitioned tables, partitioned indexes and foreign
3431 : * tables.
3432 : */
3433 278 : if (!RELKIND_HAS_STORAGE(partkind))
3434 160 : continue;
3435 :
3436 : Assert(partkind == RELKIND_INDEX ||
3437 : partkind == RELKIND_RELATION);
3438 :
3439 : /* Save partition OID */
3440 118 : old_context = MemoryContextSwitchTo(reindex_context);
3441 118 : partitions = lappend_oid(partitions, partoid);
3442 118 : MemoryContextSwitchTo(old_context);
3443 : }
3444 :
3445 : /*
3446 : * Process each partition listed in a separate transaction. Note that
3447 : * this commits and then starts a new transaction immediately.
3448 : */
3449 96 : ReindexMultipleInternal(stmt, partitions, params);
3450 :
3451 : /*
3452 : * Clean up working storage --- note we must do this after
3453 : * StartTransactionCommand, else we might be trying to delete the active
3454 : * context!
3455 : */
3456 96 : MemoryContextDelete(reindex_context);
3457 96 : }
3458 :
3459 : /*
3460 : * ReindexMultipleInternal
3461 : *
3462 : * Reindex a list of relations, each one being processed in its own
3463 : * transaction. This commits the existing transaction immediately,
3464 : * and starts a new transaction when finished.
3465 : */
3466 : static void
3467 218 : ReindexMultipleInternal(const ReindexStmt *stmt, const List *relids, const ReindexParams *params)
3468 : {
3469 : ListCell *l;
3470 :
3471 218 : PopActiveSnapshot();
3472 218 : CommitTransactionCommand();
3473 :
3474 1628 : foreach(l, relids)
3475 : {
3476 1410 : Oid relid = lfirst_oid(l);
3477 : char relkind;
3478 : char relpersistence;
3479 :
3480 1410 : StartTransactionCommand();
3481 :
3482 : /* functions in indexes may want a snapshot set */
3483 1410 : PushActiveSnapshot(GetTransactionSnapshot());
3484 :
3485 : /* check if the relation still exists */
3486 1410 : if (!SearchSysCacheExists1(RELOID, ObjectIdGetDatum(relid)))
3487 : {
3488 4 : PopActiveSnapshot();
3489 4 : CommitTransactionCommand();
3490 4 : continue;
3491 : }
3492 :
3493 : /*
3494 : * Check permissions except when moving to database's default if a new
3495 : * tablespace is chosen. Note that this check also happens in
3496 : * ExecReindex(), but we do an extra check here as this runs across
3497 : * multiple transactions.
3498 : */
3499 1406 : if (OidIsValid(params->tablespaceOid) &&
3500 12 : params->tablespaceOid != MyDatabaseTableSpace)
3501 : {
3502 : AclResult aclresult;
3503 :
3504 12 : aclresult = object_aclcheck(TableSpaceRelationId, params->tablespaceOid,
3505 : GetUserId(), ACL_CREATE);
3506 12 : if (aclresult != ACLCHECK_OK)
3507 0 : aclcheck_error(aclresult, OBJECT_TABLESPACE,
3508 0 : get_tablespace_name(params->tablespaceOid));
3509 : }
3510 :
3511 1406 : relkind = get_rel_relkind(relid);
3512 1406 : relpersistence = get_rel_persistence(relid);
3513 :
3514 : /*
3515 : * Partitioned tables and indexes can never be processed directly, and
3516 : * a list of their leaves should be built first.
3517 : */
3518 : Assert(!RELKIND_HAS_PARTITIONS(relkind));
3519 :
3520 1406 : if ((params->options & REINDEXOPT_CONCURRENTLY) != 0 &&
3521 : relpersistence != RELPERSISTENCE_TEMP)
3522 128 : {
3523 128 : ReindexParams newparams = *params;
3524 :
3525 128 : newparams.options |= REINDEXOPT_MISSING_OK;
3526 128 : (void) ReindexRelationConcurrently(stmt, relid, &newparams);
3527 128 : if (ActiveSnapshotSet())
3528 26 : PopActiveSnapshot();
3529 : /* ReindexRelationConcurrently() does the verbose output */
3530 : }
3531 1278 : else if (relkind == RELKIND_INDEX)
3532 : {
3533 18 : ReindexParams newparams = *params;
3534 :
3535 18 : newparams.options |=
3536 : REINDEXOPT_REPORT_PROGRESS | REINDEXOPT_MISSING_OK;
3537 18 : reindex_index(stmt, relid, false, relpersistence, &newparams);
3538 18 : PopActiveSnapshot();
3539 : /* reindex_index() does the verbose output */
3540 : }
3541 : else
3542 : {
3543 : bool result;
3544 1260 : ReindexParams newparams = *params;
3545 :
3546 1260 : newparams.options |=
3547 : REINDEXOPT_REPORT_PROGRESS | REINDEXOPT_MISSING_OK;
3548 1260 : result = reindex_relation(stmt, relid,
3549 : REINDEX_REL_PROCESS_TOAST |
3550 : REINDEX_REL_CHECK_CONSTRAINTS,
3551 : &newparams);
3552 :
3553 1260 : if (result && (params->options & REINDEXOPT_VERBOSE) != 0)
3554 0 : ereport(INFO,
3555 : (errmsg("table \"%s.%s\" was reindexed",
3556 : get_namespace_name(get_rel_namespace(relid)),
3557 : get_rel_name(relid))));
3558 :
3559 1260 : PopActiveSnapshot();
3560 : }
3561 :
3562 1406 : CommitTransactionCommand();
3563 : }
3564 :
3565 218 : StartTransactionCommand();
3566 218 : }
3567 :
3568 :
3569 : /*
3570 : * ReindexRelationConcurrently - process REINDEX CONCURRENTLY for given
3571 : * relation OID
3572 : *
3573 : * 'relationOid' can either belong to an index, a table or a materialized
3574 : * view. For tables and materialized views, all its indexes will be rebuilt,
3575 : * excluding invalid indexes and any indexes used in exclusion constraints,
3576 : * but including its associated toast table indexes. For indexes, the index
3577 : * itself will be rebuilt.
3578 : *
3579 : * The locks taken on parent tables and involved indexes are kept until the
3580 : * transaction is committed, at which point a session lock is taken on each
3581 : * relation. Both of these protect against concurrent schema changes.
3582 : *
3583 : * Returns true if any indexes have been rebuilt (including toast table's
3584 : * indexes, when relevant), otherwise returns false.
