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, const ItemPointerData *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, const ItemPointerData *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(FullTransactionId fxid, 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 321464 : PredicateLockingNeededForRelation(Relation relation)
499 : {
500 435906 : return !(relation->rd_id < FirstUnpinnedObjectId ||
501 114442 : 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 123466794 : SerializationNeededForRead(Relation relation, Snapshot snapshot)
517 : {
518 : /* Nothing to do if this is not a serializable transaction */
519 123466794 : if (MySerializableXact == InvalidSerializableXact)
520 123157590 : 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 309204 : if (!IsMVCCSnapshot(snapshot))
530 3770 : 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 305434 : 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 305368 : if (!PredicateLockingNeededForRelation(relation))
550 200306 : return false;
551 :
552 105062 : 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 33394450 : SerializationNeededForWrite(Relation relation)
561 : {
562 : /* Nothing to do if this is not a serializable transaction */
563 33394450 : if (MySerializableXact == InvalidSerializableXact)
564 33378504 : return false;
565 :
566 : /* Check if the relation doesn't participate in predicate locking */
567 15946 : if (!PredicateLockingNeededForRelation(relation))
568 6974 : return false;
569 :
570 8972 : 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 5570 : CreatePredXact(void)
583 : {
584 : SERIALIZABLEXACT *sxact;
585 :
586 5570 : if (dlist_is_empty(&PredXact->availableList))
587 0 : return NULL;
588 :
589 5570 : sxact = dlist_container(SERIALIZABLEXACT, xactLink,
590 : dlist_pop_head_node(&PredXact->availableList));
591 5570 : dlist_push_tail(&PredXact->activeList, &sxact->xactLink);
592 5570 : return sxact;
593 : }
594 :
595 : static void
596 3338 : ReleasePredXact(SERIALIZABLEXACT *sxact)
597 : {
598 : Assert(ShmemAddrIsValid(sxact));
599 :
600 3338 : dlist_delete(&sxact->xactLink);
601 3338 : dlist_push_tail(&PredXact->availableList, &sxact->xactLink);
602 3338 : }
603 :
604 : /*------------------------------------------------------------------------*/
605 :
606 : /*
607 : * These functions manage primitive access to the RWConflict pool and lists.
608 : */
609 : static bool
610 3780 : 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 3780 : if (SxactIsDoomed(reader)
618 3780 : || SxactIsDoomed(writer)
619 3780 : || dlist_is_empty(&reader->outConflicts)
620 1138 : || dlist_is_empty(&writer->inConflicts))
621 2722 : 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 1090 : dlist_foreach(iter, &unconstify(SERIALIZABLEXACT *, reader)->outConflicts)
630 : {
631 1058 : RWConflict conflict =
632 : dlist_container(RWConflictData, outLink, iter.cur);
633 :
634 1058 : if (conflict->sxactIn == writer)
635 1026 : return true;
636 : }
637 :
638 : /* No conflict found. */
639 32 : return false;
640 : }
641 :
642 : static void
643 1584 : SetRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
644 : {
645 : RWConflict conflict;
646 :
647 : Assert(reader != writer);
648 : Assert(!RWConflictExists(reader, writer));
649 :
650 1584 : 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 1584 : conflict = dlist_head_element(RWConflictData, outLink, &RWConflictPool->availableList);
657 1584 : dlist_delete(&conflict->outLink);
658 :
659 1584 : conflict->sxactOut = reader;
660 1584 : conflict->sxactIn = writer;
661 1584 : dlist_push_tail(&reader->outConflicts, &conflict->outLink);
662 1584 : dlist_push_tail(&writer->inConflicts, &conflict->inLink);
663 1584 : }
664 :
665 : static void
666 264 : SetPossibleUnsafeConflict(SERIALIZABLEXACT *roXact,
667 : SERIALIZABLEXACT *activeXact)
668 : {
669 : RWConflict conflict;
670 :
671 : Assert(roXact != activeXact);
672 : Assert(SxactIsReadOnly(roXact));
673 : Assert(!SxactIsReadOnly(activeXact));
674 :
675 264 : 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 :
681 264 : conflict = dlist_head_element(RWConflictData, outLink, &RWConflictPool->availableList);
682 264 : dlist_delete(&conflict->outLink);
683 :
684 264 : conflict->sxactOut = activeXact;
685 264 : conflict->sxactIn = roXact;
686 264 : dlist_push_tail(&activeXact->possibleUnsafeConflicts, &conflict->outLink);
687 264 : dlist_push_tail(&roXact->possibleUnsafeConflicts, &conflict->inLink);
688 264 : }
689 :
690 : static void
691 1848 : ReleaseRWConflict(RWConflict conflict)
692 : {
693 1848 : dlist_delete(&conflict->inLink);
694 1848 : dlist_delete(&conflict->outLink);
695 1848 : dlist_push_tail(&RWConflictPool->availableList, &conflict->outLink);
696 1848 : }
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 2232 : SerialInit(void)
807 : {
808 : bool found;
809 :
810 : /*
811 : * Set up SLRU management of the pg_serial data.
812 : */
813 2232 : SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
814 2232 : 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 2232 : serialControl = (SerialControl)
827 2232 : ShmemInitStruct("SerialControlData", sizeof(SerialControlData), &found);
828 :
829 : Assert(found == IsUnderPostmaster);
830 2232 : if (!found)
831 : {
832 : /*
833 : * Set control information to reflect empty SLRU.
834 : */
835 2232 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
836 2232 : serialControl->headPage = -1;
837 2232 : serialControl->headXid = InvalidTransactionId;
838 2232 : serialControl->tailXid = InvalidTransactionId;
839 2232 : LWLockRelease(SerialControlLock);
840 : }
841 2232 : }
842 :
843 : /*
844 : * GUC check_hook for serializable_buffers
845 : */
846 : bool
847 2306 : check_serial_buffers(int *newval, void **extra, GucSource source)
848 : {
849 2306 : 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 3484 : SerialSetActiveSerXmin(TransactionId xid)
991 : {
992 3484 : 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 3484 : if (!TransactionIdIsValid(xid))
1001 : {
1002 1724 : serialControl->tailXid = InvalidTransactionId;
1003 1724 : serialControl->headXid = InvalidTransactionId;
1004 1724 : LWLockRelease(SerialControlLock);
1005 1724 : 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 1760 : 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 1760 : serialControl->tailXid = xid;
1030 :
1031 1760 : 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 3494 : CheckPointPredicate(void)
1042 : {
1043 : int64 truncateCutoffPage;
1044 :
1045 3494 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
1046 :
1047 : /* Exit quickly if the SLRU is currently not in use. */
1048 3494 : if (serialControl->headPage < 0)
1049 : {
1050 3494 : LWLockRelease(SerialControlLock);
1051 3494 : 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 2232 : PredicateLockShmemInit(void)
1146 : {
1147 : HASHCTL info;
1148 : int64 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 2232 : max_table_size = NPREDICATELOCKTARGETENTS();
1161 :
1162 : /*
1163 : * Allocate hash table for PREDICATELOCKTARGET structs. This stores
1164 : * per-predicate-lock-target information.
1165 : */
1166 2232 : info.keysize = sizeof(PREDICATELOCKTARGETTAG);
1167 2232 : info.entrysize = sizeof(PREDICATELOCKTARGET);
1168 2232 : info.num_partitions = NUM_PREDICATELOCK_PARTITIONS;
1169 :
1170 2232 : 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 2232 : if (!IsUnderPostmaster)
1184 : {
1185 2232 : (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 2232 : ScratchTargetTagHash = PredicateLockTargetTagHashCode(&ScratchTargetTag);
1192 2232 : ScratchPartitionLock = PredicateLockHashPartitionLock(ScratchTargetTagHash);
1193 :
1194 : /*
1195 : * Allocate hash table for PREDICATELOCK structs. This stores per
1196 : * xact-lock-of-a-target information.
1197 : */
1198 2232 : info.keysize = sizeof(PREDICATELOCKTAG);
1199 2232 : info.entrysize = sizeof(PREDICATELOCK);
1200 2232 : info.hash = predicatelock_hash;
1201 2232 : info.num_partitions = NUM_PREDICATELOCK_PARTITIONS;
1202 :
1203 : /* Assume an average of 2 xacts per target */
1204 2232 : max_table_size *= 2;
1205 :
1206 2232 : 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 2232 : 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 2232 : max_table_size *= 10;
1228 :
1229 2232 : requestSize = add_size(PredXactListDataSize,
1230 : (mul_size((Size) max_table_size,
1231 : sizeof(SERIALIZABLEXACT))));
1232 :
1233 2232 : PredXact = ShmemInitStruct("PredXactList",
1234 : requestSize,
1235 : &found);
1236 : Assert(found == IsUnderPostmaster);
1237 2232 : if (!found)
1238 : {
1239 : int i;
1240 :
1241 : /* clean everything, both the header and the element */
1242 2232 : memset(PredXact, 0, requestSize);
1243 :
1244 2232 : dlist_init(&PredXact->availableList);
1245 2232 : dlist_init(&PredXact->activeList);
1246 2232 : PredXact->SxactGlobalXmin = InvalidTransactionId;
1247 2232 : PredXact->SxactGlobalXminCount = 0;
1248 2232 : PredXact->WritableSxactCount = 0;
1249 2232 : PredXact->LastSxactCommitSeqNo = FirstNormalSerCommitSeqNo - 1;
1250 2232 : PredXact->CanPartialClearThrough = 0;
1251 2232 : PredXact->HavePartialClearedThrough = 0;
1252 2232 : PredXact->element
1253 2232 : = (SERIALIZABLEXACT *) ((char *) PredXact + PredXactListDataSize);
1254 : /* Add all elements to available list, clean. */
1255 2090092 : for (i = 0; i < max_table_size; i++)
1256 : {
1257 2087860 : LWLockInitialize(&PredXact->element[i].perXactPredicateListLock,
1258 : LWTRANCHE_PER_XACT_PREDICATE_LIST);
1259 2087860 : dlist_push_tail(&PredXact->availableList, &PredXact->element[i].xactLink);
1260 : }
1261 2232 : PredXact->OldCommittedSxact = CreatePredXact();
1262 2232 : SetInvalidVirtualTransactionId(PredXact->OldCommittedSxact->vxid);
1263 2232 : PredXact->OldCommittedSxact->prepareSeqNo = 0;
1264 2232 : PredXact->OldCommittedSxact->commitSeqNo = 0;
1265 2232 : PredXact->OldCommittedSxact->SeqNo.lastCommitBeforeSnapshot = 0;
1266 2232 : dlist_init(&PredXact->OldCommittedSxact->outConflicts);
1267 2232 : dlist_init(&PredXact->OldCommittedSxact->inConflicts);
1268 2232 : dlist_init(&PredXact->OldCommittedSxact->predicateLocks);
1269 2232 : dlist_node_init(&PredXact->OldCommittedSxact->finishedLink);
1270 2232 : dlist_init(&PredXact->OldCommittedSxact->possibleUnsafeConflicts);
1271 2232 : PredXact->OldCommittedSxact->topXid = InvalidTransactionId;
1272 2232 : PredXact->OldCommittedSxact->finishedBefore = InvalidTransactionId;
1273 2232 : PredXact->OldCommittedSxact->xmin = InvalidTransactionId;
1274 2232 : PredXact->OldCommittedSxact->flags = SXACT_FLAG_COMMITTED;
1275 2232 : PredXact->OldCommittedSxact->pid = 0;
1276 2232 : PredXact->OldCommittedSxact->pgprocno = INVALID_PROC_NUMBER;
1277 : }
1278 : /* This never changes, so let's keep a local copy. */
1279 2232 : OldCommittedSxact = PredXact->OldCommittedSxact;
1280 :
1281 : /*
1282 : * Allocate hash table for SERIALIZABLEXID structs. This stores per-xid
1283 : * information for serializable transactions which have accessed data.
1284 : */
1285 2232 : info.keysize = sizeof(SERIALIZABLEXIDTAG);
1286 2232 : info.entrysize = sizeof(SERIALIZABLEXID);
1287 :
1288 2232 : SerializableXidHash = ShmemInitHash("SERIALIZABLEXID hash",
1289 : max_table_size,
1290 : max_table_size,
1291 : &info,
1292 : HASH_ELEM | HASH_BLOBS |
1293 : HASH_FIXED_SIZE);
1294 :
1295 : /*
1296 : * Allocate space for tracking rw-conflicts in lists attached to the
1297 : * transactions.
1298 : *
1299 : * Assume an average of 5 conflicts per transaction. Calculations suggest
1300 : * that this will prevent resource exhaustion in even the most pessimal
1301 : * loads up to max_connections = 200 with all 200 connections pounding the
1302 : * database with serializable transactions. Beyond that, there may be
1303 : * occasional transactions canceled when trying to flag conflicts. That's
1304 : * probably OK.
1305 : */
1306 2232 : max_table_size *= 5;
1307 :
1308 2232 : requestSize = RWConflictPoolHeaderDataSize +
1309 2232 : mul_size((Size) max_table_size,
1310 : RWConflictDataSize);
1311 :
1312 2232 : RWConflictPool = ShmemInitStruct("RWConflictPool",
1313 : requestSize,
1314 : &found);
1315 : Assert(found == IsUnderPostmaster);
1316 2232 : if (!found)
1317 : {
1318 : int i;
1319 :
1320 : /* clean everything, including the elements */
1321 2232 : memset(RWConflictPool, 0, requestSize);
1322 :
1323 2232 : dlist_init(&RWConflictPool->availableList);
1324 2232 : RWConflictPool->element = (RWConflict) ((char *) RWConflictPool +
1325 : RWConflictPoolHeaderDataSize);
1326 : /* Add all elements to available list, clean. */
1327 10441532 : for (i = 0; i < max_table_size; i++)
1328 : {
1329 10439300 : dlist_push_tail(&RWConflictPool->availableList,
1330 10439300 : &RWConflictPool->element[i].outLink);
1331 : }
1332 : }
1333 :
1334 : /*
1335 : * Create or attach to the header for the list of finished serializable
1336 : * transactions.
1337 : */
1338 2232 : FinishedSerializableTransactions = (dlist_head *)
1339 2232 : ShmemInitStruct("FinishedSerializableTransactions",
1340 : sizeof(dlist_head),
1341 : &found);
1342 : Assert(found == IsUnderPostmaster);
1343 2232 : if (!found)
1344 2232 : dlist_init(FinishedSerializableTransactions);
1345 :
1346 : /*
1347 : * Initialize the SLRU storage for old committed serializable
1348 : * transactions.
1349 : */
1350 2232 : SerialInit();
1351 2232 : }
1352 :
1353 : /*
1354 : * Estimate shared-memory space used for predicate lock table
1355 : */
1356 : Size
1357 4158 : PredicateLockShmemSize(void)
1358 : {
1359 4158 : Size size = 0;
1360 : long max_table_size;
1361 :
1362 : /* predicate lock target hash table */
1363 4158 : max_table_size = NPREDICATELOCKTARGETENTS();
1364 4158 : size = add_size(size, hash_estimate_size(max_table_size,
1365 : sizeof(PREDICATELOCKTARGET)));
1366 :
1367 : /* predicate lock hash table */
1368 4158 : max_table_size *= 2;
1369 4158 : size = add_size(size, hash_estimate_size(max_table_size,
1370 : sizeof(PREDICATELOCK)));
1371 :
1372 : /*
1373 : * Since NPREDICATELOCKTARGETENTS is only an estimate, add 10% safety
1374 : * margin.
1375 : */
1376 4158 : size = add_size(size, size / 10);
1377 :
1378 : /* transaction list */
1379 4158 : max_table_size = MaxBackends + max_prepared_xacts;
1380 4158 : max_table_size *= 10;
1381 4158 : size = add_size(size, PredXactListDataSize);
1382 4158 : size = add_size(size, mul_size((Size) max_table_size,
1383 : sizeof(SERIALIZABLEXACT)));
1384 :
1385 : /* transaction xid table */
1386 4158 : size = add_size(size, hash_estimate_size(max_table_size,
1387 : sizeof(SERIALIZABLEXID)));
1388 :
1389 : /* rw-conflict pool */
1390 4158 : max_table_size *= 5;
1391 4158 : size = add_size(size, RWConflictPoolHeaderDataSize);
1392 4158 : size = add_size(size, mul_size((Size) max_table_size,
1393 : RWConflictDataSize));
1394 :
1395 : /* Head for list of finished serializable transactions. */
1396 4158 : size = add_size(size, sizeof(dlist_head));
1397 :
1398 : /* Shared memory structures for SLRU tracking of old committed xids. */
1399 4158 : size = add_size(size, sizeof(SerialControlData));
1400 4158 : size = add_size(size, SimpleLruShmemSize(serializable_buffers, 0));
1401 :
1402 4158 : return size;
1403 : }
1404 :
1405 :
1406 : /*
1407 : * Compute the hash code associated with a PREDICATELOCKTAG.
1408 : *
1409 : * Because we want to use just one set of partition locks for both the
1410 : * PREDICATELOCKTARGET and PREDICATELOCK hash tables, we have to make sure
1411 : * that PREDICATELOCKs fall into the same partition number as their
1412 : * associated PREDICATELOCKTARGETs. dynahash.c expects the partition number
1413 : * to be the low-order bits of the hash code, and therefore a
1414 : * PREDICATELOCKTAG's hash code must have the same low-order bits as the
1415 : * associated PREDICATELOCKTARGETTAG's hash code. We achieve this with this
1416 : * specialized hash function.
1417 : */
1418 : static uint32
1419 0 : predicatelock_hash(const void *key, Size keysize)
1420 : {
1421 0 : const PREDICATELOCKTAG *predicatelocktag = (const PREDICATELOCKTAG *) key;
1422 : uint32 targethash;
1423 :
1424 : Assert(keysize == sizeof(PREDICATELOCKTAG));
1425 :
1426 : /* Look into the associated target object, and compute its hash code */
1427 0 : targethash = PredicateLockTargetTagHashCode(&predicatelocktag->myTarget->tag);
1428 :
1429 0 : return PredicateLockHashCodeFromTargetHashCode(predicatelocktag, targethash);
1430 : }
1431 :
1432 :
1433 : /*
1434 : * GetPredicateLockStatusData
1435 : * Return a table containing the internal state of the predicate
1436 : * lock manager for use in pg_lock_status.
1437 : *
1438 : * Like GetLockStatusData, this function tries to hold the partition LWLocks
1439 : * for as short a time as possible by returning two arrays that simply
1440 : * contain the PREDICATELOCKTARGETTAG and SERIALIZABLEXACT for each lock
1441 : * table entry. Multiple copies of the same PREDICATELOCKTARGETTAG and
1442 : * SERIALIZABLEXACT will likely appear.
1443 : */
1444 : PredicateLockData *
1445 464 : GetPredicateLockStatusData(void)
1446 : {
1447 : PredicateLockData *data;
1448 : int i;
1449 : int els,
1450 : el;
1451 : HASH_SEQ_STATUS seqstat;
1452 : PREDICATELOCK *predlock;
1453 :
1454 464 : data = palloc_object(PredicateLockData);
1455 :
1456 : /*
1457 : * To ensure consistency, take simultaneous locks on all partition locks
1458 : * in ascending order, then SerializableXactHashLock.
1459 : */
1460 7888 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
1461 7424 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_SHARED);
1462 464 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1463 :
1464 : /* Get number of locks and allocate appropriately-sized arrays. */
1465 464 : els = hash_get_num_entries(PredicateLockHash);
1466 464 : data->nelements = els;
1467 464 : data->locktags = palloc_array(PREDICATELOCKTARGETTAG, els);
1468 464 : data->xacts = palloc_array(SERIALIZABLEXACT, els);
1469 :
1470 :
1471 : /* Scan through PredicateLockHash and copy contents */
1472 464 : hash_seq_init(&seqstat, PredicateLockHash);
1473 :
1474 464 : el = 0;
1475 :
1476 470 : while ((predlock = (PREDICATELOCK *) hash_seq_search(&seqstat)))
1477 : {
1478 6 : data->locktags[el] = predlock->tag.myTarget->tag;
1479 6 : data->xacts[el] = *predlock->tag.myXact;
1480 6 : el++;
1481 : }
1482 :
1483 : Assert(el == els);
1484 :
1485 : /* Release locks in reverse order */
1486 464 : LWLockRelease(SerializableXactHashLock);
1487 7888 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
1488 7424 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
1489 :
1490 464 : return data;
1491 : }
1492 :
1493 : /*
1494 : * Free up shared memory structures by pushing the oldest sxact (the one at
1495 : * the front of the SummarizeOldestCommittedSxact queue) into summary form.
1496 : * Each call will free exactly one SERIALIZABLEXACT structure and may also
1497 : * free one or more of these structures: SERIALIZABLEXID, PREDICATELOCK,
1498 : * PREDICATELOCKTARGET, RWConflictData.
