Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * predicate.c
4 : * POSTGRES predicate locking
5 : * to support full serializable transaction isolation
6 : *
7 : *
8 : * The approach taken is to implement Serializable Snapshot Isolation (SSI)
9 : * as initially described in this paper:
10 : *
11 : * Michael J. Cahill, Uwe Röhm, and Alan D. Fekete. 2008.
12 : * Serializable isolation for snapshot databases.
13 : * In SIGMOD '08: Proceedings of the 2008 ACM SIGMOD
14 : * international conference on Management of data,
15 : * pages 729-738, New York, NY, USA. ACM.
16 : * http://doi.acm.org/10.1145/1376616.1376690
17 : *
18 : * and further elaborated in Cahill's doctoral thesis:
19 : *
20 : * Michael James Cahill. 2009.
21 : * Serializable Isolation for Snapshot Databases.
22 : * Sydney Digital Theses.
23 : * University of Sydney, School of Information Technologies.
24 : * http://hdl.handle.net/2123/5353
25 : *
26 : *
27 : * Predicate locks for Serializable Snapshot Isolation (SSI) are SIREAD
28 : * locks, which are so different from normal locks that a distinct set of
29 : * structures is required to handle them. They are needed to detect
30 : * rw-conflicts when the read happens before the write. (When the write
31 : * occurs first, the reading transaction can check for a conflict by
32 : * examining the MVCC data.)
33 : *
34 : * (1) Besides tuples actually read, they must cover ranges of tuples
35 : * which would have been read based on the predicate. This will
36 : * require modelling the predicates through locks against database
37 : * objects such as pages, index ranges, or entire tables.
38 : *
39 : * (2) They must be kept in RAM for quick access. Because of this, it
40 : * isn't possible to always maintain tuple-level granularity -- when
41 : * the space allocated to store these approaches exhaustion, a
42 : * request for a lock may need to scan for situations where a single
43 : * transaction holds many fine-grained locks which can be coalesced
44 : * into a single coarser-grained lock.
45 : *
46 : * (3) They never block anything; they are more like flags than locks
47 : * in that regard; although they refer to database objects and are
48 : * used to identify rw-conflicts with normal write locks.
49 : *
50 : * (4) While they are associated with a transaction, they must survive
51 : * a successful COMMIT of that transaction, and remain until all
52 : * overlapping transactions complete. This even means that they
53 : * must survive termination of the transaction's process. If a
54 : * top level transaction is rolled back, however, it is immediately
55 : * flagged so that it can be ignored, and its SIREAD locks can be
56 : * released any time after that.
57 : *
58 : * (5) The only transactions which create SIREAD locks or check for
59 : * conflicts with them are serializable transactions.
60 : *
61 : * (6) When a write lock for a top level transaction is found to cover
62 : * an existing SIREAD lock for the same transaction, the SIREAD lock
63 : * can be deleted.
64 : *
65 : * (7) A write from a serializable transaction must ensure that an xact
66 : * record exists for the transaction, with the same lifespan (until
67 : * all concurrent transaction complete or the transaction is rolled
68 : * back) so that rw-dependencies to that transaction can be
69 : * detected.
70 : *
71 : * We use an optimization for read-only transactions. Under certain
72 : * circumstances, a read-only transaction's snapshot can be shown to
73 : * never have conflicts with other transactions. This is referred to
74 : * as a "safe" snapshot (and one known not to be is "unsafe").
75 : * However, it can't be determined whether a snapshot is safe until
76 : * all concurrent read/write transactions complete.
77 : *
78 : * Once a read-only transaction is known to have a safe snapshot, it
79 : * can release its predicate locks and exempt itself from further
80 : * predicate lock tracking. READ ONLY DEFERRABLE transactions run only
81 : * on safe snapshots, waiting as necessary for one to be available.
82 : *
83 : *
84 : * Lightweight locks to manage access to the predicate locking shared
85 : * memory objects must be taken in this order, and should be released in
86 : * reverse order:
87 : *
88 : * SerializableFinishedListLock
89 : * - Protects the list of transactions which have completed but which
90 : * may yet matter because they overlap still-active transactions.
91 : *
92 : * SerializablePredicateListLock
93 : * - Protects the linked list of locks held by a transaction. Note
94 : * that the locks themselves are also covered by the partition
95 : * locks of their respective lock targets; this lock only affects
96 : * the linked list connecting the locks related to a transaction.
97 : * - All transactions share this single lock (with no partitioning).
98 : * - There is never a need for a process other than the one running
99 : * an active transaction to walk the list of locks held by that
100 : * transaction, except parallel query workers sharing the leader's
101 : * transaction. In the parallel case, an extra per-sxact lock is
102 : * taken; see below.
103 : * - It is relatively infrequent that another process needs to
104 : * modify the list for a transaction, but it does happen for such
105 : * things as index page splits for pages with predicate locks and
106 : * freeing of predicate locked pages by a vacuum process. When
107 : * removing a lock in such cases, the lock itself contains the
108 : * pointers needed to remove it from the list. When adding a
109 : * lock in such cases, the lock can be added using the anchor in
110 : * the transaction structure. Neither requires walking the list.
111 : * - Cleaning up the list for a terminated transaction is sometimes
112 : * not done on a retail basis, in which case no lock is required.
113 : * - Due to the above, a process accessing its active transaction's
114 : * list always uses a shared lock, regardless of whether it is
115 : * walking or maintaining the list. This improves concurrency
116 : * for the common access patterns.
117 : * - A process which needs to alter the list of a transaction other
118 : * than its own active transaction must acquire an exclusive
119 : * lock.
120 : *
121 : * SERIALIZABLEXACT's member 'perXactPredicateListLock'
122 : * - Protects the linked list of predicate locks held by a transaction.
123 : * Only needed for parallel mode, where multiple backends share the
124 : * same SERIALIZABLEXACT object. Not needed if
125 : * SerializablePredicateListLock is held exclusively.
126 : *
127 : * PredicateLockHashPartitionLock(hashcode)
128 : * - The same lock protects a target, all locks on that target, and
129 : * the linked list of locks on the target.
130 : * - When more than one is needed, acquire in ascending address order.
131 : * - When all are needed (rare), acquire in ascending index order with
132 : * PredicateLockHashPartitionLockByIndex(index).
133 : *
134 : * SerializableXactHashLock
135 : * - Protects both PredXact and SerializableXidHash.
136 : *
137 : * SerialControlLock
138 : * - Protects SerialControlData members
139 : *
140 : * SLRU per-bank locks
141 : * - Protects SerialSlruCtl
142 : *
143 : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
144 : * Portions Copyright (c) 1994, Regents of the University of California
145 : *
146 : *
147 : * IDENTIFICATION
148 : * src/backend/storage/lmgr/predicate.c
149 : *
150 : *-------------------------------------------------------------------------
151 : */
152 : /*
153 : * INTERFACE ROUTINES
154 : *
155 : * housekeeping for setting up shared memory predicate lock structures
156 : * PredicateLockShmemInit(void)
157 : * PredicateLockShmemSize(void)
158 : *
159 : * predicate lock reporting
160 : * GetPredicateLockStatusData(void)
161 : * PageIsPredicateLocked(Relation relation, BlockNumber blkno)
162 : *
163 : * predicate lock maintenance
164 : * GetSerializableTransactionSnapshot(Snapshot snapshot)
165 : * SetSerializableTransactionSnapshot(Snapshot snapshot,
166 : * VirtualTransactionId *sourcevxid)
167 : * RegisterPredicateLockingXid(void)
168 : * PredicateLockRelation(Relation relation, Snapshot snapshot)
169 : * PredicateLockPage(Relation relation, BlockNumber blkno,
170 : * Snapshot snapshot)
171 : * PredicateLockTID(Relation relation, ItemPointer tid, Snapshot snapshot,
172 : * TransactionId tuple_xid)
173 : * PredicateLockPageSplit(Relation relation, BlockNumber oldblkno,
174 : * BlockNumber newblkno)
175 : * PredicateLockPageCombine(Relation relation, BlockNumber oldblkno,
176 : * BlockNumber newblkno)
177 : * TransferPredicateLocksToHeapRelation(Relation relation)
178 : * ReleasePredicateLocks(bool isCommit, bool isReadOnlySafe)
179 : *
180 : * conflict detection (may also trigger rollback)
181 : * CheckForSerializableConflictOut(Relation relation, TransactionId xid,
182 : * Snapshot snapshot)
183 : * CheckForSerializableConflictIn(Relation relation, ItemPointer tid,
184 : * BlockNumber blkno)
185 : * CheckTableForSerializableConflictIn(Relation relation)
186 : *
187 : * final rollback checking
188 : * PreCommit_CheckForSerializationFailure(void)
189 : *
190 : * two-phase commit support
191 : * AtPrepare_PredicateLocks(void);
192 : * PostPrepare_PredicateLocks(TransactionId xid);
193 : * PredicateLockTwoPhaseFinish(TransactionId xid, bool isCommit);
194 : * predicatelock_twophase_recover(TransactionId xid, uint16 info,
195 : * void *recdata, uint32 len);
196 : */
197 :
198 : #include "postgres.h"
199 :
200 : #include "access/parallel.h"
201 : #include "access/slru.h"
202 : #include "access/transam.h"
203 : #include "access/twophase.h"
204 : #include "access/twophase_rmgr.h"
205 : #include "access/xact.h"
206 : #include "access/xlog.h"
207 : #include "miscadmin.h"
208 : #include "pgstat.h"
209 : #include "port/pg_lfind.h"
210 : #include "storage/predicate.h"
211 : #include "storage/predicate_internals.h"
212 : #include "storage/proc.h"
213 : #include "storage/procarray.h"
214 : #include "utils/guc_hooks.h"
215 : #include "utils/rel.h"
216 : #include "utils/snapmgr.h"
217 :
218 : /* Uncomment the next line to test the graceful degradation code. */
219 : /* #define TEST_SUMMARIZE_SERIAL */
220 :
221 : /*
222 : * Test the most selective fields first, for performance.
223 : *
224 : * a is covered by b if all of the following hold:
225 : * 1) a.database = b.database
226 : * 2) a.relation = b.relation
227 : * 3) b.offset is invalid (b is page-granularity or higher)
228 : * 4) either of the following:
229 : * 4a) a.offset is valid (a is tuple-granularity) and a.page = b.page
230 : * or 4b) a.offset is invalid and b.page is invalid (a is
231 : * page-granularity and b is relation-granularity
232 : */
233 : #define TargetTagIsCoveredBy(covered_target, covering_target) \
234 : ((GET_PREDICATELOCKTARGETTAG_RELATION(covered_target) == /* (2) */ \
235 : GET_PREDICATELOCKTARGETTAG_RELATION(covering_target)) \
236 : && (GET_PREDICATELOCKTARGETTAG_OFFSET(covering_target) == \
237 : InvalidOffsetNumber) /* (3) */ \
238 : && (((GET_PREDICATELOCKTARGETTAG_OFFSET(covered_target) != \
239 : InvalidOffsetNumber) /* (4a) */ \
240 : && (GET_PREDICATELOCKTARGETTAG_PAGE(covering_target) == \
241 : GET_PREDICATELOCKTARGETTAG_PAGE(covered_target))) \
242 : || ((GET_PREDICATELOCKTARGETTAG_PAGE(covering_target) == \
243 : InvalidBlockNumber) /* (4b) */ \
244 : && (GET_PREDICATELOCKTARGETTAG_PAGE(covered_target) \
245 : != InvalidBlockNumber))) \
246 : && (GET_PREDICATELOCKTARGETTAG_DB(covered_target) == /* (1) */ \
247 : GET_PREDICATELOCKTARGETTAG_DB(covering_target)))
248 :
249 : /*
250 : * The predicate locking target and lock shared hash tables are partitioned to
251 : * reduce contention. To determine which partition a given target belongs to,
252 : * compute the tag's hash code with PredicateLockTargetTagHashCode(), then
253 : * apply one of these macros.
254 : * NB: NUM_PREDICATELOCK_PARTITIONS must be a power of 2!
255 : */
256 : #define PredicateLockHashPartition(hashcode) \
257 : ((hashcode) % NUM_PREDICATELOCK_PARTITIONS)
258 : #define PredicateLockHashPartitionLock(hashcode) \
259 : (&MainLWLockArray[PREDICATELOCK_MANAGER_LWLOCK_OFFSET + \
260 : PredicateLockHashPartition(hashcode)].lock)
261 : #define PredicateLockHashPartitionLockByIndex(i) \
262 : (&MainLWLockArray[PREDICATELOCK_MANAGER_LWLOCK_OFFSET + (i)].lock)
263 :
264 : #define NPREDICATELOCKTARGETENTS() \
265 : mul_size(max_predicate_locks_per_xact, add_size(MaxBackends, max_prepared_xacts))
266 :
267 : #define SxactIsOnFinishedList(sxact) (!dlist_node_is_detached(&(sxact)->finishedLink))
268 :
269 : /*
270 : * Note that a sxact is marked "prepared" once it has passed
271 : * PreCommit_CheckForSerializationFailure, even if it isn't using
272 : * 2PC. This is the point at which it can no longer be aborted.
273 : *
274 : * The PREPARED flag remains set after commit, so SxactIsCommitted
275 : * implies SxactIsPrepared.
276 : */
277 : #define SxactIsCommitted(sxact) (((sxact)->flags & SXACT_FLAG_COMMITTED) != 0)
278 : #define SxactIsPrepared(sxact) (((sxact)->flags & SXACT_FLAG_PREPARED) != 0)
279 : #define SxactIsRolledBack(sxact) (((sxact)->flags & SXACT_FLAG_ROLLED_BACK) != 0)
280 : #define SxactIsDoomed(sxact) (((sxact)->flags & SXACT_FLAG_DOOMED) != 0)
281 : #define SxactIsReadOnly(sxact) (((sxact)->flags & SXACT_FLAG_READ_ONLY) != 0)
282 : #define SxactHasSummaryConflictIn(sxact) (((sxact)->flags & SXACT_FLAG_SUMMARY_CONFLICT_IN) != 0)
283 : #define SxactHasSummaryConflictOut(sxact) (((sxact)->flags & SXACT_FLAG_SUMMARY_CONFLICT_OUT) != 0)
284 : /*
285 : * The following macro actually means that the specified transaction has a
286 : * conflict out *to a transaction which committed ahead of it*. It's hard
287 : * to get that into a name of a reasonable length.
288 : */
289 : #define SxactHasConflictOut(sxact) (((sxact)->flags & SXACT_FLAG_CONFLICT_OUT) != 0)
290 : #define SxactIsDeferrableWaiting(sxact) (((sxact)->flags & SXACT_FLAG_DEFERRABLE_WAITING) != 0)
291 : #define SxactIsROSafe(sxact) (((sxact)->flags & SXACT_FLAG_RO_SAFE) != 0)
292 : #define SxactIsROUnsafe(sxact) (((sxact)->flags & SXACT_FLAG_RO_UNSAFE) != 0)
293 : #define SxactIsPartiallyReleased(sxact) (((sxact)->flags & SXACT_FLAG_PARTIALLY_RELEASED) != 0)
294 :
295 : /*
296 : * Compute the hash code associated with a PREDICATELOCKTARGETTAG.
297 : *
298 : * To avoid unnecessary recomputations of the hash code, we try to do this
299 : * just once per function, and then pass it around as needed. Aside from
300 : * passing the hashcode to hash_search_with_hash_value(), we can extract
301 : * the lock partition number from the hashcode.
302 : */
303 : #define PredicateLockTargetTagHashCode(predicatelocktargettag) \
304 : get_hash_value(PredicateLockTargetHash, predicatelocktargettag)
305 :
306 : /*
307 : * Given a predicate lock tag, and the hash for its target,
308 : * compute the lock hash.
309 : *
310 : * To make the hash code also depend on the transaction, we xor the sxid
311 : * struct's address into the hash code, left-shifted so that the
312 : * partition-number bits don't change. Since this is only a hash, we
313 : * don't care if we lose high-order bits of the address; use an
314 : * intermediate variable to suppress cast-pointer-to-int warnings.
315 : */
316 : #define PredicateLockHashCodeFromTargetHashCode(predicatelocktag, targethash) \
317 : ((targethash) ^ ((uint32) PointerGetDatum((predicatelocktag)->myXact)) \
318 : << LOG2_NUM_PREDICATELOCK_PARTITIONS)
319 :
320 :
321 : /*
322 : * The SLRU buffer area through which we access the old xids.
323 : */
324 : static SlruCtlData SerialSlruCtlData;
325 :
326 : #define SerialSlruCtl (&SerialSlruCtlData)
327 :
328 : #define SERIAL_PAGESIZE BLCKSZ
329 : #define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo)
330 : #define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE)
331 :
332 : /*
333 : * Set maximum pages based on the number needed to track all transactions.
334 : */
335 : #define SERIAL_MAX_PAGE (MaxTransactionId / SERIAL_ENTRIESPERPAGE)
336 :
337 : #define SerialNextPage(page) (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1)
338 :
339 : #define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
340 : (SerialSlruCtl->shared->page_buffer[slotno] + \
341 : ((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
342 :
343 : #define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
344 :
345 : typedef struct SerialControlData
346 : {
347 : int64 headPage; /* newest initialized page */
348 : TransactionId headXid; /* newest valid Xid in the SLRU */
349 : TransactionId tailXid; /* oldest xmin we might be interested in */
350 : } SerialControlData;
351 :
352 : typedef struct SerialControlData *SerialControl;
353 :
354 : static SerialControl serialControl;
355 :
356 : /*
357 : * When the oldest committed transaction on the "finished" list is moved to
358 : * SLRU, its predicate locks will be moved to this "dummy" transaction,
359 : * collapsing duplicate targets. When a duplicate is found, the later
360 : * commitSeqNo is used.
361 : */
362 : static SERIALIZABLEXACT *OldCommittedSxact;
363 :
364 :
365 : /*
366 : * These configuration variables are used to set the predicate lock table size
367 : * and to control promotion of predicate locks to coarser granularity in an
368 : * attempt to degrade performance (mostly as false positive serialization
369 : * failure) gracefully in the face of memory pressure.
370 : */
371 : int max_predicate_locks_per_xact; /* in guc_tables.c */
372 : int max_predicate_locks_per_relation; /* in guc_tables.c */
373 : int max_predicate_locks_per_page; /* in guc_tables.c */
374 :
375 : /*
376 : * This provides a list of objects in order to track transactions
377 : * participating in predicate locking. Entries in the list are fixed size,
378 : * and reside in shared memory. The memory address of an entry must remain
379 : * fixed during its lifetime. The list will be protected from concurrent
380 : * update externally; no provision is made in this code to manage that. The
381 : * number of entries in the list, and the size allowed for each entry is
382 : * fixed upon creation.
383 : */
384 : static PredXactList PredXact;
385 :
386 : /*
387 : * This provides a pool of RWConflict data elements to use in conflict lists
388 : * between transactions.
389 : */
390 : static RWConflictPoolHeader RWConflictPool;
391 :
392 : /*
393 : * The predicate locking hash tables are in shared memory.
394 : * Each backend keeps pointers to them.
395 : */
396 : static HTAB *SerializableXidHash;
397 : static HTAB *PredicateLockTargetHash;
398 : static HTAB *PredicateLockHash;
399 : static dlist_head *FinishedSerializableTransactions;
400 :
401 : /*
402 : * Tag for a dummy entry in PredicateLockTargetHash. By temporarily removing
403 : * this entry, you can ensure that there's enough scratch space available for
404 : * inserting one entry in the hash table. This is an otherwise-invalid tag.
405 : */
406 : static const PREDICATELOCKTARGETTAG ScratchTargetTag = {0, 0, 0, 0};
407 : static uint32 ScratchTargetTagHash;
408 : static LWLock *ScratchPartitionLock;
409 :
410 : /*
411 : * The local hash table used to determine when to combine multiple fine-
412 : * grained locks into a single courser-grained lock.
413 : */
414 : static HTAB *LocalPredicateLockHash = NULL;
415 :
416 : /*
417 : * Keep a pointer to the currently-running serializable transaction (if any)
418 : * for quick reference. Also, remember if we have written anything that could
419 : * cause a rw-conflict.
420 : */
421 : static SERIALIZABLEXACT *MySerializableXact = InvalidSerializableXact;
422 : static bool MyXactDidWrite = false;
423 :
424 : /*
425 : * The SXACT_FLAG_RO_UNSAFE optimization might lead us to release
426 : * MySerializableXact early. If that happens in a parallel query, the leader
427 : * needs to defer the destruction of the SERIALIZABLEXACT until end of
428 : * transaction, because the workers still have a reference to it. In that
429 : * case, the leader stores it here.
430 : */
431 : static SERIALIZABLEXACT *SavedSerializableXact = InvalidSerializableXact;
432 :
433 : /* local functions */
434 :
435 : static SERIALIZABLEXACT *CreatePredXact(void);
436 : static void ReleasePredXact(SERIALIZABLEXACT *sxact);
437 :
438 : static bool RWConflictExists(const SERIALIZABLEXACT *reader, const SERIALIZABLEXACT *writer);
439 : static void SetRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer);
440 : static void SetPossibleUnsafeConflict(SERIALIZABLEXACT *roXact, SERIALIZABLEXACT *activeXact);
441 : static void ReleaseRWConflict(RWConflict conflict);
442 : static void FlagSxactUnsafe(SERIALIZABLEXACT *sxact);
443 :
444 : static bool SerialPagePrecedesLogically(int64 page1, int64 page2);
445 : static void SerialInit(void);
446 : static void SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo);
447 : static SerCommitSeqNo SerialGetMinConflictCommitSeqNo(TransactionId xid);
448 : static void SerialSetActiveSerXmin(TransactionId xid);
449 :
450 : static uint32 predicatelock_hash(const void *key, Size keysize);
451 : static void SummarizeOldestCommittedSxact(void);
452 : static Snapshot GetSafeSnapshot(Snapshot origSnapshot);
453 : static Snapshot GetSerializableTransactionSnapshotInt(Snapshot snapshot,
454 : VirtualTransactionId *sourcevxid,
455 : int sourcepid);
456 : static bool PredicateLockExists(const PREDICATELOCKTARGETTAG *targettag);
457 : static bool GetParentPredicateLockTag(const PREDICATELOCKTARGETTAG *tag,
458 : PREDICATELOCKTARGETTAG *parent);
459 : static bool CoarserLockCovers(const PREDICATELOCKTARGETTAG *newtargettag);
460 : static void RemoveScratchTarget(bool lockheld);
461 : static void RestoreScratchTarget(bool lockheld);
462 : static void RemoveTargetIfNoLongerUsed(PREDICATELOCKTARGET *target,
463 : uint32 targettaghash);
464 : static void DeleteChildTargetLocks(const PREDICATELOCKTARGETTAG *newtargettag);
465 : static int MaxPredicateChildLocks(const PREDICATELOCKTARGETTAG *tag);
466 : static bool CheckAndPromotePredicateLockRequest(const PREDICATELOCKTARGETTAG *reqtag);
467 : static void DecrementParentLocks(const PREDICATELOCKTARGETTAG *targettag);
468 : static void CreatePredicateLock(const PREDICATELOCKTARGETTAG *targettag,
469 : uint32 targettaghash,
470 : SERIALIZABLEXACT *sxact);
471 : static void DeleteLockTarget(PREDICATELOCKTARGET *target, uint32 targettaghash);
472 : static bool TransferPredicateLocksToNewTarget(PREDICATELOCKTARGETTAG oldtargettag,
473 : PREDICATELOCKTARGETTAG newtargettag,
474 : bool removeOld);
475 : static void PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag);
476 : static void DropAllPredicateLocksFromTable(Relation relation,
477 : bool transfer);
478 : static void SetNewSxactGlobalXmin(void);
479 : static void ClearOldPredicateLocks(void);
480 : static void ReleaseOneSerializableXact(SERIALIZABLEXACT *sxact, bool partial,
481 : bool summarize);
482 : static bool XidIsConcurrent(TransactionId xid);
483 : static void CheckTargetForConflictsIn(PREDICATELOCKTARGETTAG *targettag);
484 : static void FlagRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer);
485 : static void OnConflict_CheckForSerializationFailure(const SERIALIZABLEXACT *reader,
486 : SERIALIZABLEXACT *writer);
487 : static void CreateLocalPredicateLockHash(void);
488 : static void ReleasePredicateLocksLocal(void);
489 :
490 :
491 : /*------------------------------------------------------------------------*/
492 :
493 : /*
494 : * Does this relation participate in predicate locking? Temporary and system
495 : * relations are exempt.
496 : */
497 : static inline bool
498 285400 : PredicateLockingNeededForRelation(Relation relation)
499 : {
500 399616 : return !(relation->rd_id < FirstUnpinnedObjectId ||
501 114216 : RelationUsesLocalBuffers(relation));
502 : }
503 :
504 : /*
505 : * When a public interface method is called for a read, this is the test to
506 : * see if we should do a quick return.
507 : *
508 : * Note: this function has side-effects! If this transaction has been flagged
509 : * as RO-safe since the last call, we release all predicate locks and reset
510 : * MySerializableXact. That makes subsequent calls to return quickly.
511 : *
512 : * This is marked as 'inline' to eliminate the function call overhead in the
513 : * common case that serialization is not needed.
514 : */
515 : static inline bool
516 113835800 : SerializationNeededForRead(Relation relation, Snapshot snapshot)
517 : {
518 : /* Nothing to do if this is not a serializable transaction */
519 113835800 : if (MySerializableXact == InvalidSerializableXact)
520 113561570 : return false;
521 :
522 : /*
523 : * Don't acquire locks or conflict when scanning with a special snapshot.
524 : * This excludes things like CLUSTER and REINDEX. They use the wholesale
525 : * functions TransferPredicateLocksToHeapRelation() and
526 : * CheckTableForSerializableConflictIn() to participate in serialization,
527 : * but the scans involved don't need serialization.
528 : */
529 274230 : if (!IsMVCCSnapshot(snapshot))
530 3214 : return false;
531 :
532 : /*
533 : * Check if we have just become "RO-safe". If we have, immediately release
534 : * all locks as they're not needed anymore. This also resets
535 : * MySerializableXact, so that subsequent calls to this function can exit
536 : * quickly.
537 : *
538 : * A transaction is flagged as RO_SAFE if all concurrent R/W transactions
539 : * commit without having conflicts out to an earlier snapshot, thus
540 : * ensuring that no conflicts are possible for this transaction.