3585 : *
3586 : * NOTE: This cannot be used on temporary relations. A concurrent build would
3587 : * cause issues with ON COMMIT actions triggered by the transactions of the
3588 : * concurrent build. Temporary relations are not subject to concurrent
3589 : * concerns, so there's no need for the more complicated concurrent build,
3590 : * anyway, and a non-concurrent reindex is more efficient.
3591 : */
3592 : static bool
3593 554 : ReindexRelationConcurrently(const ReindexStmt *stmt, Oid relationOid, const ReindexParams *params)
3594 : {
3595 : typedef struct ReindexIndexInfo
3596 : {
3597 : Oid indexId;
3598 : Oid tableId;
3599 : Oid amId;
3600 : bool safe; /* for set_indexsafe_procflags */
3601 : } ReindexIndexInfo;
3602 554 : List *heapRelationIds = NIL;
3603 554 : List *indexIds = NIL;
3604 554 : List *newIndexIds = NIL;
3605 554 : List *relationLocks = NIL;
3606 554 : List *lockTags = NIL;
3607 : ListCell *lc,
3608 : *lc2;
3609 : MemoryContext private_context;
3610 : MemoryContext oldcontext;
3611 : char relkind;
3612 554 : char *relationName = NULL;
3613 554 : char *relationNamespace = NULL;
3614 : PGRUsage ru0;
3615 554 : const int progress_index[] = {
3616 : PROGRESS_CREATEIDX_COMMAND,
3617 : PROGRESS_CREATEIDX_PHASE,
3618 : PROGRESS_CREATEIDX_INDEX_OID,
3619 : PROGRESS_CREATEIDX_ACCESS_METHOD_OID
3620 : };
3621 : int64 progress_vals[4];
3622 :
3623 : /*
3624 : * Create a memory context that will survive forced transaction commits we
3625 : * do below. Since it is a child of PortalContext, it will go away
3626 : * eventually even if we suffer an error; there's no need for special
3627 : * abort cleanup logic.
3628 : */
3629 554 : private_context = AllocSetContextCreate(PortalContext,
3630 : "ReindexConcurrent",
3631 : ALLOCSET_SMALL_SIZES);
3632 :
3633 554 : if ((params->options & REINDEXOPT_VERBOSE) != 0)
3634 : {
3635 : /* Save data needed by REINDEX VERBOSE in private context */
3636 4 : oldcontext = MemoryContextSwitchTo(private_context);
3637 :
3638 4 : relationName = get_rel_name(relationOid);
3639 4 : relationNamespace = get_namespace_name(get_rel_namespace(relationOid));
3640 :
3641 4 : pg_rusage_init(&ru0);
3642 :
3643 4 : MemoryContextSwitchTo(oldcontext);
3644 : }
3645 :
3646 554 : relkind = get_rel_relkind(relationOid);
3647 :
3648 : /*
3649 : * Extract the list of indexes that are going to be rebuilt based on the
3650 : * relation Oid given by caller.
3651 : */
3652 554 : switch (relkind)
3653 : {
3654 332 : case RELKIND_RELATION:
3655 : case RELKIND_MATVIEW:
3656 : case RELKIND_TOASTVALUE:
3657 : {
3658 : /*
3659 : * In the case of a relation, find all its indexes including
3660 : * toast indexes.
3661 : */
3662 : Relation heapRelation;
3663 :
3664 : /* Save the list of relation OIDs in private context */
3665 332 : oldcontext = MemoryContextSwitchTo(private_context);
3666 :
3667 : /* Track this relation for session locks */
3668 332 : heapRelationIds = lappend_oid(heapRelationIds, relationOid);
3669 :
3670 332 : MemoryContextSwitchTo(oldcontext);
3671 :
3672 332 : if (IsCatalogRelationOid(relationOid))
3673 36 : ereport(ERROR,
3674 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3675 : errmsg("cannot reindex system catalogs concurrently")));
3676 :
3677 : /* Open relation to get its indexes */
3678 296 : if ((params->options & REINDEXOPT_MISSING_OK) != 0)
3679 : {
3680 104 : heapRelation = try_table_open(relationOid,
3681 : ShareUpdateExclusiveLock);
3682 : /* leave if relation does not exist */
3683 104 : if (!heapRelation)
3684 0 : break;
3685 : }
3686 : else
3687 192 : heapRelation = table_open(relationOid,
3688 : ShareUpdateExclusiveLock);
3689 :
3690 318 : if (OidIsValid(params->tablespaceOid) &&
3691 22 : IsSystemRelation(heapRelation))
3692 2 : ereport(ERROR,
3693 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3694 : errmsg("cannot move system relation \"%s\"",
3695 : RelationGetRelationName(heapRelation))));
3696 :
3697 : /* Add all the valid indexes of relation to list */
3698 568 : foreach(lc, RelationGetIndexList(heapRelation))
3699 : {
3700 274 : Oid cellOid = lfirst_oid(lc);
3701 274 : Relation indexRelation = index_open(cellOid,
3702 : ShareUpdateExclusiveLock);
3703 :
3704 274 : if (!indexRelation->rd_index->indisvalid)
3705 6 : ereport(WARNING,
3706 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
3707 : errmsg("skipping reindex of invalid index \"%s.%s\"",
3708 : get_namespace_name(get_rel_namespace(cellOid)),
3709 : get_rel_name(cellOid)),
3710 : errhint("Use DROP INDEX or REINDEX INDEX.")));
3711 268 : else if (indexRelation->rd_index->indisexclusion)
3712 6 : ereport(WARNING,
3713 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3714 : errmsg("cannot reindex exclusion constraint index \"%s.%s\" concurrently, skipping",
3715 : get_namespace_name(get_rel_namespace(cellOid)),
3716 : get_rel_name(cellOid))));
3717 : else
3718 : {
3719 : ReindexIndexInfo *idx;
3720 :
3721 : /* Save the list of relation OIDs in private context */
3722 262 : oldcontext = MemoryContextSwitchTo(private_context);
3723 :
3724 262 : idx = palloc_object(ReindexIndexInfo);
3725 262 : idx->indexId = cellOid;
3726 : /* other fields set later */
3727 :
3728 262 : indexIds = lappend(indexIds, idx);
3729 :
3730 262 : MemoryContextSwitchTo(oldcontext);
3731 : }
3732 :
3733 274 : index_close(indexRelation, NoLock);
3734 : }
3735 :
3736 : /* Also add the toast indexes */
3737 294 : if (OidIsValid(heapRelation->rd_rel->reltoastrelid))
3738 : {
3739 88 : Oid toastOid = heapRelation->rd_rel->reltoastrelid;
3740 88 : Relation toastRelation = table_open(toastOid,
3741 : ShareUpdateExclusiveLock);
3742 :
3743 : /* Save the list of relation OIDs in private context */