1499 : */
1500 : static void
1501 0 : SummarizeOldestCommittedSxact(void)
1502 : {
1503 : SERIALIZABLEXACT *sxact;
1504 :
1505 0 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
1506 :
1507 : /*
1508 : * This function is only called if there are no sxact slots available.
1509 : * Some of them must belong to old, already-finished transactions, so
1510 : * there should be something in FinishedSerializableTransactions list that
1511 : * we can summarize. However, there's a race condition: while we were not
1512 : * holding any locks, a transaction might have ended and cleaned up all
1513 : * the finished sxact entries already, freeing up their sxact slots. In
1514 : * that case, we have nothing to do here. The caller will find one of the
1515 : * slots released by the other backend when it retries.
1516 : */
1517 0 : if (dlist_is_empty(FinishedSerializableTransactions))
1518 : {
1519 0 : LWLockRelease(SerializableFinishedListLock);
1520 0 : return;
1521 : }
1522 :
1523 : /*
1524 : * Grab the first sxact off the finished list -- this will be the earliest
1525 : * commit. Remove it from the list.
1526 : */
1527 0 : sxact = dlist_head_element(SERIALIZABLEXACT, finishedLink,
1528 : FinishedSerializableTransactions);
1529 0 : dlist_delete_thoroughly(&sxact->finishedLink);
1530 :
1531 : /* Add to SLRU summary information. */
1532 0 : if (TransactionIdIsValid(sxact->topXid) && !SxactIsReadOnly(sxact))
1533 0 : SerialAdd(sxact->topXid, SxactHasConflictOut(sxact)
1534 : ? sxact->SeqNo.earliestOutConflictCommit : InvalidSerCommitSeqNo);
1535 :
1536 : /* Summarize and release the detail. */
1537 0 : ReleaseOneSerializableXact(sxact, false, true);
1538 :
1539 0 : LWLockRelease(SerializableFinishedListLock);
1540 : }
1541 :
1542 : /*
1543 : * GetSafeSnapshot
1544 : * Obtain and register a snapshot for a READ ONLY DEFERRABLE
1545 : * transaction. Ensures that the snapshot is "safe", i.e. a
1546 : * read-only transaction running on it can execute serializably
1547 : * without further checks. This requires waiting for concurrent
1548 : * transactions to complete, and retrying with a new snapshot if
1549 : * one of them could possibly create a conflict.
1550 : *
1551 : * As with GetSerializableTransactionSnapshot (which this is a subroutine
1552 : * for), the passed-in Snapshot pointer should reference a static data
1553 : * area that can safely be passed to GetSnapshotData.
1554 : */
1555 : static Snapshot
1556 8 : GetSafeSnapshot(Snapshot origSnapshot)
1557 : {
1558 : Snapshot snapshot;
1559 :
1560 : Assert(XactReadOnly && XactDeferrable);
1561 :
1562 : while (true)
1563 : {
1564 : /*
1565 : * GetSerializableTransactionSnapshotInt is going to call
1566 : * GetSnapshotData, so we need to provide it the static snapshot area
1567 : * our caller passed to us. The pointer returned is actually the same
1568 : * one passed to it, but we avoid assuming that here.
1569 : */
1570 10 : snapshot = GetSerializableTransactionSnapshotInt(origSnapshot,
1571 : NULL, InvalidPid);
1572 :
1573 10 : if (MySerializableXact == InvalidSerializableXact)
1574 8 : return snapshot; /* no concurrent r/w xacts; it's safe */
1575 :
1576 2 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1577 :
1578 : /*
1579 : * Wait for concurrent transactions to finish. Stop early if one of
1580 : * them marked us as conflicted.
1581 : */
1582 2 : MySerializableXact->flags |= SXACT_FLAG_DEFERRABLE_WAITING;
1583 4 : while (!(dlist_is_empty(&MySerializableXact->possibleUnsafeConflicts) ||
1584 2 : SxactIsROUnsafe(MySerializableXact)))
1585 : {
1586 2 : LWLockRelease(SerializableXactHashLock);
1587 2 : ProcWaitForSignal(WAIT_EVENT_SAFE_SNAPSHOT);
1588 2 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1589 : }
1590 2 : MySerializableXact->flags &= ~SXACT_FLAG_DEFERRABLE_WAITING;
1591 :
1592 2 : if (!SxactIsROUnsafe(MySerializableXact))
1593 : {
1594 0 : LWLockRelease(SerializableXactHashLock);
1595 0 : break; /* success */
1596 : }
1597 :
1598 2 : LWLockRelease(SerializableXactHashLock);
1599 :
1600 : /* else, need to retry... */
1601 2 : ereport(DEBUG2,
1602 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
1603 : errmsg_internal("deferrable snapshot was unsafe; trying a new one")));
1604 2 : ReleasePredicateLocks(false, false);
1605 : }
1606 :
1607 : /*
1608 : * Now we have a safe snapshot, so we don't need to do any further checks.
1609 : */
1610 : Assert(SxactIsROSafe(MySerializableXact));
1611 0 : ReleasePredicateLocks(false, true);
1612 :
1613 0 : return snapshot;
1614 : }
1615 :
1616 : /*
1617 : * GetSafeSnapshotBlockingPids
1618 : * If the specified process is currently blocked in GetSafeSnapshot,
1619 : * write the process IDs of all processes that it is blocked by
1620 : * into the caller-supplied buffer output[]. The list is truncated at
1621 : * output_size, and the number of PIDs written into the buffer is
1622 : * returned. Returns zero if the given PID is not currently blocked
1623 : * in GetSafeSnapshot.
1624 : */
1625 : int
1626 1368 : GetSafeSnapshotBlockingPids(int blocked_pid, int *output, int output_size)
1627 : {
1628 1368 : int num_written = 0;
1629 : dlist_iter iter;
1630 1368 : SERIALIZABLEXACT *blocking_sxact = NULL;
1631 :
1632 1368 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1633 :
1634 : /* Find blocked_pid's SERIALIZABLEXACT by linear search. */
1635 3062 : dlist_foreach(iter, &PredXact->activeList)
1636 : {
1637 1878 : SERIALIZABLEXACT *sxact =
1638 1878 : dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
1639 :
1640 1878 : if (sxact->pid == blocked_pid)
1641 : {
1642 184 : blocking_sxact = sxact;
1643 184 : break;
1644 : }
1645 : }
1646 :
1647 : /* Did we find it, and is it currently waiting in GetSafeSnapshot? */
1648 1368 : if (blocking_sxact != NULL && SxactIsDeferrableWaiting(blocking_sxact))
1649 : {
1650 : /* Traverse the list of possible unsafe conflicts collecting PIDs. */
1651 4 : dlist_foreach(iter, &blocking_sxact->possibleUnsafeConflicts)
1652 : {
1653 4 : RWConflict possibleUnsafeConflict =
1654 4 : dlist_container(RWConflictData, inLink, iter.cur);
1655 :
1656 4 : output[num_written++] = possibleUnsafeConflict->sxactOut->pid;
1657 :
1658 4 : if (num_written >= output_size)
1659 4 : break;
1660 : }
1661 : }
1662 :
1663 1368 : LWLockRelease(SerializableXactHashLock);
1664 :
1665 1368 : return num_written;
1666 : }
1667 :
1668 : /*
1669 : * Acquire a snapshot that can be used for the current transaction.
1670 : *
1671 : * Make sure we have a SERIALIZABLEXACT reference in MySerializableXact.
1672 : * It should be current for this process and be contained in PredXact.
1673 : *
1674 : * The passed-in Snapshot pointer should reference a static data area that
1675 : * can safely be passed to GetSnapshotData. The return value is actually
1676 : * always this same pointer; no new snapshot data structure is allocated
1677 : * within this function.
1678 : */
1679 : Snapshot
1680 3336 : GetSerializableTransactionSnapshot(Snapshot snapshot)
1681 : {
1682 : Assert(IsolationIsSerializable());
1683 :
1684 : /*
1685 : * Can't use serializable mode while recovery is still active, as it is,
1686 : * for example, on a hot standby. We could get here despite the check in
1687 : * check_transaction_isolation() if default_transaction_isolation is set
1688 : * to serializable, so phrase the hint accordingly.
1689 : */
1690 3336 : if (RecoveryInProgress())
1691 0 : ereport(ERROR,
1692 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1693 : errmsg("cannot use serializable mode in a hot standby"),
1694 : errdetail("\"default_transaction_isolation\" is set to \"serializable\"."),
1695 : errhint("You can use \"SET default_transaction_isolation = 'repeatable read'\" to change the default.")));
1696 :
1697 : /*
1698 : * A special optimization is available for SERIALIZABLE READ ONLY
1699 : * DEFERRABLE transactions -- we can wait for a suitable snapshot and
1700 : * thereby avoid all SSI overhead once it's running.
1701 : */
1702 3336 : if (XactReadOnly && XactDeferrable)
1703 8 : return GetSafeSnapshot(snapshot);
1704 :
1705 3328 : return GetSerializableTransactionSnapshotInt(snapshot,
1706 : NULL, InvalidPid);
1707 : }
1708 :
1709 : /*
1710 : * Import a snapshot to be used for the current transaction.
1711 : *
1712 : * This is nearly the same as GetSerializableTransactionSnapshot, except that
1713 : * we don't take a new snapshot, but rather use the data we're handed.
1714 : *
1715 : * The caller must have verified that the snapshot came from a serializable
1716 : * transaction; and if we're read-write, the source transaction must not be
1717 : * read-only.
1718 : */
1719 : void
1720 26 : SetSerializableTransactionSnapshot(Snapshot snapshot,
1721 : VirtualTransactionId *sourcevxid,
1722 : int sourcepid)
1723 : {
1724 : Assert(IsolationIsSerializable());
1725 :
1726 : /*
1727 : * If this is called by parallel.c in a parallel worker, we don't want to
1728 : * create a SERIALIZABLEXACT just yet because the leader's
1729 : * SERIALIZABLEXACT will be installed with AttachSerializableXact(). We
1730 : * also don't want to reject SERIALIZABLE READ ONLY DEFERRABLE in this
1731 : * case, because the leader has already determined that the snapshot it
1732 : * has passed us is safe. So there is nothing for us to do.
1733 : */
1734 26 : if (IsParallelWorker())
1735 26 : return;
1736 :
1737 : /*
1738 : * We do not allow SERIALIZABLE READ ONLY DEFERRABLE transactions to
1739 : * import snapshots, since there's no way to wait for a safe snapshot when
1740 : * we're using the snap we're told to. (XXX instead of throwing an error,
1741 : * we could just ignore the XactDeferrable flag?)
1742 : */
1743 0 : if (XactReadOnly && XactDeferrable)
1744 0 : ereport(ERROR,
1745 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1746 : errmsg("a snapshot-importing transaction must not be READ ONLY DEFERRABLE")));
1747 :
1748 0 : (void) GetSerializableTransactionSnapshotInt(snapshot, sourcevxid,
1749 : sourcepid);
1750 : }
1751 :
1752 : /*
1753 : * Guts of GetSerializableTransactionSnapshot
1754 : *
1755 : * If sourcevxid is valid, this is actually an import operation and we should
1756 : * skip calling GetSnapshotData, because the snapshot contents are already
1757 : * loaded up. HOWEVER: to avoid race conditions, we must check that the
1758 : * source xact is still running after we acquire SerializableXactHashLock.
1759 : * We do that by calling ProcArrayInstallImportedXmin.
1760 : */
1761 : static Snapshot
1762 3338 : GetSerializableTransactionSnapshotInt(Snapshot snapshot,
1763 : VirtualTransactionId *sourcevxid,
1764 : int sourcepid)
1765 : {
1766 : PGPROC *proc;
1767 : VirtualTransactionId vxid;
1768 : SERIALIZABLEXACT *sxact,
1769 : *othersxact;
1770 :
1771 : /* We only do this for serializable transactions. Once. */
1772 : Assert(MySerializableXact == InvalidSerializableXact);
1773 :
1774 : Assert(!RecoveryInProgress());
1775 :
1776 : /*
1777 : * Since all parts of a serializable transaction must use the same
1778 : * snapshot, it is too late to establish one after a parallel operation
1779 : * has begun.
1780 : */
1781 3338 : if (IsInParallelMode())
1782 0 : elog(ERROR, "cannot establish serializable snapshot during a parallel operation");
1783 :
1784 3338 : proc = MyProc;
1785 : Assert(proc != NULL);
1786 3338 : GET_VXID_FROM_PGPROC(vxid, *proc);
1787 :
1788 : /*
1789 : * First we get the sxact structure, which may involve looping and access
1790 : * to the "finished" list to free a structure for use.
1791 : *
1792 : * We must hold SerializableXactHashLock when taking/checking the snapshot
1793 : * to avoid race conditions, for much the same reasons that
1794 : * GetSnapshotData takes the ProcArrayLock. Since we might have to
1795 : * release SerializableXactHashLock to call SummarizeOldestCommittedSxact,
1796 : * this means we have to create the sxact first, which is a bit annoying
1797 : * (in particular, an elog(ERROR) in procarray.c would cause us to leak
1798 : * the sxact). Consider refactoring to avoid this.
1799 : */
1800 : #ifdef TEST_SUMMARIZE_SERIAL
1801 : SummarizeOldestCommittedSxact();
1802 : #endif
1803 3338 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1804 : do
1805 : {
1806 3338 : sxact = CreatePredXact();
1807 : /* If null, push out committed sxact to SLRU summary & retry. */
1808 3338 : if (!sxact)
1809 : {
1810 0 : LWLockRelease(SerializableXactHashLock);
1811 0 : SummarizeOldestCommittedSxact();
1812 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1813 : }
1814 3338 : } while (!sxact);
1815 :
1816 : /* Get the snapshot, or check that it's safe to use */
1817 3338 : if (!sourcevxid)
1818 3338 : snapshot = GetSnapshotData(snapshot);
1819 0 : else if (!ProcArrayInstallImportedXmin(snapshot->xmin, sourcevxid))
1820 : {
1821 0 : ReleasePredXact(sxact);
1822 0 : LWLockRelease(SerializableXactHashLock);
1823 0 : ereport(ERROR,
1824 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1825 : errmsg("could not import the requested snapshot"),
1826 : errdetail("The source process with PID %d is not running anymore.",
1827 : sourcepid)));
1828 : }
1829 :
1830 : /*
1831 : * If there are no serializable transactions which are not read-only, we
1832 : * can "opt out" of predicate locking and conflict checking for a
1833 : * read-only transaction.
1834 : *
1835 : * The reason this is safe is that a read-only transaction can only become
1836 : * part of a dangerous structure if it overlaps a writable transaction
1837 : * which in turn overlaps a writable transaction which committed before
1838 : * the read-only transaction started. A new writable transaction can
1839 : * overlap this one, but it can't meet the other condition of overlapping
1840 : * a transaction which committed before this one started.
1841 : */
1842 3338 : if (XactReadOnly && PredXact->WritableSxactCount == 0)
1843 : {
1844 224 : ReleasePredXact(sxact);
1845 224 : LWLockRelease(SerializableXactHashLock);
1846 224 : return snapshot;
1847 : }
1848 :
1849 : /* Initialize the structure. */
1850 3114 : sxact->vxid = vxid;
1851 3114 : sxact->SeqNo.lastCommitBeforeSnapshot = PredXact->LastSxactCommitSeqNo;
1852 3114 : sxact->prepareSeqNo = InvalidSerCommitSeqNo;
1853 3114 : sxact->commitSeqNo = InvalidSerCommitSeqNo;
1854 3114 : dlist_init(&(sxact->outConflicts));
1855 3114 : dlist_init(&(sxact->inConflicts));
1856 3114 : dlist_init(&(sxact->possibleUnsafeConflicts));
1857 3114 : sxact->topXid = GetTopTransactionIdIfAny();
1858 3114 : sxact->finishedBefore = InvalidTransactionId;
1859 3114 : sxact->xmin = snapshot->xmin;
1860 3114 : sxact->pid = MyProcPid;
1861 3114 : sxact->pgprocno = MyProcNumber;
1862 3114 : dlist_init(&sxact->predicateLocks);
1863 3114 : dlist_node_init(&sxact->finishedLink);
1864 3114 : sxact->flags = 0;
1865 3114 : if (XactReadOnly)
1866 : {
1867 : dlist_iter iter;
1868 :
1869 212 : sxact->flags |= SXACT_FLAG_READ_ONLY;
1870 :
1871 : /*
1872 : * Register all concurrent r/w transactions as possible conflicts; if
1873 : * all of them commit without any outgoing conflicts to earlier
1874 : * transactions then this snapshot can be deemed safe (and we can run
1875 : * without tracking predicate locks).
1876 : */
1877 928 : dlist_foreach(iter, &PredXact->activeList)
1878 : {
1879 716 : othersxact = dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
1880 :
1881 716 : if (!SxactIsCommitted(othersxact)
1882 478 : && !SxactIsDoomed(othersxact)
1883 478 : && !SxactIsReadOnly(othersxact))
1884 : {
1885 264 : SetPossibleUnsafeConflict(sxact, othersxact);
1886 : }
1887 : }
1888 :
1889 : /*
1890 : * If we didn't find any possibly unsafe conflicts because every
1891 : * uncommitted writable transaction turned out to be doomed, then we
1892 : * can "opt out" immediately. See comments above the earlier check
1893 : * for PredXact->WritableSxactCount == 0.
1894 : */
1895 212 : if (dlist_is_empty(&sxact->possibleUnsafeConflicts))
1896 : {
1897 0 : ReleasePredXact(sxact);
1898 0 : LWLockRelease(SerializableXactHashLock);
1899 0 : return snapshot;
1900 : }
1901 : }
1902 : else
1903 : {
1904 2902 : ++(PredXact->WritableSxactCount);
1905 : Assert(PredXact->WritableSxactCount <=
1906 : (MaxBackends + max_prepared_xacts));
1907 : }
1908 :
1909 : /* Maintain serializable global xmin info. */
1910 3114 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
1911 : {
1912 : Assert(PredXact->SxactGlobalXminCount == 0);
1913 1724 : PredXact->SxactGlobalXmin = snapshot->xmin;
1914 1724 : PredXact->SxactGlobalXminCount = 1;
1915 1724 : SerialSetActiveSerXmin(snapshot->xmin);
1916 : }
1917 1390 : else if (TransactionIdEquals(snapshot->xmin, PredXact->SxactGlobalXmin))
1918 : {
1919 : Assert(PredXact->SxactGlobalXminCount > 0);
1920 1316 : PredXact->SxactGlobalXminCount++;
1921 : }
1922 : else
1923 : {
1924 : Assert(TransactionIdFollows(snapshot->xmin, PredXact->SxactGlobalXmin));
1925 : }
1926 :
1927 3114 : MySerializableXact = sxact;
1928 3114 : MyXactDidWrite = false; /* haven't written anything yet */
1929 :
1930 3114 : LWLockRelease(SerializableXactHashLock);
1931 :
1932 3114 : CreateLocalPredicateLockHash();
1933 :
1934 3114 : return snapshot;
1935 : }
1936 :
1937 : static void
1938 3140 : CreateLocalPredicateLockHash(void)
1939 : {
1940 : HASHCTL hash_ctl;
1941 :
1942 : /* Initialize the backend-local hash table of parent locks */
1943 : Assert(LocalPredicateLockHash == NULL);
1944 3140 : hash_ctl.keysize = sizeof(PREDICATELOCKTARGETTAG);
1945 3140 : hash_ctl.entrysize = sizeof(LOCALPREDICATELOCK);
1946 3140 : LocalPredicateLockHash = hash_create("Local predicate lock",
1947 : max_predicate_locks_per_xact,
1948 : &hash_ctl,
1949 : HASH_ELEM | HASH_BLOBS);
1950 3140 : }
1951 :
1952 : /*
1953 : * Register the top level XID in SerializableXidHash.
1954 : * Also store it for easy reference in MySerializableXact.
1955 : */
1956 : void
1957 272452 : RegisterPredicateLockingXid(TransactionId xid)
1958 : {
1959 : SERIALIZABLEXIDTAG sxidtag;
1960 : SERIALIZABLEXID *sxid;
1961 : bool found;
1962 :
1963 : /*
1964 : * If we're not tracking predicate lock data for this transaction, we
1965 : * should ignore the request and return quickly.
1966 : */
1967 272452 : if (MySerializableXact == InvalidSerializableXact)
1968 269870 : return;
1969 :
1970 : /* We should have a valid XID and be at the top level. */
1971 : Assert(TransactionIdIsValid(xid));
1972 :
1973 2582 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1974 :
1975 : /* This should only be done once per transaction. */
1976 : Assert(MySerializableXact->topXid == InvalidTransactionId);
1977 :
1978 2582 : MySerializableXact->topXid = xid;
1979 :
1980 2582 : sxidtag.xid = xid;
1981 2582 : sxid = (SERIALIZABLEXID *) hash_search(SerializableXidHash,
1982 : &sxidtag,
1983 : HASH_ENTER, &found);
1984 : Assert(!found);
1985 :
1986 : /* Initialize the structure. */
1987 2582 : sxid->myXact = MySerializableXact;
1988 2582 : LWLockRelease(SerializableXactHashLock);
1989 : }
1990 :
1991 :
1992 : /*
1993 : * Check whether there are any predicate locks held by any transaction
1994 : * for the page at the given block number.