541 : */
542 271016 : if (SxactIsROSafe(MySerializableXact))
543 : {
544 66 : ReleasePredicateLocks(false, true);
545 66 : return false;
546 : }
547 :
548 : /* Check if the relation doesn't participate in predicate locking */
549 270950 : if (!PredicateLockingNeededForRelation(relation))
550 166112 : return false;
551 :
552 104838 : return true; /* no excuse to skip predicate locking */
553 : }
554 :
555 : /*
556 : * Like SerializationNeededForRead(), but called on writes.
557 : * The logic is the same, but there is no snapshot and we can't be RO-safe.
558 : */
559 : static inline bool
560 31831068 : SerializationNeededForWrite(Relation relation)
561 : {
562 : /* Nothing to do if this is not a serializable transaction */
563 31831068 : if (MySerializableXact == InvalidSerializableXact)
564 31816894 : return false;
565 :
566 : /* Check if the relation doesn't participate in predicate locking */
567 14174 : if (!PredicateLockingNeededForRelation(relation))
568 5318 : return false;
569 :
570 8856 : return true; /* no excuse to skip predicate locking */
571 : }
572 :
573 :
574 : /*------------------------------------------------------------------------*/
575 :
576 : /*
577 : * These functions are a simple implementation of a list for this specific
578 : * type of struct. If there is ever a generalized shared memory list, we
579 : * should probably switch to that.
580 : */
581 : static SERIALIZABLEXACT *
582 5314 : CreatePredXact(void)
583 : {
584 : SERIALIZABLEXACT *sxact;
585 :
586 5314 : if (dlist_is_empty(&PredXact->availableList))
587 0 : return NULL;
588 :
589 5314 : sxact = dlist_container(SERIALIZABLEXACT, xactLink,
590 : dlist_pop_head_node(&PredXact->availableList));
591 5314 : dlist_push_tail(&PredXact->activeList, &sxact->xactLink);
592 5314 : return sxact;
593 : }
594 :
595 : static void
596 3282 : ReleasePredXact(SERIALIZABLEXACT *sxact)
597 : {
598 : Assert(ShmemAddrIsValid(sxact));
599 :
600 3282 : dlist_delete(&sxact->xactLink);
601 3282 : dlist_push_tail(&PredXact->availableList, &sxact->xactLink);
602 3282 : }
603 :
604 : /*------------------------------------------------------------------------*/
605 :
606 : /*
607 : * These functions manage primitive access to the RWConflict pool and lists.
608 : */
609 : static bool
610 3736 : RWConflictExists(const SERIALIZABLEXACT *reader, const SERIALIZABLEXACT *writer)
611 : {
612 : dlist_iter iter;
613 :
614 : Assert(reader != writer);
615 :
616 : /* Check the ends of the purported conflict first. */
617 3736 : if (SxactIsDoomed(reader)
618 3736 : || SxactIsDoomed(writer)
619 3736 : || dlist_is_empty(&reader->outConflicts)
620 1134 : || dlist_is_empty(&writer->inConflicts))
621 2682 : return false;
622 :
623 : /*
624 : * A conflict is possible; walk the list to find out.
625 : *
626 : * The unconstify is needed as we have no const version of
627 : * dlist_foreach().
628 : */
629 1086 : dlist_foreach(iter, &unconstify(SERIALIZABLEXACT *, reader)->outConflicts)
630 : {
631 1054 : RWConflict conflict =
632 1054 : dlist_container(RWConflictData, outLink, iter.cur);
633 :
634 1054 : if (conflict->sxactIn == writer)
635 1022 : return true;
636 : }
637 :
638 : /* No conflict found. */
639 32 : return false;
640 : }
641 :
642 : static void
643 1560 : SetRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
644 : {
645 : RWConflict conflict;
646 :
647 : Assert(reader != writer);
648 : Assert(!RWConflictExists(reader, writer));
649 :
650 1560 : if (dlist_is_empty(&RWConflictPool->availableList))
651 0 : ereport(ERROR,
652 : (errcode(ERRCODE_OUT_OF_MEMORY),
653 : errmsg("not enough elements in RWConflictPool to record a read/write conflict"),
654 : errhint("You might need to run fewer transactions at a time or increase \"max_connections\".")));
655 :
656 1560 : conflict = dlist_head_element(RWConflictData, outLink, &RWConflictPool->availableList);
657 1560 : dlist_delete(&conflict->outLink);
658 :
659 1560 : conflict->sxactOut = reader;
660 1560 : conflict->sxactIn = writer;
661 1560 : dlist_push_tail(&reader->outConflicts, &conflict->outLink);
662 1560 : dlist_push_tail(&writer->inConflicts, &conflict->inLink);
663 1560 : }
664 :
665 : static void
666 268 : SetPossibleUnsafeConflict(SERIALIZABLEXACT *roXact,
667 : SERIALIZABLEXACT *activeXact)
668 : {
669 : RWConflict conflict;
670 :
671 : Assert(roXact != activeXact);
672 : Assert(SxactIsReadOnly(roXact));
673 : Assert(!SxactIsReadOnly(activeXact));
674 :
675 268 : if (dlist_is_empty(&RWConflictPool->availableList))
676 0 : ereport(ERROR,
677 : (errcode(ERRCODE_OUT_OF_MEMORY),
678 : errmsg("not enough elements in RWConflictPool to record a potential read/write conflict"),
679 : errhint("You might need to run fewer transactions at a time or increase \"max_connections\".")));
680 :
681 268 : conflict = dlist_head_element(RWConflictData, outLink, &RWConflictPool->availableList);
682 268 : dlist_delete(&conflict->outLink);
683 :
684 268 : conflict->sxactOut = activeXact;
685 268 : conflict->sxactIn = roXact;
686 268 : dlist_push_tail(&activeXact->possibleUnsafeConflicts, &conflict->outLink);
687 268 : dlist_push_tail(&roXact->possibleUnsafeConflicts, &conflict->inLink);
688 268 : }
689 :
690 : static void
691 1828 : ReleaseRWConflict(RWConflict conflict)
692 : {
693 1828 : dlist_delete(&conflict->inLink);
694 1828 : dlist_delete(&conflict->outLink);
695 1828 : dlist_push_tail(&RWConflictPool->availableList, &conflict->outLink);
696 1828 : }
697 :
698 : static void
699 6 : FlagSxactUnsafe(SERIALIZABLEXACT *sxact)
700 : {
701 : dlist_mutable_iter iter;
702 :
703 : Assert(SxactIsReadOnly(sxact));
704 : Assert(!SxactIsROSafe(sxact));
705 :
706 6 : sxact->flags |= SXACT_FLAG_RO_UNSAFE;
707 :
708 : /*
709 : * We know this isn't a safe snapshot, so we can stop looking for other
710 : * potential conflicts.
711 : */
712 12 : dlist_foreach_modify(iter, &sxact->possibleUnsafeConflicts)
713 : {
714 6 : RWConflict conflict =
715 6 : dlist_container(RWConflictData, inLink, iter.cur);
716 :
717 : Assert(!SxactIsReadOnly(conflict->sxactOut));
718 : Assert(sxact == conflict->sxactIn);
719 :
720 6 : ReleaseRWConflict(conflict);
721 : }
722 6 : }
723 :
724 : /*------------------------------------------------------------------------*/
725 :
726 : /*
727 : * Decide whether a Serial page number is "older" for truncation purposes.
728 : * Analogous to CLOGPagePrecedes().
729 : */
730 : static bool
731 0 : SerialPagePrecedesLogically(int64 page1, int64 page2)
732 : {
733 : TransactionId xid1;
734 : TransactionId xid2;
735 :
736 0 : xid1 = ((TransactionId) page1) * SERIAL_ENTRIESPERPAGE;
737 0 : xid1 += FirstNormalTransactionId + 1;
738 0 : xid2 = ((TransactionId) page2) * SERIAL_ENTRIESPERPAGE;
739 0 : xid2 += FirstNormalTransactionId + 1;
740 :
741 0 : return (TransactionIdPrecedes(xid1, xid2) &&
742 0 : TransactionIdPrecedes(xid1, xid2 + SERIAL_ENTRIESPERPAGE - 1));
743 : }
744 :
745 : #ifdef USE_ASSERT_CHECKING
746 : static void
747 : SerialPagePrecedesLogicallyUnitTests(void)
748 : {
749 : int per_page = SERIAL_ENTRIESPERPAGE,
750 : offset = per_page / 2;
751 : int64 newestPage,
752 : oldestPage,
753 : headPage,
754 : targetPage;
755 : TransactionId newestXact,
756 : oldestXact;
757 :
758 : /* GetNewTransactionId() has assigned the last XID it can safely use. */
759 : newestPage = 2 * SLRU_PAGES_PER_SEGMENT - 1; /* nothing special */
760 : newestXact = newestPage * per_page + offset;
761 : Assert(newestXact / per_page == newestPage);
762 : oldestXact = newestXact + 1;
763 : oldestXact -= 1U << 31;
764 : oldestPage = oldestXact / per_page;
765 :
766 : /*
767 : * In this scenario, the SLRU headPage pertains to the last ~1000 XIDs
768 : * assigned. oldestXact finishes, ~2B XIDs having elapsed since it
769 : * started. Further transactions cause us to summarize oldestXact to
770 : * tailPage. Function must return false so SerialAdd() doesn't zero
771 : * tailPage (which may contain entries for other old, recently-finished
772 : * XIDs) and half the SLRU. Reaching this requires burning ~2B XIDs in
773 : * single-user mode, a negligible possibility.
774 : */
775 : headPage = newestPage;
776 : targetPage = oldestPage;
777 : Assert(!SerialPagePrecedesLogically(headPage, targetPage));
778 :
779 : /*
780 : * In this scenario, the SLRU headPage pertains to oldestXact. We're
781 : * summarizing an XID near newestXact. (Assume few other XIDs used
782 : * SERIALIZABLE, hence the minimal headPage advancement. Assume
783 : * oldestXact was long-running and only recently reached the SLRU.)
784 : * Function must return true to make SerialAdd() create targetPage.
785 : *
786 : * Today's implementation mishandles this case, but it doesn't matter
787 : * enough to fix. Verify that the defect affects just one page by
788 : * asserting correct treatment of its prior page. Reaching this case
789 : * requires burning ~2B XIDs in single-user mode, a negligible
790 : * possibility. Moreover, if it does happen, the consequence would be
791 : * mild, namely a new transaction failing in SimpleLruReadPage().
792 : */
793 : headPage = oldestPage;
794 : targetPage = newestPage;
795 : Assert(SerialPagePrecedesLogically(headPage, targetPage - 1));
796 : #if 0
797 : Assert(SerialPagePrecedesLogically(headPage, targetPage));
798 : #endif
799 : }
800 : #endif
801 :
802 : /*
803 : * Initialize for the tracking of old serializable committed xids.
804 : */
805 : static void
806 2032 : SerialInit(void)
807 : {
808 : bool found;
809 :
810 : /*
811 : * Set up SLRU management of the pg_serial data.
812 : */
813 2032 : SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
814 2032 : SimpleLruInit(SerialSlruCtl, "serializable",
815 : serializable_buffers, 0, "pg_serial",
816 : LWTRANCHE_SERIAL_BUFFER, LWTRANCHE_SERIAL_SLRU,
817 : SYNC_HANDLER_NONE, false);
818 : #ifdef USE_ASSERT_CHECKING
819 : SerialPagePrecedesLogicallyUnitTests();
820 : #endif
821 : SlruPagePrecedesUnitTests(SerialSlruCtl, SERIAL_ENTRIESPERPAGE);
822 :
823 : /*
824 : * Create or attach to the SerialControl structure.
825 : */
826 2032 : serialControl = (SerialControl)
827 2032 : ShmemInitStruct("SerialControlData", sizeof(SerialControlData), &found);
828 :
829 : Assert(found == IsUnderPostmaster);
830 2032 : if (!found)
831 : {
832 : /*
833 : * Set control information to reflect empty SLRU.
834 : */
835 2032 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
836 2032 : serialControl->headPage = -1;
837 2032 : serialControl->headXid = InvalidTransactionId;
838 2032 : serialControl->tailXid = InvalidTransactionId;
839 2032 : LWLockRelease(SerialControlLock);
840 : }
841 2032 : }
842 :
843 : /*
844 : * GUC check_hook for serializable_buffers
845 : */
846 : bool
847 2098 : check_serial_buffers(int *newval, void **extra, GucSource source)
848 : {
849 2098 : return check_slru_buffers("serializable_buffers", newval);
850 : }
851 :
852 : /*
853 : * Record a committed read write serializable xid and the minimum
854 : * commitSeqNo of any transactions to which this xid had a rw-conflict out.
855 : * An invalid commitSeqNo means that there were no conflicts out from xid.
856 : */
857 : static void
858 0 : SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
859 : {
860 : TransactionId tailXid;
861 : int64 targetPage;
862 : int slotno;
863 : int64 firstZeroPage;
864 : bool isNewPage;
865 : LWLock *lock;
866 :
867 : Assert(TransactionIdIsValid(xid));
868 :
869 0 : targetPage = SerialPage(xid);
870 0 : lock = SimpleLruGetBankLock(SerialSlruCtl, targetPage);
871 :
872 : /*
873 : * In this routine, we must hold both SerialControlLock and the SLRU bank
874 : * lock simultaneously while making the SLRU data catch up with the new
875 : * state that we determine.
876 : */
877 0 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
878 :
879 : /*
880 : * If 'xid' is older than the global xmin (== tailXid), there's no need to
881 : * store it, after all. This can happen if the oldest transaction holding
882 : * back the global xmin just finished, making 'xid' uninteresting, but
883 : * ClearOldPredicateLocks() has not yet run.
884 : */
885 0 : tailXid = serialControl->tailXid;
886 0 : if (!TransactionIdIsValid(tailXid) || TransactionIdPrecedes(xid, tailXid))
887 : {
888 0 : LWLockRelease(SerialControlLock);
889 0 : return;
890 : }
891 :
892 : /*
893 : * If the SLRU is currently unused, zero out the whole active region from
894 : * tailXid to headXid before taking it into use. Otherwise zero out only
895 : * any new pages that enter the tailXid-headXid range as we advance
896 : * headXid.
897 : */
898 0 : if (serialControl->headPage < 0)
899 : {
900 0 : firstZeroPage = SerialPage(tailXid);
901 0 : isNewPage = true;
902 : }
903 : else
904 : {
905 0 : firstZeroPage = SerialNextPage(serialControl->headPage);
906 0 : isNewPage = SerialPagePrecedesLogically(serialControl->headPage,
907 : targetPage);
908 : }
909 :
910 0 : if (!TransactionIdIsValid(serialControl->headXid)
911 0 : || TransactionIdFollows(xid, serialControl->headXid))
912 0 : serialControl->headXid = xid;
913 0 : if (isNewPage)
914 0 : serialControl->headPage = targetPage;
915 :
916 0 : if (isNewPage)
917 : {
918 : /* Initialize intervening pages; might involve trading locks */
919 : for (;;)
920 : {
921 0 : lock = SimpleLruGetBankLock(SerialSlruCtl, firstZeroPage);
922 0 : LWLockAcquire(lock, LW_EXCLUSIVE);
923 0 : slotno = SimpleLruZeroPage(SerialSlruCtl, firstZeroPage);
924 0 : if (firstZeroPage == targetPage)
925 0 : break;
926 0 : firstZeroPage = SerialNextPage(firstZeroPage);
927 0 : LWLockRelease(lock);
928 : }
929 : }
930 : else
931 : {
932 0 : LWLockAcquire(lock, LW_EXCLUSIVE);
933 0 : slotno = SimpleLruReadPage(SerialSlruCtl, targetPage, true, xid);
934 : }
935 :
936 0 : SerialValue(slotno, xid) = minConflictCommitSeqNo;
937 0 : SerialSlruCtl->shared->page_dirty[slotno] = true;
938 :
939 0 : LWLockRelease(lock);
940 0 : LWLockRelease(SerialControlLock);
941 : }
942 :
943 : /*
944 : * Get the minimum commitSeqNo for any conflict out for the given xid. For
945 : * a transaction which exists but has no conflict out, InvalidSerCommitSeqNo
946 : * will be returned.
947 : */
948 : static SerCommitSeqNo
949 50 : SerialGetMinConflictCommitSeqNo(TransactionId xid)
950 : {
951 : TransactionId headXid;
952 : TransactionId tailXid;
953 : SerCommitSeqNo val;
954 : int slotno;
955 :
956 : Assert(TransactionIdIsValid(xid));
957 :
958 50 : LWLockAcquire(SerialControlLock, LW_SHARED);
959 50 : headXid = serialControl->headXid;
960 50 : tailXid = serialControl->tailXid;
961 50 : LWLockRelease(SerialControlLock);
962 :
963 50 : if (!TransactionIdIsValid(headXid))
964 50 : return 0;
965 :
966 : Assert(TransactionIdIsValid(tailXid));
967 :
968 0 : if (TransactionIdPrecedes(xid, tailXid)
969 0 : || TransactionIdFollows(xid, headXid))
970 0 : return 0;
971 :
972 : /*
973 : * The following function must be called without holding SLRU bank lock,
974 : * but will return with that lock held, which must then be released.
975 : */
976 0 : slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
977 : SerialPage(xid), xid);
978 0 : val = SerialValue(slotno, xid);
979 0 : LWLockRelease(SimpleLruGetBankLock(SerialSlruCtl, SerialPage(xid)));
980 0 : return val;
981 : }
982 :
983 : /*
984 : * Call this whenever there is a new xmin for active serializable
985 : * transactions. We don't need to keep information on transactions which
986 : * precede that. InvalidTransactionId means none active, so everything in
987 : * the SLRU can be discarded.
988 : */
989 : static void
990 3340 : SerialSetActiveSerXmin(TransactionId xid)
991 : {
992 3340 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
993 :
994 : /*
995 : * When no sxacts are active, nothing overlaps, set the xid values to
996 : * invalid to show that there are no valid entries. Don't clear headPage,
997 : * though. A new xmin might still land on that page, and we don't want to
998 : * repeatedly zero out the same page.
999 : */
1000 3340 : if (!TransactionIdIsValid(xid))
1001 : {
1002 1652 : serialControl->tailXid = InvalidTransactionId;
1003 1652 : serialControl->headXid = InvalidTransactionId;
1004 1652 : LWLockRelease(SerialControlLock);
1005 1652 : return;
1006 : }
1007 :
1008 : /*
1009 : * When we're recovering prepared transactions, the global xmin might move
1010 : * backwards depending on the order they're recovered. Normally that's not
1011 : * OK, but during recovery no serializable transactions will commit, so
1012 : * the SLRU is empty and we can get away with it.
1013 : */
1014 1688 : if (RecoveryInProgress())
1015 : {
1016 : Assert(serialControl->headPage < 0);
1017 0 : if (!TransactionIdIsValid(serialControl->tailXid)
1018 0 : || TransactionIdPrecedes(xid, serialControl->tailXid))
1019 : {
1020 0 : serialControl->tailXid = xid;
1021 : }
1022 0 : LWLockRelease(SerialControlLock);
1023 0 : return;
1024 : }
1025 :
1026 : Assert(!TransactionIdIsValid(serialControl->tailXid)
1027 : || TransactionIdFollows(xid, serialControl->tailXid));
1028 :
1029 1688 : serialControl->tailXid = xid;
1030 :
1031 1688 : LWLockRelease(SerialControlLock);
1032 : }
1033 :
1034 : /*
1035 : * Perform a checkpoint --- either during shutdown, or on-the-fly
1036 : *
1037 : * We don't have any data that needs to survive a restart, but this is a
1038 : * convenient place to truncate the SLRU.
1039 : */
1040 : void
1041 2602 : CheckPointPredicate(void)
1042 : {
1043 : int64 truncateCutoffPage;
1044 :
1045 2602 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
1046 :
1047 : /* Exit quickly if the SLRU is currently not in use. */
1048 2602 : if (serialControl->headPage < 0)
1049 : {
1050 2602 : LWLockRelease(SerialControlLock);
1051 2602 : return;
1052 : }
1053 :
1054 0 : if (TransactionIdIsValid(serialControl->tailXid))
1055 : {
1056 : int64 tailPage;
1057 :
1058 0 : tailPage = SerialPage(serialControl->tailXid);
1059 :
1060 : /*
1061 : * It is possible for the tailXid to be ahead of the headXid. This
1062 : * occurs if we checkpoint while there are in-progress serializable
1063 : * transaction(s) advancing the tail but we are yet to summarize the
1064 : * transactions. In this case, we cutoff up to the headPage and the
1065 : * next summary will advance the headXid.
1066 : */
1067 0 : if (SerialPagePrecedesLogically(tailPage, serialControl->headPage))
1068 : {
1069 : /* We can truncate the SLRU up to the page containing tailXid */
1070 0 : truncateCutoffPage = tailPage;
1071 : }
1072 : else
1073 0 : truncateCutoffPage = serialControl->headPage;
1074 : }
1075 : else
1076 : {
1077 : /*----------
1078 : * The SLRU is no longer needed. Truncate to head before we set head
1079 : * invalid.
1080 : *
1081 : * XXX: It's possible that the SLRU is not needed again until XID
1082 : * wrap-around has happened, so that the segment containing headPage
1083 : * that we leave behind will appear to be new again. In that case it
1084 : * won't be removed until XID horizon advances enough to make it
1085 : * current again.
1086 : *
1087 : * XXX: This should happen in vac_truncate_clog(), not in checkpoints.
1088 : * Consider this scenario, starting from a system with no in-progress
1089 : * transactions and VACUUM FREEZE having maximized oldestXact:
1090 : * - Start a SERIALIZABLE transaction.
1091 : * - Start, finish, and summarize a SERIALIZABLE transaction, creating
1092 : * one SLRU page.
1093 : * - Consume XIDs to reach xidStopLimit.
1094 : * - Finish all transactions. Due to the long-running SERIALIZABLE
1095 : * transaction, earlier checkpoints did not touch headPage. The
1096 : * next checkpoint will change it, but that checkpoint happens after
1097 : * the end of the scenario.
1098 : * - VACUUM to advance XID limits.
1099 : * - Consume ~2M XIDs, crossing the former xidWrapLimit.
1100 : * - Start, finish, and summarize a SERIALIZABLE transaction.
1101 : * SerialAdd() declines to create the targetPage, because headPage
1102 : * is not regarded as in the past relative to that targetPage. The
1103 : * transaction instigating the summarize fails in
1104 : * SimpleLruReadPage().
1105 : */
1106 0 : truncateCutoffPage = serialControl->headPage;
1107 0 : serialControl->headPage = -1;
1108 : }
1109 :
1110 0 : LWLockRelease(SerialControlLock);
1111 :
1112 : /*
1113 : * Truncate away pages that are no longer required. Note that no
1114 : * additional locking is required, because this is only called as part of
1115 : * a checkpoint, and the validity limits have already been determined.
1116 : */
1117 0 : SimpleLruTruncate(SerialSlruCtl, truncateCutoffPage);
1118 :
1119 : /*
1120 : * Write dirty SLRU pages to disk
1121 : *
1122 : * This is not actually necessary from a correctness point of view. We do
1123 : * it merely as a debugging aid.
1124 : *
1125 : * We're doing this after the truncation to avoid writing pages right
1126 : * before deleting the file in which they sit, which would be completely
1127 : * pointless.
1128 : */
1129 0 : SimpleLruWriteAll(SerialSlruCtl, true);
1130 : }
1131 :
1132 : /*------------------------------------------------------------------------*/
1133 :
1134 : /*
1135 : * PredicateLockShmemInit -- Initialize the predicate locking data structures.
1136 : *
1137 : * This is called from CreateSharedMemoryAndSemaphores(), which see for
1138 : * more comments. In the normal postmaster case, the shared hash tables
1139 : * are created here. Backends inherit the pointers
1140 : * to the shared tables via fork(). In the EXEC_BACKEND case, each
1141 : * backend re-executes this code to obtain pointers to the already existing
1142 : * shared hash tables.
1143 : */
1144 : void
1145 2032 : PredicateLockShmemInit(void)
1146 : {
1147 : HASHCTL info;
1148 : long max_table_size;
1149 : Size requestSize;
1150 : bool found;
1151 :
1152 : #ifndef EXEC_BACKEND
1153 : Assert(!IsUnderPostmaster);
1154 : #endif
1155 :
1156 : /*
1157 : * Compute size of predicate lock target hashtable. Note these
1158 : * calculations must agree with PredicateLockShmemSize!
1159 : */
1160 2032 : max_table_size = NPREDICATELOCKTARGETENTS();
1161 :
1162 : /*
1163 : * Allocate hash table for PREDICATELOCKTARGET structs. This stores
1164 : * per-predicate-lock-target information.
1165 : */
1166 2032 : info.keysize = sizeof(PREDICATELOCKTARGETTAG);
1167 2032 : info.entrysize = sizeof(PREDICATELOCKTARGET);
1168 2032 : info.num_partitions = NUM_PREDICATELOCK_PARTITIONS;
1169 :
1170 2032 : PredicateLockTargetHash = ShmemInitHash("PREDICATELOCKTARGET hash",
1171 : max_table_size,
1172 : max_table_size,
1173 : &info,
1174 : HASH_ELEM | HASH_BLOBS |
1175 : HASH_PARTITION | HASH_FIXED_SIZE);
1176 :
1177 : /*
1178 : * Reserve a dummy entry in the hash table; we use it to make sure there's
1179 : * always one entry available when we need to split or combine a page,
1180 : * because running out of space there could mean aborting a
1181 : * non-serializable transaction.
1182 : */
1183 2032 : if (!IsUnderPostmaster)
1184 : {
1185 2032 : (void) hash_search(PredicateLockTargetHash, &ScratchTargetTag,
1186 : HASH_ENTER, &found);
1187 : Assert(!found);
1188 : }
1189 :
1190 : /* Pre-calculate the hash and partition lock of the scratch entry */
1191 2032 : ScratchTargetTagHash = PredicateLockTargetTagHashCode(&ScratchTargetTag);
1192 2032 : ScratchPartitionLock = PredicateLockHashPartitionLock(ScratchTargetTagHash);
1193 :
1194 : /*
1195 : * Allocate hash table for PREDICATELOCK structs. This stores per
1196 : * xact-lock-of-a-target information.
1197 : */
1198 2032 : info.keysize = sizeof(PREDICATELOCKTAG);
1199 2032 : info.entrysize = sizeof(PREDICATELOCK);
1200 2032 : info.hash = predicatelock_hash;
1201 2032 : info.num_partitions = NUM_PREDICATELOCK_PARTITIONS;
1202 :
1203 : /* Assume an average of 2 xacts per target */
1204 2032 : max_table_size *= 2;
1205 :
1206 2032 : PredicateLockHash = ShmemInitHash("PREDICATELOCK hash",
1207 : max_table_size,
1208 : max_table_size,
1209 : &info,
1210 : HASH_ELEM | HASH_FUNCTION |
1211 : HASH_PARTITION | HASH_FIXED_SIZE);
1212 :
1213 : /*
1214 : * Compute size for serializable transaction hashtable. Note these
1215 : * calculations must agree with PredicateLockShmemSize!