3744 88 : oldcontext = MemoryContextSwitchTo(private_context);
3745 :
3746 : /* Track this relation for session locks */
3747 88 : heapRelationIds = lappend_oid(heapRelationIds, toastOid);
3748 :
3749 88 : MemoryContextSwitchTo(oldcontext);
3750 :
3751 176 : foreach(lc2, RelationGetIndexList(toastRelation))
3752 : {
3753 88 : Oid cellOid = lfirst_oid(lc2);
3754 88 : Relation indexRelation = index_open(cellOid,
3755 : ShareUpdateExclusiveLock);
3756 :
3757 88 : if (!indexRelation->rd_index->indisvalid)
3758 0 : ereport(WARNING,
3759 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
3760 : errmsg("skipping reindex of invalid index \"%s.%s\"",
3761 : get_namespace_name(get_rel_namespace(cellOid)),
3762 : get_rel_name(cellOid)),
3763 : errhint("Use DROP INDEX or REINDEX INDEX.")));
3764 : else
3765 : {
3766 : ReindexIndexInfo *idx;
3767 :
3768 : /*
3769 : * Save the list of relation OIDs in private
3770 : * context
3771 : */
3772 88 : oldcontext = MemoryContextSwitchTo(private_context);
3773 :
3774 88 : idx = palloc_object(ReindexIndexInfo);
3775 88 : idx->indexId = cellOid;
3776 88 : indexIds = lappend(indexIds, idx);
3777 : /* other fields set later */
3778 :
3779 88 : MemoryContextSwitchTo(oldcontext);
3780 : }
3781 :
3782 88 : index_close(indexRelation, NoLock);
3783 : }
3784 :
3785 88 : table_close(toastRelation, NoLock);
3786 : }
3787 :
3788 294 : table_close(heapRelation, NoLock);
3789 294 : break;
3790 : }
3791 222 : case RELKIND_INDEX:
3792 : {
3793 222 : Oid heapId = IndexGetRelation(relationOid,
3794 222 : (params->options & REINDEXOPT_MISSING_OK) != 0);
3795 : Relation heapRelation;
3796 : ReindexIndexInfo *idx;
3797 :
3798 : /* if relation is missing, leave */
3799 222 : if (!OidIsValid(heapId))
3800 0 : break;
3801 :
3802 222 : if (IsCatalogRelationOid(heapId))
3803 18 : ereport(ERROR,
3804 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3805 : errmsg("cannot reindex system catalogs concurrently")));
3806 :
3807 : /*
3808 : * Don't allow reindex for an invalid index on TOAST table, as
3809 : * if rebuilt it would not be possible to drop it. Match
3810 : * error message in reindex_index().
3811 : */
3812 204 : if (IsToastNamespace(get_rel_namespace(relationOid)) &&
3813 56 : !get_index_isvalid(relationOid))
3814 0 : ereport(ERROR,
3815 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3816 : errmsg("cannot reindex invalid index on TOAST table")));
3817 :
3818 : /*
3819 : * Check if parent relation can be locked and if it exists,
3820 : * this needs to be done at this stage as the list of indexes
3821 : * to rebuild is not complete yet, and REINDEXOPT_MISSING_OK
3822 : * should not be used once all the session locks are taken.
3823 : */
3824 204 : if ((params->options & REINDEXOPT_MISSING_OK) != 0)
3825 : {
3826 24 : heapRelation = try_table_open(heapId,
3827 : ShareUpdateExclusiveLock);
3828 : /* leave if relation does not exist */
3829 24 : if (!heapRelation)
3830 0 : break;
3831 : }
3832 : else
3833 180 : heapRelation = table_open(heapId,
3834 : ShareUpdateExclusiveLock);
3835 :
3836 212 : if (OidIsValid(params->tablespaceOid) &&
3837 8 : IsSystemRelation(heapRelation))
3838 2 : ereport(ERROR,
3839 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3840 : errmsg("cannot move system relation \"%s\"",
3841 : get_rel_name(relationOid))));
3842 :
3843 202 : table_close(heapRelation, NoLock);
3844 :
3845 : /* Save the list of relation OIDs in private context */
3846 202 : oldcontext = MemoryContextSwitchTo(private_context);
3847 :
3848 : /* Track the heap relation of this index for session locks */
3849 202 : heapRelationIds = list_make1_oid(heapId);
3850 :
3851 : /*
3852 : * Save the list of relation OIDs in private context. Note
3853 : * that invalid indexes are allowed here.
3854 : */
3855 202 : idx = palloc_object(ReindexIndexInfo);
3856 202 : idx->indexId = relationOid;
3857 202 : indexIds = lappend(indexIds, idx);
3858 : /* other fields set later */
3859 :
3860 202 : MemoryContextSwitchTo(oldcontext);
3861 202 : break;
3862 : }
3863 :
3864 0 : case RELKIND_PARTITIONED_TABLE:
3865 : case RELKIND_PARTITIONED_INDEX:
3866 : default:
3867 : /* Return error if type of relation is not supported */
3868 0 : ereport(ERROR,
3869 : (errcode(ERRCODE_WRONG_OBJECT_TYPE),
3870 : errmsg("cannot reindex this type of relation concurrently")));
3871 : break;
3872 : }
3873 :
3874 : /*
3875 : * Definitely no indexes, so leave. Any checks based on
3876 : * REINDEXOPT_MISSING_OK should be done only while the list of indexes to
3877 : * work on is built as the session locks taken before this transaction
3878 : * commits will make sure that they cannot be dropped by a concurrent
3879 : * session until this operation completes.
3880 : */
3881 496 : if (indexIds == NIL)
3882 44 : return false;
3883 :
3884 : /* It's not a shared catalog, so refuse to move it to shared tablespace */
3885 452 : if (params->tablespaceOid == GLOBALTABLESPACE_OID)
3886 6 : ereport(ERROR,
3887 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
3888 : errmsg("cannot move non-shared relation to tablespace \"%s\"",
3889 : get_tablespace_name(params->tablespaceOid))));
3890 :
3891 : Assert(heapRelationIds != NIL);
3892 :
3893 : /*-----
3894 : * Now we have all the indexes we want to process in indexIds.
3895 : *
3896 : * The phases now are:
3897 : *
3898 : * 1. create new indexes in the catalog
3899 : * 2. build new indexes
3900 : * 3. let new indexes catch up with tuples inserted in the meantime
3901 : * 4. swap index names
3902 : * 5. mark old indexes as dead
3903 : * 6. drop old indexes
3904 : *
3905 : * We process each phase for all indexes before moving to the next phase,
3906 : * for efficiency.