1995 : *
1996 : * Note that the transaction may be completed but not yet subject to
1997 : * cleanup due to overlapping serializable transactions. This must
1998 : * return valid information regardless of transaction isolation level.
1999 : *
2000 : * Also note that this doesn't check for a conflicting relation lock,
2001 : * just a lock specifically on the given page.
2002 : *
2003 : * One use is to support proper behavior during GiST index vacuum.
2004 : */
2005 : bool
2006 0 : PageIsPredicateLocked(Relation relation, BlockNumber blkno)
2007 : {
2008 : PREDICATELOCKTARGETTAG targettag;
2009 : uint32 targettaghash;
2010 : LWLock *partitionLock;
2011 : PREDICATELOCKTARGET *target;
2012 :
2013 0 : SET_PREDICATELOCKTARGETTAG_PAGE(targettag,
2014 : relation->rd_locator.dbOid,
2015 : relation->rd_id,
2016 : blkno);
2017 :
2018 0 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
2019 0 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
2020 0 : LWLockAcquire(partitionLock, LW_SHARED);
2021 : target = (PREDICATELOCKTARGET *)
2022 0 : hash_search_with_hash_value(PredicateLockTargetHash,
2023 : &targettag, targettaghash,
2024 : HASH_FIND, NULL);
2025 0 : LWLockRelease(partitionLock);
2026 :
2027 0 : return (target != NULL);
2028 : }
2029 :
2030 :
2031 : /*
2032 : * Check whether a particular lock is held by this transaction.
2033 : *
2034 : * Important note: this function may return false even if the lock is
2035 : * being held, because it uses the local lock table which is not
2036 : * updated if another transaction modifies our lock list (e.g. to
2037 : * split an index page). It can also return true when a coarser
2038 : * granularity lock that covers this target is being held. Be careful
2039 : * to only use this function in circumstances where such errors are
2040 : * acceptable!
2041 : */
2042 : static bool
2043 154616 : PredicateLockExists(const PREDICATELOCKTARGETTAG *targettag)
2044 : {
2045 : LOCALPREDICATELOCK *lock;
2046 :
2047 : /* check local hash table */
2048 154616 : lock = (LOCALPREDICATELOCK *) hash_search(LocalPredicateLockHash,
2049 : targettag,
2050 : HASH_FIND, NULL);
2051 :
2052 154616 : if (!lock)
2053 60404 : return false;
2054 :
2055 : /*
2056 : * Found entry in the table, but still need to check whether it's actually
2057 : * held -- it could just be a parent of some held lock.
2058 : */
2059 94212 : return lock->held;
2060 : }
2061 :
2062 : /*
2063 : * Return the parent lock tag in the lock hierarchy: the next coarser
2064 : * lock that covers the provided tag.
2065 : *
2066 : * Returns true and sets *parent to the parent tag if one exists,
2067 : * returns false if none exists.
2068 : */
2069 : static bool
2070 90734 : GetParentPredicateLockTag(const PREDICATELOCKTARGETTAG *tag,
2071 : PREDICATELOCKTARGETTAG *parent)
2072 : {
2073 90734 : switch (GET_PREDICATELOCKTARGETTAG_TYPE(*tag))
2074 : {
2075 19812 : case PREDLOCKTAG_RELATION:
2076 : /* relation locks have no parent lock */
2077 19812 : return false;
2078 :
2079 16976 : case PREDLOCKTAG_PAGE:
2080 : /* parent lock is relation lock */
2081 16976 : SET_PREDICATELOCKTARGETTAG_RELATION(*parent,
2082 : GET_PREDICATELOCKTARGETTAG_DB(*tag),
2083 : GET_PREDICATELOCKTARGETTAG_RELATION(*tag));
2084 :
2085 16976 : return true;
2086 :
2087 53946 : case PREDLOCKTAG_TUPLE:
2088 : /* parent lock is page lock */
2089 53946 : SET_PREDICATELOCKTARGETTAG_PAGE(*parent,
2090 : GET_PREDICATELOCKTARGETTAG_DB(*tag),
2091 : GET_PREDICATELOCKTARGETTAG_RELATION(*tag),
2092 : GET_PREDICATELOCKTARGETTAG_PAGE(*tag));
2093 53946 : return true;
2094 : }
2095 :
2096 : /* not reachable */
2097 : Assert(false);
2098 0 : return false;
2099 : }
2100 :
2101 : /*
2102 : * Check whether the lock we are considering is already covered by a
2103 : * coarser lock for our transaction.
2104 : *
2105 : * Like PredicateLockExists, this function might return a false
2106 : * negative, but it will never return a false positive.
2107 : */
2108 : static bool
2109 52160 : CoarserLockCovers(const PREDICATELOCKTARGETTAG *newtargettag)
2110 : {
2111 : PREDICATELOCKTARGETTAG targettag,
2112 : parenttag;
2113 :
2114 52160 : targettag = *newtargettag;
2115 :
2116 : /* check parents iteratively until no more */
2117 63024 : while (GetParentPredicateLockTag(&targettag, &parenttag))
2118 : {
2119 54506 : targettag = parenttag;
2120 54506 : if (PredicateLockExists(&targettag))
2121 43642 : return true;
2122 : }
2123 :
2124 : /* no more parents to check; lock is not covered */
2125 8518 : return false;
2126 : }
2127 :
2128 : /*
2129 : * Remove the dummy entry from the predicate lock target hash, to free up some
2130 : * scratch space. The caller must be holding SerializablePredicateListLock,
2131 : * and must restore the entry with RestoreScratchTarget() before releasing the
2132 : * lock.
2133 : *
2134 : * If lockheld is true, the caller is already holding the partition lock
2135 : * of the partition containing the scratch entry.
2136 : */
2137 : static void
2138 88 : RemoveScratchTarget(bool lockheld)
2139 : {
2140 : bool found;
2141 :
2142 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2143 :
2144 88 : if (!lockheld)
2145 0 : LWLockAcquire(ScratchPartitionLock, LW_EXCLUSIVE);
2146 88 : hash_search_with_hash_value(PredicateLockTargetHash,
2147 : &ScratchTargetTag,
2148 : ScratchTargetTagHash,
2149 : HASH_REMOVE, &found);
2150 : Assert(found);
2151 88 : if (!lockheld)
2152 0 : LWLockRelease(ScratchPartitionLock);
2153 88 : }
2154 :
2155 : /*
2156 : * Re-insert the dummy entry in predicate lock target hash.
2157 : */
2158 : static void
2159 88 : RestoreScratchTarget(bool lockheld)
2160 : {
2161 : bool found;
2162 :
2163 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2164 :
2165 88 : if (!lockheld)
2166 0 : LWLockAcquire(ScratchPartitionLock, LW_EXCLUSIVE);
2167 88 : hash_search_with_hash_value(PredicateLockTargetHash,
2168 : &ScratchTargetTag,
2169 : ScratchTargetTagHash,
2170 : HASH_ENTER, &found);
2171 : Assert(!found);
2172 88 : if (!lockheld)
2173 0 : LWLockRelease(ScratchPartitionLock);
2174 88 : }
2175 :
2176 : /*
2177 : * Check whether the list of related predicate locks is empty for a
2178 : * predicate lock target, and remove the target if it is.
2179 : */
2180 : static void
2181 8506 : RemoveTargetIfNoLongerUsed(PREDICATELOCKTARGET *target, uint32 targettaghash)
2182 : {
2183 : PREDICATELOCKTARGET *rmtarget PG_USED_FOR_ASSERTS_ONLY;
2184 :
2185 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2186 :
2187 : /* Can't remove it until no locks at this target. */
2188 8506 : if (!dlist_is_empty(&target->predicateLocks))
2189 1946 : return;
2190 :
2191 : /* Actually remove the target. */
2192 6560 : rmtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2193 6560 : &target->tag,
2194 : targettaghash,
2195 : HASH_REMOVE, NULL);
2196 : Assert(rmtarget == target);
2197 : }
2198 :
2199 : /*
2200 : * Delete child target locks owned by this process.
2201 : * This implementation is assuming that the usage of each target tag field
2202 : * is uniform. No need to make this hard if we don't have to.
2203 : *
2204 : * We acquire an LWLock in the case of parallel mode, because worker
2205 : * backends have access to the leader's SERIALIZABLEXACT. Otherwise,
2206 : * we aren't acquiring LWLocks for the predicate lock or lock
2207 : * target structures associated with this transaction unless we're going
2208 : * to modify them, because no other process is permitted to modify our
2209 : * locks.
2210 : */
2211 : static void
2212 4754 : DeleteChildTargetLocks(const PREDICATELOCKTARGETTAG *newtargettag)
2213 : {
2214 : SERIALIZABLEXACT *sxact;
2215 : PREDICATELOCK *predlock;
2216 : dlist_mutable_iter iter;
2217 :
2218 4754 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2219 4754 : sxact = MySerializableXact;
2220 4754 : if (IsInParallelMode())
2221 22 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
2222 :
2223 15686 : dlist_foreach_modify(iter, &sxact->predicateLocks)
2224 : {
2225 : PREDICATELOCKTAG oldlocktag;
2226 : PREDICATELOCKTARGET *oldtarget;
2227 : PREDICATELOCKTARGETTAG oldtargettag;
2228 :
2229 10932 : predlock = dlist_container(PREDICATELOCK, xactLink, iter.cur);
2230 :
2231 10932 : oldlocktag = predlock->tag;
2232 : Assert(oldlocktag.myXact == sxact);
2233 10932 : oldtarget = oldlocktag.myTarget;
2234 10932 : oldtargettag = oldtarget->tag;
2235 :
2236 10932 : if (TargetTagIsCoveredBy(oldtargettag, *newtargettag))
2237 : {
2238 : uint32 oldtargettaghash;
2239 : LWLock *partitionLock;
2240 : PREDICATELOCK *rmpredlock PG_USED_FOR_ASSERTS_ONLY;
2241 :
2242 1998 : oldtargettaghash = PredicateLockTargetTagHashCode(&oldtargettag);
2243 1998 : partitionLock = PredicateLockHashPartitionLock(oldtargettaghash);
2244 :
2245 1998 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
2246 :
2247 1998 : dlist_delete(&predlock->xactLink);
2248 1998 : dlist_delete(&predlock->targetLink);
2249 1998 : rmpredlock = hash_search_with_hash_value
2250 : (PredicateLockHash,
2251 : &oldlocktag,
2252 1998 : PredicateLockHashCodeFromTargetHashCode(&oldlocktag,
2253 : oldtargettaghash),
2254 : HASH_REMOVE, NULL);
2255 : Assert(rmpredlock == predlock);
2256 :
2257 1998 : RemoveTargetIfNoLongerUsed(oldtarget, oldtargettaghash);
2258 :
2259 1998 : LWLockRelease(partitionLock);
2260 :
2261 1998 : DecrementParentLocks(&oldtargettag);
2262 : }
2263 : }
2264 4754 : if (IsInParallelMode())
2265 22 : LWLockRelease(&sxact->perXactPredicateListLock);
2266 4754 : LWLockRelease(SerializablePredicateListLock);
2267 4754 : }
2268 :
2269 : /*
2270 : * Returns the promotion limit for a given predicate lock target. This is the
2271 : * max number of descendant locks allowed before promoting to the specified
2272 : * tag. Note that the limit includes non-direct descendants (e.g., both tuples
2273 : * and pages for a relation lock).
2274 : *
2275 : * Currently the default limit is 2 for a page lock, and half of the value of
2276 : * max_pred_locks_per_transaction - 1 for a relation lock, to match behavior
2277 : * of earlier releases when upgrading.
2278 : *
2279 : * TODO SSI: We should probably add additional GUCs to allow a maximum ratio
2280 : * of page and tuple locks based on the pages in a relation, and the maximum
2281 : * ratio of tuple locks to tuples in a page. This would provide more
2282 : * generally "balanced" allocation of locks to where they are most useful,
2283 : * while still allowing the absolute numbers to prevent one relation from
2284 : * tying up all predicate lock resources.
2285 : */
2286 : static int
2287 10864 : MaxPredicateChildLocks(const PREDICATELOCKTARGETTAG *tag)
2288 : {
2289 10864 : switch (GET_PREDICATELOCKTARGETTAG_TYPE(*tag))
2290 : {
2291 7100 : case PREDLOCKTAG_RELATION:
2292 7100 : return max_predicate_locks_per_relation < 0
2293 : ? (max_predicate_locks_per_xact
2294 7100 : / (-max_predicate_locks_per_relation)) - 1
2295 7100 : : max_predicate_locks_per_relation;
2296 :
2297 3764 : case PREDLOCKTAG_PAGE:
2298 3764 : return max_predicate_locks_per_page;
2299 :
2300 0 : case PREDLOCKTAG_TUPLE:
2301 :
2302 : /*
2303 : * not reachable: nothing is finer-granularity than a tuple, so we
2304 : * should never try to promote to it.
2305 : */
2306 : Assert(false);
2307 0 : return 0;
2308 : }
2309 :
2310 : /* not reachable */
2311 : Assert(false);
2312 0 : return 0;
2313 : }
2314 :
2315 : /*
2316 : * For all ancestors of a newly-acquired predicate lock, increment
2317 : * their child count in the parent hash table. If any of them have
2318 : * more descendants than their promotion threshold, acquire the
2319 : * coarsest such lock.
2320 : *
2321 : * Returns true if a parent lock was acquired and false otherwise.
2322 : */
2323 : static bool
2324 8518 : CheckAndPromotePredicateLockRequest(const PREDICATELOCKTARGETTAG *reqtag)
2325 : {
2326 : PREDICATELOCKTARGETTAG targettag,
2327 : nexttag,
2328 : promotiontag;
2329 : LOCALPREDICATELOCK *parentlock;
2330 : bool found,
2331 : promote;
2332 :
2333 8518 : promote = false;
2334 :
2335 8518 : targettag = *reqtag;
2336 :
2337 : /* check parents iteratively */
2338 27900 : while (GetParentPredicateLockTag(&targettag, &nexttag))
2339 : {
2340 10864 : targettag = nexttag;
2341 10864 : parentlock = (LOCALPREDICATELOCK *) hash_search(LocalPredicateLockHash,
2342 : &targettag,
2343 : HASH_ENTER,
2344 : &found);
2345 10864 : if (!found)
2346 : {
2347 6742 : parentlock->held = false;
2348 6742 : parentlock->childLocks = 1;
2349 : }
2350 : else
2351 4122 : parentlock->childLocks++;
2352 :
2353 10864 : if (parentlock->childLocks >
2354 10864 : MaxPredicateChildLocks(&targettag))
2355 : {
2356 : /*
2357 : * We should promote to this parent lock. Continue to check its
2358 : * ancestors, however, both to get their child counts right and to
2359 : * check whether we should just go ahead and promote to one of
2360 : * them.
2361 : */
2362 666 : promotiontag = targettag;
2363 666 : promote = true;
2364 : }
2365 : }
2366 :
2367 8518 : if (promote)
2368 : {
2369 : /* acquire coarsest ancestor eligible for promotion */
2370 666 : PredicateLockAcquire(&promotiontag);
2371 666 : return true;
2372 : }
2373 : else
2374 7852 : return false;
2375 : }
2376 :
2377 : /*
2378 : * When releasing a lock, decrement the child count on all ancestor
2379 : * locks.
2380 : *
2381 : * This is called only when releasing a lock via
2382 : * DeleteChildTargetLocks (i.e. when a lock becomes redundant because
2383 : * we've acquired its parent, possibly due to promotion) or when a new
2384 : * MVCC write lock makes the predicate lock unnecessary. There's no
2385 : * point in calling it when locks are released at transaction end, as
2386 : * this information is no longer needed.
2387 : */
2388 : static void
2389 2776 : DecrementParentLocks(const PREDICATELOCKTARGETTAG *targettag)
2390 : {
2391 : PREDICATELOCKTARGETTAG parenttag,
2392 : nexttag;
2393 :
2394 2776 : parenttag = *targettag;
2395 :
2396 8328 : while (GetParentPredicateLockTag(&parenttag, &nexttag))
2397 : {
2398 : uint32 targettaghash;
2399 : LOCALPREDICATELOCK *parentlock,
2400 : *rmlock PG_USED_FOR_ASSERTS_ONLY;
2401 :
2402 5552 : parenttag = nexttag;
2403 5552 : targettaghash = PredicateLockTargetTagHashCode(&parenttag);
2404 : parentlock = (LOCALPREDICATELOCK *)
2405 5552 : hash_search_with_hash_value(LocalPredicateLockHash,
2406 : &parenttag, targettaghash,
2407 : HASH_FIND, NULL);
2408 :
2409 : /*
2410 : * There's a small chance the parent lock doesn't exist in the lock
2411 : * table. This can happen if we prematurely removed it because an
2412 : * index split caused the child refcount to be off.
2413 : */
2414 5552 : if (parentlock == NULL)
2415 0 : continue;
2416 :
2417 5552 : parentlock->childLocks--;
2418 :
2419 : /*
2420 : * Under similar circumstances the parent lock's refcount might be
2421 : * zero. This only happens if we're holding that lock (otherwise we
2422 : * would have removed the entry).
2423 : */
2424 5552 : if (parentlock->childLocks < 0)
2425 : {
2426 : Assert(parentlock->held);
2427 0 : parentlock->childLocks = 0;
2428 : }
2429 :
2430 5552 : if ((parentlock->childLocks == 0) && (!parentlock->held))
2431 : {
2432 : rmlock = (LOCALPREDICATELOCK *)
2433 1532 : hash_search_with_hash_value(LocalPredicateLockHash,
2434 : &parenttag, targettaghash,
2435 : HASH_REMOVE, NULL);
2436 : Assert(rmlock == parentlock);
2437 : }
2438 : }
2439 2776 : }
2440 :
2441 : /*
2442 : * Indicate that a predicate lock on the given target is held by the
2443 : * specified transaction. Has no effect if the lock is already held.
2444 : *
2445 : * This updates the lock table and the sxact's lock list, and creates
2446 : * the lock target if necessary, but does *not* do anything related to
2447 : * granularity promotion or the local lock table. See
2448 : * PredicateLockAcquire for that.
2449 : */
2450 : static void
2451 8518 : CreatePredicateLock(const PREDICATELOCKTARGETTAG *targettag,
2452 : uint32 targettaghash,
2453 : SERIALIZABLEXACT *sxact)
2454 : {
2455 : PREDICATELOCKTARGET *target;
2456 : PREDICATELOCKTAG locktag;
2457 : PREDICATELOCK *lock;
2458 : LWLock *partitionLock;
2459 : bool found;
2460 :
2461 8518 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
2462 :
2463 8518 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2464 8518 : if (IsInParallelMode())
2465 32 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
2466 8518 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
2467 :
2468 : /* Make sure that the target is represented. */
2469 : target = (PREDICATELOCKTARGET *)
2470 8518 : hash_search_with_hash_value(PredicateLockTargetHash,
2471 : targettag, targettaghash,
2472 : HASH_ENTER_NULL, &found);
2473 8518 : if (!target)
2474 0 : ereport(ERROR,
2475 : (errcode(ERRCODE_OUT_OF_MEMORY),
2476 : errmsg("out of shared memory"),
2477 : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
2478 8518 : if (!found)
2479 6560 : dlist_init(&target->predicateLocks);
2480 :
2481 : /* We've got the sxact and target, make sure they're joined. */
2482 8518 : locktag.myTarget = target;
2483 8518 : locktag.myXact = sxact;
2484 : lock = (PREDICATELOCK *)
2485 8518 : hash_search_with_hash_value(PredicateLockHash, &locktag,
2486 8518 : PredicateLockHashCodeFromTargetHashCode(&locktag, targettaghash),
2487 : HASH_ENTER_NULL, &found);
2488 8518 : if (!lock)
2489 0 : ereport(ERROR,
2490 : (errcode(ERRCODE_OUT_OF_MEMORY),
2491 : errmsg("out of shared memory"),
2492 : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
2493 :
2494 8518 : if (!found)
2495 : {
2496 8506 : dlist_push_tail(&target->predicateLocks, &lock->targetLink);
2497 8506 : dlist_push_tail(&sxact->predicateLocks, &lock->xactLink);
2498 8506 : lock->commitSeqNo = InvalidSerCommitSeqNo;
2499 : }
2500 :
2501 8518 : LWLockRelease(partitionLock);
2502 8518 : if (IsInParallelMode())
2503 32 : LWLockRelease(&sxact->perXactPredicateListLock);
2504 8518 : LWLockRelease(SerializablePredicateListLock);
2505 8518 : }
2506 :
2507 : /*
2508 : * Acquire a predicate lock on the specified target for the current
2509 : * connection if not already held. This updates the local lock table
2510 : * and uses it to implement granularity promotion. It will consolidate
2511 : * multiple locks into a coarser lock if warranted, and will release
2512 : * any finer-grained locks covered by the new one.