1216 : */
1217 2032 : max_table_size = (MaxBackends + max_prepared_xacts);
1218 :
1219 : /*
1220 : * Allocate a list to hold information on transactions participating in
1221 : * predicate locking.
1222 : *
1223 : * Assume an average of 10 predicate locking transactions per backend.
1224 : * This allows aggressive cleanup while detail is present before data must
1225 : * be summarized for storage in SLRU and the "dummy" transaction.
1226 : */
1227 2032 : max_table_size *= 10;
1228 :
1229 2032 : PredXact = ShmemInitStruct("PredXactList",
1230 : PredXactListDataSize,
1231 : &found);
1232 : Assert(found == IsUnderPostmaster);
1233 2032 : if (!found)
1234 : {
1235 : int i;
1236 :
1237 2032 : dlist_init(&PredXact->availableList);
1238 2032 : dlist_init(&PredXact->activeList);
1239 2032 : PredXact->SxactGlobalXmin = InvalidTransactionId;
1240 2032 : PredXact->SxactGlobalXminCount = 0;
1241 2032 : PredXact->WritableSxactCount = 0;
1242 2032 : PredXact->LastSxactCommitSeqNo = FirstNormalSerCommitSeqNo - 1;
1243 2032 : PredXact->CanPartialClearThrough = 0;
1244 2032 : PredXact->HavePartialClearedThrough = 0;
1245 2032 : requestSize = mul_size((Size) max_table_size,
1246 : sizeof(SERIALIZABLEXACT));
1247 2032 : PredXact->element = ShmemAlloc(requestSize);
1248 : /* Add all elements to available list, clean. */
1249 2032 : memset(PredXact->element, 0, requestSize);
1250 1910972 : for (i = 0; i < max_table_size; i++)
1251 : {
1252 1908940 : LWLockInitialize(&PredXact->element[i].perXactPredicateListLock,
1253 : LWTRANCHE_PER_XACT_PREDICATE_LIST);
1254 1908940 : dlist_push_tail(&PredXact->availableList, &PredXact->element[i].xactLink);
1255 : }
1256 2032 : PredXact->OldCommittedSxact = CreatePredXact();
1257 2032 : SetInvalidVirtualTransactionId(PredXact->OldCommittedSxact->vxid);
1258 2032 : PredXact->OldCommittedSxact->prepareSeqNo = 0;
1259 2032 : PredXact->OldCommittedSxact->commitSeqNo = 0;
1260 2032 : PredXact->OldCommittedSxact->SeqNo.lastCommitBeforeSnapshot = 0;
1261 2032 : dlist_init(&PredXact->OldCommittedSxact->outConflicts);
1262 2032 : dlist_init(&PredXact->OldCommittedSxact->inConflicts);
1263 2032 : dlist_init(&PredXact->OldCommittedSxact->predicateLocks);
1264 2032 : dlist_node_init(&PredXact->OldCommittedSxact->finishedLink);
1265 2032 : dlist_init(&PredXact->OldCommittedSxact->possibleUnsafeConflicts);
1266 2032 : PredXact->OldCommittedSxact->topXid = InvalidTransactionId;
1267 2032 : PredXact->OldCommittedSxact->finishedBefore = InvalidTransactionId;
1268 2032 : PredXact->OldCommittedSxact->xmin = InvalidTransactionId;
1269 2032 : PredXact->OldCommittedSxact->flags = SXACT_FLAG_COMMITTED;
1270 2032 : PredXact->OldCommittedSxact->pid = 0;
1271 2032 : PredXact->OldCommittedSxact->pgprocno = INVALID_PROC_NUMBER;
1272 : }
1273 : /* This never changes, so let's keep a local copy. */
1274 2032 : OldCommittedSxact = PredXact->OldCommittedSxact;
1275 :
1276 : /*
1277 : * Allocate hash table for SERIALIZABLEXID structs. This stores per-xid
1278 : * information for serializable transactions which have accessed data.
1279 : */
1280 2032 : info.keysize = sizeof(SERIALIZABLEXIDTAG);
1281 2032 : info.entrysize = sizeof(SERIALIZABLEXID);
1282 :
1283 2032 : SerializableXidHash = ShmemInitHash("SERIALIZABLEXID hash",
1284 : max_table_size,
1285 : max_table_size,
1286 : &info,
1287 : HASH_ELEM | HASH_BLOBS |
1288 : HASH_FIXED_SIZE);
1289 :
1290 : /*
1291 : * Allocate space for tracking rw-conflicts in lists attached to the
1292 : * transactions.
1293 : *
1294 : * Assume an average of 5 conflicts per transaction. Calculations suggest
1295 : * that this will prevent resource exhaustion in even the most pessimal
1296 : * loads up to max_connections = 200 with all 200 connections pounding the
1297 : * database with serializable transactions. Beyond that, there may be
1298 : * occasional transactions canceled when trying to flag conflicts. That's
1299 : * probably OK.
1300 : */
1301 2032 : max_table_size *= 5;
1302 :
1303 2032 : RWConflictPool = ShmemInitStruct("RWConflictPool",
1304 : RWConflictPoolHeaderDataSize,
1305 : &found);
1306 : Assert(found == IsUnderPostmaster);
1307 2032 : if (!found)
1308 : {
1309 : int i;
1310 :
1311 2032 : dlist_init(&RWConflictPool->availableList);
1312 2032 : requestSize = mul_size((Size) max_table_size,
1313 : RWConflictDataSize);
1314 2032 : RWConflictPool->element = ShmemAlloc(requestSize);
1315 : /* Add all elements to available list, clean. */
1316 2032 : memset(RWConflictPool->element, 0, requestSize);
1317 9546732 : for (i = 0; i < max_table_size; i++)
1318 : {
1319 9544700 : dlist_push_tail(&RWConflictPool->availableList,
1320 9544700 : &RWConflictPool->element[i].outLink);
1321 : }
1322 : }
1323 :
1324 : /*
1325 : * Create or attach to the header for the list of finished serializable
1326 : * transactions.
1327 : */
1328 2032 : FinishedSerializableTransactions = (dlist_head *)
1329 2032 : ShmemInitStruct("FinishedSerializableTransactions",
1330 : sizeof(dlist_head),
1331 : &found);
1332 : Assert(found == IsUnderPostmaster);
1333 2032 : if (!found)
1334 2032 : dlist_init(FinishedSerializableTransactions);
1335 :
1336 : /*
1337 : * Initialize the SLRU storage for old committed serializable
1338 : * transactions.
1339 : */
1340 2032 : SerialInit();
1341 2032 : }
1342 :
1343 : /*
1344 : * Estimate shared-memory space used for predicate lock table
1345 : */
1346 : Size
1347 3794 : PredicateLockShmemSize(void)
1348 : {
1349 3794 : Size size = 0;
1350 : long max_table_size;
1351 :
1352 : /* predicate lock target hash table */
1353 3794 : max_table_size = NPREDICATELOCKTARGETENTS();
1354 3794 : size = add_size(size, hash_estimate_size(max_table_size,
1355 : sizeof(PREDICATELOCKTARGET)));
1356 :
1357 : /* predicate lock hash table */
1358 3794 : max_table_size *= 2;
1359 3794 : size = add_size(size, hash_estimate_size(max_table_size,
1360 : sizeof(PREDICATELOCK)));
1361 :
1362 : /*
1363 : * Since NPREDICATELOCKTARGETENTS is only an estimate, add 10% safety
1364 : * margin.
1365 : */
1366 3794 : size = add_size(size, size / 10);
1367 :
1368 : /* transaction list */
1369 3794 : max_table_size = MaxBackends + max_prepared_xacts;
1370 3794 : max_table_size *= 10;
1371 3794 : size = add_size(size, PredXactListDataSize);
1372 3794 : size = add_size(size, mul_size((Size) max_table_size,
1373 : sizeof(SERIALIZABLEXACT)));
1374 :
1375 : /* transaction xid table */
1376 3794 : size = add_size(size, hash_estimate_size(max_table_size,
1377 : sizeof(SERIALIZABLEXID)));
1378 :
1379 : /* rw-conflict pool */
1380 3794 : max_table_size *= 5;
1381 3794 : size = add_size(size, RWConflictPoolHeaderDataSize);
1382 3794 : size = add_size(size, mul_size((Size) max_table_size,
1383 : RWConflictDataSize));
1384 :
1385 : /* Head for list of finished serializable transactions. */
1386 3794 : size = add_size(size, sizeof(dlist_head));
1387 :
1388 : /* Shared memory structures for SLRU tracking of old committed xids. */
1389 3794 : size = add_size(size, sizeof(SerialControlData));
1390 3794 : size = add_size(size, SimpleLruShmemSize(serializable_buffers, 0));
1391 :
1392 3794 : return size;
1393 : }
1394 :
1395 :
1396 : /*
1397 : * Compute the hash code associated with a PREDICATELOCKTAG.
1398 : *
1399 : * Because we want to use just one set of partition locks for both the
1400 : * PREDICATELOCKTARGET and PREDICATELOCK hash tables, we have to make sure
1401 : * that PREDICATELOCKs fall into the same partition number as their
1402 : * associated PREDICATELOCKTARGETs. dynahash.c expects the partition number
1403 : * to be the low-order bits of the hash code, and therefore a
1404 : * PREDICATELOCKTAG's hash code must have the same low-order bits as the
1405 : * associated PREDICATELOCKTARGETTAG's hash code. We achieve this with this
1406 : * specialized hash function.
1407 : */
1408 : static uint32
1409 0 : predicatelock_hash(const void *key, Size keysize)
1410 : {
1411 0 : const PREDICATELOCKTAG *predicatelocktag = (const PREDICATELOCKTAG *) key;
1412 : uint32 targethash;
1413 :
1414 : Assert(keysize == sizeof(PREDICATELOCKTAG));
1415 :
1416 : /* Look into the associated target object, and compute its hash code */
1417 0 : targethash = PredicateLockTargetTagHashCode(&predicatelocktag->myTarget->tag);
1418 :
1419 0 : return PredicateLockHashCodeFromTargetHashCode(predicatelocktag, targethash);
1420 : }
1421 :
1422 :
1423 : /*
1424 : * GetPredicateLockStatusData
1425 : * Return a table containing the internal state of the predicate
1426 : * lock manager for use in pg_lock_status.
1427 : *
1428 : * Like GetLockStatusData, this function tries to hold the partition LWLocks
1429 : * for as short a time as possible by returning two arrays that simply
1430 : * contain the PREDICATELOCKTARGETTAG and SERIALIZABLEXACT for each lock
1431 : * table entry. Multiple copies of the same PREDICATELOCKTARGETTAG and
1432 : * SERIALIZABLEXACT will likely appear.
1433 : */
1434 : PredicateLockData *
1435 516 : GetPredicateLockStatusData(void)
1436 : {
1437 : PredicateLockData *data;
1438 : int i;
1439 : int els,
1440 : el;
1441 : HASH_SEQ_STATUS seqstat;
1442 : PREDICATELOCK *predlock;
1443 :
1444 516 : data = (PredicateLockData *) palloc(sizeof(PredicateLockData));
1445 :
1446 : /*
1447 : * To ensure consistency, take simultaneous locks on all partition locks
1448 : * in ascending order, then SerializableXactHashLock.
1449 : */
1450 8772 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
1451 8256 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_SHARED);
1452 516 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1453 :
1454 : /* Get number of locks and allocate appropriately-sized arrays. */
1455 516 : els = hash_get_num_entries(PredicateLockHash);
1456 516 : data->nelements = els;
1457 516 : data->locktags = (PREDICATELOCKTARGETTAG *)
1458 516 : palloc(sizeof(PREDICATELOCKTARGETTAG) * els);
1459 516 : data->xacts = (SERIALIZABLEXACT *)
1460 516 : palloc(sizeof(SERIALIZABLEXACT) * els);
1461 :
1462 :
1463 : /* Scan through PredicateLockHash and copy contents */
1464 516 : hash_seq_init(&seqstat, PredicateLockHash);
1465 :
1466 516 : el = 0;
1467 :
1468 522 : while ((predlock = (PREDICATELOCK *) hash_seq_search(&seqstat)))
1469 : {
1470 6 : data->locktags[el] = predlock->tag.myTarget->tag;
1471 6 : data->xacts[el] = *predlock->tag.myXact;
1472 6 : el++;
1473 : }
1474 :
1475 : Assert(el == els);
1476 :
1477 : /* Release locks in reverse order */
1478 516 : LWLockRelease(SerializableXactHashLock);
1479 8772 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
1480 8256 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
1481 :
1482 516 : return data;
1483 : }
1484 :
1485 : /*
1486 : * Free up shared memory structures by pushing the oldest sxact (the one at
1487 : * the front of the SummarizeOldestCommittedSxact queue) into summary form.
1488 : * Each call will free exactly one SERIALIZABLEXACT structure and may also
1489 : * free one or more of these structures: SERIALIZABLEXID, PREDICATELOCK,
1490 : * PREDICATELOCKTARGET, RWConflictData.
1491 : */
1492 : static void
1493 0 : SummarizeOldestCommittedSxact(void)
1494 : {
1495 : SERIALIZABLEXACT *sxact;
1496 :
1497 0 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
1498 :
1499 : /*
1500 : * This function is only called if there are no sxact slots available.
1501 : * Some of them must belong to old, already-finished transactions, so
1502 : * there should be something in FinishedSerializableTransactions list that
1503 : * we can summarize. However, there's a race condition: while we were not
1504 : * holding any locks, a transaction might have ended and cleaned up all
1505 : * the finished sxact entries already, freeing up their sxact slots. In
1506 : * that case, we have nothing to do here. The caller will find one of the
1507 : * slots released by the other backend when it retries.
1508 : */
1509 0 : if (dlist_is_empty(FinishedSerializableTransactions))
1510 : {
1511 0 : LWLockRelease(SerializableFinishedListLock);
1512 0 : return;
1513 : }
1514 :
1515 : /*
1516 : * Grab the first sxact off the finished list -- this will be the earliest
1517 : * commit. Remove it from the list.
1518 : */
1519 0 : sxact = dlist_head_element(SERIALIZABLEXACT, finishedLink,
1520 : FinishedSerializableTransactions);
1521 0 : dlist_delete_thoroughly(&sxact->finishedLink);
1522 :
1523 : /* Add to SLRU summary information. */
1524 0 : if (TransactionIdIsValid(sxact->topXid) && !SxactIsReadOnly(sxact))
1525 0 : SerialAdd(sxact->topXid, SxactHasConflictOut(sxact)
1526 : ? sxact->SeqNo.earliestOutConflictCommit : InvalidSerCommitSeqNo);
1527 :
1528 : /* Summarize and release the detail. */
1529 0 : ReleaseOneSerializableXact(sxact, false, true);
1530 :
1531 0 : LWLockRelease(SerializableFinishedListLock);
1532 : }
1533 :
1534 : /*
1535 : * GetSafeSnapshot
1536 : * Obtain and register a snapshot for a READ ONLY DEFERRABLE
1537 : * transaction. Ensures that the snapshot is "safe", i.e. a
1538 : * read-only transaction running on it can execute serializably
1539 : * without further checks. This requires waiting for concurrent
1540 : * transactions to complete, and retrying with a new snapshot if
1541 : * one of them could possibly create a conflict.
1542 : *
1543 : * As with GetSerializableTransactionSnapshot (which this is a subroutine
1544 : * for), the passed-in Snapshot pointer should reference a static data
1545 : * area that can safely be passed to GetSnapshotData.
1546 : */
1547 : static Snapshot
1548 10 : GetSafeSnapshot(Snapshot origSnapshot)
1549 : {
1550 : Snapshot snapshot;
1551 :
1552 : Assert(XactReadOnly && XactDeferrable);
1553 :
1554 : while (true)
1555 : {
1556 : /*
1557 : * GetSerializableTransactionSnapshotInt is going to call
1558 : * GetSnapshotData, so we need to provide it the static snapshot area
1559 : * our caller passed to us. The pointer returned is actually the same
1560 : * one passed to it, but we avoid assuming that here.
1561 : */
1562 10 : snapshot = GetSerializableTransactionSnapshotInt(origSnapshot,
1563 : NULL, InvalidPid);
1564 :
1565 10 : if (MySerializableXact == InvalidSerializableXact)
1566 6 : return snapshot; /* no concurrent r/w xacts; it's safe */
1567 :
1568 4 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1569 :
1570 : /*
1571 : * Wait for concurrent transactions to finish. Stop early if one of
1572 : * them marked us as conflicted.
1573 : */
1574 4 : MySerializableXact->flags |= SXACT_FLAG_DEFERRABLE_WAITING;
1575 10 : while (!(dlist_is_empty(&MySerializableXact->possibleUnsafeConflicts) ||
1576 6 : SxactIsROUnsafe(MySerializableXact)))
1577 : {
1578 6 : LWLockRelease(SerializableXactHashLock);
1579 6 : ProcWaitForSignal(WAIT_EVENT_SAFE_SNAPSHOT);
1580 6 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1581 : }
1582 4 : MySerializableXact->flags &= ~SXACT_FLAG_DEFERRABLE_WAITING;
1583 :
1584 4 : if (!SxactIsROUnsafe(MySerializableXact))
1585 : {
1586 2 : LWLockRelease(SerializableXactHashLock);
1587 2 : break; /* success */
1588 : }
1589 :
1590 2 : LWLockRelease(SerializableXactHashLock);
1591 :
1592 : /* else, need to retry... */
1593 2 : ereport(DEBUG2,
1594 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
1595 : errmsg_internal("deferrable snapshot was unsafe; trying a new one")));
1596 2 : ReleasePredicateLocks(false, false);
1597 : }
1598 :
1599 : /*
1600 : * Now we have a safe snapshot, so we don't need to do any further checks.
1601 : */
1602 : Assert(SxactIsROSafe(MySerializableXact));
1603 2 : ReleasePredicateLocks(false, true);
1604 :
1605 2 : return snapshot;
1606 : }
1607 :
1608 : /*
1609 : * GetSafeSnapshotBlockingPids
1610 : * If the specified process is currently blocked in GetSafeSnapshot,
1611 : * write the process IDs of all processes that it is blocked by
1612 : * into the caller-supplied buffer output[]. The list is truncated at
1613 : * output_size, and the number of PIDs written into the buffer is
1614 : * returned. Returns zero if the given PID is not currently blocked
1615 : * in GetSafeSnapshot.
1616 : */
1617 : int
1618 864 : GetSafeSnapshotBlockingPids(int blocked_pid, int *output, int output_size)
1619 : {
1620 864 : int num_written = 0;
1621 : dlist_iter iter;
1622 864 : SERIALIZABLEXACT *blocking_sxact = NULL;
1623 :
1624 864 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1625 :
1626 : /* Find blocked_pid's SERIALIZABLEXACT by linear search. */
1627 2006 : dlist_foreach(iter, &PredXact->activeList)
1628 : {
1629 1350 : SERIALIZABLEXACT *sxact =
1630 1350 : dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
1631 :
1632 1350 : if (sxact->pid == blocked_pid)
1633 : {
1634 208 : blocking_sxact = sxact;
1635 208 : break;
1636 : }
1637 : }
1638 :
1639 : /* Did we find it, and is it currently waiting in GetSafeSnapshot? */
1640 864 : if (blocking_sxact != NULL && SxactIsDeferrableWaiting(blocking_sxact))
1641 : {
1642 : /* Traverse the list of possible unsafe conflicts collecting PIDs. */
1643 4 : dlist_foreach(iter, &blocking_sxact->possibleUnsafeConflicts)
1644 : {
1645 4 : RWConflict possibleUnsafeConflict =
1646 4 : dlist_container(RWConflictData, inLink, iter.cur);
1647 :
1648 4 : output[num_written++] = possibleUnsafeConflict->sxactOut->pid;
1649 :
1650 4 : if (num_written >= output_size)
1651 4 : break;
1652 : }
1653 : }
1654 :
1655 864 : LWLockRelease(SerializableXactHashLock);
1656 :
1657 864 : return num_written;
1658 : }
1659 :
1660 : /*
1661 : * Acquire a snapshot that can be used for the current transaction.
1662 : *
1663 : * Make sure we have a SERIALIZABLEXACT reference in MySerializableXact.
1664 : * It should be current for this process and be contained in PredXact.
1665 : *
1666 : * The passed-in Snapshot pointer should reference a static data area that
1667 : * can safely be passed to GetSnapshotData. The return value is actually
1668 : * always this same pointer; no new snapshot data structure is allocated
1669 : * within this function.
1670 : */
1671 : Snapshot
1672 3280 : GetSerializableTransactionSnapshot(Snapshot snapshot)
1673 : {
1674 : Assert(IsolationIsSerializable());
1675 :
1676 : /*
1677 : * Can't use serializable mode while recovery is still active, as it is,
1678 : * for example, on a hot standby. We could get here despite the check in
1679 : * check_transaction_isolation() if default_transaction_isolation is set
1680 : * to serializable, so phrase the hint accordingly.
1681 : */
1682 3280 : if (RecoveryInProgress())
1683 0 : ereport(ERROR,
1684 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1685 : errmsg("cannot use serializable mode in a hot standby"),
1686 : errdetail("\"default_transaction_isolation\" is set to \"serializable\"."),
1687 : errhint("You can use \"SET default_transaction_isolation = 'repeatable read'\" to change the default.")));
1688 :
1689 : /*
1690 : * A special optimization is available for SERIALIZABLE READ ONLY
1691 : * DEFERRABLE transactions -- we can wait for a suitable snapshot and
1692 : * thereby avoid all SSI overhead once it's running.
1693 : */
1694 3280 : if (XactReadOnly && XactDeferrable)
1695 8 : return GetSafeSnapshot(snapshot);
1696 :
1697 3272 : return GetSerializableTransactionSnapshotInt(snapshot,
1698 : NULL, InvalidPid);
1699 : }
1700 :
1701 : /*
1702 : * Import a snapshot to be used for the current transaction.
1703 : *
1704 : * This is nearly the same as GetSerializableTransactionSnapshot, except that
1705 : * we don't take a new snapshot, but rather use the data we're handed.
1706 : *
1707 : * The caller must have verified that the snapshot came from a serializable
1708 : * transaction; and if we're read-write, the source transaction must not be
1709 : * read-only.
1710 : */
1711 : void
1712 26 : SetSerializableTransactionSnapshot(Snapshot snapshot,
1713 : VirtualTransactionId *sourcevxid,
1714 : int sourcepid)
1715 : {
1716 : Assert(IsolationIsSerializable());
1717 :
1718 : /*
1719 : * If this is called by parallel.c in a parallel worker, we don't want to
1720 : * create a SERIALIZABLEXACT just yet because the leader's
1721 : * SERIALIZABLEXACT will be installed with AttachSerializableXact(). We
1722 : * also don't want to reject SERIALIZABLE READ ONLY DEFERRABLE in this
1723 : * case, because the leader has already determined that the snapshot it
1724 : * has passed us is safe. So there is nothing for us to do.
1725 : */
1726 26 : if (IsParallelWorker())
1727 26 : return;
1728 :
1729 : /*
1730 : * We do not allow SERIALIZABLE READ ONLY DEFERRABLE transactions to
1731 : * import snapshots, since there's no way to wait for a safe snapshot when
1732 : * we're using the snap we're told to. (XXX instead of throwing an error,
1733 : * we could just ignore the XactDeferrable flag?)
1734 : */
1735 0 : if (XactReadOnly && XactDeferrable)
1736 0 : ereport(ERROR,
1737 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1738 : errmsg("a snapshot-importing transaction must not be READ ONLY DEFERRABLE")));
1739 :
1740 0 : (void) GetSerializableTransactionSnapshotInt(snapshot, sourcevxid,
1741 : sourcepid);
1742 : }
1743 :
1744 : /*
1745 : * Guts of GetSerializableTransactionSnapshot
1746 : *
1747 : * If sourcevxid is valid, this is actually an import operation and we should
1748 : * skip calling GetSnapshotData, because the snapshot contents are already
1749 : * loaded up. HOWEVER: to avoid race conditions, we must check that the
1750 : * source xact is still running after we acquire SerializableXactHashLock.
1751 : * We do that by calling ProcArrayInstallImportedXmin.
1752 : */
1753 : static Snapshot
1754 3282 : GetSerializableTransactionSnapshotInt(Snapshot snapshot,
1755 : VirtualTransactionId *sourcevxid,
1756 : int sourcepid)
1757 : {
1758 : PGPROC *proc;
1759 : VirtualTransactionId vxid;
1760 : SERIALIZABLEXACT *sxact,
1761 : *othersxact;
1762 :
1763 : /* We only do this for serializable transactions. Once. */
1764 : Assert(MySerializableXact == InvalidSerializableXact);
1765 :
1766 : Assert(!RecoveryInProgress());
1767 :
1768 : /*
1769 : * Since all parts of a serializable transaction must use the same
1770 : * snapshot, it is too late to establish one after a parallel operation
1771 : * has begun.
1772 : */
1773 3282 : if (IsInParallelMode())
1774 0 : elog(ERROR, "cannot establish serializable snapshot during a parallel operation");
1775 :
1776 3282 : proc = MyProc;
1777 : Assert(proc != NULL);
1778 3282 : GET_VXID_FROM_PGPROC(vxid, *proc);
1779 :
1780 : /*
1781 : * First we get the sxact structure, which may involve looping and access
1782 : * to the "finished" list to free a structure for use.
1783 : *
1784 : * We must hold SerializableXactHashLock when taking/checking the snapshot
1785 : * to avoid race conditions, for much the same reasons that
1786 : * GetSnapshotData takes the ProcArrayLock. Since we might have to
1787 : * release SerializableXactHashLock to call SummarizeOldestCommittedSxact,
1788 : * this means we have to create the sxact first, which is a bit annoying
1789 : * (in particular, an elog(ERROR) in procarray.c would cause us to leak
1790 : * the sxact). Consider refactoring to avoid this.