3907 : */
3908 :
3909 : /*
3910 : * Phase 1 of REINDEX CONCURRENTLY
3911 : *
3912 : * Create a new index with the same properties as the old one, but it is
3913 : * only registered in catalogs and will be built later. Then get session
3914 : * locks on all involved tables. See analogous code in DefineIndex() for
3915 : * more detailed comments.
3916 : */
3917 :
3918 986 : foreach(lc, indexIds)
3919 : {
3920 : char *concurrentName;
3921 546 : ReindexIndexInfo *idx = lfirst(lc);
3922 : ReindexIndexInfo *newidx;
3923 : Oid newIndexId;
3924 : Relation indexRel;
3925 : Relation heapRel;
3926 : Oid save_userid;
3927 : int save_sec_context;
3928 : int save_nestlevel;
3929 : Relation newIndexRel;
3930 : LockRelId *lockrelid;
3931 : Oid tablespaceid;
3932 :
3933 546 : indexRel = index_open(idx->indexId, ShareUpdateExclusiveLock);
3934 546 : heapRel = table_open(indexRel->rd_index->indrelid,
3935 : ShareUpdateExclusiveLock);
3936 :
3937 : /*
3938 : * Switch to the table owner's userid, so that any index functions are
3939 : * run as that user. Also lock down security-restricted operations
3940 : * and arrange to make GUC variable changes local to this command.
3941 : */
3942 546 : GetUserIdAndSecContext(&save_userid, &save_sec_context);
3943 546 : SetUserIdAndSecContext(heapRel->rd_rel->relowner,
3944 : save_sec_context | SECURITY_RESTRICTED_OPERATION);
3945 546 : save_nestlevel = NewGUCNestLevel();
3946 546 : RestrictSearchPath();
3947 :
3948 : /* determine safety of this index for set_indexsafe_procflags */
3949 1052 : idx->safe = (RelationGetIndexExpressions(indexRel) == NIL &&
3950 506 : RelationGetIndexPredicate(indexRel) == NIL);
3951 :
3952 : #ifdef USE_INJECTION_POINTS
3953 546 : if (idx->safe)
3954 498 : INJECTION_POINT("reindex-conc-index-safe", NULL);
3955 : else
3956 48 : INJECTION_POINT("reindex-conc-index-not-safe", NULL);
3957 : #endif
3958 :
3959 546 : idx->tableId = RelationGetRelid(heapRel);
3960 546 : idx->amId = indexRel->rd_rel->relam;
3961 :
3962 : /* This function shouldn't be called for temporary relations. */
3963 546 : if (indexRel->rd_rel->relpersistence == RELPERSISTENCE_TEMP)
3964 0 : elog(ERROR, "cannot reindex a temporary table concurrently");
3965 :
3966 546 : pgstat_progress_start_command(PROGRESS_COMMAND_CREATE_INDEX, idx->tableId);
3967 :
3968 546 : progress_vals[0] = PROGRESS_CREATEIDX_COMMAND_REINDEX_CONCURRENTLY;
3969 546 : progress_vals[1] = 0; /* initializing */
3970 546 : progress_vals[2] = idx->indexId;
3971 546 : progress_vals[3] = idx->amId;
3972 546 : pgstat_progress_update_multi_param(4, progress_index, progress_vals);
3973 :
3974 : /* Choose a temporary relation name for the new index */
3975 546 : concurrentName = ChooseRelationName(get_rel_name(idx->indexId),
3976 : NULL,
3977 : "ccnew",
3978 546 : get_rel_namespace(indexRel->rd_index->indrelid),
3979 : false);
3980 :
3981 : /* Choose the new tablespace, indexes of toast tables are not moved */
3982 546 : if (OidIsValid(params->tablespaceOid) &&
3983 28 : heapRel->rd_rel->relkind != RELKIND_TOASTVALUE)
3984 20 : tablespaceid = params->tablespaceOid;
3985 : else
3986 526 : tablespaceid = indexRel->rd_rel->reltablespace;
3987 :
3988 : /* Create new index definition based on given index */
3989 546 : newIndexId = index_concurrently_create_copy(heapRel,
3990 : idx->indexId,
3991 : tablespaceid,
3992 : concurrentName);
3993 :
3994 : /*
3995 : * Now open the relation of the new index, a session-level lock is
3996 : * also needed on it.
3997 : */
3998 540 : newIndexRel = index_open(newIndexId, ShareUpdateExclusiveLock);
3999 :
4000 : /*
4001 : * Save the list of OIDs and locks in private context
4002 : */
4003 540 : oldcontext = MemoryContextSwitchTo(private_context);
4004 :
4005 540 : newidx = palloc_object(ReindexIndexInfo);
4006 540 : newidx->indexId = newIndexId;
4007 540 : newidx->safe = idx->safe;
4008 540 : newidx->tableId = idx->tableId;
4009 540 : newidx->amId = idx->amId;
4010 :
4011 540 : newIndexIds = lappend(newIndexIds, newidx);
4012 :
4013 : /*
4014 : * Save lockrelid to protect each relation from drop then close
4015 : * relations. The lockrelid on parent relation is not taken here to
4016 : * avoid multiple locks taken on the same relation, instead we rely on
4017 : * parentRelationIds built earlier.
4018 : */
4019 540 : lockrelid = palloc_object(LockRelId);
4020 540 : *lockrelid = indexRel->rd_lockInfo.lockRelId;
4021 540 : relationLocks = lappend(relationLocks, lockrelid);
4022 540 : lockrelid = palloc_object(LockRelId);
4023 540 : *lockrelid = newIndexRel->rd_lockInfo.lockRelId;
4024 540 : relationLocks = lappend(relationLocks, lockrelid);
4025 :
4026 540 : MemoryContextSwitchTo(oldcontext);
4027 :
4028 540 : index_close(indexRel, NoLock);
4029 540 : index_close(newIndexRel, NoLock);
4030 :
4031 : /* Roll back any GUC changes executed by index functions */
4032 540 : AtEOXact_GUC(false, save_nestlevel);
4033 :
4034 : /* Restore userid and security context */
4035 540 : SetUserIdAndSecContext(save_userid, save_sec_context);
4036 :
4037 540 : table_close(heapRel, NoLock);
4038 :
4039 : /*
4040 : * If a statement is available, telling that this comes from a REINDEX
4041 : * command, collect the new index for event triggers.