2513 : */
2514 : static void
2515 52570 : PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag)
2516 : {
2517 : uint32 targettaghash;
2518 : bool found;
2519 : LOCALPREDICATELOCK *locallock;
2520 :
2521 : /* Do we have the lock already, or a covering lock? */
2522 52570 : if (PredicateLockExists(targettag))
2523 44052 : return;
2524 :
2525 52160 : if (CoarserLockCovers(targettag))
2526 43642 : return;
2527 :
2528 : /* the same hash and LW lock apply to the lock target and the local lock. */
2529 8518 : targettaghash = PredicateLockTargetTagHashCode(targettag);
2530 :
2531 : /* Acquire lock in local table */
2532 : locallock = (LOCALPREDICATELOCK *)
2533 8518 : hash_search_with_hash_value(LocalPredicateLockHash,
2534 : targettag, targettaghash,
2535 : HASH_ENTER, &found);
2536 8518 : locallock->held = true;
2537 8518 : if (!found)
2538 7852 : locallock->childLocks = 0;
2539 :
2540 : /* Actually create the lock */
2541 8518 : CreatePredicateLock(targettag, targettaghash, MySerializableXact);
2542 :
2543 : /*
2544 : * Lock has been acquired. Check whether it should be promoted to a
2545 : * coarser granularity, or whether there are finer-granularity locks to
2546 : * clean up.
2547 : */
2548 8518 : if (CheckAndPromotePredicateLockRequest(targettag))
2549 : {
2550 : /*
2551 : * Lock request was promoted to a coarser-granularity lock, and that
2552 : * lock was acquired. It will delete this lock and any of its
2553 : * children, so we're done.
2554 : */
2555 : }
2556 : else
2557 : {
2558 : /* Clean up any finer-granularity locks */
2559 7852 : if (GET_PREDICATELOCKTARGETTAG_TYPE(*targettag) != PREDLOCKTAG_TUPLE)
2560 4754 : DeleteChildTargetLocks(targettag);
2561 : }
2562 : }
2563 :
2564 :
2565 : /*
2566 : * PredicateLockRelation
2567 : *
2568 : * Gets a predicate lock at the relation level.
2569 : * Skip if not in full serializable transaction isolation level.
2570 : * Skip if this is a temporary table.
2571 : * Clear any finer-grained predicate locks this session has on the relation.
2572 : */
2573 : void
2574 768884 : PredicateLockRelation(Relation relation, Snapshot snapshot)
2575 : {
2576 : PREDICATELOCKTARGETTAG tag;
2577 :
2578 768884 : if (!SerializationNeededForRead(relation, snapshot))
2579 767424 : return;
2580 :
2581 1460 : SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2582 : relation->rd_locator.dbOid,
2583 : relation->rd_id);
2584 1460 : PredicateLockAcquire(&tag);
2585 : }
2586 :
2587 : /*
2588 : * PredicateLockPage
2589 : *
2590 : * Gets a predicate lock at the page level.
2591 : * Skip if not in full serializable transaction isolation level.
2592 : * Skip if this is a temporary table.
2593 : * Skip if a coarser predicate lock already covers this page.
2594 : * Clear any finer-grained predicate locks this session has on the relation.
2595 : */
2596 : void
2597 19460696 : PredicateLockPage(Relation relation, BlockNumber blkno, Snapshot snapshot)
2598 : {
2599 : PREDICATELOCKTARGETTAG tag;
2600 :
2601 19460696 : if (!SerializationNeededForRead(relation, snapshot))
2602 19457792 : return;
2603 :
2604 2904 : SET_PREDICATELOCKTARGETTAG_PAGE(tag,
2605 : relation->rd_locator.dbOid,
2606 : relation->rd_id,
2607 : blkno);
2608 2904 : PredicateLockAcquire(&tag);
2609 : }
2610 :
2611 : /*
2612 : * PredicateLockTID
2613 : *
2614 : * Gets a predicate lock at the tuple level.
2615 : * Skip if not in full serializable transaction isolation level.
2616 : * Skip if this is a temporary table.
2617 : */
2618 : void
2619 33974530 : PredicateLockTID(Relation relation, const ItemPointerData *tid, Snapshot snapshot,
2620 : TransactionId tuple_xid)
2621 : {
2622 : PREDICATELOCKTARGETTAG tag;
2623 :
2624 33974530 : if (!SerializationNeededForRead(relation, snapshot))
2625 33926990 : return;
2626 :
2627 : /*
2628 : * Return if this xact wrote it.
2629 : */
2630 47544 : if (relation->rd_index == NULL)
2631 : {
2632 : /* If we wrote it; we already have a write lock. */
2633 47544 : if (TransactionIdIsCurrentTransactionId(tuple_xid))
2634 4 : return;
2635 : }
2636 :
2637 : /*
2638 : * Do quick-but-not-definitive test for a relation lock first. This will
2639 : * never cause a return when the relation is *not* locked, but will
2640 : * occasionally let the check continue when there really *is* a relation
2641 : * level lock.
2642 : */
2643 47540 : SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2644 : relation->rd_locator.dbOid,
2645 : relation->rd_id);
2646 47540 : if (PredicateLockExists(&tag))
2647 0 : return;
2648 :
2649 47540 : SET_PREDICATELOCKTARGETTAG_TUPLE(tag,
2650 : relation->rd_locator.dbOid,
2651 : relation->rd_id,
2652 : ItemPointerGetBlockNumber(tid),
2653 : ItemPointerGetOffsetNumber(tid));
2654 47540 : PredicateLockAcquire(&tag);
2655 : }
2656 :
2657 :
2658 : /*
2659 : * DeleteLockTarget
2660 : *
2661 : * Remove a predicate lock target along with any locks held for it.
2662 : *
2663 : * Caller must hold SerializablePredicateListLock and the
2664 : * appropriate hash partition lock for the target.
2665 : */
2666 : static void
2667 0 : DeleteLockTarget(PREDICATELOCKTARGET *target, uint32 targettaghash)
2668 : {
2669 : dlist_mutable_iter iter;
2670 :
2671 : Assert(LWLockHeldByMeInMode(SerializablePredicateListLock,
2672 : LW_EXCLUSIVE));
2673 : Assert(LWLockHeldByMe(PredicateLockHashPartitionLock(targettaghash)));
2674 :
2675 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2676 :
2677 0 : dlist_foreach_modify(iter, &target->predicateLocks)
2678 : {
2679 0 : PREDICATELOCK *predlock =
2680 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
2681 : bool found;
2682 :
2683 0 : dlist_delete(&(predlock->xactLink));
2684 0 : dlist_delete(&(predlock->targetLink));
2685 :
2686 0 : hash_search_with_hash_value
2687 : (PredicateLockHash,
2688 0 : &predlock->tag,
2689 0 : PredicateLockHashCodeFromTargetHashCode(&predlock->tag,
2690 : targettaghash),
2691 : HASH_REMOVE, &found);
2692 : Assert(found);
2693 : }
2694 0 : LWLockRelease(SerializableXactHashLock);
2695 :
2696 : /* Remove the target itself, if possible. */
2697 0 : RemoveTargetIfNoLongerUsed(target, targettaghash);
2698 0 : }
2699 :
2700 :
2701 : /*
2702 : * TransferPredicateLocksToNewTarget
2703 : *
2704 : * Move or copy all the predicate locks for a lock target, for use by
2705 : * index page splits/combines and other things that create or replace
2706 : * lock targets. If 'removeOld' is true, the old locks and the target
2707 : * will be removed.
2708 : *
2709 : * Returns true on success, or false if we ran out of shared memory to
2710 : * allocate the new target or locks. Guaranteed to always succeed if
2711 : * removeOld is set (by using the scratch entry in PredicateLockTargetHash
2712 : * for scratch space).
2713 : *
2714 : * Warning: the "removeOld" option should be used only with care,
2715 : * because this function does not (indeed, can not) update other
2716 : * backends' LocalPredicateLockHash. If we are only adding new
2717 : * entries, this is not a problem: the local lock table is used only
2718 : * as a hint, so missing entries for locks that are held are
2719 : * OK. Having entries for locks that are no longer held, as can happen
2720 : * when using "removeOld", is not in general OK. We can only use it
2721 : * safely when replacing a lock with a coarser-granularity lock that
2722 : * covers it, or if we are absolutely certain that no one will need to
2723 : * refer to that lock in the future.
2724 : *
2725 : * Caller must hold SerializablePredicateListLock exclusively.
2726 : */
2727 : static bool
2728 0 : TransferPredicateLocksToNewTarget(PREDICATELOCKTARGETTAG oldtargettag,
2729 : PREDICATELOCKTARGETTAG newtargettag,
2730 : bool removeOld)
2731 : {
2732 : uint32 oldtargettaghash;
2733 : LWLock *oldpartitionLock;
2734 : PREDICATELOCKTARGET *oldtarget;
2735 : uint32 newtargettaghash;
2736 : LWLock *newpartitionLock;
2737 : bool found;
2738 0 : bool outOfShmem = false;
2739 :
2740 : Assert(LWLockHeldByMeInMode(SerializablePredicateListLock,
2741 : LW_EXCLUSIVE));
2742 :
2743 0 : oldtargettaghash = PredicateLockTargetTagHashCode(&oldtargettag);
2744 0 : newtargettaghash = PredicateLockTargetTagHashCode(&newtargettag);
2745 0 : oldpartitionLock = PredicateLockHashPartitionLock(oldtargettaghash);
2746 0 : newpartitionLock = PredicateLockHashPartitionLock(newtargettaghash);
2747 :
2748 0 : if (removeOld)
2749 : {
2750 : /*
2751 : * Remove the dummy entry to give us scratch space, so we know we'll
2752 : * be able to create the new lock target.
2753 : */
2754 0 : RemoveScratchTarget(false);
2755 : }
2756 :
2757 : /*
2758 : * We must get the partition locks in ascending sequence to avoid
2759 : * deadlocks. If old and new partitions are the same, we must request the
2760 : * lock only once.
2761 : */
2762 0 : if (oldpartitionLock < newpartitionLock)
2763 : {
2764 0 : LWLockAcquire(oldpartitionLock,
2765 0 : (removeOld ? LW_EXCLUSIVE : LW_SHARED));
2766 0 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2767 : }
2768 0 : else if (oldpartitionLock > newpartitionLock)
2769 : {
2770 0 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2771 0 : LWLockAcquire(oldpartitionLock,
2772 0 : (removeOld ? LW_EXCLUSIVE : LW_SHARED));
2773 : }
2774 : else
2775 0 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2776 :
2777 : /*
2778 : * Look for the old target. If not found, that's OK; no predicate locks
2779 : * are affected, so we can just clean up and return. If it does exist,
2780 : * walk its list of predicate locks and move or copy them to the new
2781 : * target.
2782 : */
2783 0 : oldtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2784 : &oldtargettag,
2785 : oldtargettaghash,
2786 : HASH_FIND, NULL);
2787 :
2788 0 : if (oldtarget)
2789 : {
2790 : PREDICATELOCKTARGET *newtarget;
2791 : PREDICATELOCKTAG newpredlocktag;
2792 : dlist_mutable_iter iter;
2793 :
2794 0 : newtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2795 : &newtargettag,
2796 : newtargettaghash,
2797 : HASH_ENTER_NULL, &found);
2798 :
2799 0 : if (!newtarget)
2800 : {
2801 : /* Failed to allocate due to insufficient shmem */
2802 0 : outOfShmem = true;
2803 0 : goto exit;
2804 : }
2805 :
2806 : /* If we created a new entry, initialize it */
2807 0 : if (!found)
2808 0 : dlist_init(&newtarget->predicateLocks);
2809 :
2810 0 : newpredlocktag.myTarget = newtarget;
2811 :
2812 : /*
2813 : * Loop through all the locks on the old target, replacing them with
2814 : * locks on the new target.
2815 : */
2816 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2817 :
2818 0 : dlist_foreach_modify(iter, &oldtarget->predicateLocks)
2819 : {
2820 0 : PREDICATELOCK *oldpredlock =
2821 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
2822 : PREDICATELOCK *newpredlock;
2823 0 : SerCommitSeqNo oldCommitSeqNo = oldpredlock->commitSeqNo;
2824 :
2825 0 : newpredlocktag.myXact = oldpredlock->tag.myXact;
2826 :
2827 0 : if (removeOld)
2828 : {
2829 0 : dlist_delete(&(oldpredlock->xactLink));
2830 0 : dlist_delete(&(oldpredlock->targetLink));
2831 :
2832 0 : hash_search_with_hash_value
2833 : (PredicateLockHash,
2834 0 : &oldpredlock->tag,
2835 0 : PredicateLockHashCodeFromTargetHashCode(&oldpredlock->tag,
2836 : oldtargettaghash),
2837 : HASH_REMOVE, &found);
2838 : Assert(found);
2839 : }
2840 :
2841 : newpredlock = (PREDICATELOCK *)
2842 0 : hash_search_with_hash_value(PredicateLockHash,
2843 : &newpredlocktag,
2844 0 : PredicateLockHashCodeFromTargetHashCode(&newpredlocktag,
2845 : newtargettaghash),
2846 : HASH_ENTER_NULL,
2847 : &found);
2848 0 : if (!newpredlock)
2849 : {
2850 : /* Out of shared memory. Undo what we've done so far. */
2851 0 : LWLockRelease(SerializableXactHashLock);
2852 0 : DeleteLockTarget(newtarget, newtargettaghash);
2853 0 : outOfShmem = true;
2854 0 : goto exit;
2855 : }
2856 0 : if (!found)
2857 : {
2858 0 : dlist_push_tail(&(newtarget->predicateLocks),
2859 : &(newpredlock->targetLink));
2860 0 : dlist_push_tail(&(newpredlocktag.myXact->predicateLocks),
2861 : &(newpredlock->xactLink));
2862 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
2863 : }
2864 : else
2865 : {
2866 0 : if (newpredlock->commitSeqNo < oldCommitSeqNo)
2867 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
2868 : }
2869 :
2870 : Assert(newpredlock->commitSeqNo != 0);
2871 : Assert((newpredlock->commitSeqNo == InvalidSerCommitSeqNo)
2872 : || (newpredlock->tag.myXact == OldCommittedSxact));
2873 : }
2874 0 : LWLockRelease(SerializableXactHashLock);
2875 :
2876 0 : if (removeOld)
2877 : {
2878 : Assert(dlist_is_empty(&oldtarget->predicateLocks));
2879 0 : RemoveTargetIfNoLongerUsed(oldtarget, oldtargettaghash);
2880 : }
2881 : }
2882 :
2883 :
2884 0 : exit:
2885 : /* Release partition locks in reverse order of acquisition. */
2886 0 : if (oldpartitionLock < newpartitionLock)
2887 : {
2888 0 : LWLockRelease(newpartitionLock);
2889 0 : LWLockRelease(oldpartitionLock);
2890 : }
2891 0 : else if (oldpartitionLock > newpartitionLock)
2892 : {
2893 0 : LWLockRelease(oldpartitionLock);
2894 0 : LWLockRelease(newpartitionLock);
2895 : }
2896 : else
2897 0 : LWLockRelease(newpartitionLock);
2898 :
2899 0 : if (removeOld)
2900 : {
2901 : /* We shouldn't run out of memory if we're moving locks */
2902 : Assert(!outOfShmem);
2903 :
2904 : /* Put the scratch entry back */
2905 0 : RestoreScratchTarget(false);
2906 : }
2907 :
2908 0 : return !outOfShmem;
2909 : }
2910 :
2911 : /*
2912 : * Drop all predicate locks of any granularity from the specified relation,
2913 : * which can be a heap relation or an index relation. If 'transfer' is true,
2914 : * acquire a relation lock on the heap for any transactions with any lock(s)
2915 : * on the specified relation.
2916 : *
2917 : * This requires grabbing a lot of LW locks and scanning the entire lock
2918 : * target table for matches. That makes this more expensive than most
2919 : * predicate lock management functions, but it will only be called for DDL
2920 : * type commands that are expensive anyway, and there are fast returns when
2921 : * no serializable transactions are active or the relation is temporary.
2922 : *
2923 : * We don't use the TransferPredicateLocksToNewTarget function because it
2924 : * acquires its own locks on the partitions of the two targets involved,
2925 : * and we'll already be holding all partition locks.
2926 : *
2927 : * We can't throw an error from here, because the call could be from a
2928 : * transaction which is not serializable.
2929 : *
2930 : * NOTE: This is currently only called with transfer set to true, but that may
2931 : * change. If we decide to clean up the locks from a table on commit of a
2932 : * transaction which executed DROP TABLE, the false condition will be useful.
2933 : */
2934 : static void
2935 35232 : DropAllPredicateLocksFromTable(Relation relation, bool transfer)
2936 : {
2937 : HASH_SEQ_STATUS seqstat;
2938 : PREDICATELOCKTARGET *oldtarget;
2939 : PREDICATELOCKTARGET *heaptarget;
2940 : Oid dbId;
2941 : Oid relId;
2942 : Oid heapId;
2943 : int i;
2944 : bool isIndex;
2945 : bool found;
2946 : uint32 heaptargettaghash;
2947 :
2948 : /*
2949 : * Bail out quickly if there are no serializable transactions running.
2950 : * It's safe to check this without taking locks because the caller is
2951 : * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
2952 : * would matter here can be acquired while that is held.
2953 : */
2954 35232 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
2955 35144 : return;
2956 :
2957 120 : if (!PredicateLockingNeededForRelation(relation))
2958 32 : return;
2959 :
2960 88 : dbId = relation->rd_locator.dbOid;
2961 88 : relId = relation->rd_id;
2962 88 : if (relation->rd_index == NULL)
2963 : {
2964 10 : isIndex = false;
2965 10 : heapId = relId;
2966 : }
2967 : else
2968 : {
2969 78 : isIndex = true;
2970 78 : heapId = relation->rd_index->indrelid;
2971 : }
2972 : Assert(heapId != InvalidOid);
2973 : Assert(transfer || !isIndex); /* index OID only makes sense with
2974 : * transfer */
2975 :
2976 : /* Retrieve first time needed, then keep. */
2977 88 : heaptargettaghash = 0;
2978 88 : heaptarget = NULL;
2979 :
2980 : /* Acquire locks on all lock partitions */
2981 88 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
2982 1496 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
2983 1408 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_EXCLUSIVE);
2984 88 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2985 :
2986 : /*
2987 : * Remove the dummy entry to give us scratch space, so we know we'll be
2988 : * able to create the new lock target.
2989 : */
2990 88 : if (transfer)
2991 88 : RemoveScratchTarget(true);
2992 :
2993 : /* Scan through target map */
2994 88 : hash_seq_init(&seqstat, PredicateLockTargetHash);
2995 :
2996 176 : while ((oldtarget = (PREDICATELOCKTARGET *) hash_seq_search(&seqstat)))
2997 : {
2998 : dlist_mutable_iter iter;
2999 :
3000 : /*
3001 : * Check whether this is a target which needs attention.
3002 : */
3003 88 : if (GET_PREDICATELOCKTARGETTAG_RELATION(oldtarget->tag) != relId)
3004 88 : continue; /* wrong relation id */
3005 0 : if (GET_PREDICATELOCKTARGETTAG_DB(oldtarget->tag) != dbId)
3006 0 : continue; /* wrong database id */
3007 0 : if (transfer && !isIndex
3008 0 : && GET_PREDICATELOCKTARGETTAG_TYPE(oldtarget->tag) == PREDLOCKTAG_RELATION)
3009 0 : continue; /* already the right lock */
3010 :
3011 : /*
3012 : * If we made it here, we have work to do. We make sure the heap
3013 : * relation lock exists, then we walk the list of predicate locks for
3014 : * the old target we found, moving all locks to the heap relation lock
3015 : * -- unless they already hold that.
3016 : */
3017 :
3018 : /*
3019 : * First make sure we have the heap relation target. We only need to
3020 : * do this once.
3021 : */
3022 0 : if (transfer && heaptarget == NULL)
3023 : {
3024 : PREDICATELOCKTARGETTAG heaptargettag;
3025 :
3026 0 : SET_PREDICATELOCKTARGETTAG_RELATION(heaptargettag, dbId, heapId);
3027 0 : heaptargettaghash = PredicateLockTargetTagHashCode(&heaptargettag);
3028 0 : heaptarget = hash_search_with_hash_value(PredicateLockTargetHash,
3029 : &heaptargettag,
3030 : heaptargettaghash,
3031 : HASH_ENTER, &found);
3032 0 : if (!found)
3033 0 : dlist_init(&heaptarget->predicateLocks);
3034 : }
3035 :
3036 : /*
3037 : * Loop through all the locks on the old target, replacing them with
3038 : * locks on the new target.