1791 : */
1792 : #ifdef TEST_SUMMARIZE_SERIAL
1793 : SummarizeOldestCommittedSxact();
1794 : #endif
1795 3282 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1796 : do
1797 : {
1798 3282 : sxact = CreatePredXact();
1799 : /* If null, push out committed sxact to SLRU summary & retry. */
1800 3282 : if (!sxact)
1801 : {
1802 0 : LWLockRelease(SerializableXactHashLock);
1803 0 : SummarizeOldestCommittedSxact();
1804 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1805 : }
1806 3282 : } while (!sxact);
1807 :
1808 : /* Get the snapshot, or check that it's safe to use */
1809 3282 : if (!sourcevxid)
1810 3282 : snapshot = GetSnapshotData(snapshot);
1811 0 : else if (!ProcArrayInstallImportedXmin(snapshot->xmin, sourcevxid))
1812 : {
1813 0 : ReleasePredXact(sxact);
1814 0 : LWLockRelease(SerializableXactHashLock);
1815 0 : ereport(ERROR,
1816 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1817 : errmsg("could not import the requested snapshot"),
1818 : errdetail("The source process with PID %d is not running anymore.",
1819 : sourcepid)));
1820 : }
1821 :
1822 : /*
1823 : * If there are no serializable transactions which are not read-only, we
1824 : * can "opt out" of predicate locking and conflict checking for a
1825 : * read-only transaction.
1826 : *
1827 : * The reason this is safe is that a read-only transaction can only become
1828 : * part of a dangerous structure if it overlaps a writable transaction
1829 : * which in turn overlaps a writable transaction which committed before
1830 : * the read-only transaction started. A new writable transaction can
1831 : * overlap this one, but it can't meet the other condition of overlapping
1832 : * a transaction which committed before this one started.
1833 : */
1834 3282 : if (XactReadOnly && PredXact->WritableSxactCount == 0)
1835 : {
1836 222 : ReleasePredXact(sxact);
1837 222 : LWLockRelease(SerializableXactHashLock);
1838 222 : return snapshot;
1839 : }
1840 :
1841 : /* Initialize the structure. */
1842 3060 : sxact->vxid = vxid;
1843 3060 : sxact->SeqNo.lastCommitBeforeSnapshot = PredXact->LastSxactCommitSeqNo;
1844 3060 : sxact->prepareSeqNo = InvalidSerCommitSeqNo;
1845 3060 : sxact->commitSeqNo = InvalidSerCommitSeqNo;
1846 3060 : dlist_init(&(sxact->outConflicts));
1847 3060 : dlist_init(&(sxact->inConflicts));
1848 3060 : dlist_init(&(sxact->possibleUnsafeConflicts));
1849 3060 : sxact->topXid = GetTopTransactionIdIfAny();
1850 3060 : sxact->finishedBefore = InvalidTransactionId;
1851 3060 : sxact->xmin = snapshot->xmin;
1852 3060 : sxact->pid = MyProcPid;
1853 3060 : sxact->pgprocno = MyProcNumber;
1854 3060 : dlist_init(&sxact->predicateLocks);
1855 3060 : dlist_node_init(&sxact->finishedLink);
1856 3060 : sxact->flags = 0;
1857 3060 : if (XactReadOnly)
1858 : {
1859 : dlist_iter iter;
1860 :
1861 214 : sxact->flags |= SXACT_FLAG_READ_ONLY;
1862 :
1863 : /*
1864 : * Register all concurrent r/w transactions as possible conflicts; if
1865 : * all of them commit without any outgoing conflicts to earlier
1866 : * transactions then this snapshot can be deemed safe (and we can run
1867 : * without tracking predicate locks).
1868 : */
1869 938 : dlist_foreach(iter, &PredXact->activeList)
1870 : {
1871 724 : othersxact = dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
1872 :
1873 724 : if (!SxactIsCommitted(othersxact)
1874 484 : && !SxactIsDoomed(othersxact)
1875 484 : && !SxactIsReadOnly(othersxact))
1876 : {
1877 268 : SetPossibleUnsafeConflict(sxact, othersxact);
1878 : }
1879 : }
1880 :
1881 : /*
1882 : * If we didn't find any possibly unsafe conflicts because every
1883 : * uncommitted writable transaction turned out to be doomed, then we
1884 : * can "opt out" immediately. See comments above the earlier check
1885 : * for PredXact->WritableSxactCount == 0.
1886 : */
1887 214 : if (dlist_is_empty(&sxact->possibleUnsafeConflicts))
1888 : {
1889 0 : ReleasePredXact(sxact);
1890 0 : LWLockRelease(SerializableXactHashLock);
1891 0 : return snapshot;
1892 : }
1893 : }
1894 : else
1895 : {
1896 2846 : ++(PredXact->WritableSxactCount);
1897 : Assert(PredXact->WritableSxactCount <=
1898 : (MaxBackends + max_prepared_xacts));
1899 : }
1900 :
1901 : /* Maintain serializable global xmin info. */
1902 3060 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
1903 : {
1904 : Assert(PredXact->SxactGlobalXminCount == 0);
1905 1652 : PredXact->SxactGlobalXmin = snapshot->xmin;
1906 1652 : PredXact->SxactGlobalXminCount = 1;
1907 1652 : SerialSetActiveSerXmin(snapshot->xmin);
1908 : }
1909 1408 : else if (TransactionIdEquals(snapshot->xmin, PredXact->SxactGlobalXmin))
1910 : {
1911 : Assert(PredXact->SxactGlobalXminCount > 0);
1912 1336 : PredXact->SxactGlobalXminCount++;
1913 : }
1914 : else
1915 : {
1916 : Assert(TransactionIdFollows(snapshot->xmin, PredXact->SxactGlobalXmin));
1917 : }
1918 :
1919 3060 : MySerializableXact = sxact;
1920 3060 : MyXactDidWrite = false; /* haven't written anything yet */
1921 :
1922 3060 : LWLockRelease(SerializableXactHashLock);
1923 :
1924 3060 : CreateLocalPredicateLockHash();
1925 :
1926 3060 : return snapshot;
1927 : }
1928 :
1929 : static void
1930 3086 : CreateLocalPredicateLockHash(void)
1931 : {
1932 : HASHCTL hash_ctl;
1933 :
1934 : /* Initialize the backend-local hash table of parent locks */
1935 : Assert(LocalPredicateLockHash == NULL);
1936 3086 : hash_ctl.keysize = sizeof(PREDICATELOCKTARGETTAG);
1937 3086 : hash_ctl.entrysize = sizeof(LOCALPREDICATELOCK);
1938 3086 : LocalPredicateLockHash = hash_create("Local predicate lock",
1939 : max_predicate_locks_per_xact,
1940 : &hash_ctl,
1941 : HASH_ELEM | HASH_BLOBS);
1942 3086 : }
1943 :
1944 : /*
1945 : * Register the top level XID in SerializableXidHash.
1946 : * Also store it for easy reference in MySerializableXact.
1947 : */
1948 : void
1949 249950 : RegisterPredicateLockingXid(TransactionId xid)
1950 : {
1951 : SERIALIZABLEXIDTAG sxidtag;
1952 : SERIALIZABLEXID *sxid;
1953 : bool found;
1954 :
1955 : /*
1956 : * If we're not tracking predicate lock data for this transaction, we
1957 : * should ignore the request and return quickly.
1958 : */
1959 249950 : if (MySerializableXact == InvalidSerializableXact)
1960 247424 : return;
1961 :
1962 : /* We should have a valid XID and be at the top level. */
1963 : Assert(TransactionIdIsValid(xid));
1964 :
1965 2526 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1966 :
1967 : /* This should only be done once per transaction. */
1968 : Assert(MySerializableXact->topXid == InvalidTransactionId);
1969 :
1970 2526 : MySerializableXact->topXid = xid;
1971 :
1972 2526 : sxidtag.xid = xid;
1973 2526 : sxid = (SERIALIZABLEXID *) hash_search(SerializableXidHash,
1974 : &sxidtag,
1975 : HASH_ENTER, &found);
1976 : Assert(!found);
1977 :
1978 : /* Initialize the structure. */
1979 2526 : sxid->myXact = MySerializableXact;
1980 2526 : LWLockRelease(SerializableXactHashLock);
1981 : }
1982 :
1983 :
1984 : /*
1985 : * Check whether there are any predicate locks held by any transaction
1986 : * for the page at the given block number.
1987 : *
1988 : * Note that the transaction may be completed but not yet subject to
1989 : * cleanup due to overlapping serializable transactions. This must
1990 : * return valid information regardless of transaction isolation level.
1991 : *
1992 : * Also note that this doesn't check for a conflicting relation lock,
1993 : * just a lock specifically on the given page.
1994 : *
1995 : * One use is to support proper behavior during GiST index vacuum.
1996 : */
1997 : bool
1998 0 : PageIsPredicateLocked(Relation relation, BlockNumber blkno)
1999 : {
2000 : PREDICATELOCKTARGETTAG targettag;
2001 : uint32 targettaghash;
2002 : LWLock *partitionLock;
2003 : PREDICATELOCKTARGET *target;
2004 :
2005 0 : SET_PREDICATELOCKTARGETTAG_PAGE(targettag,
2006 : relation->rd_locator.dbOid,
2007 : relation->rd_id,
2008 : blkno);
2009 :
2010 0 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
2011 0 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
2012 0 : LWLockAcquire(partitionLock, LW_SHARED);
2013 : target = (PREDICATELOCKTARGET *)
2014 0 : hash_search_with_hash_value(PredicateLockTargetHash,
2015 : &targettag, targettaghash,
2016 : HASH_FIND, NULL);
2017 0 : LWLockRelease(partitionLock);
2018 :
2019 0 : return (target != NULL);
2020 : }
2021 :
2022 :
2023 : /*
2024 : * Check whether a particular lock is held by this transaction.
2025 : *
2026 : * Important note: this function may return false even if the lock is
2027 : * being held, because it uses the local lock table which is not
2028 : * updated if another transaction modifies our lock list (e.g. to
2029 : * split an index page). It can also return true when a coarser
2030 : * granularity lock that covers this target is being held. Be careful
2031 : * to only use this function in circumstances where such errors are
2032 : * acceptable!
2033 : */
2034 : static bool
2035 154380 : PredicateLockExists(const PREDICATELOCKTARGETTAG *targettag)
2036 : {
2037 : LOCALPREDICATELOCK *lock;
2038 :
2039 : /* check local hash table */
2040 154380 : lock = (LOCALPREDICATELOCK *) hash_search(LocalPredicateLockHash,
2041 : targettag,
2042 : HASH_FIND, NULL);
2043 :
2044 154380 : if (!lock)
2045 60178 : return false;
2046 :
2047 : /*
2048 : * Found entry in the table, but still need to check whether it's actually
2049 : * held -- it could just be a parent of some held lock.
2050 : */
2051 94202 : return lock->held;
2052 : }
2053 :
2054 : /*
2055 : * Return the parent lock tag in the lock hierarchy: the next coarser
2056 : * lock that covers the provided tag.
2057 : *
2058 : * Returns true and sets *parent to the parent tag if one exists,
2059 : * returns false if none exists.
2060 : */
2061 : static bool
2062 90298 : GetParentPredicateLockTag(const PREDICATELOCKTARGETTAG *tag,
2063 : PREDICATELOCKTARGETTAG *parent)
2064 : {
2065 90298 : switch (GET_PREDICATELOCKTARGETTAG_TYPE(*tag))
2066 : {
2067 19600 : case PREDLOCKTAG_RELATION:
2068 : /* relation locks have no parent lock */
2069 19600 : return false;
2070 :
2071 16832 : case PREDLOCKTAG_PAGE:
2072 : /* parent lock is relation lock */
2073 16832 : SET_PREDICATELOCKTARGETTAG_RELATION(*parent,
2074 : GET_PREDICATELOCKTARGETTAG_DB(*tag),
2075 : GET_PREDICATELOCKTARGETTAG_RELATION(*tag));
2076 :
2077 16832 : return true;
2078 :
2079 53866 : case PREDLOCKTAG_TUPLE:
2080 : /* parent lock is page lock */
2081 53866 : SET_PREDICATELOCKTARGETTAG_PAGE(*parent,
2082 : GET_PREDICATELOCKTARGETTAG_DB(*tag),
2083 : GET_PREDICATELOCKTARGETTAG_RELATION(*tag),
2084 : GET_PREDICATELOCKTARGETTAG_PAGE(*tag));
2085 53866 : return true;
2086 : }
2087 :
2088 : /* not reachable */
2089 : Assert(false);
2090 0 : return false;
2091 : }
2092 :
2093 : /*
2094 : * Check whether the lock we are considering is already covered by a
2095 : * coarser lock for our transaction.
2096 : *
2097 : * Like PredicateLockExists, this function might return a false
2098 : * negative, but it will never return a false positive.
2099 : */
2100 : static bool
2101 52062 : CoarserLockCovers(const PREDICATELOCKTARGETTAG *newtargettag)
2102 : {
2103 : PREDICATELOCKTARGETTAG targettag,
2104 : parenttag;
2105 :
2106 52062 : targettag = *newtargettag;
2107 :
2108 : /* check parents iteratively until no more */
2109 62830 : while (GetParentPredicateLockTag(&targettag, &parenttag))
2110 : {
2111 54410 : targettag = parenttag;
2112 54410 : if (PredicateLockExists(&targettag))
2113 43642 : return true;
2114 : }
2115 :
2116 : /* no more parents to check; lock is not covered */
2117 8420 : return false;
2118 : }
2119 :
2120 : /*
2121 : * Remove the dummy entry from the predicate lock target hash, to free up some
2122 : * scratch space. The caller must be holding SerializablePredicateListLock,
2123 : * and must restore the entry with RestoreScratchTarget() before releasing the
2124 : * lock.
2125 : *
2126 : * If lockheld is true, the caller is already holding the partition lock
2127 : * of the partition containing the scratch entry.
2128 : */
2129 : static void
2130 94 : RemoveScratchTarget(bool lockheld)
2131 : {
2132 : bool found;
2133 :
2134 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2135 :
2136 94 : if (!lockheld)
2137 0 : LWLockAcquire(ScratchPartitionLock, LW_EXCLUSIVE);
2138 94 : hash_search_with_hash_value(PredicateLockTargetHash,
2139 : &ScratchTargetTag,
2140 : ScratchTargetTagHash,
2141 : HASH_REMOVE, &found);
2142 : Assert(found);
2143 94 : if (!lockheld)
2144 0 : LWLockRelease(ScratchPartitionLock);
2145 94 : }
2146 :
2147 : /*
2148 : * Re-insert the dummy entry in predicate lock target hash.
2149 : */
2150 : static void
2151 94 : RestoreScratchTarget(bool lockheld)
2152 : {
2153 : bool found;
2154 :
2155 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2156 :
2157 94 : if (!lockheld)
2158 0 : LWLockAcquire(ScratchPartitionLock, LW_EXCLUSIVE);
2159 94 : hash_search_with_hash_value(PredicateLockTargetHash,
2160 : &ScratchTargetTag,
2161 : ScratchTargetTagHash,
2162 : HASH_ENTER, &found);
2163 : Assert(!found);
2164 94 : if (!lockheld)
2165 0 : LWLockRelease(ScratchPartitionLock);
2166 94 : }
2167 :
2168 : /*
2169 : * Check whether the list of related predicate locks is empty for a
2170 : * predicate lock target, and remove the target if it is.
2171 : */
2172 : static void
2173 8408 : RemoveTargetIfNoLongerUsed(PREDICATELOCKTARGET *target, uint32 targettaghash)
2174 : {
2175 : PREDICATELOCKTARGET *rmtarget PG_USED_FOR_ASSERTS_ONLY;
2176 :
2177 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2178 :
2179 : /* Can't remove it until no locks at this target. */
2180 8408 : if (!dlist_is_empty(&target->predicateLocks))
2181 1922 : return;
2182 :
2183 : /* Actually remove the target. */
2184 6486 : rmtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2185 6486 : &target->tag,
2186 : targettaghash,
2187 : HASH_REMOVE, NULL);
2188 : Assert(rmtarget == target);
2189 : }
2190 :
2191 : /*
2192 : * Delete child target locks owned by this process.
2193 : * This implementation is assuming that the usage of each target tag field
2194 : * is uniform. No need to make this hard if we don't have to.
2195 : *
2196 : * We acquire an LWLock in the case of parallel mode, because worker
2197 : * backends have access to the leader's SERIALIZABLEXACT. Otherwise,
2198 : * we aren't acquiring LWLocks for the predicate lock or lock
2199 : * target structures associated with this transaction unless we're going
2200 : * to modify them, because no other process is permitted to modify our
2201 : * locks.
2202 : */
2203 : static void
2204 4688 : DeleteChildTargetLocks(const PREDICATELOCKTARGETTAG *newtargettag)
2205 : {
2206 : SERIALIZABLEXACT *sxact;
2207 : PREDICATELOCK *predlock;
2208 : dlist_mutable_iter iter;
2209 :
2210 4688 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2211 4688 : sxact = MySerializableXact;
2212 4688 : if (IsInParallelMode())
2213 22 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
2214 :
2215 15532 : dlist_foreach_modify(iter, &sxact->predicateLocks)
2216 : {
2217 : PREDICATELOCKTAG oldlocktag;
2218 : PREDICATELOCKTARGET *oldtarget;
2219 : PREDICATELOCKTARGETTAG oldtargettag;
2220 :
2221 10844 : predlock = dlist_container(PREDICATELOCK, xactLink, iter.cur);
2222 :
2223 10844 : oldlocktag = predlock->tag;
2224 : Assert(oldlocktag.myXact == sxact);
2225 10844 : oldtarget = oldlocktag.myTarget;
2226 10844 : oldtargettag = oldtarget->tag;
2227 :
2228 10844 : if (TargetTagIsCoveredBy(oldtargettag, *newtargettag))
2229 : {
2230 : uint32 oldtargettaghash;
2231 : LWLock *partitionLock;
2232 : PREDICATELOCK *rmpredlock PG_USED_FOR_ASSERTS_ONLY;
2233 :
2234 1998 : oldtargettaghash = PredicateLockTargetTagHashCode(&oldtargettag);
2235 1998 : partitionLock = PredicateLockHashPartitionLock(oldtargettaghash);
2236 :
2237 1998 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
2238 :
2239 1998 : dlist_delete(&predlock->xactLink);
2240 1998 : dlist_delete(&predlock->targetLink);
2241 1998 : rmpredlock = hash_search_with_hash_value
2242 : (PredicateLockHash,
2243 : &oldlocktag,
2244 1998 : PredicateLockHashCodeFromTargetHashCode(&oldlocktag,
2245 : oldtargettaghash),
2246 : HASH_REMOVE, NULL);
2247 : Assert(rmpredlock == predlock);
2248 :
2249 1998 : RemoveTargetIfNoLongerUsed(oldtarget, oldtargettaghash);
2250 :
2251 1998 : LWLockRelease(partitionLock);
2252 :
2253 1998 : DecrementParentLocks(&oldtargettag);
2254 : }
2255 : }
2256 4688 : if (IsInParallelMode())
2257 22 : LWLockRelease(&sxact->perXactPredicateListLock);
2258 4688 : LWLockRelease(SerializablePredicateListLock);
2259 4688 : }
2260 :
2261 : /*
2262 : * Returns the promotion limit for a given predicate lock target. This is the
2263 : * max number of descendant locks allowed before promoting to the specified
2264 : * tag. Note that the limit includes non-direct descendants (e.g., both tuples
2265 : * and pages for a relation lock).
2266 : *
2267 : * Currently the default limit is 2 for a page lock, and half of the value of
2268 : * max_pred_locks_per_transaction - 1 for a relation lock, to match behavior
2269 : * of earlier releases when upgrading.
2270 : *
2271 : * TODO SSI: We should probably add additional GUCs to allow a maximum ratio
2272 : * of page and tuple locks based on the pages in a relation, and the maximum
2273 : * ratio of tuple locks to tuples in a page. This would provide more
2274 : * generally "balanced" allocation of locks to where they are most useful,
2275 : * while still allowing the absolute numbers to prevent one relation from
2276 : * tying up all predicate lock resources.
2277 : */
2278 : static int
2279 10768 : MaxPredicateChildLocks(const PREDICATELOCKTARGETTAG *tag)
2280 : {
2281 10768 : switch (GET_PREDICATELOCKTARGETTAG_TYPE(*tag))
2282 : {
2283 7036 : case PREDLOCKTAG_RELATION:
2284 7036 : return max_predicate_locks_per_relation < 0
2285 : ? (max_predicate_locks_per_xact
2286 7036 : / (-max_predicate_locks_per_relation)) - 1
2287 7036 : : max_predicate_locks_per_relation;
2288 :
2289 3732 : case PREDLOCKTAG_PAGE:
2290 3732 : return max_predicate_locks_per_page;
2291 :
2292 0 : case PREDLOCKTAG_TUPLE:
2293 :
2294 : /*
2295 : * not reachable: nothing is finer-granularity than a tuple, so we
2296 : * should never try to promote to it.
2297 : */
2298 : Assert(false);
2299 0 : return 0;
2300 : }
2301 :
2302 : /* not reachable */
2303 : Assert(false);
2304 0 : return 0;
2305 : }
2306 :
2307 : /*
2308 : * For all ancestors of a newly-acquired predicate lock, increment
2309 : * their child count in the parent hash table. If any of them have
2310 : * more descendants than their promotion threshold, acquire the
2311 : * coarsest such lock.
2312 : *
2313 : * Returns true if a parent lock was acquired and false otherwise.
2314 : */
2315 : static bool
2316 8420 : CheckAndPromotePredicateLockRequest(const PREDICATELOCKTARGETTAG *reqtag)
2317 : {
2318 : PREDICATELOCKTARGETTAG targettag,
2319 : nexttag,
2320 : promotiontag;
2321 : LOCALPREDICATELOCK *parentlock;
2322 : bool found,
2323 : promote;
2324 :
2325 8420 : promote = false;
2326 :
2327 8420 : targettag = *reqtag;
2328 :
2329 : /* check parents iteratively */
2330 19188 : while (GetParentPredicateLockTag(&targettag, &nexttag))
2331 : {
2332 10768 : targettag = nexttag;
2333 10768 : parentlock = (LOCALPREDICATELOCK *) hash_search(LocalPredicateLockHash,
2334 : &targettag,
2335 : HASH_ENTER,
2336 : &found);
2337 10768 : if (!found)
2338 : {
2339 6646 : parentlock->held = false;
2340 6646 : parentlock->childLocks = 1;
2341 : }
2342 : else
2343 4122 : parentlock->childLocks++;
2344 :
2345 10768 : if (parentlock->childLocks >
2346 10768 : MaxPredicateChildLocks(&targettag))
2347 : {
2348 : /*
2349 : * We should promote to this parent lock. Continue to check its
2350 : * ancestors, however, both to get their child counts right and to
2351 : * check whether we should just go ahead and promote to one of
2352 : * them.
2353 : */
2354 666 : promotiontag = targettag;
2355 666 : promote = true;
2356 : }
2357 : }
2358 :
2359 8420 : if (promote)
2360 : {
2361 : /* acquire coarsest ancestor eligible for promotion */
2362 666 : PredicateLockAcquire(&promotiontag);
2363 666 : return true;
2364 : }
2365 : else
2366 7754 : return false;
2367 : }
2368 :
2369 : /*
2370 : * When releasing a lock, decrement the child count on all ancestor
2371 : * locks.
2372 : *
2373 : * This is called only when releasing a lock via
2374 : * DeleteChildTargetLocks (i.e. when a lock becomes redundant because
2375 : * we've acquired its parent, possibly due to promotion) or when a new
2376 : * MVCC write lock makes the predicate lock unnecessary. There's no
2377 : * point in calling it when locks are released at transaction end, as
2378 : * this information is no longer needed.
2379 : */
2380 : static void
2381 2760 : DecrementParentLocks(const PREDICATELOCKTARGETTAG *targettag)
2382 : {
2383 : PREDICATELOCKTARGETTAG parenttag,
2384 : nexttag;
2385 :
2386 2760 : parenttag = *targettag;
2387 :
2388 8280 : while (GetParentPredicateLockTag(&parenttag, &nexttag))
2389 : {
2390 : uint32 targettaghash;
2391 : LOCALPREDICATELOCK *parentlock,
2392 : *rmlock PG_USED_FOR_ASSERTS_ONLY;
2393 :
2394 5520 : parenttag = nexttag;
2395 5520 : targettaghash = PredicateLockTargetTagHashCode(&parenttag);
2396 : parentlock = (LOCALPREDICATELOCK *)
2397 5520 : hash_search_with_hash_value(LocalPredicateLockHash,
2398 : &parenttag, targettaghash,
2399 : HASH_FIND, NULL);
2400 :
2401 : /*
2402 : * There's a small chance the parent lock doesn't exist in the lock
2403 : * table. This can happen if we prematurely removed it because an
2404 : * index split caused the child refcount to be off.
2405 : */
2406 5520 : if (parentlock == NULL)
2407 0 : continue;
2408 :
2409 5520 : parentlock->childLocks--;
2410 :
2411 : /*
2412 : * Under similar circumstances the parent lock's refcount might be
2413 : * zero. This only happens if we're holding that lock (otherwise we
2414 : * would have removed the entry).
2415 : */
2416 5520 : if (parentlock->childLocks < 0)
2417 : {
2418 : Assert(parentlock->held);
2419 0 : parentlock->childLocks = 0;
2420 : }
2421 :
2422 5520 : if ((parentlock->childLocks == 0) && (!parentlock->held))
2423 : {
2424 : rmlock = (LOCALPREDICATELOCK *)
2425 1500 : hash_search_with_hash_value(LocalPredicateLockHash,
2426 : &parenttag, targettaghash,
2427 : HASH_REMOVE, NULL);
2428 : Assert(rmlock == parentlock);
2429 : }
2430 : }
2431 2760 : }
2432 :
2433 : /*
2434 : * Indicate that a predicate lock on the given target is held by the
2435 : * specified transaction. Has no effect if the lock is already held.
2436 : *
2437 : * This updates the lock table and the sxact's lock list, and creates
2438 : * the lock target if necessary, but does *not* do anything related to
2439 : * granularity promotion or the local lock table. See
2440 : * PredicateLockAcquire for that.