4042 : */
4043 540 : if (stmt)
4044 : {
4045 : ObjectAddress address;
4046 :
4047 540 : ObjectAddressSet(address, RelationRelationId, newIndexId);
4048 540 : EventTriggerCollectSimpleCommand(address,
4049 : InvalidObjectAddress,
4050 : (Node *) stmt);
4051 : }
4052 : }
4053 :
4054 : /*
4055 : * Save the heap lock for following visibility checks with other backends
4056 : * might conflict with this session.
4057 : */
4058 968 : foreach(lc, heapRelationIds)
4059 : {
4060 528 : Relation heapRelation = table_open(lfirst_oid(lc), ShareUpdateExclusiveLock);
4061 : LockRelId *lockrelid;
4062 : LOCKTAG *heaplocktag;
4063 :
4064 : /* Save the list of locks in private context */
4065 528 : oldcontext = MemoryContextSwitchTo(private_context);
4066 :
4067 : /* Add lockrelid of heap relation to the list of locked relations */
4068 528 : lockrelid = palloc_object(LockRelId);
4069 528 : *lockrelid = heapRelation->rd_lockInfo.lockRelId;
4070 528 : relationLocks = lappend(relationLocks, lockrelid);
4071 :
4072 528 : heaplocktag = palloc_object(LOCKTAG);
4073 :
4074 : /* Save the LOCKTAG for this parent relation for the wait phase */
4075 528 : SET_LOCKTAG_RELATION(*heaplocktag, lockrelid->dbId, lockrelid->relId);
4076 528 : lockTags = lappend(lockTags, heaplocktag);
4077 :
4078 528 : MemoryContextSwitchTo(oldcontext);
4079 :
4080 : /* Close heap relation */
4081 528 : table_close(heapRelation, NoLock);
4082 : }
4083 :
4084 : /* Get a session-level lock on each table. */
4085 2048 : foreach(lc, relationLocks)
4086 : {
4087 1608 : LockRelId *lockrelid = (LockRelId *) lfirst(lc);
4088 :
4089 1608 : LockRelationIdForSession(lockrelid, ShareUpdateExclusiveLock);
4090 : }
4091 :
4092 440 : PopActiveSnapshot();
4093 440 : CommitTransactionCommand();
4094 440 : StartTransactionCommand();
4095 :
4096 : /*
4097 : * Because we don't take a snapshot in this transaction, there's no need
4098 : * to set the PROC_IN_SAFE_IC flag here.
4099 : */
4100 :
4101 : /*
4102 : * Phase 2 of REINDEX CONCURRENTLY
4103 : *
4104 : * Build the new indexes in a separate transaction for each index to avoid
4105 : * having open transactions for an unnecessary long time. But before
4106 : * doing that, wait until no running transactions could have the table of
4107 : * the index open with the old list of indexes. See "phase 2" in
4108 : * DefineIndex() for more details.
4109 : */
4110 :
4111 440 : pgstat_progress_update_param(PROGRESS_CREATEIDX_PHASE,
4112 : PROGRESS_CREATEIDX_PHASE_WAIT_1);
4113 440 : WaitForLockersMultiple(lockTags, ShareLock, true);
4114 440 : CommitTransactionCommand();
4115 :
4116 974 : foreach(lc, newIndexIds)
4117 : {
4118 540 : ReindexIndexInfo *newidx = lfirst(lc);
4119 :
4120 : /* Start new transaction for this index's concurrent build */
4121 540 : StartTransactionCommand();
4122 :
4123 : /*
4124 : * Check for user-requested abort. This is inside a transaction so as
4125 : * xact.c does not issue a useless WARNING, and ensures that
4126 : * session-level locks are cleaned up on abort.
4127 : */
4128 540 : CHECK_FOR_INTERRUPTS();
4129 :
4130 : /* Tell concurrent indexing to ignore us, if index qualifies */
4131 540 : if (newidx->safe)
4132 492 : set_indexsafe_procflags();
4133 :
4134 : /* Set ActiveSnapshot since functions in the indexes may need it */
4135 540 : PushActiveSnapshot(GetTransactionSnapshot());
4136 :
4137 : /*
4138 : * Update progress for the index to build, with the correct parent
4139 : * table involved.
4140 : */
4141 540 : pgstat_progress_start_command(PROGRESS_COMMAND_CREATE_INDEX, newidx->tableId);
4142 540 : progress_vals[0] = PROGRESS_CREATEIDX_COMMAND_REINDEX_CONCURRENTLY;
4143 540 : progress_vals[1] = PROGRESS_CREATEIDX_PHASE_BUILD;
4144 540 : progress_vals[2] = newidx->indexId;
4145 540 : progress_vals[3] = newidx->amId;
4146 540 : pgstat_progress_update_multi_param(4, progress_index, progress_vals);
4147 :
4148 : /* Perform concurrent build of new index */
4149 540 : index_concurrently_build(newidx->tableId, newidx->indexId);
4150 :
4151 534 : PopActiveSnapshot();
4152 534 : CommitTransactionCommand();
4153 : }
4154 :
4155 434 : StartTransactionCommand();
4156 :
4157 : /*
4158 : * Because we don't take a snapshot or Xid in this transaction, there's no
4159 : * need to set the PROC_IN_SAFE_IC flag here.
4160 : */
4161 :
4162 : /*
4163 : * Phase 3 of REINDEX CONCURRENTLY
4164 : *
4165 : * During this phase the old indexes catch up with any new tuples that
4166 : * were created during the previous phase. See "phase 3" in DefineIndex()
4167 : * for more details.
4168 : */
4169 :
4170 434 : pgstat_progress_update_param(PROGRESS_CREATEIDX_PHASE,
4171 : PROGRESS_CREATEIDX_PHASE_WAIT_2);
4172 434 : WaitForLockersMultiple(lockTags, ShareLock, true);
4173 434 : CommitTransactionCommand();
4174 :
4175 968 : foreach(lc, newIndexIds)
4176 : {
4177 534 : ReindexIndexInfo *newidx = lfirst(lc);
4178 : TransactionId limitXmin;
4179 : Snapshot snapshot;
4180 :
4181 534 : StartTransactionCommand();
4182 :
4183 : /*
4184 : * Check for user-requested abort. This is inside a transaction so as
4185 : * xact.c does not issue a useless WARNING, and ensures that
4186 : * session-level locks are cleaned up on abort.
4187 : */
4188 534 : CHECK_FOR_INTERRUPTS();
4189 :
4190 : /* Tell concurrent indexing to ignore us, if index qualifies */
4191 534 : if (newidx->safe)
4192 486 : set_indexsafe_procflags();
4193 :
4194 : /*
4195 : * Take the "reference snapshot" that will be used by validate_index()
4196 : * to filter candidate tuples.