3039 : */
3040 0 : dlist_foreach_modify(iter, &oldtarget->predicateLocks)
3041 : {
3042 0 : PREDICATELOCK *oldpredlock =
3043 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
3044 : PREDICATELOCK *newpredlock;
3045 : SerCommitSeqNo oldCommitSeqNo;
3046 : SERIALIZABLEXACT *oldXact;
3047 :
3048 : /*
3049 : * Remove the old lock first. This avoids the chance of running
3050 : * out of lock structure entries for the hash table.
3051 : */
3052 0 : oldCommitSeqNo = oldpredlock->commitSeqNo;
3053 0 : oldXact = oldpredlock->tag.myXact;
3054 :
3055 0 : dlist_delete(&(oldpredlock->xactLink));
3056 :
3057 : /*
3058 : * No need for retail delete from oldtarget list, we're removing
3059 : * the whole target anyway.
3060 : */
3061 0 : hash_search(PredicateLockHash,
3062 0 : &oldpredlock->tag,
3063 : HASH_REMOVE, &found);
3064 : Assert(found);
3065 :
3066 0 : if (transfer)
3067 : {
3068 : PREDICATELOCKTAG newpredlocktag;
3069 :
3070 0 : newpredlocktag.myTarget = heaptarget;
3071 0 : newpredlocktag.myXact = oldXact;
3072 : newpredlock = (PREDICATELOCK *)
3073 0 : hash_search_with_hash_value(PredicateLockHash,
3074 : &newpredlocktag,
3075 0 : PredicateLockHashCodeFromTargetHashCode(&newpredlocktag,
3076 : heaptargettaghash),
3077 : HASH_ENTER,
3078 : &found);
3079 0 : if (!found)
3080 : {
3081 0 : dlist_push_tail(&(heaptarget->predicateLocks),
3082 : &(newpredlock->targetLink));
3083 0 : dlist_push_tail(&(newpredlocktag.myXact->predicateLocks),
3084 : &(newpredlock->xactLink));
3085 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
3086 : }
3087 : else
3088 : {
3089 0 : if (newpredlock->commitSeqNo < oldCommitSeqNo)
3090 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
3091 : }
3092 :
3093 : Assert(newpredlock->commitSeqNo != 0);
3094 : Assert((newpredlock->commitSeqNo == InvalidSerCommitSeqNo)
3095 : || (newpredlock->tag.myXact == OldCommittedSxact));
3096 : }
3097 : }
3098 :
3099 0 : hash_search(PredicateLockTargetHash, &oldtarget->tag, HASH_REMOVE,
3100 : &found);
3101 : Assert(found);
3102 : }
3103 :
3104 : /* Put the scratch entry back */
3105 88 : if (transfer)
3106 88 : RestoreScratchTarget(true);
3107 :
3108 : /* Release locks in reverse order */
3109 88 : LWLockRelease(SerializableXactHashLock);
3110 1496 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
3111 1408 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
3112 88 : LWLockRelease(SerializablePredicateListLock);
3113 : }
3114 :
3115 : /*
3116 : * TransferPredicateLocksToHeapRelation
3117 : * For all transactions, transfer all predicate locks for the given
3118 : * relation to a single relation lock on the heap.
3119 : */
3120 : void
3121 35232 : TransferPredicateLocksToHeapRelation(Relation relation)
3122 : {
3123 35232 : DropAllPredicateLocksFromTable(relation, true);
3124 35232 : }
3125 :
3126 :
3127 : /*
3128 : * PredicateLockPageSplit
3129 : *
3130 : * Copies any predicate locks for the old page to the new page.
3131 : * Skip if this is a temporary table or toast table.
3132 : *
3133 : * NOTE: A page split (or overflow) affects all serializable transactions,
3134 : * even if it occurs in the context of another transaction isolation level.
3135 : *
3136 : * NOTE: This currently leaves the local copy of the locks without
3137 : * information on the new lock which is in shared memory. This could cause
3138 : * problems if enough page splits occur on locked pages without the processes
3139 : * which hold the locks getting in and noticing.
3140 : */
3141 : void
3142 60412 : PredicateLockPageSplit(Relation relation, BlockNumber oldblkno,
3143 : BlockNumber newblkno)
3144 : {
3145 : PREDICATELOCKTARGETTAG oldtargettag;
3146 : PREDICATELOCKTARGETTAG newtargettag;
3147 : bool success;
3148 :
3149 : /*
3150 : * Bail out quickly if there are no serializable transactions running.
3151 : *
3152 : * It's safe to do this check without taking any additional locks. Even if
3153 : * a serializable transaction starts concurrently, we know it can't take
3154 : * any SIREAD locks on the page being split because the caller is holding
3155 : * the associated buffer page lock. Memory reordering isn't an issue; the
3156 : * memory barrier in the LWLock acquisition guarantees that this read
3157 : * occurs while the buffer page lock is held.
3158 : */
3159 60412 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
3160 60412 : return;
3161 :
3162 30 : if (!PredicateLockingNeededForRelation(relation))
3163 30 : return;
3164 :
3165 : Assert(oldblkno != newblkno);
3166 : Assert(BlockNumberIsValid(oldblkno));
3167 : Assert(BlockNumberIsValid(newblkno));
3168 :
3169 0 : SET_PREDICATELOCKTARGETTAG_PAGE(oldtargettag,
3170 : relation->rd_locator.dbOid,
3171 : relation->rd_id,
3172 : oldblkno);
3173 0 : SET_PREDICATELOCKTARGETTAG_PAGE(newtargettag,
3174 : relation->rd_locator.dbOid,
3175 : relation->rd_id,
3176 : newblkno);
3177 :
3178 0 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
3179 :
3180 : /*
3181 : * Try copying the locks over to the new page's tag, creating it if
3182 : * necessary.
3183 : */
3184 0 : success = TransferPredicateLocksToNewTarget(oldtargettag,
3185 : newtargettag,
3186 : false);
3187 :
3188 0 : if (!success)
3189 : {
3190 : /*
3191 : * No more predicate lock entries are available. Failure isn't an
3192 : * option here, so promote the page lock to a relation lock.
3193 : */
3194 :
3195 : /* Get the parent relation lock's lock tag */
3196 0 : success = GetParentPredicateLockTag(&oldtargettag,
3197 : &newtargettag);
3198 : Assert(success);
3199 :
3200 : /*
3201 : * Move the locks to the parent. This shouldn't fail.
3202 : *
3203 : * Note that here we are removing locks held by other backends,
3204 : * leading to a possible inconsistency in their local lock hash table.
3205 : * This is OK because we're replacing it with a lock that covers the
3206 : * old one.
3207 : */
3208 0 : success = TransferPredicateLocksToNewTarget(oldtargettag,
3209 : newtargettag,
3210 : true);
3211 : Assert(success);
3212 : }
3213 :
3214 0 : LWLockRelease(SerializablePredicateListLock);
3215 : }
3216 :
3217 : /*
3218 : * PredicateLockPageCombine
3219 : *
3220 : * Combines predicate locks for two existing pages.
3221 : * Skip if this is a temporary table or toast table.
3222 : *
3223 : * NOTE: A page combine affects all serializable transactions, even if it
3224 : * occurs in the context of another transaction isolation level.
3225 : */
3226 : void
3227 5468 : PredicateLockPageCombine(Relation relation, BlockNumber oldblkno,
3228 : BlockNumber newblkno)
3229 : {
3230 : /*
3231 : * Page combines differ from page splits in that we ought to be able to
3232 : * remove the locks on the old page after transferring them to the new
3233 : * page, instead of duplicating them. However, because we can't edit other
3234 : * backends' local lock tables, removing the old lock would leave them
3235 : * with an entry in their LocalPredicateLockHash for a lock they're not
3236 : * holding, which isn't acceptable. So we wind up having to do the same
3237 : * work as a page split, acquiring a lock on the new page and keeping the
3238 : * old page locked too. That can lead to some false positives, but should
3239 : * be rare in practice.
3240 : */
3241 5468 : PredicateLockPageSplit(relation, oldblkno, newblkno);
3242 5468 : }
3243 :
3244 : /*
3245 : * Walk the list of in-progress serializable transactions and find the new
3246 : * xmin.
3247 : */
3248 : static void
3249 1760 : SetNewSxactGlobalXmin(void)
3250 : {
3251 : dlist_iter iter;
3252 :
3253 : Assert(LWLockHeldByMe(SerializableXactHashLock));
3254 :
3255 1760 : PredXact->SxactGlobalXmin = InvalidTransactionId;
3256 1760 : PredXact->SxactGlobalXminCount = 0;
3257 :
3258 6660 : dlist_foreach(iter, &PredXact->activeList)
3259 : {
3260 4900 : SERIALIZABLEXACT *sxact =
3261 4900 : dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
3262 :
3263 4900 : if (!SxactIsRolledBack(sxact)
3264 4294 : && !SxactIsCommitted(sxact)
3265 36 : && sxact != OldCommittedSxact)
3266 : {
3267 : Assert(sxact->xmin != InvalidTransactionId);
3268 36 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)
3269 0 : || TransactionIdPrecedes(sxact->xmin,
3270 0 : PredXact->SxactGlobalXmin))
3271 : {
3272 36 : PredXact->SxactGlobalXmin = sxact->xmin;
3273 36 : PredXact->SxactGlobalXminCount = 1;
3274 : }
3275 0 : else if (TransactionIdEquals(sxact->xmin,
3276 : PredXact->SxactGlobalXmin))
3277 0 : PredXact->SxactGlobalXminCount++;
3278 : }
3279 : }
3280 :
3281 1760 : SerialSetActiveSerXmin(PredXact->SxactGlobalXmin);
3282 1760 : }
3283 :
3284 : /*
3285 : * ReleasePredicateLocks
3286 : *
3287 : * Releases predicate locks based on completion of the current transaction,
3288 : * whether committed or rolled back. It can also be called for a read only
3289 : * transaction when it becomes impossible for the transaction to become
3290 : * part of a dangerous structure.
3291 : *
3292 : * We do nothing unless this is a serializable transaction.
3293 : *
3294 : * This method must ensure that shared memory hash tables are cleaned
3295 : * up in some relatively timely fashion.
3296 : *
3297 : * If this transaction is committing and is holding any predicate locks,
3298 : * it must be added to a list of completed serializable transactions still
3299 : * holding locks.
3300 : *
3301 : * If isReadOnlySafe is true, then predicate locks are being released before
3302 : * the end of the transaction because MySerializableXact has been determined
3303 : * to be RO_SAFE. In non-parallel mode we can release it completely, but it
3304 : * in parallel mode we partially release the SERIALIZABLEXACT and keep it
3305 : * around until the end of the transaction, allowing each backend to clear its
3306 : * MySerializableXact variable and benefit from the optimization in its own
3307 : * time.
3308 : */
3309 : void
3310 1042368 : ReleasePredicateLocks(bool isCommit, bool isReadOnlySafe)
3311 : {
3312 1042368 : bool partiallyReleasing = false;
3313 : bool needToClear;
3314 : SERIALIZABLEXACT *roXact;
3315 : dlist_mutable_iter iter;
3316 :
3317 : /*
3318 : * We can't trust XactReadOnly here, because a transaction which started
3319 : * as READ WRITE can show as READ ONLY later, e.g., within
3320 : * subtransactions. We want to flag a transaction as READ ONLY if it
3321 : * commits without writing so that de facto READ ONLY transactions get the
3322 : * benefit of some RO optimizations, so we will use this local variable to
3323 : * get some cleanup logic right which is based on whether the transaction
3324 : * was declared READ ONLY at the top level.
3325 : */
3326 : bool topLevelIsDeclaredReadOnly;
3327 :
3328 : /* We can't be both committing and releasing early due to RO_SAFE. */
3329 : Assert(!(isCommit && isReadOnlySafe));
3330 :
3331 : /* Are we at the end of a transaction, that is, a commit or abort? */
3332 1042368 : if (!isReadOnlySafe)
3333 : {
3334 : /*
3335 : * Parallel workers mustn't release predicate locks at the end of
3336 : * their transaction. The leader will do that at the end of its
3337 : * transaction.
3338 : */
3339 1042302 : if (IsParallelWorker())
3340 : {
3341 8628 : ReleasePredicateLocksLocal();
3342 1039250 : return;
3343 : }
3344 :
3345 : /*
3346 : * By the time the leader in a parallel query reaches end of
3347 : * transaction, it has waited for all workers to exit.
3348 : */
3349 : Assert(!ParallelContextActive());
3350 :
3351 : /*
3352 : * If the leader in a parallel query earlier stashed a partially
3353 : * released SERIALIZABLEXACT for final clean-up at end of transaction
3354 : * (because workers might still have been accessing it), then it's
3355 : * time to restore it.
3356 : */
3357 1033674 : if (SavedSerializableXact != InvalidSerializableXact)
3358 : {
3359 : Assert(MySerializableXact == InvalidSerializableXact);
3360 2 : MySerializableXact = SavedSerializableXact;
3361 2 : SavedSerializableXact = InvalidSerializableXact;
3362 : Assert(SxactIsPartiallyReleased(MySerializableXact));
3363 : }
3364 : }
3365 :
3366 1033740 : if (MySerializableXact == InvalidSerializableXact)
3367 : {
3368 : Assert(LocalPredicateLockHash == NULL);
3369 1030616 : return;
3370 : }
3371 :
3372 3124 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
3373 :
3374 : /*
3375 : * If the transaction is committing, but it has been partially released
3376 : * already, then treat this as a roll back. It was marked as rolled back.
3377 : */
3378 3124 : if (isCommit && SxactIsPartiallyReleased(MySerializableXact))
3379 4 : isCommit = false;
3380 :
3381 : /*
3382 : * If we're called in the middle of a transaction because we discovered
3383 : * that the SXACT_FLAG_RO_SAFE flag was set, then we'll partially release
3384 : * it (that is, release the predicate locks and conflicts, but not the
3385 : * SERIALIZABLEXACT itself) if we're the first backend to have noticed.
3386 : */
3387 3124 : if (isReadOnlySafe && IsInParallelMode())
3388 : {
3389 : /*
3390 : * The leader needs to stash a pointer to it, so that it can
3391 : * completely release it at end-of-transaction.
3392 : */
3393 10 : if (!IsParallelWorker())
3394 2 : SavedSerializableXact = MySerializableXact;
3395 :
3396 : /*
3397 : * The first backend to reach this condition will partially release
3398 : * the SERIALIZABLEXACT. All others will just clear their
3399 : * backend-local state so that they stop doing SSI checks for the rest
3400 : * of the transaction.
3401 : */
3402 10 : if (SxactIsPartiallyReleased(MySerializableXact))
3403 : {
3404 6 : LWLockRelease(SerializableXactHashLock);
3405 6 : ReleasePredicateLocksLocal();
3406 6 : return;
3407 : }
3408 : else
3409 : {
3410 4 : MySerializableXact->flags |= SXACT_FLAG_PARTIALLY_RELEASED;
3411 4 : partiallyReleasing = true;
3412 : /* ... and proceed to perform the partial release below. */
3413 : }
3414 : }
3415 : Assert(!isCommit || SxactIsPrepared(MySerializableXact));
3416 : Assert(!isCommit || !SxactIsDoomed(MySerializableXact));
3417 : Assert(!SxactIsCommitted(MySerializableXact));
3418 : Assert(SxactIsPartiallyReleased(MySerializableXact)
3419 : || !SxactIsRolledBack(MySerializableXact));
3420 :
3421 : /* may not be serializable during COMMIT/ROLLBACK PREPARED */
3422 : Assert(MySerializableXact->pid == 0 || IsolationIsSerializable());
3423 :
3424 : /* We'd better not already be on the cleanup list. */
3425 : Assert(!SxactIsOnFinishedList(MySerializableXact));
3426 :
3427 3118 : topLevelIsDeclaredReadOnly = SxactIsReadOnly(MySerializableXact);
3428 :
3429 : /*
3430 : * We don't hold XidGenLock lock here, assuming that TransactionId is
3431 : * atomic!
3432 : *
3433 : * If this value is changing, we don't care that much whether we get the
3434 : * old or new value -- it is just used to determine how far
3435 : * SxactGlobalXmin must advance before this transaction can be fully
3436 : * cleaned up. The worst that could happen is we wait for one more
3437 : * transaction to complete before freeing some RAM; correctness of visible
3438 : * behavior is not affected.
3439 : */
3440 3118 : MySerializableXact->finishedBefore = XidFromFullTransactionId(TransamVariables->nextXid);
3441 :
3442 : /*
3443 : * If it's not a commit it's either a rollback or a read-only transaction
3444 : * flagged SXACT_FLAG_RO_SAFE, and we can clear our locks immediately.
3445 : */
3446 3118 : if (isCommit)
3447 : {
3448 2462 : MySerializableXact->flags |= SXACT_FLAG_COMMITTED;
3449 2462 : MySerializableXact->commitSeqNo = ++(PredXact->LastSxactCommitSeqNo);
3450 : /* Recognize implicit read-only transaction (commit without write). */
3451 2462 : if (!MyXactDidWrite)
3452 466 : MySerializableXact->flags |= SXACT_FLAG_READ_ONLY;
3453 : }
3454 : else
3455 : {
3456 : /*
3457 : * The DOOMED flag indicates that we intend to roll back this
3458 : * transaction and so it should not cause serialization failures for
3459 : * other transactions that conflict with it. Note that this flag might
3460 : * already be set, if another backend marked this transaction for
3461 : * abort.
3462 : *
3463 : * The ROLLED_BACK flag further indicates that ReleasePredicateLocks
3464 : * has been called, and so the SerializableXact is eligible for
3465 : * cleanup. This means it should not be considered when calculating
3466 : * SxactGlobalXmin.
3467 : */
3468 656 : MySerializableXact->flags |= SXACT_FLAG_DOOMED;
3469 656 : MySerializableXact->flags |= SXACT_FLAG_ROLLED_BACK;
3470 :
3471 : /*
3472 : * If the transaction was previously prepared, but is now failing due
3473 : * to a ROLLBACK PREPARED or (hopefully very rare) error after the
3474 : * prepare, clear the prepared flag. This simplifies conflict
3475 : * checking.
3476 : */
3477 656 : MySerializableXact->flags &= ~SXACT_FLAG_PREPARED;
3478 : }
3479 :
3480 3118 : if (!topLevelIsDeclaredReadOnly)
3481 : {
3482 : Assert(PredXact->WritableSxactCount > 0);
3483 2902 : if (--(PredXact->WritableSxactCount) == 0)
3484 : {
3485 : /*
3486 : * Release predicate locks and rw-conflicts in for all committed
3487 : * transactions. There are no longer any transactions which might
3488 : * conflict with the locks and no chance for new transactions to
3489 : * overlap. Similarly, existing conflicts in can't cause pivots,
3490 : * and any conflicts in which could have completed a dangerous
3491 : * structure would already have caused a rollback, so any
3492 : * remaining ones must be benign.
3493 : */
3494 1744 : PredXact->CanPartialClearThrough = PredXact->LastSxactCommitSeqNo;
3495 : }
3496 : }
3497 : else
3498 : {
3499 : /*
3500 : * Read-only transactions: clear the list of transactions that might
3501 : * make us unsafe. Note that we use 'inLink' for the iteration as
3502 : * opposed to 'outLink' for the r/w xacts.
3503 : */
3504 300 : dlist_foreach_modify(iter, &MySerializableXact->possibleUnsafeConflicts)
3505 : {
3506 84 : RWConflict possibleUnsafeConflict =
3507 84 : dlist_container(RWConflictData, inLink, iter.cur);
3508 :
3509 : Assert(!SxactIsReadOnly(possibleUnsafeConflict->sxactOut));
3510 : Assert(MySerializableXact == possibleUnsafeConflict->sxactIn);
3511 :
3512 84 : ReleaseRWConflict(possibleUnsafeConflict);
3513 : }
3514 : }
3515 :
3516 : /* Check for conflict out to old committed transactions. */
3517 3118 : if (isCommit
3518 2462 : && !SxactIsReadOnly(MySerializableXact)
3519 1996 : && SxactHasSummaryConflictOut(MySerializableXact))
3520 : {
3521 : /*
3522 : * we don't know which old committed transaction we conflicted with,
3523 : * so be conservative and use FirstNormalSerCommitSeqNo here
3524 : */
3525 0 : MySerializableXact->SeqNo.earliestOutConflictCommit =
3526 : FirstNormalSerCommitSeqNo;
3527 0 : MySerializableXact->flags |= SXACT_FLAG_CONFLICT_OUT;
3528 : }
3529 :
3530 : /*
3531 : * Release all outConflicts to committed transactions. If we're rolling
3532 : * back clear them all. Set SXACT_FLAG_CONFLICT_OUT if any point to
3533 : * previously committed transactions.