2441 : */
2442 : static void
2443 8420 : CreatePredicateLock(const PREDICATELOCKTARGETTAG *targettag,
2444 : uint32 targettaghash,
2445 : SERIALIZABLEXACT *sxact)
2446 : {
2447 : PREDICATELOCKTARGET *target;
2448 : PREDICATELOCKTAG locktag;
2449 : PREDICATELOCK *lock;
2450 : LWLock *partitionLock;
2451 : bool found;
2452 :
2453 8420 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
2454 :
2455 8420 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2456 8420 : if (IsInParallelMode())
2457 32 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
2458 8420 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
2459 :
2460 : /* Make sure that the target is represented. */
2461 : target = (PREDICATELOCKTARGET *)
2462 8420 : hash_search_with_hash_value(PredicateLockTargetHash,
2463 : targettag, targettaghash,
2464 : HASH_ENTER_NULL, &found);
2465 8420 : if (!target)
2466 0 : ereport(ERROR,
2467 : (errcode(ERRCODE_OUT_OF_MEMORY),
2468 : errmsg("out of shared memory"),
2469 : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
2470 8420 : if (!found)
2471 6486 : dlist_init(&target->predicateLocks);
2472 :
2473 : /* We've got the sxact and target, make sure they're joined. */
2474 8420 : locktag.myTarget = target;
2475 8420 : locktag.myXact = sxact;
2476 : lock = (PREDICATELOCK *)
2477 8420 : hash_search_with_hash_value(PredicateLockHash, &locktag,
2478 8420 : PredicateLockHashCodeFromTargetHashCode(&locktag, targettaghash),
2479 : HASH_ENTER_NULL, &found);
2480 8420 : if (!lock)
2481 0 : ereport(ERROR,
2482 : (errcode(ERRCODE_OUT_OF_MEMORY),
2483 : errmsg("out of shared memory"),
2484 : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
2485 :
2486 8420 : if (!found)
2487 : {
2488 8408 : dlist_push_tail(&target->predicateLocks, &lock->targetLink);
2489 8408 : dlist_push_tail(&sxact->predicateLocks, &lock->xactLink);
2490 8408 : lock->commitSeqNo = InvalidSerCommitSeqNo;
2491 : }
2492 :
2493 8420 : LWLockRelease(partitionLock);
2494 8420 : if (IsInParallelMode())
2495 32 : LWLockRelease(&sxact->perXactPredicateListLock);
2496 8420 : LWLockRelease(SerializablePredicateListLock);
2497 8420 : }
2498 :
2499 : /*
2500 : * Acquire a predicate lock on the specified target for the current
2501 : * connection if not already held. This updates the local lock table
2502 : * and uses it to implement granularity promotion. It will consolidate
2503 : * multiple locks into a coarser lock if warranted, and will release
2504 : * any finer-grained locks covered by the new one.
2505 : */
2506 : static void
2507 52462 : PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag)
2508 : {
2509 : uint32 targettaghash;
2510 : bool found;
2511 : LOCALPREDICATELOCK *locallock;
2512 :
2513 : /* Do we have the lock already, or a covering lock? */
2514 52462 : if (PredicateLockExists(targettag))
2515 44042 : return;
2516 :
2517 52062 : if (CoarserLockCovers(targettag))
2518 43642 : return;
2519 :
2520 : /* the same hash and LW lock apply to the lock target and the local lock. */
2521 8420 : targettaghash = PredicateLockTargetTagHashCode(targettag);
2522 :
2523 : /* Acquire lock in local table */
2524 : locallock = (LOCALPREDICATELOCK *)
2525 8420 : hash_search_with_hash_value(LocalPredicateLockHash,
2526 : targettag, targettaghash,
2527 : HASH_ENTER, &found);
2528 8420 : locallock->held = true;
2529 8420 : if (!found)
2530 7754 : locallock->childLocks = 0;
2531 :
2532 : /* Actually create the lock */
2533 8420 : CreatePredicateLock(targettag, targettaghash, MySerializableXact);
2534 :
2535 : /*
2536 : * Lock has been acquired. Check whether it should be promoted to a
2537 : * coarser granularity, or whether there are finer-granularity locks to
2538 : * clean up.
2539 : */
2540 8420 : if (CheckAndPromotePredicateLockRequest(targettag))
2541 : {
2542 : /*
2543 : * Lock request was promoted to a coarser-granularity lock, and that
2544 : * lock was acquired. It will delete this lock and any of its
2545 : * children, so we're done.
2546 : */
2547 : }
2548 : else
2549 : {
2550 : /* Clean up any finer-granularity locks */
2551 7754 : if (GET_PREDICATELOCKTARGETTAG_TYPE(*targettag) != PREDLOCKTAG_TUPLE)
2552 4688 : DeleteChildTargetLocks(targettag);
2553 : }
2554 : }
2555 :
2556 :
2557 : /*
2558 : * PredicateLockRelation
2559 : *
2560 : * Gets a predicate lock at the relation level.
2561 : * Skip if not in full serializable transaction isolation level.
2562 : * Skip if this is a temporary table.
2563 : * Clear any finer-grained predicate locks this session has on the relation.
2564 : */
2565 : void
2566 697724 : PredicateLockRelation(Relation relation, Snapshot snapshot)
2567 : {
2568 : PREDICATELOCKTARGETTAG tag;
2569 :
2570 697724 : if (!SerializationNeededForRead(relation, snapshot))
2571 696298 : return;
2572 :
2573 1426 : SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2574 : relation->rd_locator.dbOid,
2575 : relation->rd_id);
2576 1426 : PredicateLockAcquire(&tag);
2577 : }
2578 :
2579 : /*
2580 : * PredicateLockPage
2581 : *
2582 : * Gets a predicate lock at the page level.
2583 : * Skip if not in full serializable transaction isolation level.
2584 : * Skip if this is a temporary table.
2585 : * Skip if a coarser predicate lock already covers this page.
2586 : * Clear any finer-grained predicate locks this session has on the relation.
2587 : */
2588 : void
2589 17612760 : PredicateLockPage(Relation relation, BlockNumber blkno, Snapshot snapshot)
2590 : {
2591 : PREDICATELOCKTARGETTAG tag;
2592 :
2593 17612760 : if (!SerializationNeededForRead(relation, snapshot))
2594 17609898 : return;
2595 :
2596 2862 : SET_PREDICATELOCKTARGETTAG_PAGE(tag,
2597 : relation->rd_locator.dbOid,
2598 : relation->rd_id,
2599 : blkno);
2600 2862 : PredicateLockAcquire(&tag);
2601 : }
2602 :
2603 : /*
2604 : * PredicateLockTID
2605 : *
2606 : * Gets a predicate lock at the tuple level.
2607 : * Skip if not in full serializable transaction isolation level.
2608 : * Skip if this is a temporary table.
2609 : */
2610 : void
2611 30177484 : PredicateLockTID(Relation relation, ItemPointer tid, Snapshot snapshot,
2612 : TransactionId tuple_xid)
2613 : {
2614 : PREDICATELOCKTARGETTAG tag;
2615 :
2616 30177484 : if (!SerializationNeededForRead(relation, snapshot))
2617 30129976 : return;
2618 :
2619 : /*
2620 : * Return if this xact wrote it.
2621 : */
2622 47508 : if (relation->rd_index == NULL)
2623 : {
2624 : /* If we wrote it; we already have a write lock. */
2625 47508 : if (TransactionIdIsCurrentTransactionId(tuple_xid))
2626 0 : return;
2627 : }
2628 :
2629 : /*
2630 : * Do quick-but-not-definitive test for a relation lock first. This will
2631 : * never cause a return when the relation is *not* locked, but will
2632 : * occasionally let the check continue when there really *is* a relation
2633 : * level lock.
2634 : */
2635 47508 : SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2636 : relation->rd_locator.dbOid,
2637 : relation->rd_id);
2638 47508 : if (PredicateLockExists(&tag))
2639 0 : return;
2640 :
2641 47508 : SET_PREDICATELOCKTARGETTAG_TUPLE(tag,
2642 : relation->rd_locator.dbOid,
2643 : relation->rd_id,
2644 : ItemPointerGetBlockNumber(tid),
2645 : ItemPointerGetOffsetNumber(tid));
2646 47508 : PredicateLockAcquire(&tag);
2647 : }
2648 :
2649 :
2650 : /*
2651 : * DeleteLockTarget
2652 : *
2653 : * Remove a predicate lock target along with any locks held for it.
2654 : *
2655 : * Caller must hold SerializablePredicateListLock and the
2656 : * appropriate hash partition lock for the target.
2657 : */
2658 : static void
2659 0 : DeleteLockTarget(PREDICATELOCKTARGET *target, uint32 targettaghash)
2660 : {
2661 : dlist_mutable_iter iter;
2662 :
2663 : Assert(LWLockHeldByMeInMode(SerializablePredicateListLock,
2664 : LW_EXCLUSIVE));
2665 : Assert(LWLockHeldByMe(PredicateLockHashPartitionLock(targettaghash)));
2666 :
2667 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2668 :
2669 0 : dlist_foreach_modify(iter, &target->predicateLocks)
2670 : {
2671 0 : PREDICATELOCK *predlock =
2672 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
2673 : bool found;
2674 :
2675 0 : dlist_delete(&(predlock->xactLink));
2676 0 : dlist_delete(&(predlock->targetLink));
2677 :
2678 0 : hash_search_with_hash_value
2679 : (PredicateLockHash,
2680 0 : &predlock->tag,
2681 0 : PredicateLockHashCodeFromTargetHashCode(&predlock->tag,
2682 : targettaghash),
2683 : HASH_REMOVE, &found);
2684 : Assert(found);
2685 : }
2686 0 : LWLockRelease(SerializableXactHashLock);
2687 :
2688 : /* Remove the target itself, if possible. */
2689 0 : RemoveTargetIfNoLongerUsed(target, targettaghash);
2690 0 : }
2691 :
2692 :
2693 : /*
2694 : * TransferPredicateLocksToNewTarget
2695 : *
2696 : * Move or copy all the predicate locks for a lock target, for use by
2697 : * index page splits/combines and other things that create or replace
2698 : * lock targets. If 'removeOld' is true, the old locks and the target
2699 : * will be removed.
2700 : *
2701 : * Returns true on success, or false if we ran out of shared memory to
2702 : * allocate the new target or locks. Guaranteed to always succeed if
2703 : * removeOld is set (by using the scratch entry in PredicateLockTargetHash
2704 : * for scratch space).
2705 : *
2706 : * Warning: the "removeOld" option should be used only with care,
2707 : * because this function does not (indeed, can not) update other
2708 : * backends' LocalPredicateLockHash. If we are only adding new
2709 : * entries, this is not a problem: the local lock table is used only
2710 : * as a hint, so missing entries for locks that are held are
2711 : * OK. Having entries for locks that are no longer held, as can happen
2712 : * when using "removeOld", is not in general OK. We can only use it
2713 : * safely when replacing a lock with a coarser-granularity lock that
2714 : * covers it, or if we are absolutely certain that no one will need to
2715 : * refer to that lock in the future.
2716 : *
2717 : * Caller must hold SerializablePredicateListLock exclusively.
2718 : */
2719 : static bool
2720 6 : TransferPredicateLocksToNewTarget(PREDICATELOCKTARGETTAG oldtargettag,
2721 : PREDICATELOCKTARGETTAG newtargettag,
2722 : bool removeOld)
2723 : {
2724 : uint32 oldtargettaghash;
2725 : LWLock *oldpartitionLock;
2726 : PREDICATELOCKTARGET *oldtarget;
2727 : uint32 newtargettaghash;
2728 : LWLock *newpartitionLock;
2729 : bool found;
2730 6 : bool outOfShmem = false;
2731 :
2732 : Assert(LWLockHeldByMeInMode(SerializablePredicateListLock,
2733 : LW_EXCLUSIVE));
2734 :
2735 6 : oldtargettaghash = PredicateLockTargetTagHashCode(&oldtargettag);
2736 6 : newtargettaghash = PredicateLockTargetTagHashCode(&newtargettag);
2737 6 : oldpartitionLock = PredicateLockHashPartitionLock(oldtargettaghash);
2738 6 : newpartitionLock = PredicateLockHashPartitionLock(newtargettaghash);
2739 :
2740 6 : if (removeOld)
2741 : {
2742 : /*
2743 : * Remove the dummy entry to give us scratch space, so we know we'll
2744 : * be able to create the new lock target.
2745 : */
2746 0 : RemoveScratchTarget(false);
2747 : }
2748 :
2749 : /*
2750 : * We must get the partition locks in ascending sequence to avoid
2751 : * deadlocks. If old and new partitions are the same, we must request the
2752 : * lock only once.
2753 : */
2754 6 : if (oldpartitionLock < newpartitionLock)
2755 : {
2756 4 : LWLockAcquire(oldpartitionLock,
2757 4 : (removeOld ? LW_EXCLUSIVE : LW_SHARED));
2758 4 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2759 : }
2760 2 : else if (oldpartitionLock > newpartitionLock)
2761 : {
2762 2 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2763 2 : LWLockAcquire(oldpartitionLock,
2764 2 : (removeOld ? LW_EXCLUSIVE : LW_SHARED));
2765 : }
2766 : else
2767 0 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2768 :
2769 : /*
2770 : * Look for the old target. If not found, that's OK; no predicate locks
2771 : * are affected, so we can just clean up and return. If it does exist,
2772 : * walk its list of predicate locks and move or copy them to the new
2773 : * target.
2774 : */
2775 6 : oldtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2776 : &oldtargettag,
2777 : oldtargettaghash,
2778 : HASH_FIND, NULL);
2779 :
2780 6 : if (oldtarget)
2781 : {
2782 : PREDICATELOCKTARGET *newtarget;
2783 : PREDICATELOCKTAG newpredlocktag;
2784 : dlist_mutable_iter iter;
2785 :
2786 0 : newtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2787 : &newtargettag,
2788 : newtargettaghash,
2789 : HASH_ENTER_NULL, &found);
2790 :
2791 0 : if (!newtarget)
2792 : {
2793 : /* Failed to allocate due to insufficient shmem */
2794 0 : outOfShmem = true;
2795 0 : goto exit;
2796 : }
2797 :
2798 : /* If we created a new entry, initialize it */
2799 0 : if (!found)
2800 0 : dlist_init(&newtarget->predicateLocks);
2801 :
2802 0 : newpredlocktag.myTarget = newtarget;
2803 :
2804 : /*
2805 : * Loop through all the locks on the old target, replacing them with
2806 : * locks on the new target.
2807 : */
2808 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2809 :
2810 0 : dlist_foreach_modify(iter, &oldtarget->predicateLocks)
2811 : {
2812 0 : PREDICATELOCK *oldpredlock =
2813 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
2814 : PREDICATELOCK *newpredlock;
2815 0 : SerCommitSeqNo oldCommitSeqNo = oldpredlock->commitSeqNo;
2816 :
2817 0 : newpredlocktag.myXact = oldpredlock->tag.myXact;
2818 :
2819 0 : if (removeOld)
2820 : {
2821 0 : dlist_delete(&(oldpredlock->xactLink));
2822 0 : dlist_delete(&(oldpredlock->targetLink));
2823 :
2824 0 : hash_search_with_hash_value
2825 : (PredicateLockHash,
2826 0 : &oldpredlock->tag,
2827 0 : PredicateLockHashCodeFromTargetHashCode(&oldpredlock->tag,
2828 : oldtargettaghash),
2829 : HASH_REMOVE, &found);
2830 : Assert(found);
2831 : }
2832 :
2833 : newpredlock = (PREDICATELOCK *)
2834 0 : hash_search_with_hash_value(PredicateLockHash,
2835 : &newpredlocktag,
2836 0 : PredicateLockHashCodeFromTargetHashCode(&newpredlocktag,
2837 : newtargettaghash),
2838 : HASH_ENTER_NULL,
2839 : &found);
2840 0 : if (!newpredlock)
2841 : {
2842 : /* Out of shared memory. Undo what we've done so far. */
2843 0 : LWLockRelease(SerializableXactHashLock);
2844 0 : DeleteLockTarget(newtarget, newtargettaghash);
2845 0 : outOfShmem = true;
2846 0 : goto exit;
2847 : }
2848 0 : if (!found)
2849 : {
2850 0 : dlist_push_tail(&(newtarget->predicateLocks),
2851 : &(newpredlock->targetLink));
2852 0 : dlist_push_tail(&(newpredlocktag.myXact->predicateLocks),
2853 : &(newpredlock->xactLink));
2854 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
2855 : }
2856 : else
2857 : {
2858 0 : if (newpredlock->commitSeqNo < oldCommitSeqNo)
2859 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
2860 : }
2861 :
2862 : Assert(newpredlock->commitSeqNo != 0);
2863 : Assert((newpredlock->commitSeqNo == InvalidSerCommitSeqNo)
2864 : || (newpredlock->tag.myXact == OldCommittedSxact));
2865 : }
2866 0 : LWLockRelease(SerializableXactHashLock);
2867 :
2868 0 : if (removeOld)
2869 : {
2870 : Assert(dlist_is_empty(&oldtarget->predicateLocks));
2871 0 : RemoveTargetIfNoLongerUsed(oldtarget, oldtargettaghash);
2872 : }
2873 : }
2874 :
2875 :
2876 6 : exit:
2877 : /* Release partition locks in reverse order of acquisition. */
2878 6 : if (oldpartitionLock < newpartitionLock)
2879 : {
2880 4 : LWLockRelease(newpartitionLock);
2881 4 : LWLockRelease(oldpartitionLock);
2882 : }
2883 2 : else if (oldpartitionLock > newpartitionLock)
2884 : {
2885 2 : LWLockRelease(oldpartitionLock);
2886 2 : LWLockRelease(newpartitionLock);
2887 : }
2888 : else
2889 0 : LWLockRelease(newpartitionLock);
2890 :
2891 6 : if (removeOld)
2892 : {
2893 : /* We shouldn't run out of memory if we're moving locks */
2894 : Assert(!outOfShmem);
2895 :
2896 : /* Put the scratch entry back */
2897 0 : RestoreScratchTarget(false);
2898 : }
2899 :
2900 6 : return !outOfShmem;
2901 : }
2902 :
2903 : /*
2904 : * Drop all predicate locks of any granularity from the specified relation,
2905 : * which can be a heap relation or an index relation. If 'transfer' is true,
2906 : * acquire a relation lock on the heap for any transactions with any lock(s)
2907 : * on the specified relation.
2908 : *
2909 : * This requires grabbing a lot of LW locks and scanning the entire lock
2910 : * target table for matches. That makes this more expensive than most
2911 : * predicate lock management functions, but it will only be called for DDL
2912 : * type commands that are expensive anyway, and there are fast returns when
2913 : * no serializable transactions are active or the relation is temporary.
2914 : *
2915 : * We don't use the TransferPredicateLocksToNewTarget function because it
2916 : * acquires its own locks on the partitions of the two targets involved,
2917 : * and we'll already be holding all partition locks.
2918 : *
2919 : * We can't throw an error from here, because the call could be from a
2920 : * transaction which is not serializable.
2921 : *
2922 : * NOTE: This is currently only called with transfer set to true, but that may
2923 : * change. If we decide to clean up the locks from a table on commit of a
2924 : * transaction which executed DROP TABLE, the false condition will be useful.
2925 : */
2926 : static void
2927 33322 : DropAllPredicateLocksFromTable(Relation relation, bool transfer)
2928 : {
2929 : HASH_SEQ_STATUS seqstat;
2930 : PREDICATELOCKTARGET *oldtarget;
2931 : PREDICATELOCKTARGET *heaptarget;
2932 : Oid dbId;
2933 : Oid relId;
2934 : Oid heapId;
2935 : int i;
2936 : bool isIndex;
2937 : bool found;
2938 : uint32 heaptargettaghash;
2939 :
2940 : /*
2941 : * Bail out quickly if there are no serializable transactions running.
2942 : * It's safe to check this without taking locks because the caller is
2943 : * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
2944 : * would matter here can be acquired while that is held.
2945 : */
2946 33322 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
2947 33228 : return;
2948 :
2949 222 : if (!PredicateLockingNeededForRelation(relation))
2950 128 : return;
2951 :
2952 94 : dbId = relation->rd_locator.dbOid;
2953 94 : relId = relation->rd_id;
2954 94 : if (relation->rd_index == NULL)
2955 : {
2956 4 : isIndex = false;
2957 4 : heapId = relId;
2958 : }
2959 : else
2960 : {
2961 90 : isIndex = true;
2962 90 : heapId = relation->rd_index->indrelid;
2963 : }
2964 : Assert(heapId != InvalidOid);
2965 : Assert(transfer || !isIndex); /* index OID only makes sense with
2966 : * transfer */
2967 :
2968 : /* Retrieve first time needed, then keep. */
2969 94 : heaptargettaghash = 0;
2970 94 : heaptarget = NULL;
2971 :
2972 : /* Acquire locks on all lock partitions */
2973 94 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
2974 1598 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
2975 1504 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_EXCLUSIVE);
2976 94 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2977 :
2978 : /*
2979 : * Remove the dummy entry to give us scratch space, so we know we'll be
2980 : * able to create the new lock target.
2981 : */
2982 94 : if (transfer)
2983 94 : RemoveScratchTarget(true);
2984 :
2985 : /* Scan through target map */
2986 94 : hash_seq_init(&seqstat, PredicateLockTargetHash);
2987 :
2988 230 : while ((oldtarget = (PREDICATELOCKTARGET *) hash_seq_search(&seqstat)))
2989 : {
2990 : dlist_mutable_iter iter;
2991 :
2992 : /*
2993 : * Check whether this is a target which needs attention.
2994 : */
2995 136 : if (GET_PREDICATELOCKTARGETTAG_RELATION(oldtarget->tag) != relId)
2996 136 : continue; /* wrong relation id */
2997 0 : if (GET_PREDICATELOCKTARGETTAG_DB(oldtarget->tag) != dbId)
2998 0 : continue; /* wrong database id */
2999 0 : if (transfer && !isIndex
3000 0 : && GET_PREDICATELOCKTARGETTAG_TYPE(oldtarget->tag) == PREDLOCKTAG_RELATION)
3001 0 : continue; /* already the right lock */
3002 :
3003 : /*
3004 : * If we made it here, we have work to do. We make sure the heap
3005 : * relation lock exists, then we walk the list of predicate locks for
3006 : * the old target we found, moving all locks to the heap relation lock
3007 : * -- unless they already hold that.
3008 : */
3009 :
3010 : /*
3011 : * First make sure we have the heap relation target. We only need to
3012 : * do this once.
3013 : */
3014 0 : if (transfer && heaptarget == NULL)
3015 : {
3016 : PREDICATELOCKTARGETTAG heaptargettag;
3017 :
3018 0 : SET_PREDICATELOCKTARGETTAG_RELATION(heaptargettag, dbId, heapId);
3019 0 : heaptargettaghash = PredicateLockTargetTagHashCode(&heaptargettag);
3020 0 : heaptarget = hash_search_with_hash_value(PredicateLockTargetHash,
3021 : &heaptargettag,
3022 : heaptargettaghash,
3023 : HASH_ENTER, &found);
3024 0 : if (!found)
3025 0 : dlist_init(&heaptarget->predicateLocks);
3026 : }
3027 :
3028 : /*
3029 : * Loop through all the locks on the old target, replacing them with
3030 : * locks on the new target.
3031 : */
3032 0 : dlist_foreach_modify(iter, &oldtarget->predicateLocks)
3033 : {
3034 0 : PREDICATELOCK *oldpredlock =
3035 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
3036 : PREDICATELOCK *newpredlock;
3037 : SerCommitSeqNo oldCommitSeqNo;
3038 : SERIALIZABLEXACT *oldXact;
3039 :
3040 : /*
3041 : * Remove the old lock first. This avoids the chance of running
3042 : * out of lock structure entries for the hash table.
3043 : */
3044 0 : oldCommitSeqNo = oldpredlock->commitSeqNo;
3045 0 : oldXact = oldpredlock->tag.myXact;
3046 :
3047 0 : dlist_delete(&(oldpredlock->xactLink));
3048 :
3049 : /*
3050 : * No need for retail delete from oldtarget list, we're removing
3051 : * the whole target anyway.
3052 : */
3053 0 : hash_search(PredicateLockHash,
3054 0 : &oldpredlock->tag,
3055 : HASH_REMOVE, &found);
3056 : Assert(found);
3057 :
3058 0 : if (transfer)
3059 : {
3060 : PREDICATELOCKTAG newpredlocktag;
3061 :
3062 0 : newpredlocktag.myTarget = heaptarget;
3063 0 : newpredlocktag.myXact = oldXact;
3064 : newpredlock = (PREDICATELOCK *)
3065 0 : hash_search_with_hash_value(PredicateLockHash,
3066 : &newpredlocktag,
3067 0 : PredicateLockHashCodeFromTargetHashCode(&newpredlocktag,
3068 : heaptargettaghash),
3069 : HASH_ENTER,
3070 : &found);
3071 0 : if (!found)
3072 : {
3073 0 : dlist_push_tail(&(heaptarget->predicateLocks),
3074 : &(newpredlock->targetLink));
3075 0 : dlist_push_tail(&(newpredlocktag.myXact->predicateLocks),
3076 : &(newpredlock->xactLink));
3077 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
3078 : }
3079 : else
3080 : {
3081 0 : if (newpredlock->commitSeqNo < oldCommitSeqNo)
3082 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
3083 : }
3084 :
3085 : Assert(newpredlock->commitSeqNo != 0);
3086 : Assert((newpredlock->commitSeqNo == InvalidSerCommitSeqNo)
3087 : || (newpredlock->tag.myXact == OldCommittedSxact));
3088 : }
3089 : }
3090 :
3091 0 : hash_search(PredicateLockTargetHash, &oldtarget->tag, HASH_REMOVE,
3092 : &found);
3093 : Assert(found);
3094 : }
3095 :
3096 : /* Put the scratch entry back */
3097 94 : if (transfer)
3098 94 : RestoreScratchTarget(true);
3099 :
3100 : /* Release locks in reverse order */
3101 94 : LWLockRelease(SerializableXactHashLock);
3102 1598 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
3103 1504 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
3104 94 : LWLockRelease(SerializablePredicateListLock);
3105 : }
3106 :
3107 : /*
3108 : * TransferPredicateLocksToHeapRelation
3109 : * For all transactions, transfer all predicate locks for the given
3110 : * relation to a single relation lock on the heap.
3111 : */
3112 : void
3113 33322 : TransferPredicateLocksToHeapRelation(Relation relation)
3114 : {
3115 33322 : DropAllPredicateLocksFromTable(relation, true);
3116 33322 : }
3117 :
3118 :
3119 : /*
3120 : * PredicateLockPageSplit
3121 : *
3122 : * Copies any predicate locks for the old page to the new page.
3123 : * Skip if this is a temporary table or toast table.
3124 : *
3125 : * NOTE: A page split (or overflow) affects all serializable transactions,
3126 : * even if it occurs in the context of another transaction isolation level.
3127 : *
3128 : * NOTE: This currently leaves the local copy of the locks without
3129 : * information on the new lock which is in shared memory. This could cause
3130 : * problems if enough page splits occur on locked pages without the processes
3131 : * which hold the locks getting in and noticing.