4197 : */
4198 534 : snapshot = RegisterSnapshot(GetTransactionSnapshot());
4199 534 : PushActiveSnapshot(snapshot);
4200 :
4201 : /*
4202 : * Update progress for the index to build, with the correct parent
4203 : * table involved.
4204 : */
4205 534 : pgstat_progress_start_command(PROGRESS_COMMAND_CREATE_INDEX, newidx->tableId);
4206 534 : progress_vals[0] = PROGRESS_CREATEIDX_COMMAND_REINDEX_CONCURRENTLY;
4207 534 : progress_vals[1] = PROGRESS_CREATEIDX_PHASE_VALIDATE_IDXSCAN;
4208 534 : progress_vals[2] = newidx->indexId;
4209 534 : progress_vals[3] = newidx->amId;
4210 534 : pgstat_progress_update_multi_param(4, progress_index, progress_vals);
4211 :
4212 534 : validate_index(newidx->tableId, newidx->indexId, snapshot);
4213 :
4214 : /*
4215 : * We can now do away with our active snapshot, we still need to save
4216 : * the xmin limit to wait for older snapshots.
4217 : */
4218 534 : limitXmin = snapshot->xmin;
4219 :
4220 534 : PopActiveSnapshot();
4221 534 : UnregisterSnapshot(snapshot);
4222 :
4223 : /*
4224 : * To ensure no deadlocks, we must commit and start yet another
4225 : * transaction, and do our wait before any snapshot has been taken in
4226 : * it.
4227 : */
4228 534 : CommitTransactionCommand();
4229 534 : StartTransactionCommand();
4230 :
4231 : /*
4232 : * The index is now valid in the sense that it contains all currently
4233 : * interesting tuples. But since it might not contain tuples deleted
4234 : * just before the reference snap was taken, we have to wait out any
4235 : * transactions that might have older snapshots.
4236 : *
4237 : * Because we don't take a snapshot or Xid in this transaction,
4238 : * there's no need to set the PROC_IN_SAFE_IC flag here.
4239 : */
4240 534 : pgstat_progress_update_param(PROGRESS_CREATEIDX_PHASE,
4241 : PROGRESS_CREATEIDX_PHASE_WAIT_3);
4242 534 : WaitForOlderSnapshots(limitXmin, true);
4243 :
4244 534 : CommitTransactionCommand();
4245 : }
4246 :
4247 : /*
4248 : * Phase 4 of REINDEX CONCURRENTLY
4249 : *
4250 : * Now that the new indexes have been validated, swap each new index with
4251 : * its corresponding old index.
4252 : *
4253 : * We mark the new indexes as valid and the old indexes as not valid at
4254 : * the same time to make sure we only get constraint violations from the
4255 : * indexes with the correct names.
4256 : */
4257 :
4258 434 : INJECTION_POINT("reindex-relation-concurrently-before-swap", NULL);
4259 434 : StartTransactionCommand();
4260 :
4261 : /*
4262 : * Because this transaction only does catalog manipulations and doesn't do
4263 : * any index operations, we can set the PROC_IN_SAFE_IC flag here
4264 : * unconditionally.
4265 : */
4266 434 : set_indexsafe_procflags();
4267 :
4268 968 : forboth(lc, indexIds, lc2, newIndexIds)
4269 : {
4270 534 : ReindexIndexInfo *oldidx = lfirst(lc);
4271 534 : ReindexIndexInfo *newidx = lfirst(lc2);
4272 : char *oldName;
4273 :
4274 : /*
4275 : * Check for user-requested abort. This is inside a transaction so as
4276 : * xact.c does not issue a useless WARNING, and ensures that
4277 : * session-level locks are cleaned up on abort.
4278 : */
4279 534 : CHECK_FOR_INTERRUPTS();
4280 :
4281 : /* Choose a relation name for old index */
4282 534 : oldName = ChooseRelationName(get_rel_name(oldidx->indexId),
4283 : NULL,
4284 : "ccold",
4285 : get_rel_namespace(oldidx->tableId),
4286 : false);
4287 :
4288 : /*
4289 : * Swapping the indexes might involve TOAST table access, so ensure we
4290 : * have a valid snapshot.
4291 : */
4292 534 : PushActiveSnapshot(GetTransactionSnapshot());
4293 :
4294 : /*
4295 : * Swap old index with the new one. This also marks the new one as
4296 : * valid and the old one as not valid.
4297 : */
4298 534 : index_concurrently_swap(newidx->indexId, oldidx->indexId, oldName);
4299 :
4300 534 : PopActiveSnapshot();
4301 :
4302 : /*
4303 : * Invalidate the relcache for the table, so that after this commit
4304 : * all sessions will refresh any cached plans that might reference the
4305 : * index.
4306 : */
4307 534 : CacheInvalidateRelcacheByRelid(oldidx->tableId);
4308 :
4309 : /*
4310 : * CCI here so that subsequent iterations see the oldName in the
4311 : * catalog and can choose a nonconflicting name for their oldName.
4312 : * Otherwise, this could lead to conflicts if a table has two indexes
4313 : * whose names are equal for the first NAMEDATALEN-minus-a-few
4314 : * characters.
4315 : */
4316 534 : CommandCounterIncrement();
4317 : }
4318 :
4319 : /* Commit this transaction and make index swaps visible */
4320 434 : CommitTransactionCommand();
4321 434 : StartTransactionCommand();
4322 :
4323 : /*
4324 : * While we could set PROC_IN_SAFE_IC if all indexes qualified, there's no
4325 : * real need for that, because we only acquire an Xid after the wait is
4326 : * done, and that lasts for a very short period.
4327 : */
4328 :
4329 : /*
4330 : * Phase 5 of REINDEX CONCURRENTLY
4331 : *
4332 : * Mark the old indexes as dead. First we must wait until no running
4333 : * transaction could be using the index for a query. See also
4334 : * index_drop() for more details.
4335 : */
4336 :
4337 434 : INJECTION_POINT("reindex-relation-concurrently-before-set-dead", NULL);
4338 434 : pgstat_progress_update_param(PROGRESS_CREATEIDX_PHASE,
4339 : PROGRESS_CREATEIDX_PHASE_WAIT_4);
4340 434 : WaitForLockersMultiple(lockTags, AccessExclusiveLock, true);
4341 :
4342 968 : foreach(lc, indexIds)
4343 : {
4344 534 : ReindexIndexInfo *oldidx = lfirst(lc);
4345 :
4346 : /*
4347 : * Check for user-requested abort. This is inside a transaction so as
4348 : * xact.c does not issue a useless WARNING, and ensures that
4349 : * session-level locks are cleaned up on abort.