3534 : */
3535 4492 : dlist_foreach_modify(iter, &MySerializableXact->outConflicts)
3536 : {
3537 1374 : RWConflict conflict =
3538 : dlist_container(RWConflictData, outLink, iter.cur);
3539 :
3540 1374 : if (isCommit
3541 910 : && !SxactIsReadOnly(MySerializableXact)
3542 694 : && SxactIsCommitted(conflict->sxactIn))
3543 : {
3544 192 : if ((MySerializableXact->flags & SXACT_FLAG_CONFLICT_OUT) == 0
3545 0 : || conflict->sxactIn->prepareSeqNo < MySerializableXact->SeqNo.earliestOutConflictCommit)
3546 192 : MySerializableXact->SeqNo.earliestOutConflictCommit = conflict->sxactIn->prepareSeqNo;
3547 192 : MySerializableXact->flags |= SXACT_FLAG_CONFLICT_OUT;
3548 : }
3549 :
3550 1374 : if (!isCommit
3551 910 : || SxactIsCommitted(conflict->sxactIn)
3552 674 : || (conflict->sxactIn->SeqNo.lastCommitBeforeSnapshot >= PredXact->LastSxactCommitSeqNo))
3553 700 : ReleaseRWConflict(conflict);
3554 : }
3555 :
3556 : /*
3557 : * Release all inConflicts from committed and read-only transactions. If
3558 : * we're rolling back, clear them all.
3559 : */
3560 4680 : dlist_foreach_modify(iter, &MySerializableXact->inConflicts)
3561 : {
3562 1562 : RWConflict conflict =
3563 1562 : dlist_container(RWConflictData, inLink, iter.cur);
3564 :
3565 1562 : if (!isCommit
3566 1208 : || SxactIsCommitted(conflict->sxactOut)
3567 838 : || SxactIsReadOnly(conflict->sxactOut))
3568 884 : ReleaseRWConflict(conflict);
3569 : }
3570 :
3571 3118 : if (!topLevelIsDeclaredReadOnly)
3572 : {
3573 : /*
3574 : * Remove ourselves from the list of possible conflicts for concurrent
3575 : * READ ONLY transactions, flagging them as unsafe if we have a
3576 : * conflict out. If any are waiting DEFERRABLE transactions, wake them
3577 : * up if they are known safe or known unsafe.
3578 : */
3579 3082 : dlist_foreach_modify(iter, &MySerializableXact->possibleUnsafeConflicts)
3580 : {
3581 180 : RWConflict possibleUnsafeConflict =
3582 : dlist_container(RWConflictData, outLink, iter.cur);
3583 :
3584 180 : roXact = possibleUnsafeConflict->sxactIn;
3585 : Assert(MySerializableXact == possibleUnsafeConflict->sxactOut);
3586 : Assert(SxactIsReadOnly(roXact));
3587 :
3588 : /* Mark conflicted if necessary. */
3589 180 : if (isCommit
3590 176 : && MyXactDidWrite
3591 166 : && SxactHasConflictOut(MySerializableXact)
3592 26 : && (MySerializableXact->SeqNo.earliestOutConflictCommit
3593 26 : <= roXact->SeqNo.lastCommitBeforeSnapshot))
3594 : {
3595 : /*
3596 : * This releases possibleUnsafeConflict (as well as all other
3597 : * possible conflicts for roXact)
3598 : */
3599 6 : FlagSxactUnsafe(roXact);
3600 : }
3601 : else
3602 : {
3603 174 : ReleaseRWConflict(possibleUnsafeConflict);
3604 :
3605 : /*
3606 : * If we were the last possible conflict, flag it safe. The
3607 : * transaction can now safely release its predicate locks (but
3608 : * that transaction's backend has to do that itself).
3609 : */
3610 174 : if (dlist_is_empty(&roXact->possibleUnsafeConflicts))
3611 130 : roXact->flags |= SXACT_FLAG_RO_SAFE;
3612 : }
3613 :
3614 : /*
3615 : * Wake up the process for a waiting DEFERRABLE transaction if we
3616 : * now know it's either safe or conflicted.
3617 : */
3618 180 : if (SxactIsDeferrableWaiting(roXact) &&
3619 2 : (SxactIsROUnsafe(roXact) || SxactIsROSafe(roXact)))
3620 2 : ProcSendSignal(roXact->pgprocno);
3621 : }
3622 : }
3623 :
3624 : /*
3625 : * Check whether it's time to clean up old transactions. This can only be
3626 : * done when the last serializable transaction with the oldest xmin among
3627 : * serializable transactions completes. We then find the "new oldest"
3628 : * xmin and purge any transactions which finished before this transaction
3629 : * was launched.
3630 : *
3631 : * For parallel queries in read-only transactions, it might run twice. We
3632 : * only release the reference on the first call.
3633 : */
3634 3118 : needToClear = false;
3635 3118 : if ((partiallyReleasing ||
3636 3114 : !SxactIsPartiallyReleased(MySerializableXact)) &&
3637 3114 : TransactionIdEquals(MySerializableXact->xmin,
3638 : PredXact->SxactGlobalXmin))
3639 : {
3640 : Assert(PredXact->SxactGlobalXminCount > 0);
3641 3076 : if (--(PredXact->SxactGlobalXminCount) == 0)
3642 : {
3643 1760 : SetNewSxactGlobalXmin();
3644 1760 : needToClear = true;
3645 : }
3646 : }
3647 :
3648 3118 : LWLockRelease(SerializableXactHashLock);
3649 :
3650 3118 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
3651 :
3652 : /* Add this to the list of transactions to check for later cleanup. */
3653 3118 : if (isCommit)
3654 2462 : dlist_push_tail(FinishedSerializableTransactions,
3655 2462 : &MySerializableXact->finishedLink);
3656 :
3657 : /*
3658 : * If we're releasing a RO_SAFE transaction in parallel mode, we'll only
3659 : * partially release it. That's necessary because other backends may have
3660 : * a reference to it. The leader will release the SERIALIZABLEXACT itself
3661 : * at the end of the transaction after workers have stopped running.
3662 : */
3663 3118 : if (!isCommit)
3664 656 : ReleaseOneSerializableXact(MySerializableXact,
3665 656 : isReadOnlySafe && IsInParallelMode(),
3666 : false);
3667 :
3668 3118 : LWLockRelease(SerializableFinishedListLock);
3669 :
3670 3118 : if (needToClear)
3671 1760 : ClearOldPredicateLocks();
3672 :
3673 3118 : ReleasePredicateLocksLocal();
3674 : }
3675 :
3676 : static void
3677 11752 : ReleasePredicateLocksLocal(void)
3678 : {
3679 11752 : MySerializableXact = InvalidSerializableXact;
3680 11752 : MyXactDidWrite = false;
3681 :
3682 : /* Delete per-transaction lock table */
3683 11752 : if (LocalPredicateLockHash != NULL)
3684 : {
3685 3116 : hash_destroy(LocalPredicateLockHash);
3686 3116 : LocalPredicateLockHash = NULL;
3687 : }
3688 11752 : }
3689 :
3690 : /*
3691 : * Clear old predicate locks, belonging to committed transactions that are no
3692 : * longer interesting to any in-progress transaction.
3693 : */
3694 : static void
3695 1760 : ClearOldPredicateLocks(void)
3696 : {
3697 : dlist_mutable_iter iter;
3698 :
3699 : /*
3700 : * Loop through finished transactions. They are in commit order, so we can
3701 : * stop as soon as we find one that's still interesting.
3702 : */
3703 1760 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
3704 1760 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3705 4238 : dlist_foreach_modify(iter, FinishedSerializableTransactions)
3706 : {
3707 2496 : SERIALIZABLEXACT *finishedSxact =
3708 2496 : dlist_container(SERIALIZABLEXACT, finishedLink, iter.cur);
3709 :
3710 2496 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)
3711 56 : || TransactionIdPrecedesOrEquals(finishedSxact->finishedBefore,
3712 56 : PredXact->SxactGlobalXmin))
3713 : {
3714 : /*
3715 : * This transaction committed before any in-progress transaction
3716 : * took its snapshot. It's no longer interesting.
3717 : */
3718 2462 : LWLockRelease(SerializableXactHashLock);
3719 2462 : dlist_delete_thoroughly(&finishedSxact->finishedLink);
3720 2462 : ReleaseOneSerializableXact(finishedSxact, false, false);
3721 2462 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3722 : }
3723 34 : else if (finishedSxact->commitSeqNo > PredXact->HavePartialClearedThrough
3724 34 : && finishedSxact->commitSeqNo <= PredXact->CanPartialClearThrough)
3725 : {
3726 : /*
3727 : * Any active transactions that took their snapshot before this
3728 : * transaction committed are read-only, so we can clear part of
3729 : * its state.
3730 : */
3731 16 : LWLockRelease(SerializableXactHashLock);
3732 :
3733 16 : if (SxactIsReadOnly(finishedSxact))
3734 : {
3735 : /* A read-only transaction can be removed entirely */
3736 0 : dlist_delete_thoroughly(&(finishedSxact->finishedLink));
3737 0 : ReleaseOneSerializableXact(finishedSxact, false, false);
3738 : }
3739 : else
3740 : {
3741 : /*
3742 : * A read-write transaction can only be partially cleared. We
3743 : * need to keep the SERIALIZABLEXACT but can release the
3744 : * SIREAD locks and conflicts in.
3745 : */
3746 16 : ReleaseOneSerializableXact(finishedSxact, true, false);
3747 : }
3748 :
3749 16 : PredXact->HavePartialClearedThrough = finishedSxact->commitSeqNo;
3750 16 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3751 : }
3752 : else
3753 : {
3754 : /* Still interesting. */
3755 : break;
3756 : }
3757 : }
3758 1760 : LWLockRelease(SerializableXactHashLock);
3759 :
3760 : /*
3761 : * Loop through predicate locks on dummy transaction for summarized data.
3762 : */
3763 1760 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
3764 1760 : dlist_foreach_modify(iter, &OldCommittedSxact->predicateLocks)
3765 : {
3766 0 : PREDICATELOCK *predlock =
3767 0 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
3768 : bool canDoPartialCleanup;
3769 :
3770 0 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3771 : Assert(predlock->commitSeqNo != 0);
3772 : Assert(predlock->commitSeqNo != InvalidSerCommitSeqNo);
3773 0 : canDoPartialCleanup = (predlock->commitSeqNo <= PredXact->CanPartialClearThrough);
3774 0 : LWLockRelease(SerializableXactHashLock);
3775 :
3776 : /*
3777 : * If this lock originally belonged to an old enough transaction, we
3778 : * can release it.
3779 : */
3780 0 : if (canDoPartialCleanup)
3781 : {
3782 : PREDICATELOCKTAG tag;
3783 : PREDICATELOCKTARGET *target;
3784 : PREDICATELOCKTARGETTAG targettag;
3785 : uint32 targettaghash;
3786 : LWLock *partitionLock;
3787 :
3788 0 : tag = predlock->tag;
3789 0 : target = tag.myTarget;
3790 0 : targettag = target->tag;
3791 0 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
3792 0 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
3793 :
3794 0 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
3795 :
3796 0 : dlist_delete(&(predlock->targetLink));
3797 0 : dlist_delete(&(predlock->xactLink));
3798 :
3799 0 : hash_search_with_hash_value(PredicateLockHash, &tag,
3800 0 : PredicateLockHashCodeFromTargetHashCode(&tag,
3801 : targettaghash),
3802 : HASH_REMOVE, NULL);
3803 0 : RemoveTargetIfNoLongerUsed(target, targettaghash);
3804 :
3805 0 : LWLockRelease(partitionLock);
3806 : }
3807 : }
3808 :
3809 1760 : LWLockRelease(SerializablePredicateListLock);
3810 1760 : LWLockRelease(SerializableFinishedListLock);
3811 1760 : }
3812 :
3813 : /*
3814 : * This is the normal way to delete anything from any of the predicate
3815 : * locking hash tables. Given a transaction which we know can be deleted:
3816 : * delete all predicate locks held by that transaction and any predicate
3817 : * lock targets which are now unreferenced by a lock; delete all conflicts
3818 : * for the transaction; delete all xid values for the transaction; then
3819 : * delete the transaction.
3820 : *
3821 : * When the partial flag is set, we can release all predicate locks and
3822 : * in-conflict information -- we've established that there are no longer
3823 : * any overlapping read write transactions for which this transaction could
3824 : * matter -- but keep the transaction entry itself and any outConflicts.
3825 : *
3826 : * When the summarize flag is set, we've run short of room for sxact data
3827 : * and must summarize to the SLRU. Predicate locks are transferred to a
3828 : * dummy "old" transaction, with duplicate locks on a single target
3829 : * collapsing to a single lock with the "latest" commitSeqNo from among
3830 : * the conflicting locks..
3831 : */
3832 : static void
3833 3134 : ReleaseOneSerializableXact(SERIALIZABLEXACT *sxact, bool partial,
3834 : bool summarize)
3835 : {
3836 : SERIALIZABLEXIDTAG sxidtag;
3837 : dlist_mutable_iter iter;
3838 :
3839 : Assert(sxact != NULL);
3840 : Assert(SxactIsRolledBack(sxact) || SxactIsCommitted(sxact));
3841 : Assert(partial || !SxactIsOnFinishedList(sxact));
3842 : Assert(LWLockHeldByMe(SerializableFinishedListLock));
3843 :
3844 : /*
3845 : * First release all the predicate locks held by this xact (or transfer
3846 : * them to OldCommittedSxact if summarize is true)
3847 : */
3848 3134 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
3849 3134 : if (IsInParallelMode())
3850 6 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
3851 8864 : dlist_foreach_modify(iter, &sxact->predicateLocks)
3852 : {
3853 5730 : PREDICATELOCK *predlock =
3854 5730 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
3855 : PREDICATELOCKTAG tag;
3856 : PREDICATELOCKTARGET *target;
3857 : PREDICATELOCKTARGETTAG targettag;
3858 : uint32 targettaghash;
3859 : LWLock *partitionLock;
3860 :
3861 5730 : tag = predlock->tag;
3862 5730 : target = tag.myTarget;
3863 5730 : targettag = target->tag;
3864 5730 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
3865 5730 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
3866 :
3867 5730 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
3868 :
3869 5730 : dlist_delete(&predlock->targetLink);
3870 :
3871 5730 : hash_search_with_hash_value(PredicateLockHash, &tag,
3872 5730 : PredicateLockHashCodeFromTargetHashCode(&tag,
3873 : targettaghash),
3874 : HASH_REMOVE, NULL);
3875 5730 : if (summarize)
3876 : {
3877 : bool found;
3878 :
3879 : /* Fold into dummy transaction list. */
3880 0 : tag.myXact = OldCommittedSxact;
3881 0 : predlock = hash_search_with_hash_value(PredicateLockHash, &tag,
3882 0 : PredicateLockHashCodeFromTargetHashCode(&tag,
3883 : targettaghash),
3884 : HASH_ENTER_NULL, &found);
3885 0 : if (!predlock)
3886 0 : ereport(ERROR,
3887 : (errcode(ERRCODE_OUT_OF_MEMORY),
3888 : errmsg("out of shared memory"),
3889 : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
3890 0 : if (found)
3891 : {
3892 : Assert(predlock->commitSeqNo != 0);
3893 : Assert(predlock->commitSeqNo != InvalidSerCommitSeqNo);
3894 0 : if (predlock->commitSeqNo < sxact->commitSeqNo)
3895 0 : predlock->commitSeqNo = sxact->commitSeqNo;
3896 : }
3897 : else
3898 : {
3899 0 : dlist_push_tail(&target->predicateLocks,
3900 : &predlock->targetLink);
3901 0 : dlist_push_tail(&OldCommittedSxact->predicateLocks,
3902 : &predlock->xactLink);
3903 0 : predlock->commitSeqNo = sxact->commitSeqNo;
3904 : }
3905 : }
3906 : else
3907 5730 : RemoveTargetIfNoLongerUsed(target, targettaghash);
3908 :
3909 5730 : LWLockRelease(partitionLock);
3910 : }
3911 :
3912 : /*
3913 : * Rather than retail removal, just re-init the head after we've run
3914 : * through the list.
3915 : */
3916 3134 : dlist_init(&sxact->predicateLocks);
3917 :
3918 3134 : if (IsInParallelMode())
3919 6 : LWLockRelease(&sxact->perXactPredicateListLock);
3920 3134 : LWLockRelease(SerializablePredicateListLock);
3921 :
3922 3134 : sxidtag.xid = sxact->topXid;
3923 3134 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
3924 :
3925 : /* Release all outConflicts (unless 'partial' is true) */
3926 3134 : if (!partial)
3927 : {
3928 3114 : dlist_foreach_modify(iter, &sxact->outConflicts)
3929 : {
3930 0 : RWConflict conflict =
3931 : dlist_container(RWConflictData, outLink, iter.cur);
3932 :
3933 0 : if (summarize)
3934 0 : conflict->sxactIn->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
3935 0 : ReleaseRWConflict(conflict);
3936 : }
3937 : }
3938 :
3939 : /* Release all inConflicts. */
3940 3134 : dlist_foreach_modify(iter, &sxact->inConflicts)
3941 : {
3942 0 : RWConflict conflict =
3943 0 : dlist_container(RWConflictData, inLink, iter.cur);
3944 :
3945 0 : if (summarize)
3946 0 : conflict->sxactOut->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
3947 0 : ReleaseRWConflict(conflict);
3948 : }
3949 :
3950 : /* Finally, get rid of the xid and the record of the transaction itself. */
3951 3134 : if (!partial)
3952 : {
3953 3114 : if (sxidtag.xid != InvalidTransactionId)
3954 2582 : hash_search(SerializableXidHash, &sxidtag, HASH_REMOVE, NULL);
3955 3114 : ReleasePredXact(sxact);
3956 : }
3957 :
3958 3134 : LWLockRelease(SerializableXactHashLock);
3959 3134 : }
3960 :
3961 : /*
3962 : * Tests whether the given top level transaction is concurrent with
3963 : * (overlaps) our current transaction.
3964 : *
3965 : * We need to identify the top level transaction for SSI, anyway, so pass
3966 : * that to this function to save the overhead of checking the snapshot's
3967 : * subxip array.
3968 : */
3969 : static bool
3970 1072 : XidIsConcurrent(TransactionId xid)
3971 : {
3972 : Snapshot snap;
3973 :
3974 : Assert(TransactionIdIsValid(xid));
3975 : Assert(!TransactionIdEquals(xid, GetTopTransactionIdIfAny()));
3976 :
3977 1072 : snap = GetTransactionSnapshot();
3978 :
3979 1072 : if (TransactionIdPrecedes(xid, snap->xmin))
3980 0 : return false;
3981 :
3982 1072 : if (TransactionIdFollowsOrEquals(xid, snap->xmax))
3983 1052 : return true;
3984 :
3985 20 : return pg_lfind32(xid, snap->xip, snap->xcnt);
3986 : }
3987 :
3988 : bool
3989 69261542 : CheckForSerializableConflictOutNeeded(Relation relation, Snapshot snapshot)
3990 : {
3991 69261542 : if (!SerializationNeededForRead(relation, snapshot))
3992 69209530 : return false;
3993 :
3994 : /* Check if someone else has already decided that we need to die */
3995 52012 : if (SxactIsDoomed(MySerializableXact))
3996 : {
3997 0 : ereport(ERROR,
3998 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
3999 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4000 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4001 : errhint("The transaction might succeed if retried.")));
4002 : }
4003 :
4004 52012 : return true;
4005 : }
4006 :
4007 : /*
4008 : * CheckForSerializableConflictOut
4009 : * A table AM is reading a tuple that has been modified. If it determines
4010 : * that the tuple version it is reading is not visible to us, it should
4011 : * pass in the top level xid of the transaction that created it.
4012 : * Otherwise, if it determines that it is visible to us but it has been
4013 : * deleted or there is a newer version available due to an update, it
4014 : * should pass in the top level xid of the modifying transaction.
4015 : *
4016 : * This function will check for overlap with our own transaction. If the given
4017 : * xid is also serializable and the transactions overlap (i.e., they cannot see
4018 : * each other's writes), then we have a conflict out.
4019 : */
4020 : void
4021 1142 : CheckForSerializableConflictOut(Relation relation, TransactionId xid, Snapshot snapshot)
4022 : {
4023 : SERIALIZABLEXIDTAG sxidtag;
4024 : SERIALIZABLEXID *sxid;
4025 : SERIALIZABLEXACT *sxact;
4026 :
4027 1142 : if (!SerializationNeededForRead(relation, snapshot))
4028 408 : return;
4029 :
4030 : /* Check if someone else has already decided that we need to die */
4031 1142 : if (SxactIsDoomed(MySerializableXact))
4032 : {
4033 0 : ereport(ERROR,
4034 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4035 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4036 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4037 : errhint("The transaction might succeed if retried.")));
4038 : }
4039 : Assert(TransactionIdIsValid(xid));
4040 :
4041 1142 : if (TransactionIdEquals(xid, GetTopTransactionIdIfAny()))
4042 0 : return;
4043 :
4044 : /*
4045 : * Find sxact or summarized info for the top level xid.