3132 : */
3133 : void
3134 59654 : PredicateLockPageSplit(Relation relation, BlockNumber oldblkno,
3135 : BlockNumber newblkno)
3136 : {
3137 : PREDICATELOCKTARGETTAG oldtargettag;
3138 : PREDICATELOCKTARGETTAG newtargettag;
3139 : bool success;
3140 :
3141 : /*
3142 : * Bail out quickly if there are no serializable transactions running.
3143 : *
3144 : * It's safe to do this check without taking any additional locks. Even if
3145 : * a serializable transaction starts concurrently, we know it can't take
3146 : * any SIREAD locks on the page being split because the caller is holding
3147 : * the associated buffer page lock. Memory reordering isn't an issue; the
3148 : * memory barrier in the LWLock acquisition guarantees that this read
3149 : * occurs while the buffer page lock is held.
3150 : */
3151 59654 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
3152 59648 : return;
3153 :
3154 54 : if (!PredicateLockingNeededForRelation(relation))
3155 48 : return;
3156 :
3157 : Assert(oldblkno != newblkno);
3158 : Assert(BlockNumberIsValid(oldblkno));
3159 : Assert(BlockNumberIsValid(newblkno));
3160 :
3161 6 : SET_PREDICATELOCKTARGETTAG_PAGE(oldtargettag,
3162 : relation->rd_locator.dbOid,
3163 : relation->rd_id,
3164 : oldblkno);
3165 6 : SET_PREDICATELOCKTARGETTAG_PAGE(newtargettag,
3166 : relation->rd_locator.dbOid,
3167 : relation->rd_id,
3168 : newblkno);
3169 :
3170 6 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
3171 :
3172 : /*
3173 : * Try copying the locks over to the new page's tag, creating it if
3174 : * necessary.
3175 : */
3176 6 : success = TransferPredicateLocksToNewTarget(oldtargettag,
3177 : newtargettag,
3178 : false);
3179 :
3180 6 : if (!success)
3181 : {
3182 : /*
3183 : * No more predicate lock entries are available. Failure isn't an
3184 : * option here, so promote the page lock to a relation lock.
3185 : */
3186 :
3187 : /* Get the parent relation lock's lock tag */
3188 0 : success = GetParentPredicateLockTag(&oldtargettag,
3189 : &newtargettag);
3190 : Assert(success);
3191 :
3192 : /*
3193 : * Move the locks to the parent. This shouldn't fail.
3194 : *
3195 : * Note that here we are removing locks held by other backends,
3196 : * leading to a possible inconsistency in their local lock hash table.
3197 : * This is OK because we're replacing it with a lock that covers the
3198 : * old one.
3199 : */
3200 0 : success = TransferPredicateLocksToNewTarget(oldtargettag,
3201 : newtargettag,
3202 : true);
3203 : Assert(success);
3204 : }
3205 :
3206 6 : LWLockRelease(SerializablePredicateListLock);
3207 : }
3208 :
3209 : /*
3210 : * PredicateLockPageCombine
3211 : *
3212 : * Combines predicate locks for two existing pages.
3213 : * Skip if this is a temporary table or toast table.
3214 : *
3215 : * NOTE: A page combine affects all serializable transactions, even if it
3216 : * occurs in the context of another transaction isolation level.
3217 : */
3218 : void
3219 5386 : PredicateLockPageCombine(Relation relation, BlockNumber oldblkno,
3220 : BlockNumber newblkno)
3221 : {
3222 : /*
3223 : * Page combines differ from page splits in that we ought to be able to
3224 : * remove the locks on the old page after transferring them to the new
3225 : * page, instead of duplicating them. However, because we can't edit other
3226 : * backends' local lock tables, removing the old lock would leave them
3227 : * with an entry in their LocalPredicateLockHash for a lock they're not
3228 : * holding, which isn't acceptable. So we wind up having to do the same
3229 : * work as a page split, acquiring a lock on the new page and keeping the
3230 : * old page locked too. That can lead to some false positives, but should
3231 : * be rare in practice.
3232 : */
3233 5386 : PredicateLockPageSplit(relation, oldblkno, newblkno);
3234 5386 : }
3235 :
3236 : /*
3237 : * Walk the list of in-progress serializable transactions and find the new
3238 : * xmin.
3239 : */
3240 : static void
3241 1688 : SetNewSxactGlobalXmin(void)
3242 : {
3243 : dlist_iter iter;
3244 :
3245 : Assert(LWLockHeldByMe(SerializableXactHashLock));
3246 :
3247 1688 : PredXact->SxactGlobalXmin = InvalidTransactionId;
3248 1688 : PredXact->SxactGlobalXminCount = 0;
3249 :
3250 6468 : dlist_foreach(iter, &PredXact->activeList)
3251 : {
3252 4780 : SERIALIZABLEXACT *sxact =
3253 4780 : dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
3254 :
3255 4780 : if (!SxactIsRolledBack(sxact)
3256 4184 : && !SxactIsCommitted(sxact)
3257 36 : && sxact != OldCommittedSxact)
3258 : {
3259 : Assert(sxact->xmin != InvalidTransactionId);
3260 36 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)
3261 0 : || TransactionIdPrecedes(sxact->xmin,
3262 0 : PredXact->SxactGlobalXmin))
3263 : {
3264 36 : PredXact->SxactGlobalXmin = sxact->xmin;
3265 36 : PredXact->SxactGlobalXminCount = 1;
3266 : }
3267 0 : else if (TransactionIdEquals(sxact->xmin,
3268 : PredXact->SxactGlobalXmin))
3269 0 : PredXact->SxactGlobalXminCount++;
3270 : }
3271 : }
3272 :
3273 1688 : SerialSetActiveSerXmin(PredXact->SxactGlobalXmin);
3274 1688 : }
3275 :
3276 : /*
3277 : * ReleasePredicateLocks
3278 : *
3279 : * Releases predicate locks based on completion of the current transaction,
3280 : * whether committed or rolled back. It can also be called for a read only
3281 : * transaction when it becomes impossible for the transaction to become
3282 : * part of a dangerous structure.
3283 : *
3284 : * We do nothing unless this is a serializable transaction.
3285 : *
3286 : * This method must ensure that shared memory hash tables are cleaned
3287 : * up in some relatively timely fashion.
3288 : *
3289 : * If this transaction is committing and is holding any predicate locks,
3290 : * it must be added to a list of completed serializable transactions still
3291 : * holding locks.
3292 : *
3293 : * If isReadOnlySafe is true, then predicate locks are being released before
3294 : * the end of the transaction because MySerializableXact has been determined
3295 : * to be RO_SAFE. In non-parallel mode we can release it completely, but it
3296 : * in parallel mode we partially release the SERIALIZABLEXACT and keep it
3297 : * around until the end of the transaction, allowing each backend to clear its
3298 : * MySerializableXact variable and benefit from the optimization in its own
3299 : * time.
3300 : */
3301 : void
3302 818164 : ReleasePredicateLocks(bool isCommit, bool isReadOnlySafe)
3303 : {
3304 818164 : bool partiallyReleasing = false;
3305 : bool needToClear;
3306 : SERIALIZABLEXACT *roXact;
3307 : dlist_mutable_iter iter;
3308 :
3309 : /*
3310 : * We can't trust XactReadOnly here, because a transaction which started
3311 : * as READ WRITE can show as READ ONLY later, e.g., within
3312 : * subtransactions. We want to flag a transaction as READ ONLY if it
3313 : * commits without writing so that de facto READ ONLY transactions get the
3314 : * benefit of some RO optimizations, so we will use this local variable to
3315 : * get some cleanup logic right which is based on whether the transaction
3316 : * was declared READ ONLY at the top level.
3317 : */
3318 : bool topLevelIsDeclaredReadOnly;
3319 :
3320 : /* We can't be both committing and releasing early due to RO_SAFE. */
3321 : Assert(!(isCommit && isReadOnlySafe));
3322 :
3323 : /* Are we at the end of a transaction, that is, a commit or abort? */
3324 818164 : if (!isReadOnlySafe)
3325 : {
3326 : /*
3327 : * Parallel workers mustn't release predicate locks at the end of
3328 : * their transaction. The leader will do that at the end of its
3329 : * transaction.
3330 : */
3331 818096 : if (IsParallelWorker())
3332 : {
3333 8208 : ReleasePredicateLocksLocal();
3334 815100 : return;
3335 : }
3336 :
3337 : /*
3338 : * By the time the leader in a parallel query reaches end of
3339 : * transaction, it has waited for all workers to exit.
3340 : */
3341 : Assert(!ParallelContextActive());
3342 :
3343 : /*
3344 : * If the leader in a parallel query earlier stashed a partially
3345 : * released SERIALIZABLEXACT for final clean-up at end of transaction
3346 : * (because workers might still have been accessing it), then it's
3347 : * time to restore it.
3348 : */
3349 809888 : if (SavedSerializableXact != InvalidSerializableXact)
3350 : {
3351 : Assert(MySerializableXact == InvalidSerializableXact);
3352 2 : MySerializableXact = SavedSerializableXact;
3353 2 : SavedSerializableXact = InvalidSerializableXact;
3354 : Assert(SxactIsPartiallyReleased(MySerializableXact));
3355 : }
3356 : }
3357 :
3358 809956 : if (MySerializableXact == InvalidSerializableXact)
3359 : {
3360 : Assert(LocalPredicateLockHash == NULL);
3361 806886 : return;
3362 : }
3363 :
3364 3070 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
3365 :
3366 : /*
3367 : * If the transaction is committing, but it has been partially released
3368 : * already, then treat this as a roll back. It was marked as rolled back.
3369 : */
3370 3070 : if (isCommit && SxactIsPartiallyReleased(MySerializableXact))
3371 4 : isCommit = false;
3372 :
3373 : /*
3374 : * If we're called in the middle of a transaction because we discovered
3375 : * that the SXACT_FLAG_RO_SAFE flag was set, then we'll partially release
3376 : * it (that is, release the predicate locks and conflicts, but not the
3377 : * SERIALIZABLEXACT itself) if we're the first backend to have noticed.
3378 : */
3379 3070 : if (isReadOnlySafe && IsInParallelMode())
3380 : {
3381 : /*
3382 : * The leader needs to stash a pointer to it, so that it can
3383 : * completely release it at end-of-transaction.
3384 : */
3385 10 : if (!IsParallelWorker())
3386 2 : SavedSerializableXact = MySerializableXact;
3387 :
3388 : /*
3389 : * The first backend to reach this condition will partially release
3390 : * the SERIALIZABLEXACT. All others will just clear their
3391 : * backend-local state so that they stop doing SSI checks for the rest
3392 : * of the transaction.
3393 : */
3394 10 : if (SxactIsPartiallyReleased(MySerializableXact))
3395 : {
3396 6 : LWLockRelease(SerializableXactHashLock);
3397 6 : ReleasePredicateLocksLocal();
3398 6 : return;
3399 : }
3400 : else
3401 : {
3402 4 : MySerializableXact->flags |= SXACT_FLAG_PARTIALLY_RELEASED;
3403 4 : partiallyReleasing = true;
3404 : /* ... and proceed to perform the partial release below. */
3405 : }
3406 : }
3407 : Assert(!isCommit || SxactIsPrepared(MySerializableXact));
3408 : Assert(!isCommit || !SxactIsDoomed(MySerializableXact));
3409 : Assert(!SxactIsCommitted(MySerializableXact));
3410 : Assert(SxactIsPartiallyReleased(MySerializableXact)
3411 : || !SxactIsRolledBack(MySerializableXact));
3412 :
3413 : /* may not be serializable during COMMIT/ROLLBACK PREPARED */
3414 : Assert(MySerializableXact->pid == 0 || IsolationIsSerializable());
3415 :
3416 : /* We'd better not already be on the cleanup list. */
3417 : Assert(!SxactIsOnFinishedList(MySerializableXact));
3418 :
3419 3064 : topLevelIsDeclaredReadOnly = SxactIsReadOnly(MySerializableXact);
3420 :
3421 : /*
3422 : * We don't hold XidGenLock lock here, assuming that TransactionId is
3423 : * atomic!
3424 : *
3425 : * If this value is changing, we don't care that much whether we get the
3426 : * old or new value -- it is just used to determine how far
3427 : * SxactGlobalXmin must advance before this transaction can be fully
3428 : * cleaned up. The worst that could happen is we wait for one more
3429 : * transaction to complete before freeing some RAM; correctness of visible
3430 : * behavior is not affected.
3431 : */
3432 3064 : MySerializableXact->finishedBefore = XidFromFullTransactionId(TransamVariables->nextXid);
3433 :
3434 : /*
3435 : * If it's not a commit it's either a rollback or a read-only transaction
3436 : * flagged SXACT_FLAG_RO_SAFE, and we can clear our locks immediately.
3437 : */
3438 3064 : if (isCommit)
3439 : {
3440 2420 : MySerializableXact->flags |= SXACT_FLAG_COMMITTED;
3441 2420 : MySerializableXact->commitSeqNo = ++(PredXact->LastSxactCommitSeqNo);
3442 : /* Recognize implicit read-only transaction (commit without write). */
3443 2420 : if (!MyXactDidWrite)
3444 466 : MySerializableXact->flags |= SXACT_FLAG_READ_ONLY;
3445 : }
3446 : else
3447 : {
3448 : /*
3449 : * The DOOMED flag indicates that we intend to roll back this
3450 : * transaction and so it should not cause serialization failures for
3451 : * other transactions that conflict with it. Note that this flag might
3452 : * already be set, if another backend marked this transaction for
3453 : * abort.
3454 : *
3455 : * The ROLLED_BACK flag further indicates that ReleasePredicateLocks
3456 : * has been called, and so the SerializableXact is eligible for
3457 : * cleanup. This means it should not be considered when calculating
3458 : * SxactGlobalXmin.
3459 : */
3460 644 : MySerializableXact->flags |= SXACT_FLAG_DOOMED;
3461 644 : MySerializableXact->flags |= SXACT_FLAG_ROLLED_BACK;
3462 :
3463 : /*
3464 : * If the transaction was previously prepared, but is now failing due
3465 : * to a ROLLBACK PREPARED or (hopefully very rare) error after the
3466 : * prepare, clear the prepared flag. This simplifies conflict
3467 : * checking.
3468 : */
3469 644 : MySerializableXact->flags &= ~SXACT_FLAG_PREPARED;
3470 : }
3471 :
3472 3064 : if (!topLevelIsDeclaredReadOnly)
3473 : {
3474 : Assert(PredXact->WritableSxactCount > 0);
3475 2846 : if (--(PredXact->WritableSxactCount) == 0)
3476 : {
3477 : /*
3478 : * Release predicate locks and rw-conflicts in for all committed
3479 : * transactions. There are no longer any transactions which might
3480 : * conflict with the locks and no chance for new transactions to
3481 : * overlap. Similarly, existing conflicts in can't cause pivots,
3482 : * and any conflicts in which could have completed a dangerous
3483 : * structure would already have caused a rollback, so any
3484 : * remaining ones must be benign.
3485 : */
3486 1672 : PredXact->CanPartialClearThrough = PredXact->LastSxactCommitSeqNo;
3487 : }
3488 : }
3489 : else
3490 : {
3491 : /*
3492 : * Read-only transactions: clear the list of transactions that might
3493 : * make us unsafe. Note that we use 'inLink' for the iteration as
3494 : * opposed to 'outLink' for the r/w xacts.
3495 : */
3496 302 : dlist_foreach_modify(iter, &MySerializableXact->possibleUnsafeConflicts)
3497 : {
3498 84 : RWConflict possibleUnsafeConflict =
3499 84 : dlist_container(RWConflictData, inLink, iter.cur);
3500 :
3501 : Assert(!SxactIsReadOnly(possibleUnsafeConflict->sxactOut));
3502 : Assert(MySerializableXact == possibleUnsafeConflict->sxactIn);
3503 :
3504 84 : ReleaseRWConflict(possibleUnsafeConflict);
3505 : }
3506 : }
3507 :
3508 : /* Check for conflict out to old committed transactions. */
3509 3064 : if (isCommit
3510 2420 : && !SxactIsReadOnly(MySerializableXact)
3511 1954 : && SxactHasSummaryConflictOut(MySerializableXact))
3512 : {
3513 : /*
3514 : * we don't know which old committed transaction we conflicted with,
3515 : * so be conservative and use FirstNormalSerCommitSeqNo here
3516 : */
3517 0 : MySerializableXact->SeqNo.earliestOutConflictCommit =
3518 : FirstNormalSerCommitSeqNo;
3519 0 : MySerializableXact->flags |= SXACT_FLAG_CONFLICT_OUT;
3520 : }
3521 :
3522 : /*
3523 : * Release all outConflicts to committed transactions. If we're rolling
3524 : * back clear them all. Set SXACT_FLAG_CONFLICT_OUT if any point to
3525 : * previously committed transactions.
3526 : */
3527 4422 : dlist_foreach_modify(iter, &MySerializableXact->outConflicts)
3528 : {
3529 1358 : RWConflict conflict =
3530 1358 : dlist_container(RWConflictData, outLink, iter.cur);
3531 :
3532 1358 : if (isCommit
3533 902 : && !SxactIsReadOnly(MySerializableXact)
3534 686 : && SxactIsCommitted(conflict->sxactIn))
3535 : {
3536 192 : if ((MySerializableXact->flags & SXACT_FLAG_CONFLICT_OUT) == 0
3537 0 : || conflict->sxactIn->prepareSeqNo < MySerializableXact->SeqNo.earliestOutConflictCommit)
3538 192 : MySerializableXact->SeqNo.earliestOutConflictCommit = conflict->sxactIn->prepareSeqNo;
3539 192 : MySerializableXact->flags |= SXACT_FLAG_CONFLICT_OUT;
3540 : }
3541 :
3542 1358 : if (!isCommit
3543 902 : || SxactIsCommitted(conflict->sxactIn)
3544 666 : || (conflict->sxactIn->SeqNo.lastCommitBeforeSnapshot >= PredXact->LastSxactCommitSeqNo))
3545 692 : ReleaseRWConflict(conflict);
3546 : }
3547 :
3548 : /*
3549 : * Release all inConflicts from committed and read-only transactions. If
3550 : * we're rolling back, clear them all.
3551 : */
3552 4602 : dlist_foreach_modify(iter, &MySerializableXact->inConflicts)
3553 : {
3554 1538 : RWConflict conflict =
3555 1538 : dlist_container(RWConflictData, inLink, iter.cur);
3556 :
3557 1538 : if (!isCommit
3558 1198 : || SxactIsCommitted(conflict->sxactOut)
3559 830 : || SxactIsReadOnly(conflict->sxactOut))
3560 868 : ReleaseRWConflict(conflict);
3561 : }
3562 :
3563 3064 : if (!topLevelIsDeclaredReadOnly)
3564 : {
3565 : /*
3566 : * Remove ourselves from the list of possible conflicts for concurrent
3567 : * READ ONLY transactions, flagging them as unsafe if we have a
3568 : * conflict out. If any are waiting DEFERRABLE transactions, wake them
3569 : * up if they are known safe or known unsafe.
3570 : */
3571 3030 : dlist_foreach_modify(iter, &MySerializableXact->possibleUnsafeConflicts)
3572 : {
3573 184 : RWConflict possibleUnsafeConflict =
3574 184 : dlist_container(RWConflictData, outLink, iter.cur);
3575 :
3576 184 : roXact = possibleUnsafeConflict->sxactIn;
3577 : Assert(MySerializableXact == possibleUnsafeConflict->sxactOut);
3578 : Assert(SxactIsReadOnly(roXact));
3579 :
3580 : /* Mark conflicted if necessary. */
3581 184 : if (isCommit
3582 180 : && MyXactDidWrite
3583 170 : && SxactHasConflictOut(MySerializableXact)
3584 26 : && (MySerializableXact->SeqNo.earliestOutConflictCommit
3585 26 : <= roXact->SeqNo.lastCommitBeforeSnapshot))
3586 : {
3587 : /*
3588 : * This releases possibleUnsafeConflict (as well as all other
3589 : * possible conflicts for roXact)
3590 : */
3591 6 : FlagSxactUnsafe(roXact);
3592 : }
3593 : else
3594 : {
3595 178 : ReleaseRWConflict(possibleUnsafeConflict);
3596 :
3597 : /*
3598 : * If we were the last possible conflict, flag it safe. The
3599 : * transaction can now safely release its predicate locks (but
3600 : * that transaction's backend has to do that itself).
3601 : */
3602 178 : if (dlist_is_empty(&roXact->possibleUnsafeConflicts))
3603 132 : roXact->flags |= SXACT_FLAG_RO_SAFE;
3604 : }
3605 :
3606 : /*
3607 : * Wake up the process for a waiting DEFERRABLE transaction if we
3608 : * now know it's either safe or conflicted.
3609 : */
3610 184 : if (SxactIsDeferrableWaiting(roXact) &&
3611 6 : (SxactIsROUnsafe(roXact) || SxactIsROSafe(roXact)))
3612 4 : ProcSendSignal(roXact->pgprocno);
3613 : }
3614 : }
3615 :
3616 : /*
3617 : * Check whether it's time to clean up old transactions. This can only be
3618 : * done when the last serializable transaction with the oldest xmin among
3619 : * serializable transactions completes. We then find the "new oldest"
3620 : * xmin and purge any transactions which finished before this transaction
3621 : * was launched.
3622 : *
3623 : * For parallel queries in read-only transactions, it might run twice. We
3624 : * only release the reference on the first call.
3625 : */
3626 3064 : needToClear = false;
3627 3064 : if ((partiallyReleasing ||
3628 3060 : !SxactIsPartiallyReleased(MySerializableXact)) &&
3629 3060 : TransactionIdEquals(MySerializableXact->xmin,
3630 : PredXact->SxactGlobalXmin))
3631 : {
3632 : Assert(PredXact->SxactGlobalXminCount > 0);
3633 3024 : if (--(PredXact->SxactGlobalXminCount) == 0)
3634 : {
3635 1688 : SetNewSxactGlobalXmin();
3636 1688 : needToClear = true;
3637 : }
3638 : }
3639 :
3640 3064 : LWLockRelease(SerializableXactHashLock);
3641 :
3642 3064 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
3643 :
3644 : /* Add this to the list of transactions to check for later cleanup. */
3645 3064 : if (isCommit)
3646 2420 : dlist_push_tail(FinishedSerializableTransactions,
3647 2420 : &MySerializableXact->finishedLink);
3648 :
3649 : /*
3650 : * If we're releasing a RO_SAFE transaction in parallel mode, we'll only
3651 : * partially release it. That's necessary because other backends may have
3652 : * a reference to it. The leader will release the SERIALIZABLEXACT itself
3653 : * at the end of the transaction after workers have stopped running.
3654 : */
3655 3064 : if (!isCommit)
3656 644 : ReleaseOneSerializableXact(MySerializableXact,
3657 644 : isReadOnlySafe && IsInParallelMode(),
3658 : false);
3659 :
3660 3064 : LWLockRelease(SerializableFinishedListLock);
3661 :
3662 3064 : if (needToClear)
3663 1688 : ClearOldPredicateLocks();
3664 :
3665 3064 : ReleasePredicateLocksLocal();
3666 : }
3667 :
3668 : static void
3669 11278 : ReleasePredicateLocksLocal(void)
3670 : {
3671 11278 : MySerializableXact = InvalidSerializableXact;
3672 11278 : MyXactDidWrite = false;
3673 :
3674 : /* Delete per-transaction lock table */
3675 11278 : if (LocalPredicateLockHash != NULL)
3676 : {
3677 3062 : hash_destroy(LocalPredicateLockHash);
3678 3062 : LocalPredicateLockHash = NULL;
3679 : }
3680 11278 : }
3681 :
3682 : /*
3683 : * Clear old predicate locks, belonging to committed transactions that are no
3684 : * longer interesting to any in-progress transaction.
3685 : */
3686 : static void
3687 1688 : ClearOldPredicateLocks(void)
3688 : {
3689 : dlist_mutable_iter iter;
3690 :
3691 : /*
3692 : * Loop through finished transactions. They are in commit order, so we can
3693 : * stop as soon as we find one that's still interesting.
3694 : */
3695 1688 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
3696 1688 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3697 4124 : dlist_foreach_modify(iter, FinishedSerializableTransactions)
3698 : {
3699 2452 : SERIALIZABLEXACT *finishedSxact =
3700 2452 : dlist_container(SERIALIZABLEXACT, finishedLink, iter.cur);
3701 :
3702 2452 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)
3703 54 : || TransactionIdPrecedesOrEquals(finishedSxact->finishedBefore,
3704 54 : PredXact->SxactGlobalXmin))
3705 : {
3706 : /*
3707 : * This transaction committed before any in-progress transaction
3708 : * took its snapshot. It's no longer interesting.
3709 : */
3710 2420 : LWLockRelease(SerializableXactHashLock);
3711 2420 : dlist_delete_thoroughly(&finishedSxact->finishedLink);
3712 2420 : ReleaseOneSerializableXact(finishedSxact, false, false);
3713 2420 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3714 : }
3715 32 : else if (finishedSxact->commitSeqNo > PredXact->HavePartialClearedThrough
3716 32 : && finishedSxact->commitSeqNo <= PredXact->CanPartialClearThrough)
3717 : {
3718 : /*
3719 : * Any active transactions that took their snapshot before this
3720 : * transaction committed are read-only, so we can clear part of
3721 : * its state.
3722 : */
3723 16 : LWLockRelease(SerializableXactHashLock);
3724 :
3725 16 : if (SxactIsReadOnly(finishedSxact))
3726 : {
3727 : /* A read-only transaction can be removed entirely */
3728 0 : dlist_delete_thoroughly(&(finishedSxact->finishedLink));
3729 0 : ReleaseOneSerializableXact(finishedSxact, false, false);
3730 : }
3731 : else
3732 : {
3733 : /*
3734 : * A read-write transaction can only be partially cleared. We
3735 : * need to keep the SERIALIZABLEXACT but can release the
3736 : * SIREAD locks and conflicts in.
3737 : */
3738 16 : ReleaseOneSerializableXact(finishedSxact, true, false);
3739 : }
3740 :
3741 16 : PredXact->HavePartialClearedThrough = finishedSxact->commitSeqNo;
3742 16 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3743 : }
3744 : else
3745 : {
3746 : /* Still interesting. */
3747 : break;
3748 : }
3749 : }
3750 1688 : LWLockRelease(SerializableXactHashLock);
3751 :
3752 : /*
3753 : * Loop through predicate locks on dummy transaction for summarized data.