4350 : */
4351 534 : CHECK_FOR_INTERRUPTS();
4352 :
4353 : /*
4354 : * Updating pg_index might involve TOAST table access, so ensure we
4355 : * have a valid snapshot.
4356 : */
4357 534 : PushActiveSnapshot(GetTransactionSnapshot());
4358 :
4359 534 : index_concurrently_set_dead(oldidx->tableId, oldidx->indexId);
4360 :
4361 534 : PopActiveSnapshot();
4362 : }
4363 :
4364 : /* Commit this transaction to make the updates visible. */
4365 434 : CommitTransactionCommand();
4366 434 : StartTransactionCommand();
4367 :
4368 : /*
4369 : * While we could set PROC_IN_SAFE_IC if all indexes qualified, there's no
4370 : * real need for that, because we only acquire an Xid after the wait is
4371 : * done, and that lasts for a very short period.
4372 : */
4373 :
4374 : /*
4375 : * Phase 6 of REINDEX CONCURRENTLY
4376 : *
4377 : * Drop the old indexes.
4378 : */
4379 :
4380 434 : pgstat_progress_update_param(PROGRESS_CREATEIDX_PHASE,
4381 : PROGRESS_CREATEIDX_PHASE_WAIT_5);
4382 434 : WaitForLockersMultiple(lockTags, AccessExclusiveLock, true);
4383 :
4384 434 : PushActiveSnapshot(GetTransactionSnapshot());
4385 :
4386 : {
4387 434 : ObjectAddresses *objects = new_object_addresses();
4388 :
4389 968 : foreach(lc, indexIds)
4390 : {
4391 534 : ReindexIndexInfo *idx = lfirst(lc);
4392 : ObjectAddress object;
4393 :
4394 534 : object.classId = RelationRelationId;
4395 534 : object.objectId = idx->indexId;
4396 534 : object.objectSubId = 0;
4397 :
4398 534 : add_exact_object_address(&object, objects);
4399 : }
4400 :
4401 : /*
4402 : * Use PERFORM_DELETION_CONCURRENT_LOCK so that index_drop() uses the
4403 : * right lock level.
4404 : */
4405 434 : performMultipleDeletions(objects, DROP_RESTRICT,
4406 : PERFORM_DELETION_CONCURRENT_LOCK | PERFORM_DELETION_INTERNAL);
4407 : }
4408 :
4409 434 : PopActiveSnapshot();
4410 434 : CommitTransactionCommand();
4411 :
4412 : /*
4413 : * Finally, release the session-level lock on the table.
4414 : */
4415 2024 : foreach(lc, relationLocks)
4416 : {
4417 1590 : LockRelId *lockrelid = (LockRelId *) lfirst(lc);
4418 :
4419 1590 : UnlockRelationIdForSession(lockrelid, ShareUpdateExclusiveLock);
4420 : }
4421 :
4422 : /* Start a new transaction to finish process properly */
4423 434 : StartTransactionCommand();
4424 :
4425 : /* Log what we did */
4426 434 : if ((params->options & REINDEXOPT_VERBOSE) != 0)
4427 : {
4428 4 : if (relkind == RELKIND_INDEX)
4429 0 : ereport(INFO,
4430 : (errmsg("index \"%s.%s\" was reindexed",
4431 : relationNamespace, relationName),
4432 : errdetail("%s.",
4433 : pg_rusage_show(&ru0))));
4434 : else
4435 : {
4436 12 : foreach(lc, newIndexIds)
4437 : {
4438 8 : ReindexIndexInfo *idx = lfirst(lc);
4439 8 : Oid indOid = idx->indexId;
4440 :
4441 8 : ereport(INFO,
4442 : (errmsg("index \"%s.%s\" was reindexed",
4443 : get_namespace_name(get_rel_namespace(indOid)),
4444 : get_rel_name(indOid))));
4445 : /* Don't show rusage here, since it's not per index. */
4446 : }
4447 :
4448 4 : ereport(INFO,
4449 : (errmsg("table \"%s.%s\" was reindexed",
4450 : relationNamespace, relationName),
4451 : errdetail("%s.",
4452 : pg_rusage_show(&ru0))));
4453 : }
4454 : }
4455 :
4456 434 : MemoryContextDelete(private_context);
4457 :
4458 434 : pgstat_progress_end_command();
4459 :
4460 434 : return true;
4461 : }
4462 :
4463 : /*
4464 : * Insert or delete an appropriate pg_inherits tuple to make the given index
4465 : * be a partition of the indicated parent index.
4466 : *
4467 : * This also corrects the pg_depend information for the affected index.
4468 : */
4469 : void
4470 1032 : IndexSetParentIndex(Relation partitionIdx, Oid parentOid)
4471 : {
4472 : Relation pg_inherits;
4473 : ScanKeyData key[2];
4474 : SysScanDesc scan;
4475 1032 : Oid partRelid = RelationGetRelid(partitionIdx);
4476 : HeapTuple tuple;
4477 : bool fix_dependencies;
4478 :
4479 : /* Make sure this is an index */
4480 : Assert(partitionIdx->rd_rel->relkind == RELKIND_INDEX ||
4481 : partitionIdx->rd_rel->relkind == RELKIND_PARTITIONED_INDEX);
4482 :
4483 : /*
4484 : * Scan pg_inherits for rows linking our index to some parent.
4485 : */
4486 1032 : pg_inherits = relation_open(InheritsRelationId, RowExclusiveLock);
4487 1032 : ScanKeyInit(&key[0],
4488 : Anum_pg_inherits_inhrelid,
4489 : BTEqualStrategyNumber, F_OIDEQ,
4490 : ObjectIdGetDatum(partRelid));
4491 1032 : ScanKeyInit(&key[1],
4492 : Anum_pg_inherits_inhseqno,
4493 : BTEqualStrategyNumber, F_INT4EQ,
4494 : Int32GetDatum(1));
4495 1032 : scan = systable_beginscan(pg_inherits, InheritsRelidSeqnoIndexId, true,
4496 : NULL, 2, key);
4497 1032 : tuple = systable_getnext(scan);
4498 :
4499 1032 : if (!HeapTupleIsValid(tuple))
4500 : {
4501 604 : if (parentOid == InvalidOid)
4502 : {
4503 : /*
4504 : * No pg_inherits row, and no parent wanted: nothing to do in this
4505 : * case.