4046 : */
4047 1142 : sxidtag.xid = xid;
4048 1142 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4049 : sxid = (SERIALIZABLEXID *)
4050 1142 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
4051 1142 : if (!sxid)
4052 : {
4053 : /*
4054 : * Transaction not found in "normal" SSI structures. Check whether it
4055 : * got pushed out to SLRU storage for "old committed" transactions.
4056 : */
4057 : SerCommitSeqNo conflictCommitSeqNo;
4058 :
4059 50 : conflictCommitSeqNo = SerialGetMinConflictCommitSeqNo(xid);
4060 50 : if (conflictCommitSeqNo != 0)
4061 : {
4062 0 : if (conflictCommitSeqNo != InvalidSerCommitSeqNo
4063 0 : && (!SxactIsReadOnly(MySerializableXact)
4064 0 : || conflictCommitSeqNo
4065 0 : <= MySerializableXact->SeqNo.lastCommitBeforeSnapshot))
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 conflict out to old pivot %u.", xid),
4070 : errhint("The transaction might succeed if retried.")));
4071 :
4072 0 : if (SxactHasSummaryConflictIn(MySerializableXact)
4073 0 : || !dlist_is_empty(&MySerializableXact->inConflicts))
4074 0 : ereport(ERROR,
4075 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4076 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4077 : errdetail_internal("Reason code: Canceled on identification as a pivot, with conflict out to old committed transaction %u.", xid),
4078 : errhint("The transaction might succeed if retried.")));
4079 :
4080 0 : MySerializableXact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4081 : }
4082 :
4083 : /* It's not serializable or otherwise not important. */
4084 50 : LWLockRelease(SerializableXactHashLock);
4085 50 : return;
4086 : }
4087 1092 : sxact = sxid->myXact;
4088 : Assert(TransactionIdEquals(sxact->topXid, xid));
4089 1092 : if (sxact == MySerializableXact || SxactIsDoomed(sxact))
4090 : {
4091 : /* Can't conflict with ourself or a transaction that will roll back. */
4092 8 : LWLockRelease(SerializableXactHashLock);
4093 8 : return;
4094 : }
4095 :
4096 : /*
4097 : * We have a conflict out to a transaction which has a conflict out to a
4098 : * summarized transaction. That summarized transaction must have
4099 : * committed first, and we can't tell when it committed in relation to our
4100 : * snapshot acquisition, so something needs to be canceled.
4101 : */
4102 1084 : if (SxactHasSummaryConflictOut(sxact))
4103 : {
4104 0 : if (!SxactIsPrepared(sxact))
4105 : {
4106 0 : sxact->flags |= SXACT_FLAG_DOOMED;
4107 0 : LWLockRelease(SerializableXactHashLock);
4108 0 : return;
4109 : }
4110 : else
4111 : {
4112 0 : LWLockRelease(SerializableXactHashLock);
4113 0 : ereport(ERROR,
4114 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4115 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4116 : errdetail_internal("Reason code: Canceled on conflict out to old pivot."),
4117 : errhint("The transaction might succeed if retried.")));
4118 : }
4119 : }
4120 :
4121 : /*
4122 : * If this is a read-only transaction and the writing transaction has
4123 : * committed, and it doesn't have a rw-conflict to a transaction which
4124 : * committed before it, no conflict.
4125 : */
4126 1084 : if (SxactIsReadOnly(MySerializableXact)
4127 238 : && SxactIsCommitted(sxact)
4128 16 : && !SxactHasSummaryConflictOut(sxact)
4129 16 : && (!SxactHasConflictOut(sxact)
4130 4 : || MySerializableXact->SeqNo.lastCommitBeforeSnapshot < sxact->SeqNo.earliestOutConflictCommit))
4131 : {
4132 : /* Read-only transaction will appear to run first. No conflict. */
4133 12 : LWLockRelease(SerializableXactHashLock);
4134 12 : return;
4135 : }
4136 :
4137 1072 : if (!XidIsConcurrent(xid))
4138 : {
4139 : /* This write was already in our snapshot; no conflict. */
4140 0 : LWLockRelease(SerializableXactHashLock);
4141 0 : return;
4142 : }
4143 :
4144 1072 : if (RWConflictExists(MySerializableXact, sxact))
4145 : {
4146 : /* We don't want duplicate conflict records in the list. */
4147 338 : LWLockRelease(SerializableXactHashLock);
4148 338 : return;
4149 : }
4150 :
4151 : /*
4152 : * Flag the conflict. But first, if this conflict creates a dangerous
4153 : * structure, ereport an error.
4154 : */
4155 734 : FlagRWConflict(MySerializableXact, sxact);
4156 708 : LWLockRelease(SerializableXactHashLock);
4157 : }
4158 :
4159 : /*
4160 : * Check a particular target for rw-dependency conflict in. A subroutine of
4161 : * CheckForSerializableConflictIn().
4162 : */
4163 : static void
4164 15080 : CheckTargetForConflictsIn(PREDICATELOCKTARGETTAG *targettag)
4165 : {
4166 : uint32 targettaghash;
4167 : LWLock *partitionLock;
4168 : PREDICATELOCKTARGET *target;
4169 15080 : PREDICATELOCK *mypredlock = NULL;
4170 : PREDICATELOCKTAG mypredlocktag;
4171 : dlist_mutable_iter iter;
4172 :
4173 : Assert(MySerializableXact != InvalidSerializableXact);
4174 :
4175 : /*
4176 : * The same hash and LW lock apply to the lock target and the lock itself.
4177 : */
4178 15080 : targettaghash = PredicateLockTargetTagHashCode(targettag);
4179 15080 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
4180 15080 : LWLockAcquire(partitionLock, LW_SHARED);
4181 : target = (PREDICATELOCKTARGET *)
4182 15080 : hash_search_with_hash_value(PredicateLockTargetHash,
4183 : targettag, targettaghash,
4184 : HASH_FIND, NULL);
4185 15080 : if (!target)
4186 : {
4187 : /* Nothing has this target locked; we're done here. */
4188 11306 : LWLockRelease(partitionLock);
4189 11306 : return;
4190 : }
4191 :
4192 : /*
4193 : * Each lock for an overlapping transaction represents a conflict: a
4194 : * rw-dependency in to this transaction.
4195 : */
4196 3774 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4197 :
4198 8508 : dlist_foreach_modify(iter, &target->predicateLocks)
4199 : {
4200 4868 : PREDICATELOCK *predlock =
4201 4868 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
4202 4868 : SERIALIZABLEXACT *sxact = predlock->tag.myXact;
4203 :
4204 4868 : if (sxact == MySerializableXact)
4205 : {
4206 : /*
4207 : * If we're getting a write lock on a tuple, we don't need a
4208 : * predicate (SIREAD) lock on the same tuple. We can safely remove
4209 : * our SIREAD lock, but we'll defer doing so until after the loop
4210 : * because that requires upgrading to an exclusive partition lock.
4211 : *
4212 : * We can't use this optimization within a subtransaction because
4213 : * the subtransaction could roll back, and we would be left
4214 : * without any lock at the top level.
4215 : */
4216 3152 : if (!IsSubTransaction()
4217 3152 : && GET_PREDICATELOCKTARGETTAG_OFFSET(*targettag))
4218 : {
4219 792 : mypredlock = predlock;
4220 792 : mypredlocktag = predlock->tag;
4221 : }
4222 : }
4223 1716 : else if (!SxactIsDoomed(sxact)
4224 1716 : && (!SxactIsCommitted(sxact)
4225 178 : || TransactionIdPrecedes(GetTransactionSnapshot()->xmin,
4226 : sxact->finishedBefore))
4227 1698 : && !RWConflictExists(sxact, MySerializableXact))
4228 : {
4229 1010 : LWLockRelease(SerializableXactHashLock);
4230 1010 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4231 :
4232 : /*
4233 : * Re-check after getting exclusive lock because the other
4234 : * transaction may have flagged a conflict.
4235 : */
4236 1010 : if (!SxactIsDoomed(sxact)
4237 1010 : && (!SxactIsCommitted(sxact)
4238 156 : || TransactionIdPrecedes(GetTransactionSnapshot()->xmin,
4239 : sxact->finishedBefore))
4240 1010 : && !RWConflictExists(sxact, MySerializableXact))
4241 : {
4242 1010 : FlagRWConflict(sxact, MySerializableXact);
4243 : }
4244 :
4245 876 : LWLockRelease(SerializableXactHashLock);
4246 876 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4247 : }
4248 : }
4249 3640 : LWLockRelease(SerializableXactHashLock);
4250 3640 : LWLockRelease(partitionLock);
4251 :
4252 : /*
4253 : * If we found one of our own SIREAD locks to remove, remove it now.
4254 : *
4255 : * At this point our transaction already has a RowExclusiveLock on the
4256 : * relation, so we are OK to drop the predicate lock on the tuple, if
4257 : * found, without fearing that another write against the tuple will occur
4258 : * before the MVCC information makes it to the buffer.
4259 : */
4260 3640 : if (mypredlock != NULL)
4261 : {
4262 : uint32 predlockhashcode;
4263 : PREDICATELOCK *rmpredlock;
4264 :
4265 778 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
4266 778 : if (IsInParallelMode())
4267 0 : LWLockAcquire(&MySerializableXact->perXactPredicateListLock, LW_EXCLUSIVE);
4268 778 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
4269 778 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4270 :
4271 : /*
4272 : * Remove the predicate lock from shared memory, if it wasn't removed
4273 : * while the locks were released. One way that could happen is from
4274 : * autovacuum cleaning up an index.
4275 : */
4276 778 : predlockhashcode = PredicateLockHashCodeFromTargetHashCode
4277 : (&mypredlocktag, targettaghash);
4278 : rmpredlock = (PREDICATELOCK *)
4279 778 : hash_search_with_hash_value(PredicateLockHash,
4280 : &mypredlocktag,
4281 : predlockhashcode,
4282 : HASH_FIND, NULL);
4283 778 : if (rmpredlock != NULL)
4284 : {
4285 : Assert(rmpredlock == mypredlock);
4286 :
4287 778 : dlist_delete(&(mypredlock->targetLink));
4288 778 : dlist_delete(&(mypredlock->xactLink));
4289 :
4290 : rmpredlock = (PREDICATELOCK *)
4291 778 : hash_search_with_hash_value(PredicateLockHash,
4292 : &mypredlocktag,
4293 : predlockhashcode,
4294 : HASH_REMOVE, NULL);
4295 : Assert(rmpredlock == mypredlock);
4296 :
4297 778 : RemoveTargetIfNoLongerUsed(target, targettaghash);
4298 : }
4299 :
4300 778 : LWLockRelease(SerializableXactHashLock);
4301 778 : LWLockRelease(partitionLock);
4302 778 : if (IsInParallelMode())
4303 0 : LWLockRelease(&MySerializableXact->perXactPredicateListLock);
4304 778 : LWLockRelease(SerializablePredicateListLock);
4305 :
4306 778 : if (rmpredlock != NULL)
4307 : {
4308 : /*
4309 : * Remove entry in local lock table if it exists. It's OK if it
4310 : * doesn't exist; that means the lock was transferred to a new
4311 : * target by a different backend.
4312 : */
4313 778 : hash_search_with_hash_value(LocalPredicateLockHash,
4314 : targettag, targettaghash,
4315 : HASH_REMOVE, NULL);
4316 :
4317 778 : DecrementParentLocks(targettag);
4318 : }
4319 : }
4320 : }
4321 :
4322 : /*
4323 : * CheckForSerializableConflictIn
4324 : * We are writing the given tuple. If that indicates a rw-conflict
4325 : * in from another serializable transaction, take appropriate action.
4326 : *
4327 : * Skip checking for any granularity for which a parameter is missing.
4328 : *
4329 : * A tuple update or delete is in conflict if we have a predicate lock
4330 : * against the relation or page in which the tuple exists, or against the
4331 : * tuple itself.
4332 : */
4333 : void
4334 33394160 : CheckForSerializableConflictIn(Relation relation, const ItemPointerData *tid, BlockNumber blkno)
4335 : {
4336 : PREDICATELOCKTARGETTAG targettag;
4337 :
4338 33394160 : if (!SerializationNeededForWrite(relation))
4339 33385230 : return;
4340 :
4341 : /* Check if someone else has already decided that we need to die */
4342 8930 : if (SxactIsDoomed(MySerializableXact))
4343 2 : ereport(ERROR,
4344 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4345 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4346 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict in checking."),
4347 : errhint("The transaction might succeed if retried.")));
4348 :
4349 : /*
4350 : * We're doing a write which might cause rw-conflicts now or later.
4351 : * Memorize that fact.
4352 : */
4353 8928 : MyXactDidWrite = true;
4354 :
4355 : /*
4356 : * It is important that we check for locks from the finest granularity to
4357 : * the coarsest granularity, so that granularity promotion doesn't cause
4358 : * us to miss a lock. The new (coarser) lock will be acquired before the
4359 : * old (finer) locks are released.
4360 : *
4361 : * It is not possible to take and hold a lock across the checks for all
4362 : * granularities because each target could be in a separate partition.
4363 : */
4364 8928 : if (tid != NULL)
4365 : {
4366 1302 : SET_PREDICATELOCKTARGETTAG_TUPLE(targettag,
4367 : relation->rd_locator.dbOid,
4368 : relation->rd_id,
4369 : ItemPointerGetBlockNumber(tid),
4370 : ItemPointerGetOffsetNumber(tid));
4371 1302 : CheckTargetForConflictsIn(&targettag);
4372 : }
4373 :
4374 8882 : if (blkno != InvalidBlockNumber)
4375 : {
4376 4956 : SET_PREDICATELOCKTARGETTAG_PAGE(targettag,
4377 : relation->rd_locator.dbOid,
4378 : relation->rd_id,
4379 : blkno);
4380 4956 : CheckTargetForConflictsIn(&targettag);
4381 : }
4382 :
4383 8822 : SET_PREDICATELOCKTARGETTAG_RELATION(targettag,
4384 : relation->rd_locator.dbOid,
4385 : relation->rd_id);
4386 8822 : CheckTargetForConflictsIn(&targettag);
4387 : }
4388 :
4389 : /*
4390 : * CheckTableForSerializableConflictIn
4391 : * The entire table is going through a DDL-style logical mass delete
4392 : * like TRUNCATE or DROP TABLE. If that causes a rw-conflict in from
4393 : * another serializable transaction, take appropriate action.
4394 : *
4395 : * While these operations do not operate entirely within the bounds of
4396 : * snapshot isolation, they can occur inside a serializable transaction, and
4397 : * will logically occur after any reads which saw rows which were destroyed
4398 : * by these operations, so we do what we can to serialize properly under
4399 : * SSI.
4400 : *
4401 : * The relation passed in must be a heap relation. Any predicate lock of any
4402 : * granularity on the heap will cause a rw-conflict in to this transaction.
4403 : * Predicate locks on indexes do not matter because they only exist to guard
4404 : * against conflicting inserts into the index, and this is a mass *delete*.
4405 : * When a table is truncated or dropped, the index will also be truncated
4406 : * or dropped, and we'll deal with locks on the index when that happens.
4407 : *
4408 : * Dropping or truncating a table also needs to drop any existing predicate
4409 : * locks on heap tuples or pages, because they're about to go away. This
4410 : * should be done before altering the predicate locks because the transaction
4411 : * could be rolled back because of a conflict, in which case the lock changes
4412 : * are not needed. (At the moment, we don't actually bother to drop the
4413 : * existing locks on a dropped or truncated table at the moment. That might
4414 : * lead to some false positives, but it doesn't seem worth the trouble.)
4415 : */
4416 : void
4417 54658 : CheckTableForSerializableConflictIn(Relation relation)
4418 : {
4419 : HASH_SEQ_STATUS seqstat;
4420 : PREDICATELOCKTARGET *target;
4421 : Oid dbId;
4422 : Oid heapId;
4423 : int i;
4424 :
4425 : /*
4426 : * Bail out quickly if there are no serializable transactions running.
4427 : * It's safe to check this without taking locks because the caller is
4428 : * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
4429 : * would matter here can be acquired while that is held.
4430 : */
4431 54658 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
4432 54616 : return;
4433 :
4434 290 : if (!SerializationNeededForWrite(relation))
4435 248 : return;
4436 :
4437 : /*
4438 : * We're doing a write which might cause rw-conflicts now or later.
4439 : * Memorize that fact.
4440 : */
4441 42 : MyXactDidWrite = true;
4442 :
4443 : Assert(relation->rd_index == NULL); /* not an index relation */
4444 :
4445 42 : dbId = relation->rd_locator.dbOid;
4446 42 : heapId = relation->rd_id;
4447 :
4448 42 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
4449 714 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
4450 672 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_SHARED);
4451 42 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4452 :
4453 : /* Scan through target list */
4454 42 : hash_seq_init(&seqstat, PredicateLockTargetHash);
4455 :
4456 144 : while ((target = (PREDICATELOCKTARGET *) hash_seq_search(&seqstat)))
4457 : {
4458 : dlist_mutable_iter iter;
4459 :
4460 : /*
4461 : * Check whether this is a target which needs attention.
4462 : */
4463 102 : if (GET_PREDICATELOCKTARGETTAG_RELATION(target->tag) != heapId)
4464 84 : continue; /* wrong relation id */
4465 18 : if (GET_PREDICATELOCKTARGETTAG_DB(target->tag) != dbId)
4466 0 : continue; /* wrong database id */
4467 :
4468 : /*
4469 : * Loop through locks for this target and flag conflicts.
4470 : */
4471 36 : dlist_foreach_modify(iter, &target->predicateLocks)
4472 : {
4473 18 : PREDICATELOCK *predlock =
4474 18 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
4475 :
4476 18 : if (predlock->tag.myXact != MySerializableXact
4477 0 : && !RWConflictExists(predlock->tag.myXact, MySerializableXact))
4478 : {
4479 0 : FlagRWConflict(predlock->tag.myXact, MySerializableXact);
4480 : }
4481 : }
4482 : }
4483 :
4484 : /* Release locks in reverse order */
4485 42 : LWLockRelease(SerializableXactHashLock);
4486 714 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
4487 672 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
4488 42 : LWLockRelease(SerializablePredicateListLock);
4489 : }
4490 :
4491 :
4492 : /*
4493 : * Flag a rw-dependency between two serializable transactions.
4494 : *
4495 : * The caller is responsible for ensuring that we have a LW lock on
4496 : * the transaction hash table.
4497 : */
4498 : static void
4499 1744 : FlagRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
4500 : {
4501 : Assert(reader != writer);
4502 :
4503 : /* First, see if this conflict causes failure. */
4504 1744 : OnConflict_CheckForSerializationFailure(reader, writer);
4505 :
4506 : /* Actually do the conflict flagging. */
4507 1584 : if (reader == OldCommittedSxact)
4508 0 : writer->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
4509 1584 : else if (writer == OldCommittedSxact)
4510 0 : reader->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4511 : else
4512 1584 : SetRWConflict(reader, writer);
4513 1584 : }
4514 :
4515 : /*----------------------------------------------------------------------------
4516 : * We are about to add a RW-edge to the dependency graph - check that we don't
4517 : * introduce a dangerous structure by doing so, and abort one of the
4518 : * transactions if so.
4519 : *
4520 : * A serialization failure can only occur if there is a dangerous structure
4521 : * in the dependency graph:
4522 : *
4523 : * Tin ------> Tpivot ------> Tout
4524 : * rw rw
4525 : *
4526 : * Furthermore, Tout must commit first.
4527 : *
4528 : * One more optimization is that if Tin is declared READ ONLY (or commits
4529 : * without writing), we can only have a problem if Tout committed before Tin
4530 : * acquired its snapshot.
4531 : *----------------------------------------------------------------------------
4532 : */
4533 : static void
4534 1744 : OnConflict_CheckForSerializationFailure(const SERIALIZABLEXACT *reader,
4535 : SERIALIZABLEXACT *writer)
4536 : {
4537 : bool failure;
4538 :
4539 : Assert(LWLockHeldByMe(SerializableXactHashLock));
4540 :
4541 1744 : failure = false;
4542 :
4543 : /*------------------------------------------------------------------------
4544 : * Check for already-committed writer with rw-conflict out flagged
4545 : * (conflict-flag on W means that T2 committed before W):
4546 : *
4547 : * R ------> W ------> T2
4548 : * rw rw
4549 : *
4550 : * That is a dangerous structure, so we must abort. (Since the writer
4551 : * has already committed, we must be the reader)
4552 : *------------------------------------------------------------------------
4553 : */
4554 1744 : if (SxactIsCommitted(writer)
4555 36 : && (SxactHasConflictOut(writer) || SxactHasSummaryConflictOut(writer)))
4556 4 : failure = true;
4557 :
4558 : /*------------------------------------------------------------------------
4559 : * Check whether the writer has become a pivot with an out-conflict
4560 : * committed transaction (T2), and T2 committed first:
4561 : *
4562 : * R ------> W ------> T2
4563 : * rw rw
4564 : *
4565 : * Because T2 must've committed first, there is no anomaly if:
4566 : * - the reader committed before T2
4567 : * - the writer committed before T2
4568 : * - the reader is a READ ONLY transaction and the reader was concurrent
4569 : * with T2 (= reader acquired its snapshot before T2 committed)
4570 : *
4571 : * We also handle the case that T2 is prepared but not yet committed
4572 : * here. In that case T2 has already checked for conflicts, so if it
4573 : * commits first, making the above conflict real, it's too late for it
4574 : * to abort.