3754 : */
3755 1688 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
3756 1688 : dlist_foreach_modify(iter, &OldCommittedSxact->predicateLocks)
3757 : {
3758 0 : PREDICATELOCK *predlock =
3759 0 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
3760 : bool canDoPartialCleanup;
3761 :
3762 0 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3763 : Assert(predlock->commitSeqNo != 0);
3764 : Assert(predlock->commitSeqNo != InvalidSerCommitSeqNo);
3765 0 : canDoPartialCleanup = (predlock->commitSeqNo <= PredXact->CanPartialClearThrough);
3766 0 : LWLockRelease(SerializableXactHashLock);
3767 :
3768 : /*
3769 : * If this lock originally belonged to an old enough transaction, we
3770 : * can release it.
3771 : */
3772 0 : if (canDoPartialCleanup)
3773 : {
3774 : PREDICATELOCKTAG tag;
3775 : PREDICATELOCKTARGET *target;
3776 : PREDICATELOCKTARGETTAG targettag;
3777 : uint32 targettaghash;
3778 : LWLock *partitionLock;
3779 :
3780 0 : tag = predlock->tag;
3781 0 : target = tag.myTarget;
3782 0 : targettag = target->tag;
3783 0 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
3784 0 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
3785 :
3786 0 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
3787 :
3788 0 : dlist_delete(&(predlock->targetLink));
3789 0 : dlist_delete(&(predlock->xactLink));
3790 :
3791 0 : hash_search_with_hash_value(PredicateLockHash, &tag,
3792 0 : PredicateLockHashCodeFromTargetHashCode(&tag,
3793 : targettaghash),
3794 : HASH_REMOVE, NULL);
3795 0 : RemoveTargetIfNoLongerUsed(target, targettaghash);
3796 :
3797 0 : LWLockRelease(partitionLock);
3798 : }
3799 : }
3800 :
3801 1688 : LWLockRelease(SerializablePredicateListLock);
3802 1688 : LWLockRelease(SerializableFinishedListLock);
3803 1688 : }
3804 :
3805 : /*
3806 : * This is the normal way to delete anything from any of the predicate
3807 : * locking hash tables. Given a transaction which we know can be deleted:
3808 : * delete all predicate locks held by that transaction and any predicate
3809 : * lock targets which are now unreferenced by a lock; delete all conflicts
3810 : * for the transaction; delete all xid values for the transaction; then
3811 : * delete the transaction.
3812 : *
3813 : * When the partial flag is set, we can release all predicate locks and
3814 : * in-conflict information -- we've established that there are no longer
3815 : * any overlapping read write transactions for which this transaction could
3816 : * matter -- but keep the transaction entry itself and any outConflicts.
3817 : *
3818 : * When the summarize flag is set, we've run short of room for sxact data
3819 : * and must summarize to the SLRU. Predicate locks are transferred to a
3820 : * dummy "old" transaction, with duplicate locks on a single target
3821 : * collapsing to a single lock with the "latest" commitSeqNo from among
3822 : * the conflicting locks..
3823 : */
3824 : static void
3825 3080 : ReleaseOneSerializableXact(SERIALIZABLEXACT *sxact, bool partial,
3826 : bool summarize)
3827 : {
3828 : SERIALIZABLEXIDTAG sxidtag;
3829 : dlist_mutable_iter iter;
3830 :
3831 : Assert(sxact != NULL);
3832 : Assert(SxactIsRolledBack(sxact) || SxactIsCommitted(sxact));
3833 : Assert(partial || !SxactIsOnFinishedList(sxact));
3834 : Assert(LWLockHeldByMe(SerializableFinishedListLock));
3835 :
3836 : /*
3837 : * First release all the predicate locks held by this xact (or transfer
3838 : * them to OldCommittedSxact if summarize is true)
3839 : */
3840 3080 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
3841 3080 : if (IsInParallelMode())
3842 6 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
3843 8728 : dlist_foreach_modify(iter, &sxact->predicateLocks)
3844 : {
3845 5648 : PREDICATELOCK *predlock =
3846 5648 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
3847 : PREDICATELOCKTAG tag;
3848 : PREDICATELOCKTARGET *target;
3849 : PREDICATELOCKTARGETTAG targettag;
3850 : uint32 targettaghash;
3851 : LWLock *partitionLock;
3852 :
3853 5648 : tag = predlock->tag;
3854 5648 : target = tag.myTarget;
3855 5648 : targettag = target->tag;
3856 5648 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
3857 5648 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
3858 :
3859 5648 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
3860 :
3861 5648 : dlist_delete(&predlock->targetLink);
3862 :
3863 5648 : hash_search_with_hash_value(PredicateLockHash, &tag,
3864 5648 : PredicateLockHashCodeFromTargetHashCode(&tag,
3865 : targettaghash),
3866 : HASH_REMOVE, NULL);
3867 5648 : if (summarize)
3868 : {
3869 : bool found;
3870 :
3871 : /* Fold into dummy transaction list. */
3872 0 : tag.myXact = OldCommittedSxact;
3873 0 : predlock = hash_search_with_hash_value(PredicateLockHash, &tag,
3874 0 : PredicateLockHashCodeFromTargetHashCode(&tag,
3875 : targettaghash),
3876 : HASH_ENTER_NULL, &found);
3877 0 : if (!predlock)
3878 0 : ereport(ERROR,
3879 : (errcode(ERRCODE_OUT_OF_MEMORY),
3880 : errmsg("out of shared memory"),
3881 : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
3882 0 : if (found)
3883 : {
3884 : Assert(predlock->commitSeqNo != 0);
3885 : Assert(predlock->commitSeqNo != InvalidSerCommitSeqNo);
3886 0 : if (predlock->commitSeqNo < sxact->commitSeqNo)
3887 0 : predlock->commitSeqNo = sxact->commitSeqNo;
3888 : }
3889 : else
3890 : {
3891 0 : dlist_push_tail(&target->predicateLocks,
3892 : &predlock->targetLink);
3893 0 : dlist_push_tail(&OldCommittedSxact->predicateLocks,
3894 : &predlock->xactLink);
3895 0 : predlock->commitSeqNo = sxact->commitSeqNo;
3896 : }
3897 : }
3898 : else
3899 5648 : RemoveTargetIfNoLongerUsed(target, targettaghash);
3900 :
3901 5648 : LWLockRelease(partitionLock);
3902 : }
3903 :
3904 : /*
3905 : * Rather than retail removal, just re-init the head after we've run
3906 : * through the list.
3907 : */
3908 3080 : dlist_init(&sxact->predicateLocks);
3909 :
3910 3080 : if (IsInParallelMode())
3911 6 : LWLockRelease(&sxact->perXactPredicateListLock);
3912 3080 : LWLockRelease(SerializablePredicateListLock);
3913 :
3914 3080 : sxidtag.xid = sxact->topXid;
3915 3080 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
3916 :
3917 : /* Release all outConflicts (unless 'partial' is true) */
3918 3080 : if (!partial)
3919 : {
3920 3060 : dlist_foreach_modify(iter, &sxact->outConflicts)
3921 : {
3922 0 : RWConflict conflict =
3923 0 : dlist_container(RWConflictData, outLink, iter.cur);
3924 :
3925 0 : if (summarize)
3926 0 : conflict->sxactIn->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
3927 0 : ReleaseRWConflict(conflict);
3928 : }
3929 : }
3930 :
3931 : /* Release all inConflicts. */
3932 3080 : dlist_foreach_modify(iter, &sxact->inConflicts)
3933 : {
3934 0 : RWConflict conflict =
3935 0 : dlist_container(RWConflictData, inLink, iter.cur);
3936 :
3937 0 : if (summarize)
3938 0 : conflict->sxactOut->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
3939 0 : ReleaseRWConflict(conflict);
3940 : }
3941 :
3942 : /* Finally, get rid of the xid and the record of the transaction itself. */
3943 3080 : if (!partial)
3944 : {
3945 3060 : if (sxidtag.xid != InvalidTransactionId)
3946 2526 : hash_search(SerializableXidHash, &sxidtag, HASH_REMOVE, NULL);
3947 3060 : ReleasePredXact(sxact);
3948 : }
3949 :
3950 3080 : LWLockRelease(SerializableXactHashLock);
3951 3080 : }
3952 :
3953 : /*
3954 : * Tests whether the given top level transaction is concurrent with
3955 : * (overlaps) our current transaction.
3956 : *
3957 : * We need to identify the top level transaction for SSI, anyway, so pass
3958 : * that to this function to save the overhead of checking the snapshot's
3959 : * subxip array.
3960 : */
3961 : static bool
3962 1064 : XidIsConcurrent(TransactionId xid)
3963 : {
3964 : Snapshot snap;
3965 :
3966 : Assert(TransactionIdIsValid(xid));
3967 : Assert(!TransactionIdEquals(xid, GetTopTransactionIdIfAny()));
3968 :
3969 1064 : snap = GetTransactionSnapshot();
3970 :
3971 1064 : if (TransactionIdPrecedes(xid, snap->xmin))
3972 0 : return false;
3973 :
3974 1064 : if (TransactionIdFollowsOrEquals(xid, snap->xmax))
3975 1048 : return true;
3976 :
3977 16 : return pg_lfind32(xid, snap->xip, snap->xcnt);
3978 : }
3979 :
3980 : bool
3981 65346698 : CheckForSerializableConflictOutNeeded(Relation relation, Snapshot snapshot)
3982 : {
3983 65346698 : if (!SerializationNeededForRead(relation, snapshot))
3984 65294790 : return false;
3985 :
3986 : /* Check if someone else has already decided that we need to die */
3987 51908 : if (SxactIsDoomed(MySerializableXact))
3988 : {
3989 0 : ereport(ERROR,
3990 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
3991 : errmsg("could not serialize access due to read/write dependencies among transactions"),
3992 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
3993 : errhint("The transaction might succeed if retried.")));
3994 : }
3995 :
3996 51908 : return true;
3997 : }
3998 :
3999 : /*
4000 : * CheckForSerializableConflictOut
4001 : * A table AM is reading a tuple that has been modified. If it determines
4002 : * that the tuple version it is reading is not visible to us, it should
4003 : * pass in the top level xid of the transaction that created it.
4004 : * Otherwise, if it determines that it is visible to us but it has been
4005 : * deleted or there is a newer version available due to an update, it
4006 : * should pass in the top level xid of the modifying transaction.
4007 : *
4008 : * This function will check for overlap with our own transaction. If the given
4009 : * xid is also serializable and the transactions overlap (i.e., they cannot see
4010 : * each other's writes), then we have a conflict out.
4011 : */
4012 : void
4013 1134 : CheckForSerializableConflictOut(Relation relation, TransactionId xid, Snapshot snapshot)
4014 : {
4015 : SERIALIZABLEXIDTAG sxidtag;
4016 : SERIALIZABLEXID *sxid;
4017 : SERIALIZABLEXACT *sxact;
4018 :
4019 1134 : if (!SerializationNeededForRead(relation, snapshot))
4020 408 : return;
4021 :
4022 : /* Check if someone else has already decided that we need to die */
4023 1134 : if (SxactIsDoomed(MySerializableXact))
4024 : {
4025 0 : ereport(ERROR,
4026 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4027 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4028 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4029 : errhint("The transaction might succeed if retried.")));
4030 : }
4031 : Assert(TransactionIdIsValid(xid));
4032 :
4033 1134 : if (TransactionIdEquals(xid, GetTopTransactionIdIfAny()))
4034 0 : return;
4035 :
4036 : /*
4037 : * Find sxact or summarized info for the top level xid.
4038 : */
4039 1134 : sxidtag.xid = xid;
4040 1134 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4041 : sxid = (SERIALIZABLEXID *)
4042 1134 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
4043 1134 : if (!sxid)
4044 : {
4045 : /*
4046 : * Transaction not found in "normal" SSI structures. Check whether it
4047 : * got pushed out to SLRU storage for "old committed" transactions.
4048 : */
4049 : SerCommitSeqNo conflictCommitSeqNo;
4050 :
4051 50 : conflictCommitSeqNo = SerialGetMinConflictCommitSeqNo(xid);
4052 50 : if (conflictCommitSeqNo != 0)
4053 : {
4054 0 : if (conflictCommitSeqNo != InvalidSerCommitSeqNo
4055 0 : && (!SxactIsReadOnly(MySerializableXact)
4056 0 : || conflictCommitSeqNo
4057 0 : <= MySerializableXact->SeqNo.lastCommitBeforeSnapshot))
4058 0 : ereport(ERROR,
4059 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4060 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4061 : errdetail_internal("Reason code: Canceled on conflict out to old pivot %u.", xid),
4062 : errhint("The transaction might succeed if retried.")));
4063 :
4064 0 : if (SxactHasSummaryConflictIn(MySerializableXact)
4065 0 : || !dlist_is_empty(&MySerializableXact->inConflicts))
4066 0 : ereport(ERROR,
4067 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4068 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4069 : errdetail_internal("Reason code: Canceled on identification as a pivot, with conflict out to old committed transaction %u.", xid),
4070 : errhint("The transaction might succeed if retried.")));
4071 :
4072 0 : MySerializableXact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4073 : }
4074 :
4075 : /* It's not serializable or otherwise not important. */
4076 50 : LWLockRelease(SerializableXactHashLock);
4077 50 : return;
4078 : }
4079 1084 : sxact = sxid->myXact;
4080 : Assert(TransactionIdEquals(sxact->topXid, xid));
4081 1084 : if (sxact == MySerializableXact || SxactIsDoomed(sxact))
4082 : {
4083 : /* Can't conflict with ourself or a transaction that will roll back. */
4084 8 : LWLockRelease(SerializableXactHashLock);
4085 8 : return;
4086 : }
4087 :
4088 : /*
4089 : * We have a conflict out to a transaction which has a conflict out to a
4090 : * summarized transaction. That summarized transaction must have
4091 : * committed first, and we can't tell when it committed in relation to our
4092 : * snapshot acquisition, so something needs to be canceled.
4093 : */
4094 1076 : if (SxactHasSummaryConflictOut(sxact))
4095 : {
4096 0 : if (!SxactIsPrepared(sxact))
4097 : {
4098 0 : sxact->flags |= SXACT_FLAG_DOOMED;
4099 0 : LWLockRelease(SerializableXactHashLock);
4100 0 : return;
4101 : }
4102 : else
4103 : {
4104 0 : LWLockRelease(SerializableXactHashLock);
4105 0 : ereport(ERROR,
4106 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4107 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4108 : errdetail_internal("Reason code: Canceled on conflict out to old pivot."),
4109 : errhint("The transaction might succeed if retried.")));
4110 : }
4111 : }
4112 :
4113 : /*
4114 : * If this is a read-only transaction and the writing transaction has
4115 : * committed, and it doesn't have a rw-conflict to a transaction which
4116 : * committed before it, no conflict.
4117 : */
4118 1076 : if (SxactIsReadOnly(MySerializableXact)
4119 238 : && SxactIsCommitted(sxact)
4120 16 : && !SxactHasSummaryConflictOut(sxact)
4121 16 : && (!SxactHasConflictOut(sxact)
4122 4 : || MySerializableXact->SeqNo.lastCommitBeforeSnapshot < sxact->SeqNo.earliestOutConflictCommit))
4123 : {
4124 : /* Read-only transaction will appear to run first. No conflict. */
4125 12 : LWLockRelease(SerializableXactHashLock);
4126 12 : return;
4127 : }
4128 :
4129 1064 : if (!XidIsConcurrent(xid))
4130 : {
4131 : /* This write was already in our snapshot; no conflict. */
4132 0 : LWLockRelease(SerializableXactHashLock);
4133 0 : return;
4134 : }
4135 :
4136 1064 : if (RWConflictExists(MySerializableXact, sxact))
4137 : {
4138 : /* We don't want duplicate conflict records in the list. */
4139 338 : LWLockRelease(SerializableXactHashLock);
4140 338 : return;
4141 : }
4142 :
4143 : /*
4144 : * Flag the conflict. But first, if this conflict creates a dangerous
4145 : * structure, ereport an error.
4146 : */
4147 726 : FlagRWConflict(MySerializableXact, sxact);
4148 700 : LWLockRelease(SerializableXactHashLock);
4149 : }
4150 :
4151 : /*
4152 : * Check a particular target for rw-dependency conflict in. A subroutine of
4153 : * CheckForSerializableConflictIn().
4154 : */
4155 : static void
4156 14956 : CheckTargetForConflictsIn(PREDICATELOCKTARGETTAG *targettag)
4157 : {
4158 : uint32 targettaghash;
4159 : LWLock *partitionLock;
4160 : PREDICATELOCKTARGET *target;
4161 14956 : PREDICATELOCK *mypredlock = NULL;
4162 : PREDICATELOCKTAG mypredlocktag;
4163 : dlist_mutable_iter iter;
4164 :
4165 : Assert(MySerializableXact != InvalidSerializableXact);
4166 :
4167 : /*
4168 : * The same hash and LW lock apply to the lock target and the lock itself.
4169 : */
4170 14956 : targettaghash = PredicateLockTargetTagHashCode(targettag);
4171 14956 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
4172 14956 : LWLockAcquire(partitionLock, LW_SHARED);
4173 : target = (PREDICATELOCKTARGET *)
4174 14956 : hash_search_with_hash_value(PredicateLockTargetHash,
4175 : targettag, targettaghash,
4176 : HASH_FIND, NULL);
4177 14956 : if (!target)
4178 : {
4179 : /* Nothing has this target locked; we're done here. */
4180 11214 : LWLockRelease(partitionLock);
4181 11214 : return;
4182 : }
4183 :
4184 : /*
4185 : * Each lock for an overlapping transaction represents a conflict: a
4186 : * rw-dependency in to this transaction.
4187 : */
4188 3742 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4189 :
4190 8432 : dlist_foreach_modify(iter, &target->predicateLocks)
4191 : {
4192 4824 : PREDICATELOCK *predlock =
4193 4824 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
4194 4824 : SERIALIZABLEXACT *sxact = predlock->tag.myXact;
4195 :
4196 4824 : if (sxact == MySerializableXact)
4197 : {
4198 : /*
4199 : * If we're getting a write lock on a tuple, we don't need a
4200 : * predicate (SIREAD) lock on the same tuple. We can safely remove
4201 : * our SIREAD lock, but we'll defer doing so until after the loop
4202 : * because that requires upgrading to an exclusive partition lock.
4203 : *
4204 : * We can't use this optimization within a subtransaction because
4205 : * the subtransaction could roll back, and we would be left
4206 : * without any lock at the top level.
4207 : */
4208 3128 : if (!IsSubTransaction()
4209 3128 : && GET_PREDICATELOCKTARGETTAG_OFFSET(*targettag))
4210 : {
4211 776 : mypredlock = predlock;
4212 776 : mypredlocktag = predlock->tag;
4213 : }
4214 : }
4215 1696 : else if (!SxactIsDoomed(sxact)
4216 1696 : && (!SxactIsCommitted(sxact)
4217 166 : || TransactionIdPrecedes(GetTransactionSnapshot()->xmin,
4218 : sxact->finishedBefore))
4219 1678 : && !RWConflictExists(sxact, MySerializableXact))
4220 : {
4221 994 : LWLockRelease(SerializableXactHashLock);
4222 994 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4223 :
4224 : /*
4225 : * Re-check after getting exclusive lock because the other
4226 : * transaction may have flagged a conflict.
4227 : */
4228 994 : if (!SxactIsDoomed(sxact)
4229 994 : && (!SxactIsCommitted(sxact)
4230 148 : || TransactionIdPrecedes(GetTransactionSnapshot()->xmin,
4231 : sxact->finishedBefore))
4232 994 : && !RWConflictExists(sxact, MySerializableXact))
4233 : {
4234 994 : FlagRWConflict(sxact, MySerializableXact);
4235 : }
4236 :
4237 860 : LWLockRelease(SerializableXactHashLock);
4238 860 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4239 : }
4240 : }
4241 3608 : LWLockRelease(SerializableXactHashLock);
4242 3608 : LWLockRelease(partitionLock);
4243 :
4244 : /*
4245 : * If we found one of our own SIREAD locks to remove, remove it now.
4246 : *
4247 : * At this point our transaction already has a RowExclusiveLock on the
4248 : * relation, so we are OK to drop the predicate lock on the tuple, if
4249 : * found, without fearing that another write against the tuple will occur
4250 : * before the MVCC information makes it to the buffer.
4251 : */
4252 3608 : if (mypredlock != NULL)
4253 : {
4254 : uint32 predlockhashcode;
4255 : PREDICATELOCK *rmpredlock;
4256 :
4257 762 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
4258 762 : if (IsInParallelMode())
4259 0 : LWLockAcquire(&MySerializableXact->perXactPredicateListLock, LW_EXCLUSIVE);
4260 762 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
4261 762 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4262 :
4263 : /*
4264 : * Remove the predicate lock from shared memory, if it wasn't removed
4265 : * while the locks were released. One way that could happen is from
4266 : * autovacuum cleaning up an index.
4267 : */
4268 762 : predlockhashcode = PredicateLockHashCodeFromTargetHashCode
4269 : (&mypredlocktag, targettaghash);
4270 : rmpredlock = (PREDICATELOCK *)
4271 762 : hash_search_with_hash_value(PredicateLockHash,
4272 : &mypredlocktag,
4273 : predlockhashcode,
4274 : HASH_FIND, NULL);
4275 762 : if (rmpredlock != NULL)
4276 : {
4277 : Assert(rmpredlock == mypredlock);
4278 :
4279 762 : dlist_delete(&(mypredlock->targetLink));
4280 762 : dlist_delete(&(mypredlock->xactLink));
4281 :
4282 : rmpredlock = (PREDICATELOCK *)
4283 762 : hash_search_with_hash_value(PredicateLockHash,
4284 : &mypredlocktag,
4285 : predlockhashcode,
4286 : HASH_REMOVE, NULL);
4287 : Assert(rmpredlock == mypredlock);
4288 :
4289 762 : RemoveTargetIfNoLongerUsed(target, targettaghash);
4290 : }
4291 :
4292 762 : LWLockRelease(SerializableXactHashLock);
4293 762 : LWLockRelease(partitionLock);
4294 762 : if (IsInParallelMode())
4295 0 : LWLockRelease(&MySerializableXact->perXactPredicateListLock);
4296 762 : LWLockRelease(SerializablePredicateListLock);
4297 :
4298 762 : if (rmpredlock != NULL)
4299 : {
4300 : /*
4301 : * Remove entry in local lock table if it exists. It's OK if it
4302 : * doesn't exist; that means the lock was transferred to a new
4303 : * target by a different backend.
4304 : */
4305 762 : hash_search_with_hash_value(LocalPredicateLockHash,
4306 : targettag, targettaghash,
4307 : HASH_REMOVE, NULL);
4308 :
4309 762 : DecrementParentLocks(targettag);
4310 : }
4311 : }
4312 : }
4313 :
4314 : /*
4315 : * CheckForSerializableConflictIn
4316 : * We are writing the given tuple. If that indicates a rw-conflict
4317 : * in from another serializable transaction, take appropriate action.
4318 : *
4319 : * Skip checking for any granularity for which a parameter is missing.
4320 : *
4321 : * A tuple update or delete is in conflict if we have a predicate lock
4322 : * against the relation or page in which the tuple exists, or against the
4323 : * tuple itself.
4324 : */
4325 : void
4326 31830646 : CheckForSerializableConflictIn(Relation relation, ItemPointer tid, BlockNumber blkno)
4327 : {
4328 : PREDICATELOCKTARGETTAG targettag;
4329 :
4330 31830646 : if (!SerializationNeededForWrite(relation))
4331 31821796 : return;
4332 :
4333 : /* Check if someone else has already decided that we need to die */
4334 8850 : if (SxactIsDoomed(MySerializableXact))
4335 2 : ereport(ERROR,
4336 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4337 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4338 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict in checking."),
4339 : errhint("The transaction might succeed if retried.")));
4340 :
4341 : /*
4342 : * We're doing a write which might cause rw-conflicts now or later.
4343 : * Memorize that fact.
4344 : */
4345 8848 : MyXactDidWrite = true;
4346 :
4347 : /*
4348 : * It is important that we check for locks from the finest granularity to
4349 : * the coarsest granularity, so that granularity promotion doesn't cause
4350 : * us to miss a lock. The new (coarser) lock will be acquired before the
4351 : * old (finer) locks are released.
4352 : *
4353 : * It is not possible to take and hold a lock across the checks for all
4354 : * granularities because each target could be in a separate partition.
4355 : */
4356 8848 : if (tid != NULL)
4357 : {
4358 1286 : SET_PREDICATELOCKTARGETTAG_TUPLE(targettag,
4359 : relation->rd_locator.dbOid,
4360 : relation->rd_id,
4361 : ItemPointerGetBlockNumber(tid),
4362 : ItemPointerGetOffsetNumber(tid));
4363 1286 : CheckTargetForConflictsIn(&targettag);
4364 : }
4365 :
4366 8802 : if (blkno != InvalidBlockNumber)
4367 : {
4368 4928 : SET_PREDICATELOCKTARGETTAG_PAGE(targettag,
4369 : relation->rd_locator.dbOid,
4370 : relation->rd_id,
4371 : blkno);
4372 4928 : CheckTargetForConflictsIn(&targettag);
4373 : }
4374 :
4375 8742 : SET_PREDICATELOCKTARGETTAG_RELATION(targettag,
4376 : relation->rd_locator.dbOid,
4377 : relation->rd_id);
4378 8742 : CheckTargetForConflictsIn(&targettag);
4379 : }
4380 :
4381 : /*
4382 : * CheckTableForSerializableConflictIn
4383 : * The entire table is going through a DDL-style logical mass delete
4384 : * like TRUNCATE or DROP TABLE. If that causes a rw-conflict in from
4385 : * another serializable transaction, take appropriate action.
4386 : *
4387 : * While these operations do not operate entirely within the bounds of
4388 : * snapshot isolation, they can occur inside a serializable transaction, and
4389 : * will logically occur after any reads which saw rows which were destroyed
4390 : * by these operations, so we do what we can to serialize properly under
4391 : * SSI.
4392 : *
4393 : * The relation passed in must be a heap relation. Any predicate lock of any
4394 : * granularity on the heap will cause a rw-conflict in to this transaction.
4395 : * Predicate locks on indexes do not matter because they only exist to guard
4396 : * against conflicting inserts into the index, and this is a mass *delete*.
4397 : * When a table is truncated or dropped, the index will also be truncated
4398 : * or dropped, and we'll deal with locks on the index when that happens.