4506 : */
4507 0 : fix_dependencies = false;
4508 : }
4509 : else
4510 : {
4511 604 : StoreSingleInheritance(partRelid, parentOid, 1);
4512 604 : fix_dependencies = true;
4513 : }
4514 : }
4515 : else
4516 : {
4517 428 : Form_pg_inherits inhForm = (Form_pg_inherits) GETSTRUCT(tuple);
4518 :
4519 428 : if (parentOid == InvalidOid)
4520 : {
4521 : /*
4522 : * There exists a pg_inherits row, which we want to clear; do so.
4523 : */
4524 428 : CatalogTupleDelete(pg_inherits, &tuple->t_self);
4525 428 : fix_dependencies = true;
4526 : }
4527 : else
4528 : {
4529 : /*
4530 : * A pg_inherits row exists. If it's the same we want, then we're
4531 : * good; if it differs, that amounts to a corrupt catalog and
4532 : * should not happen.
4533 : */
4534 0 : if (inhForm->inhparent != parentOid)
4535 : {
4536 : /* unexpected: we should not get called in this case */
4537 0 : elog(ERROR, "bogus pg_inherit row: inhrelid %u inhparent %u",
4538 : inhForm->inhrelid, inhForm->inhparent);
4539 : }
4540 :
4541 : /* already in the right state */
4542 0 : fix_dependencies = false;
4543 : }
4544 : }
4545 :
4546 : /* done with pg_inherits */
4547 1032 : systable_endscan(scan);
4548 1032 : relation_close(pg_inherits, RowExclusiveLock);
4549 :
4550 : /* set relhassubclass if an index partition has been added to the parent */
4551 1032 : if (OidIsValid(parentOid))
4552 : {
4553 604 : LockRelationOid(parentOid, ShareUpdateExclusiveLock);
4554 604 : SetRelationHasSubclass(parentOid, true);
4555 : }
4556 :
4557 : /* set relispartition correctly on the partition */
4558 1032 : update_relispartition(partRelid, OidIsValid(parentOid));
4559 :
4560 1032 : if (fix_dependencies)
4561 : {
4562 : /*
4563 : * Insert/delete pg_depend rows. If setting a parent, add PARTITION
4564 : * dependencies on the parent index and the table; if removing a
4565 : * parent, delete PARTITION dependencies.
4566 : */
4567 1032 : if (OidIsValid(parentOid))
4568 : {
4569 : ObjectAddress partIdx;
4570 : ObjectAddress parentIdx;
4571 : ObjectAddress partitionTbl;
4572 :
4573 604 : ObjectAddressSet(partIdx, RelationRelationId, partRelid);
4574 604 : ObjectAddressSet(parentIdx, RelationRelationId, parentOid);
4575 604 : ObjectAddressSet(partitionTbl, RelationRelationId,
4576 : partitionIdx->rd_index->indrelid);
4577 604 : recordDependencyOn(&partIdx, &parentIdx,
4578 : DEPENDENCY_PARTITION_PRI);
4579 604 : recordDependencyOn(&partIdx, &partitionTbl,
4580 : DEPENDENCY_PARTITION_SEC);
4581 : }
4582 : else
4583 : {
4584 428 : deleteDependencyRecordsForClass(RelationRelationId, partRelid,
4585 : RelationRelationId,
4586 : DEPENDENCY_PARTITION_PRI);
4587 428 : deleteDependencyRecordsForClass(RelationRelationId, partRelid,
4588 : RelationRelationId,
4589 : DEPENDENCY_PARTITION_SEC);
4590 : }
4591 :
4592 : /* make our updates visible */
4593 1032 : CommandCounterIncrement();
4594 : }
4595 1032 : }
4596 :
4597 : /*
4598 : * Subroutine of IndexSetParentIndex to update the relispartition flag of the
4599 : * given index to the given value.
4600 : */
4601 : static void
4602 1032 : update_relispartition(Oid relationId, bool newval)
4603 : {
4604 : HeapTuple tup;
4605 : Relation classRel;
4606 : ItemPointerData otid;
4607 :
4608 1032 : classRel = table_open(RelationRelationId, RowExclusiveLock);
4609 1032 : tup = SearchSysCacheLockedCopy1(RELOID, ObjectIdGetDatum(relationId));
4610 1032 : if (!HeapTupleIsValid(tup))
4611 0 : elog(ERROR, "cache lookup failed for relation %u", relationId);
4612 1032 : otid = tup->t_self;
4613 : Assert(((Form_pg_class) GETSTRUCT(tup))->relispartition != newval);
4614 1032 : ((Form_pg_class) GETSTRUCT(tup))->relispartition = newval;
4615 1032 : CatalogTupleUpdate(classRel, &otid, tup);
4616 1032 : UnlockTuple(classRel, &otid, InplaceUpdateTupleLock);
4617 1032 : heap_freetuple(tup);
4618 1032 : table_close(classRel, RowExclusiveLock);
4619 1032 : }
4620 :
4621 : /*
4622 : * Set the PROC_IN_SAFE_IC flag in MyProc->statusFlags.
4623 : *
4624 : * When doing concurrent index builds, we can set this flag
4625 : * to tell other processes concurrently running CREATE
4626 : * INDEX CONCURRENTLY or REINDEX CONCURRENTLY to ignore us when
4627 : * doing their waits for concurrent snapshots. On one hand it
4628 : * avoids pointlessly waiting for a process that's not interesting
4629 : * anyway; but more importantly it avoids deadlocks in some cases.
4630 : *
4631 : * This can be done safely only for indexes that don't execute any
4632 : * expressions that could access other tables, so index must not be
4633 : * expressional nor partial. Caller is responsible for only calling
4634 : * this routine when that assumption holds true.
4635 : *
4636 : * (The flag is reset automatically at transaction end, so it must be
4637 : * set for each transaction.)
4638 : */
4639 : static inline void
4640 1778 : set_indexsafe_procflags(void)
4641 : {
4642 : /*
4643 : * This should only be called before installing xid or xmin in MyProc;
4644 : * otherwise, concurrent processes could see an Xmin that moves backwards.
4645 : */
4646 : Assert(MyProc->xid == InvalidTransactionId &&
4647 : MyProc->xmin == InvalidTransactionId);
4648 :
4649 1778 : LWLockAcquire(ProcArrayLock, LW_EXCLUSIVE);
4650 1778 : MyProc->statusFlags |= PROC_IN_SAFE_IC;
4651 1778 : ProcGlobal->statusFlags[MyProc->pgxactoff] = MyProc->statusFlags;
4652 1778 : LWLockRelease(ProcArrayLock);
4653 1778 : }
|