4575 : *------------------------------------------------------------------------
4576 : */
4577 1744 : if (!failure && SxactHasSummaryConflictOut(writer))
4578 0 : failure = true;
4579 1744 : else if (!failure)
4580 : {
4581 : dlist_iter iter;
4582 :
4583 2174 : dlist_foreach(iter, &writer->outConflicts)
4584 : {
4585 584 : RWConflict conflict =
4586 : dlist_container(RWConflictData, outLink, iter.cur);
4587 584 : SERIALIZABLEXACT *t2 = conflict->sxactIn;
4588 :
4589 584 : if (SxactIsPrepared(t2)
4590 162 : && (!SxactIsCommitted(reader)
4591 130 : || t2->prepareSeqNo <= reader->commitSeqNo)
4592 162 : && (!SxactIsCommitted(writer)
4593 0 : || t2->prepareSeqNo <= writer->commitSeqNo)
4594 162 : && (!SxactIsReadOnly(reader)
4595 24 : || t2->prepareSeqNo <= reader->SeqNo.lastCommitBeforeSnapshot))
4596 : {
4597 150 : failure = true;
4598 150 : break;
4599 : }
4600 : }
4601 : }
4602 :
4603 : /*------------------------------------------------------------------------
4604 : * Check whether the reader has become a pivot with a writer
4605 : * that's committed (or prepared):
4606 : *
4607 : * T0 ------> R ------> W
4608 : * rw rw
4609 : *
4610 : * Because W must've committed first for an anomaly to occur, there is no
4611 : * anomaly if:
4612 : * - T0 committed before the writer
4613 : * - T0 is READ ONLY, and overlaps the writer
4614 : *------------------------------------------------------------------------
4615 : */
4616 1744 : if (!failure && SxactIsPrepared(writer) && !SxactIsReadOnly(reader))
4617 : {
4618 36 : if (SxactHasSummaryConflictIn(reader))
4619 : {
4620 0 : failure = true;
4621 : }
4622 : else
4623 : {
4624 : dlist_iter iter;
4625 :
4626 : /*
4627 : * The unconstify is needed as we have no const version of
4628 : * dlist_foreach().
4629 : */
4630 36 : dlist_foreach(iter, &unconstify(SERIALIZABLEXACT *, reader)->inConflicts)
4631 : {
4632 22 : const RWConflict conflict =
4633 22 : dlist_container(RWConflictData, inLink, iter.cur);
4634 22 : const SERIALIZABLEXACT *t0 = conflict->sxactOut;
4635 :
4636 22 : if (!SxactIsDoomed(t0)
4637 22 : && (!SxactIsCommitted(t0)
4638 22 : || t0->commitSeqNo >= writer->prepareSeqNo)
4639 22 : && (!SxactIsReadOnly(t0)
4640 0 : || t0->SeqNo.lastCommitBeforeSnapshot >= writer->prepareSeqNo))
4641 : {
4642 22 : failure = true;
4643 22 : break;
4644 : }
4645 : }
4646 : }
4647 : }
4648 :
4649 1744 : if (failure)
4650 : {
4651 : /*
4652 : * We have to kill a transaction to avoid a possible anomaly from
4653 : * occurring. If the writer is us, we can just ereport() to cause a
4654 : * transaction abort. Otherwise we flag the writer for termination,
4655 : * causing it to abort when it tries to commit. However, if the writer
4656 : * is a prepared transaction, already prepared, we can't abort it
4657 : * anymore, so we have to kill the reader instead.
4658 : */
4659 176 : if (MySerializableXact == writer)
4660 : {
4661 134 : LWLockRelease(SerializableXactHashLock);
4662 134 : ereport(ERROR,
4663 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4664 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4665 : errdetail_internal("Reason code: Canceled on identification as a pivot, during write."),
4666 : errhint("The transaction might succeed if retried.")));
4667 : }
4668 42 : else if (SxactIsPrepared(writer))
4669 : {
4670 26 : LWLockRelease(SerializableXactHashLock);
4671 :
4672 : /* if we're not the writer, we have to be the reader */
4673 : Assert(MySerializableXact == reader);
4674 26 : ereport(ERROR,
4675 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4676 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4677 : errdetail_internal("Reason code: Canceled on conflict out to pivot %u, during read.", writer->topXid),
4678 : errhint("The transaction might succeed if retried.")));
4679 : }
4680 16 : writer->flags |= SXACT_FLAG_DOOMED;
4681 : }
4682 1584 : }
4683 :
4684 : /*
4685 : * PreCommit_CheckForSerializationFailure
4686 : * Check for dangerous structures in a serializable transaction
4687 : * at commit.
4688 : *
4689 : * We're checking for a dangerous structure as each conflict is recorded.
4690 : * The only way we could have a problem at commit is if this is the "out"
4691 : * side of a pivot, and neither the "in" side nor the pivot has yet
4692 : * committed.
4693 : *
4694 : * If a dangerous structure is found, the pivot (the near conflict) is
4695 : * marked for death, because rolling back another transaction might mean
4696 : * that we fail without ever making progress. This transaction is
4697 : * committing writes, so letting it commit ensures progress. If we
4698 : * canceled the far conflict, it might immediately fail again on retry.
4699 : */
4700 : void
4701 987852 : PreCommit_CheckForSerializationFailure(void)
4702 : {
4703 : dlist_iter near_iter;
4704 :
4705 987852 : if (MySerializableXact == InvalidSerializableXact)
4706 985046 : return;
4707 :
4708 : Assert(IsolationIsSerializable());
4709 :
4710 2806 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4711 :
4712 : /*
4713 : * Check if someone else has already decided that we need to die. Since
4714 : * we set our own DOOMED flag when partially releasing, ignore in that
4715 : * case.
4716 : */
4717 2806 : if (SxactIsDoomed(MySerializableXact) &&
4718 312 : !SxactIsPartiallyReleased(MySerializableXact))
4719 : {
4720 310 : LWLockRelease(SerializableXactHashLock);
4721 310 : ereport(ERROR,
4722 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4723 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4724 : errdetail_internal("Reason code: Canceled on identification as a pivot, during commit attempt."),
4725 : errhint("The transaction might succeed if retried.")));
4726 : }
4727 :
4728 3708 : dlist_foreach(near_iter, &MySerializableXact->inConflicts)
4729 : {
4730 1212 : RWConflict nearConflict =
4731 1212 : dlist_container(RWConflictData, inLink, near_iter.cur);
4732 :
4733 1212 : if (!SxactIsCommitted(nearConflict->sxactOut)
4734 842 : && !SxactIsDoomed(nearConflict->sxactOut))
4735 : {
4736 : dlist_iter far_iter;
4737 :
4738 902 : dlist_foreach(far_iter, &nearConflict->sxactOut->inConflicts)
4739 : {
4740 364 : RWConflict farConflict =
4741 364 : dlist_container(RWConflictData, inLink, far_iter.cur);
4742 :
4743 364 : if (farConflict->sxactOut == MySerializableXact
4744 84 : || (!SxactIsCommitted(farConflict->sxactOut)
4745 48 : && !SxactIsReadOnly(farConflict->sxactOut)
4746 24 : && !SxactIsDoomed(farConflict->sxactOut)))
4747 : {
4748 : /*
4749 : * Normally, we kill the pivot transaction to make sure we
4750 : * make progress if the failing transaction is retried.
4751 : * However, we can't kill it if it's already prepared, so
4752 : * in that case we commit suicide instead.
4753 : */
4754 304 : if (SxactIsPrepared(nearConflict->sxactOut))
4755 : {
4756 0 : LWLockRelease(SerializableXactHashLock);
4757 0 : ereport(ERROR,
4758 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4759 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4760 : errdetail_internal("Reason code: Canceled on commit attempt with conflict in from prepared pivot."),
4761 : errhint("The transaction might succeed if retried.")));
4762 : }
4763 304 : nearConflict->sxactOut->flags |= SXACT_FLAG_DOOMED;
4764 304 : break;
4765 : }
4766 : }
4767 : }
4768 : }
4769 :
4770 2496 : MySerializableXact->prepareSeqNo = ++(PredXact->LastSxactCommitSeqNo);
4771 2496 : MySerializableXact->flags |= SXACT_FLAG_PREPARED;
4772 :
4773 2496 : LWLockRelease(SerializableXactHashLock);
4774 : }
4775 :
4776 : /*------------------------------------------------------------------------*/
4777 :
4778 : /*
4779 : * Two-phase commit support
4780 : */
4781 :
4782 : /*
4783 : * AtPrepare_Locks
4784 : * Do the preparatory work for a PREPARE: make 2PC state file
4785 : * records for all predicate locks currently held.
4786 : */
4787 : void
4788 632 : AtPrepare_PredicateLocks(void)
4789 : {
4790 : SERIALIZABLEXACT *sxact;
4791 : TwoPhasePredicateRecord record;
4792 : TwoPhasePredicateXactRecord *xactRecord;
4793 : TwoPhasePredicateLockRecord *lockRecord;
4794 : dlist_iter iter;
4795 :
4796 632 : sxact = MySerializableXact;
4797 632 : xactRecord = &(record.data.xactRecord);
4798 632 : lockRecord = &(record.data.lockRecord);
4799 :
4800 632 : if (MySerializableXact == InvalidSerializableXact)
4801 608 : return;
4802 :
4803 : /* Generate an xact record for our SERIALIZABLEXACT */
4804 24 : record.type = TWOPHASEPREDICATERECORD_XACT;
4805 24 : xactRecord->xmin = MySerializableXact->xmin;
4806 24 : xactRecord->flags = MySerializableXact->flags;
4807 :
4808 : /*
4809 : * Note that we don't include the list of conflicts in our out in the
4810 : * statefile, because new conflicts can be added even after the
4811 : * transaction prepares. We'll just make a conservative assumption during
4812 : * recovery instead.
4813 : */
4814 :
4815 24 : RegisterTwoPhaseRecord(TWOPHASE_RM_PREDICATELOCK_ID, 0,
4816 : &record, sizeof(record));
4817 :
4818 : /*
4819 : * Generate a lock record for each lock.
4820 : *
4821 : * To do this, we need to walk the predicate lock list in our sxact rather
4822 : * than using the local predicate lock table because the latter is not
4823 : * guaranteed to be accurate.
4824 : */
4825 24 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
4826 :
4827 : /*
4828 : * No need to take sxact->perXactPredicateListLock in parallel mode
4829 : * because there cannot be any parallel workers running while we are
4830 : * preparing a transaction.
4831 : */
4832 : Assert(!IsParallelWorker() && !ParallelContextActive());
4833 :
4834 44 : dlist_foreach(iter, &sxact->predicateLocks)
4835 : {
4836 20 : PREDICATELOCK *predlock =
4837 20 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
4838 :
4839 20 : record.type = TWOPHASEPREDICATERECORD_LOCK;
4840 20 : lockRecord->target = predlock->tag.myTarget->tag;
4841 :
4842 20 : RegisterTwoPhaseRecord(TWOPHASE_RM_PREDICATELOCK_ID, 0,
4843 : &record, sizeof(record));
4844 : }
4845 :
4846 24 : LWLockRelease(SerializablePredicateListLock);
4847 : }
4848 :
4849 : /*
4850 : * PostPrepare_Locks
4851 : * Clean up after successful PREPARE. Unlike the non-predicate
4852 : * lock manager, we do not need to transfer locks to a dummy
4853 : * PGPROC because our SERIALIZABLEXACT will stay around
4854 : * anyway. We only need to clean up our local state.
4855 : */
4856 : void
4857 632 : PostPrepare_PredicateLocks(FullTransactionId fxid)
4858 : {
4859 632 : if (MySerializableXact == InvalidSerializableXact)
4860 608 : return;
4861 :
4862 : Assert(SxactIsPrepared(MySerializableXact));
4863 :
4864 24 : MySerializableXact->pid = 0;
4865 24 : MySerializableXact->pgprocno = INVALID_PROC_NUMBER;
4866 :
4867 24 : hash_destroy(LocalPredicateLockHash);
4868 24 : LocalPredicateLockHash = NULL;
4869 :
4870 24 : MySerializableXact = InvalidSerializableXact;
4871 24 : MyXactDidWrite = false;
4872 : }
4873 :
4874 : /*
4875 : * PredicateLockTwoPhaseFinish
4876 : * Release a prepared transaction's predicate locks once it
4877 : * commits or aborts.
4878 : */
4879 : void
4880 646 : PredicateLockTwoPhaseFinish(FullTransactionId fxid, bool isCommit)
4881 : {
4882 : SERIALIZABLEXID *sxid;
4883 : SERIALIZABLEXIDTAG sxidtag;
4884 :
4885 646 : sxidtag.xid = XidFromFullTransactionId(fxid);
4886 :
4887 646 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4888 : sxid = (SERIALIZABLEXID *)
4889 646 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
4890 646 : LWLockRelease(SerializableXactHashLock);
4891 :
4892 : /* xid will not be found if it wasn't a serializable transaction */
4893 646 : if (sxid == NULL)
4894 622 : return;
4895 :
4896 : /* Release its locks */
4897 24 : MySerializableXact = sxid->myXact;
4898 24 : MyXactDidWrite = true; /* conservatively assume that we wrote
4899 : * something */
4900 24 : ReleasePredicateLocks(isCommit, false);
4901 : }
4902 :
4903 : /*
4904 : * Re-acquire a predicate lock belonging to a transaction that was prepared.
4905 : */
4906 : void
4907 0 : predicatelock_twophase_recover(FullTransactionId fxid, uint16 info,
4908 : void *recdata, uint32 len)
4909 : {
4910 : TwoPhasePredicateRecord *record;
4911 0 : TransactionId xid = XidFromFullTransactionId(fxid);
4912 :
4913 : Assert(len == sizeof(TwoPhasePredicateRecord));
4914 :
4915 0 : record = (TwoPhasePredicateRecord *) recdata;
4916 :
4917 : Assert((record->type == TWOPHASEPREDICATERECORD_XACT) ||
4918 : (record->type == TWOPHASEPREDICATERECORD_LOCK));
4919 :
4920 0 : if (record->type == TWOPHASEPREDICATERECORD_XACT)
4921 : {
4922 : /* Per-transaction record. Set up a SERIALIZABLEXACT. */
4923 : TwoPhasePredicateXactRecord *xactRecord;
4924 : SERIALIZABLEXACT *sxact;
4925 : SERIALIZABLEXID *sxid;
4926 : SERIALIZABLEXIDTAG sxidtag;
4927 : bool found;
4928 :
4929 0 : xactRecord = (TwoPhasePredicateXactRecord *) &record->data.xactRecord;
4930 :
4931 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4932 0 : sxact = CreatePredXact();
4933 0 : if (!sxact)
4934 0 : ereport(ERROR,
4935 : (errcode(ERRCODE_OUT_OF_MEMORY),
4936 : errmsg("out of shared memory")));
4937 :
4938 : /* vxid for a prepared xact is INVALID_PROC_NUMBER/xid; no pid */
4939 0 : sxact->vxid.procNumber = INVALID_PROC_NUMBER;
4940 0 : sxact->vxid.localTransactionId = (LocalTransactionId) xid;
4941 0 : sxact->pid = 0;
4942 0 : sxact->pgprocno = INVALID_PROC_NUMBER;
4943 :
4944 : /* a prepared xact hasn't committed yet */
4945 0 : sxact->prepareSeqNo = RecoverySerCommitSeqNo;
4946 0 : sxact->commitSeqNo = InvalidSerCommitSeqNo;
4947 0 : sxact->finishedBefore = InvalidTransactionId;
4948 :
4949 0 : sxact->SeqNo.lastCommitBeforeSnapshot = RecoverySerCommitSeqNo;
4950 :
4951 : /*
4952 : * Don't need to track this; no transactions running at the time the
4953 : * recovered xact started are still active, except possibly other
4954 : * prepared xacts and we don't care whether those are RO_SAFE or not.
4955 : */
4956 0 : dlist_init(&(sxact->possibleUnsafeConflicts));
4957 :
4958 0 : dlist_init(&(sxact->predicateLocks));
4959 0 : dlist_node_init(&sxact->finishedLink);
4960 :
4961 0 : sxact->topXid = xid;
4962 0 : sxact->xmin = xactRecord->xmin;
4963 0 : sxact->flags = xactRecord->flags;
4964 : Assert(SxactIsPrepared(sxact));
4965 0 : if (!SxactIsReadOnly(sxact))
4966 : {
4967 0 : ++(PredXact->WritableSxactCount);
4968 : Assert(PredXact->WritableSxactCount <=
4969 : (MaxBackends + max_prepared_xacts));
4970 : }
4971 :
4972 : /*
4973 : * We don't know whether the transaction had any conflicts or not, so
4974 : * we'll conservatively assume that it had both a conflict in and a
4975 : * conflict out, and represent that with the summary conflict flags.
4976 : */
4977 0 : dlist_init(&(sxact->outConflicts));
4978 0 : dlist_init(&(sxact->inConflicts));
4979 0 : sxact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
4980 0 : sxact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4981 :
4982 : /* Register the transaction's xid */
4983 0 : sxidtag.xid = xid;
4984 0 : sxid = (SERIALIZABLEXID *) hash_search(SerializableXidHash,
4985 : &sxidtag,
4986 : HASH_ENTER, &found);
4987 : Assert(sxid != NULL);
4988 : Assert(!found);
4989 0 : sxid->myXact = sxact;
4990 :
4991 : /*
4992 : * Update global xmin. Note that this is a special case compared to
4993 : * registering a normal transaction, because the global xmin might go
4994 : * backwards. That's OK, because until recovery is over we're not
4995 : * going to complete any transactions or create any non-prepared
4996 : * transactions, so there's no danger of throwing away.
4997 : */
4998 0 : if ((!TransactionIdIsValid(PredXact->SxactGlobalXmin)) ||
4999 0 : (TransactionIdFollows(PredXact->SxactGlobalXmin, sxact->xmin)))
5000 : {
5001 0 : PredXact->SxactGlobalXmin = sxact->xmin;
5002 0 : PredXact->SxactGlobalXminCount = 1;
5003 0 : SerialSetActiveSerXmin(sxact->xmin);
5004 : }
5005 0 : else if (TransactionIdEquals(sxact->xmin, PredXact->SxactGlobalXmin))
5006 : {
5007 : Assert(PredXact->SxactGlobalXminCount > 0);
5008 0 : PredXact->SxactGlobalXminCount++;
5009 : }
5010 :
5011 0 : LWLockRelease(SerializableXactHashLock);
5012 : }
5013 0 : else if (record->type == TWOPHASEPREDICATERECORD_LOCK)
5014 : {
5015 : /* Lock record. Recreate the PREDICATELOCK */
5016 : TwoPhasePredicateLockRecord *lockRecord;
5017 : SERIALIZABLEXID *sxid;
5018 : SERIALIZABLEXACT *sxact;
5019 : SERIALIZABLEXIDTAG sxidtag;
5020 : uint32 targettaghash;
5021 :
5022 0 : lockRecord = (TwoPhasePredicateLockRecord *) &record->data.lockRecord;
5023 0 : targettaghash = PredicateLockTargetTagHashCode(&lockRecord->target);
5024 :
5025 0 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
5026 0 : sxidtag.xid = xid;
5027 : sxid = (SERIALIZABLEXID *)
5028 0 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
5029 0 : LWLockRelease(SerializableXactHashLock);
5030 :
5031 : Assert(sxid != NULL);
5032 0 : sxact = sxid->myXact;
5033 : Assert(sxact != InvalidSerializableXact);
5034 :
5035 0 : CreatePredicateLock(&lockRecord->target, targettaghash, sxact);
5036 : }
5037 0 : }
5038 :
5039 : /*
5040 : * Prepare to share the current SERIALIZABLEXACT with parallel workers.
5041 : * Return a handle object that can be used by AttachSerializableXact() in a
5042 : * parallel worker.
5043 : */
5044 : SerializableXactHandle
5045 952 : ShareSerializableXact(void)
5046 : {
5047 952 : return MySerializableXact;
5048 : }
5049 :
5050 : /*
5051 : * Allow parallel workers to import the leader's SERIALIZABLEXACT.
5052 : */
5053 : void
5054 2876 : AttachSerializableXact(SerializableXactHandle handle)
5055 : {
5056 :
5057 : Assert(MySerializableXact == InvalidSerializableXact);
5058 :
5059 2876 : MySerializableXact = (SERIALIZABLEXACT *) handle;
5060 2876 : if (MySerializableXact != InvalidSerializableXact)
5061 26 : CreateLocalPredicateLockHash();
5062 2876 : }
|