4399 : *
4400 : * Dropping or truncating a table also needs to drop any existing predicate
4401 : * locks on heap tuples or pages, because they're about to go away. This
4402 : * should be done before altering the predicate locks because the transaction
4403 : * could be rolled back because of a conflict, in which case the lock changes
4404 : * are not needed. (At the moment, we don't actually bother to drop the
4405 : * existing locks on a dropped or truncated table at the moment. That might
4406 : * lead to some false positives, but it doesn't seem worth the trouble.)
4407 : */
4408 : void
4409 50350 : CheckTableForSerializableConflictIn(Relation relation)
4410 : {
4411 : HASH_SEQ_STATUS seqstat;
4412 : PREDICATELOCKTARGET *target;
4413 : Oid dbId;
4414 : Oid heapId;
4415 : int i;
4416 :
4417 : /*
4418 : * Bail out quickly if there are no serializable transactions running.
4419 : * It's safe to check this without taking locks because the caller is
4420 : * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
4421 : * would matter here can be acquired while that is held.
4422 : */
4423 50350 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
4424 50344 : return;
4425 :
4426 422 : if (!SerializationNeededForWrite(relation))
4427 416 : return;
4428 :
4429 : /*
4430 : * We're doing a write which might cause rw-conflicts now or later.
4431 : * Memorize that fact.
4432 : */
4433 6 : MyXactDidWrite = true;
4434 :
4435 : Assert(relation->rd_index == NULL); /* not an index relation */
4436 :
4437 6 : dbId = relation->rd_locator.dbOid;
4438 6 : heapId = relation->rd_id;
4439 :
4440 6 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
4441 102 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
4442 96 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_SHARED);
4443 6 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4444 :
4445 : /* Scan through target list */
4446 6 : hash_seq_init(&seqstat, PredicateLockTargetHash);
4447 :
4448 12 : while ((target = (PREDICATELOCKTARGET *) hash_seq_search(&seqstat)))
4449 : {
4450 : dlist_mutable_iter iter;
4451 :
4452 : /*
4453 : * Check whether this is a target which needs attention.
4454 : */
4455 6 : if (GET_PREDICATELOCKTARGETTAG_RELATION(target->tag) != heapId)
4456 6 : continue; /* wrong relation id */
4457 0 : if (GET_PREDICATELOCKTARGETTAG_DB(target->tag) != dbId)
4458 0 : continue; /* wrong database id */
4459 :
4460 : /*
4461 : * Loop through locks for this target and flag conflicts.
4462 : */
4463 0 : dlist_foreach_modify(iter, &target->predicateLocks)
4464 : {
4465 0 : PREDICATELOCK *predlock =
4466 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
4467 :
4468 0 : if (predlock->tag.myXact != MySerializableXact
4469 0 : && !RWConflictExists(predlock->tag.myXact, MySerializableXact))
4470 : {
4471 0 : FlagRWConflict(predlock->tag.myXact, MySerializableXact);
4472 : }
4473 : }
4474 : }
4475 :
4476 : /* Release locks in reverse order */
4477 6 : LWLockRelease(SerializableXactHashLock);
4478 102 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
4479 96 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
4480 6 : LWLockRelease(SerializablePredicateListLock);
4481 : }
4482 :
4483 :
4484 : /*
4485 : * Flag a rw-dependency between two serializable transactions.
4486 : *
4487 : * The caller is responsible for ensuring that we have a LW lock on
4488 : * the transaction hash table.
4489 : */
4490 : static void
4491 1720 : FlagRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
4492 : {
4493 : Assert(reader != writer);
4494 :
4495 : /* First, see if this conflict causes failure. */
4496 1720 : OnConflict_CheckForSerializationFailure(reader, writer);
4497 :
4498 : /* Actually do the conflict flagging. */
4499 1560 : if (reader == OldCommittedSxact)
4500 0 : writer->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
4501 1560 : else if (writer == OldCommittedSxact)
4502 0 : reader->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4503 : else
4504 1560 : SetRWConflict(reader, writer);
4505 1560 : }
4506 :
4507 : /*----------------------------------------------------------------------------
4508 : * We are about to add a RW-edge to the dependency graph - check that we don't
4509 : * introduce a dangerous structure by doing so, and abort one of the
4510 : * transactions if so.
4511 : *
4512 : * A serialization failure can only occur if there is a dangerous structure
4513 : * in the dependency graph:
4514 : *
4515 : * Tin ------> Tpivot ------> Tout
4516 : * rw rw
4517 : *
4518 : * Furthermore, Tout must commit first.
4519 : *
4520 : * One more optimization is that if Tin is declared READ ONLY (or commits
4521 : * without writing), we can only have a problem if Tout committed before Tin
4522 : * acquired its snapshot.
4523 : *----------------------------------------------------------------------------
4524 : */
4525 : static void
4526 1720 : OnConflict_CheckForSerializationFailure(const SERIALIZABLEXACT *reader,
4527 : SERIALIZABLEXACT *writer)
4528 : {
4529 : bool failure;
4530 :
4531 : Assert(LWLockHeldByMe(SerializableXactHashLock));
4532 :
4533 1720 : failure = false;
4534 :
4535 : /*------------------------------------------------------------------------
4536 : * Check for already-committed writer with rw-conflict out flagged
4537 : * (conflict-flag on W means that T2 committed before W):
4538 : *
4539 : * R ------> W ------> T2
4540 : * rw rw
4541 : *
4542 : * That is a dangerous structure, so we must abort. (Since the writer
4543 : * has already committed, we must be the reader)
4544 : *------------------------------------------------------------------------
4545 : */
4546 1720 : if (SxactIsCommitted(writer)
4547 36 : && (SxactHasConflictOut(writer) || SxactHasSummaryConflictOut(writer)))
4548 4 : failure = true;
4549 :
4550 : /*------------------------------------------------------------------------
4551 : * Check whether the writer has become a pivot with an out-conflict
4552 : * committed transaction (T2), and T2 committed first:
4553 : *
4554 : * R ------> W ------> T2
4555 : * rw rw
4556 : *
4557 : * Because T2 must've committed first, there is no anomaly if:
4558 : * - the reader committed before T2
4559 : * - the writer committed before T2
4560 : * - the reader is a READ ONLY transaction and the reader was concurrent
4561 : * with T2 (= reader acquired its snapshot before T2 committed)
4562 : *
4563 : * We also handle the case that T2 is prepared but not yet committed
4564 : * here. In that case T2 has already checked for conflicts, so if it
4565 : * commits first, making the above conflict real, it's too late for it
4566 : * to abort.
4567 : *------------------------------------------------------------------------
4568 : */
4569 1720 : if (!failure && SxactHasSummaryConflictOut(writer))
4570 0 : failure = true;
4571 1720 : else if (!failure)
4572 : {
4573 : dlist_iter iter;
4574 :
4575 2142 : dlist_foreach(iter, &writer->outConflicts)
4576 : {
4577 576 : RWConflict conflict =
4578 576 : dlist_container(RWConflictData, outLink, iter.cur);
4579 576 : SERIALIZABLEXACT *t2 = conflict->sxactIn;
4580 :
4581 576 : if (SxactIsPrepared(t2)
4582 162 : && (!SxactIsCommitted(reader)
4583 130 : || t2->prepareSeqNo <= reader->commitSeqNo)
4584 162 : && (!SxactIsCommitted(writer)
4585 0 : || t2->prepareSeqNo <= writer->commitSeqNo)
4586 162 : && (!SxactIsReadOnly(reader)
4587 24 : || t2->prepareSeqNo <= reader->SeqNo.lastCommitBeforeSnapshot))
4588 : {
4589 150 : failure = true;
4590 150 : break;
4591 : }
4592 : }
4593 : }
4594 :
4595 : /*------------------------------------------------------------------------
4596 : * Check whether the reader has become a pivot with a writer
4597 : * that's committed (or prepared):
4598 : *
4599 : * T0 ------> R ------> W
4600 : * rw rw
4601 : *
4602 : * Because W must've committed first for an anomaly to occur, there is no
4603 : * anomaly if:
4604 : * - T0 committed before the writer
4605 : * - T0 is READ ONLY, and overlaps the writer
4606 : *------------------------------------------------------------------------
4607 : */
4608 1720 : if (!failure && SxactIsPrepared(writer) && !SxactIsReadOnly(reader))
4609 : {
4610 36 : if (SxactHasSummaryConflictIn(reader))
4611 : {
4612 0 : failure = true;
4613 : }
4614 : else
4615 : {
4616 : dlist_iter iter;
4617 :
4618 : /*
4619 : * The unconstify is needed as we have no const version of
4620 : * dlist_foreach().
4621 : */
4622 36 : dlist_foreach(iter, &unconstify(SERIALIZABLEXACT *, reader)->inConflicts)
4623 : {
4624 22 : const RWConflict conflict =
4625 22 : dlist_container(RWConflictData, inLink, iter.cur);
4626 22 : const SERIALIZABLEXACT *t0 = conflict->sxactOut;
4627 :
4628 22 : if (!SxactIsDoomed(t0)
4629 22 : && (!SxactIsCommitted(t0)
4630 22 : || t0->commitSeqNo >= writer->prepareSeqNo)
4631 22 : && (!SxactIsReadOnly(t0)
4632 0 : || t0->SeqNo.lastCommitBeforeSnapshot >= writer->prepareSeqNo))
4633 : {
4634 22 : failure = true;
4635 22 : break;
4636 : }
4637 : }
4638 : }
4639 : }
4640 :
4641 1720 : if (failure)
4642 : {
4643 : /*
4644 : * We have to kill a transaction to avoid a possible anomaly from
4645 : * occurring. If the writer is us, we can just ereport() to cause a
4646 : * transaction abort. Otherwise we flag the writer for termination,
4647 : * causing it to abort when it tries to commit. However, if the writer
4648 : * is a prepared transaction, already prepared, we can't abort it
4649 : * anymore, so we have to kill the reader instead.
4650 : */
4651 176 : if (MySerializableXact == writer)
4652 : {
4653 134 : LWLockRelease(SerializableXactHashLock);
4654 134 : ereport(ERROR,
4655 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4656 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4657 : errdetail_internal("Reason code: Canceled on identification as a pivot, during write."),
4658 : errhint("The transaction might succeed if retried.")));
4659 : }
4660 42 : else if (SxactIsPrepared(writer))
4661 : {
4662 26 : LWLockRelease(SerializableXactHashLock);
4663 :
4664 : /* if we're not the writer, we have to be the reader */
4665 : Assert(MySerializableXact == reader);
4666 26 : ereport(ERROR,
4667 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4668 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4669 : errdetail_internal("Reason code: Canceled on conflict out to pivot %u, during read.", writer->topXid),
4670 : errhint("The transaction might succeed if retried.")));
4671 : }
4672 16 : writer->flags |= SXACT_FLAG_DOOMED;
4673 : }
4674 1560 : }
4675 :
4676 : /*
4677 : * PreCommit_CheckForSerializationFailure
4678 : * Check for dangerous structures in a serializable transaction
4679 : * at commit.
4680 : *
4681 : * We're checking for a dangerous structure as each conflict is recorded.
4682 : * The only way we could have a problem at commit is if this is the "out"
4683 : * side of a pivot, and neither the "in" side nor the pivot has yet
4684 : * committed.
4685 : *
4686 : * If a dangerous structure is found, the pivot (the near conflict) is
4687 : * marked for death, because rolling back another transaction might mean
4688 : * that we fail without ever making progress. This transaction is
4689 : * committing writes, so letting it commit ensures progress. If we
4690 : * canceled the far conflict, it might immediately fail again on retry.
4691 : */
4692 : void
4693 767394 : PreCommit_CheckForSerializationFailure(void)
4694 : {
4695 : dlist_iter near_iter;
4696 :
4697 767394 : if (MySerializableXact == InvalidSerializableXact)
4698 764630 : return;
4699 :
4700 : Assert(IsolationIsSerializable());
4701 :
4702 2764 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4703 :
4704 : /*
4705 : * Check if someone else has already decided that we need to die. Since
4706 : * we set our own DOOMED flag when partially releasing, ignore in that
4707 : * case.
4708 : */
4709 2764 : if (SxactIsDoomed(MySerializableXact) &&
4710 312 : !SxactIsPartiallyReleased(MySerializableXact))
4711 : {
4712 310 : LWLockRelease(SerializableXactHashLock);
4713 310 : ereport(ERROR,
4714 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4715 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4716 : errdetail_internal("Reason code: Canceled on identification as a pivot, during commit attempt."),
4717 : errhint("The transaction might succeed if retried.")));
4718 : }
4719 :
4720 3656 : dlist_foreach(near_iter, &MySerializableXact->inConflicts)
4721 : {
4722 1202 : RWConflict nearConflict =
4723 1202 : dlist_container(RWConflictData, inLink, near_iter.cur);
4724 :
4725 1202 : if (!SxactIsCommitted(nearConflict->sxactOut)
4726 834 : && !SxactIsDoomed(nearConflict->sxactOut))
4727 : {
4728 : dlist_iter far_iter;
4729 :
4730 894 : dlist_foreach(far_iter, &nearConflict->sxactOut->inConflicts)
4731 : {
4732 356 : RWConflict farConflict =
4733 356 : dlist_container(RWConflictData, inLink, far_iter.cur);
4734 :
4735 356 : if (farConflict->sxactOut == MySerializableXact
4736 84 : || (!SxactIsCommitted(farConflict->sxactOut)
4737 48 : && !SxactIsReadOnly(farConflict->sxactOut)
4738 24 : && !SxactIsDoomed(farConflict->sxactOut)))
4739 : {
4740 : /*
4741 : * Normally, we kill the pivot transaction to make sure we
4742 : * make progress if the failing transaction is retried.
4743 : * However, we can't kill it if it's already prepared, so
4744 : * in that case we commit suicide instead.
4745 : */
4746 296 : if (SxactIsPrepared(nearConflict->sxactOut))
4747 : {
4748 0 : LWLockRelease(SerializableXactHashLock);
4749 0 : ereport(ERROR,
4750 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4751 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4752 : errdetail_internal("Reason code: Canceled on commit attempt with conflict in from prepared pivot."),
4753 : errhint("The transaction might succeed if retried.")));
4754 : }
4755 296 : nearConflict->sxactOut->flags |= SXACT_FLAG_DOOMED;
4756 296 : break;
4757 : }
4758 : }
4759 : }
4760 : }
4761 :
4762 2454 : MySerializableXact->prepareSeqNo = ++(PredXact->LastSxactCommitSeqNo);
4763 2454 : MySerializableXact->flags |= SXACT_FLAG_PREPARED;
4764 :
4765 2454 : LWLockRelease(SerializableXactHashLock);
4766 : }
4767 :
4768 : /*------------------------------------------------------------------------*/
4769 :
4770 : /*
4771 : * Two-phase commit support
4772 : */
4773 :
4774 : /*
4775 : * AtPrepare_Locks
4776 : * Do the preparatory work for a PREPARE: make 2PC state file
4777 : * records for all predicate locks currently held.
4778 : */
4779 : void
4780 562 : AtPrepare_PredicateLocks(void)
4781 : {
4782 : SERIALIZABLEXACT *sxact;
4783 : TwoPhasePredicateRecord record;
4784 : TwoPhasePredicateXactRecord *xactRecord;
4785 : TwoPhasePredicateLockRecord *lockRecord;
4786 : dlist_iter iter;
4787 :
4788 562 : sxact = MySerializableXact;
4789 562 : xactRecord = &(record.data.xactRecord);
4790 562 : lockRecord = &(record.data.lockRecord);
4791 :
4792 562 : if (MySerializableXact == InvalidSerializableXact)
4793 538 : return;
4794 :
4795 : /* Generate an xact record for our SERIALIZABLEXACT */
4796 24 : record.type = TWOPHASEPREDICATERECORD_XACT;
4797 24 : xactRecord->xmin = MySerializableXact->xmin;
4798 24 : xactRecord->flags = MySerializableXact->flags;
4799 :
4800 : /*
4801 : * Note that we don't include the list of conflicts in our out in the
4802 : * statefile, because new conflicts can be added even after the
4803 : * transaction prepares. We'll just make a conservative assumption during
4804 : * recovery instead.
4805 : */
4806 :
4807 24 : RegisterTwoPhaseRecord(TWOPHASE_RM_PREDICATELOCK_ID, 0,
4808 : &record, sizeof(record));
4809 :
4810 : /*
4811 : * Generate a lock record for each lock.
4812 : *
4813 : * To do this, we need to walk the predicate lock list in our sxact rather
4814 : * than using the local predicate lock table because the latter is not
4815 : * guaranteed to be accurate.
4816 : */
4817 24 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
4818 :
4819 : /*
4820 : * No need to take sxact->perXactPredicateListLock in parallel mode
4821 : * because there cannot be any parallel workers running while we are
4822 : * preparing a transaction.
4823 : */
4824 : Assert(!IsParallelWorker() && !ParallelContextActive());
4825 :
4826 44 : dlist_foreach(iter, &sxact->predicateLocks)
4827 : {
4828 20 : PREDICATELOCK *predlock =
4829 20 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
4830 :
4831 20 : record.type = TWOPHASEPREDICATERECORD_LOCK;
4832 20 : lockRecord->target = predlock->tag.myTarget->tag;
4833 :
4834 20 : RegisterTwoPhaseRecord(TWOPHASE_RM_PREDICATELOCK_ID, 0,
4835 : &record, sizeof(record));
4836 : }
4837 :
4838 24 : LWLockRelease(SerializablePredicateListLock);
4839 : }
4840 :
4841 : /*
4842 : * PostPrepare_Locks
4843 : * Clean up after successful PREPARE. Unlike the non-predicate
4844 : * lock manager, we do not need to transfer locks to a dummy
4845 : * PGPROC because our SERIALIZABLEXACT will stay around
4846 : * anyway. We only need to clean up our local state.
4847 : */
4848 : void
4849 562 : PostPrepare_PredicateLocks(TransactionId xid)
4850 : {
4851 562 : if (MySerializableXact == InvalidSerializableXact)
4852 538 : return;
4853 :
4854 : Assert(SxactIsPrepared(MySerializableXact));
4855 :
4856 24 : MySerializableXact->pid = 0;
4857 24 : MySerializableXact->pgprocno = INVALID_PROC_NUMBER;
4858 :
4859 24 : hash_destroy(LocalPredicateLockHash);
4860 24 : LocalPredicateLockHash = NULL;
4861 :
4862 24 : MySerializableXact = InvalidSerializableXact;
4863 24 : MyXactDidWrite = false;
4864 : }
4865 :
4866 : /*
4867 : * PredicateLockTwoPhaseFinish
4868 : * Release a prepared transaction's predicate locks once it
4869 : * commits or aborts.
4870 : */
4871 : void
4872 576 : PredicateLockTwoPhaseFinish(TransactionId xid, bool isCommit)
4873 : {
4874 : SERIALIZABLEXID *sxid;
4875 : SERIALIZABLEXIDTAG sxidtag;
4876 :
4877 576 : sxidtag.xid = xid;
4878 :
4879 576 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4880 : sxid = (SERIALIZABLEXID *)
4881 576 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
4882 576 : LWLockRelease(SerializableXactHashLock);
4883 :
4884 : /* xid will not be found if it wasn't a serializable transaction */
4885 576 : if (sxid == NULL)
4886 552 : return;
4887 :
4888 : /* Release its locks */
4889 24 : MySerializableXact = sxid->myXact;
4890 24 : MyXactDidWrite = true; /* conservatively assume that we wrote
4891 : * something */
4892 24 : ReleasePredicateLocks(isCommit, false);
4893 : }
4894 :
4895 : /*
4896 : * Re-acquire a predicate lock belonging to a transaction that was prepared.
4897 : */
4898 : void
4899 0 : predicatelock_twophase_recover(TransactionId xid, uint16 info,
4900 : void *recdata, uint32 len)
4901 : {
4902 : TwoPhasePredicateRecord *record;
4903 :
4904 : Assert(len == sizeof(TwoPhasePredicateRecord));
4905 :
4906 0 : record = (TwoPhasePredicateRecord *) recdata;
4907 :
4908 : Assert((record->type == TWOPHASEPREDICATERECORD_XACT) ||
4909 : (record->type == TWOPHASEPREDICATERECORD_LOCK));
4910 :
4911 0 : if (record->type == TWOPHASEPREDICATERECORD_XACT)
4912 : {
4913 : /* Per-transaction record. Set up a SERIALIZABLEXACT. */
4914 : TwoPhasePredicateXactRecord *xactRecord;
4915 : SERIALIZABLEXACT *sxact;
4916 : SERIALIZABLEXID *sxid;
4917 : SERIALIZABLEXIDTAG sxidtag;
4918 : bool found;
4919 :
4920 0 : xactRecord = (TwoPhasePredicateXactRecord *) &record->data.xactRecord;
4921 :
4922 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4923 0 : sxact = CreatePredXact();
4924 0 : if (!sxact)
4925 0 : ereport(ERROR,
4926 : (errcode(ERRCODE_OUT_OF_MEMORY),
4927 : errmsg("out of shared memory")));
4928 :
4929 : /* vxid for a prepared xact is INVALID_PROC_NUMBER/xid; no pid */
4930 0 : sxact->vxid.procNumber = INVALID_PROC_NUMBER;
4931 0 : sxact->vxid.localTransactionId = (LocalTransactionId) xid;
4932 0 : sxact->pid = 0;
4933 0 : sxact->pgprocno = INVALID_PROC_NUMBER;
4934 :
4935 : /* a prepared xact hasn't committed yet */
4936 0 : sxact->prepareSeqNo = RecoverySerCommitSeqNo;
4937 0 : sxact->commitSeqNo = InvalidSerCommitSeqNo;
4938 0 : sxact->finishedBefore = InvalidTransactionId;
4939 :
4940 0 : sxact->SeqNo.lastCommitBeforeSnapshot = RecoverySerCommitSeqNo;
4941 :
4942 : /*
4943 : * Don't need to track this; no transactions running at the time the
4944 : * recovered xact started are still active, except possibly other
4945 : * prepared xacts and we don't care whether those are RO_SAFE or not.
4946 : */
4947 0 : dlist_init(&(sxact->possibleUnsafeConflicts));
4948 :
4949 0 : dlist_init(&(sxact->predicateLocks));
4950 0 : dlist_node_init(&sxact->finishedLink);
4951 :
4952 0 : sxact->topXid = xid;
4953 0 : sxact->xmin = xactRecord->xmin;
4954 0 : sxact->flags = xactRecord->flags;
4955 : Assert(SxactIsPrepared(sxact));
4956 0 : if (!SxactIsReadOnly(sxact))
4957 : {
4958 0 : ++(PredXact->WritableSxactCount);
4959 : Assert(PredXact->WritableSxactCount <=
4960 : (MaxBackends + max_prepared_xacts));
4961 : }
4962 :
4963 : /*
4964 : * We don't know whether the transaction had any conflicts or not, so
4965 : * we'll conservatively assume that it had both a conflict in and a
4966 : * conflict out, and represent that with the summary conflict flags.
4967 : */
4968 0 : dlist_init(&(sxact->outConflicts));
4969 0 : dlist_init(&(sxact->inConflicts));
4970 0 : sxact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
4971 0 : sxact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4972 :
4973 : /* Register the transaction's xid */
4974 0 : sxidtag.xid = xid;
4975 0 : sxid = (SERIALIZABLEXID *) hash_search(SerializableXidHash,
4976 : &sxidtag,
4977 : HASH_ENTER, &found);
4978 : Assert(sxid != NULL);
4979 : Assert(!found);
4980 0 : sxid->myXact = (SERIALIZABLEXACT *) sxact;
4981 :
4982 : /*
4983 : * Update global xmin. Note that this is a special case compared to
4984 : * registering a normal transaction, because the global xmin might go
4985 : * backwards. That's OK, because until recovery is over we're not
4986 : * going to complete any transactions or create any non-prepared
4987 : * transactions, so there's no danger of throwing away.
4988 : */
4989 0 : if ((!TransactionIdIsValid(PredXact->SxactGlobalXmin)) ||
4990 0 : (TransactionIdFollows(PredXact->SxactGlobalXmin, sxact->xmin)))
4991 : {
4992 0 : PredXact->SxactGlobalXmin = sxact->xmin;
4993 0 : PredXact->SxactGlobalXminCount = 1;
4994 0 : SerialSetActiveSerXmin(sxact->xmin);
4995 : }
4996 0 : else if (TransactionIdEquals(sxact->xmin, PredXact->SxactGlobalXmin))
4997 : {
4998 : Assert(PredXact->SxactGlobalXminCount > 0);
4999 0 : PredXact->SxactGlobalXminCount++;
5000 : }
5001 :
5002 0 : LWLockRelease(SerializableXactHashLock);
5003 : }
5004 0 : else if (record->type == TWOPHASEPREDICATERECORD_LOCK)
5005 : {
5006 : /* Lock record. Recreate the PREDICATELOCK */
5007 : TwoPhasePredicateLockRecord *lockRecord;
5008 : SERIALIZABLEXID *sxid;
5009 : SERIALIZABLEXACT *sxact;
5010 : SERIALIZABLEXIDTAG sxidtag;
5011 : uint32 targettaghash;
5012 :
5013 0 : lockRecord = (TwoPhasePredicateLockRecord *) &record->data.lockRecord;
5014 0 : targettaghash = PredicateLockTargetTagHashCode(&lockRecord->target);
5015 :
5016 0 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
5017 0 : sxidtag.xid = xid;
5018 : sxid = (SERIALIZABLEXID *)
5019 0 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
5020 0 : LWLockRelease(SerializableXactHashLock);
5021 :
5022 : Assert(sxid != NULL);
5023 0 : sxact = sxid->myXact;
5024 : Assert(sxact != InvalidSerializableXact);
5025 :
5026 0 : CreatePredicateLock(&lockRecord->target, targettaghash, sxact);
5027 : }
5028 0 : }
5029 :
5030 : /*
5031 : * Prepare to share the current SERIALIZABLEXACT with parallel workers.
5032 : * Return a handle object that can be used by AttachSerializableXact() in a
5033 : * parallel worker.
5034 : */
5035 : SerializableXactHandle
5036 910 : ShareSerializableXact(void)
5037 : {
5038 910 : return MySerializableXact;
5039 : }
5040 :
5041 : /*
5042 : * Allow parallel workers to import the leader's SERIALIZABLEXACT.
5043 : */
5044 : void
5045 2736 : AttachSerializableXact(SerializableXactHandle handle)
5046 : {
5047 :
5048 : Assert(MySerializableXact == InvalidSerializableXact);
5049 :
5050 2736 : MySerializableXact = (SERIALIZABLEXACT *) handle;
5051 2736 : if (MySerializableXact != InvalidSerializableXact)
5052 26 : CreateLocalPredicateLockHash();
5053 2736 : }
|
