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) 1996-2024, 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 : * InitPredicateLocks(void)
157 : * PredicateLockShmemSize(void)
158 : *
159 : * predicate lock reporting
160 : * GetPredicateLockStatusData(void)
161 : * PageIsPredicateLocked(Relation relation, BlockNumber blkno)
162 : *
163 : * predicate lock maintenance
164 : * GetSerializableTransactionSnapshot(Snapshot snapshot)
165 : * SetSerializableTransactionSnapshot(Snapshot snapshot,
166 : * VirtualTransactionId *sourcevxid)
167 : * RegisterPredicateLockingXid(void)
168 : * PredicateLockRelation(Relation relation, Snapshot snapshot)
169 : * PredicateLockPage(Relation relation, BlockNumber blkno,
170 : * Snapshot snapshot)
171 : * PredicateLockTID(Relation relation, ItemPointer tid, Snapshot snapshot,
172 : * TransactionId tuple_xid)
173 : * PredicateLockPageSplit(Relation relation, BlockNumber oldblkno,
174 : * BlockNumber newblkno)
175 : * PredicateLockPageCombine(Relation relation, BlockNumber oldblkno,
176 : * BlockNumber newblkno)
177 : * TransferPredicateLocksToHeapRelation(Relation relation)
178 : * ReleasePredicateLocks(bool isCommit, bool isReadOnlySafe)
179 : *
180 : * conflict detection (may also trigger rollback)
181 : * CheckForSerializableConflictOut(Relation relation, TransactionId xid,
182 : * Snapshot snapshot)
183 : * CheckForSerializableConflictIn(Relation relation, ItemPointer tid,
184 : * BlockNumber blkno)
185 : * CheckTableForSerializableConflictIn(Relation relation)
186 : *
187 : * final rollback checking
188 : * PreCommit_CheckForSerializationFailure(void)
189 : *
190 : * two-phase commit support
191 : * AtPrepare_PredicateLocks(void);
192 : * PostPrepare_PredicateLocks(TransactionId xid);
193 : * PredicateLockTwoPhaseFinish(TransactionId xid, bool isCommit);
194 : * predicatelock_twophase_recover(TransactionId xid, uint16 info,
195 : * void *recdata, uint32 len);
196 : */
197 :
198 : #include "postgres.h"
199 :
200 : #include "access/parallel.h"
201 : #include "access/slru.h"
202 : #include "access/transam.h"
203 : #include "access/twophase.h"
204 : #include "access/twophase_rmgr.h"
205 : #include "access/xact.h"
206 : #include "access/xlog.h"
207 : #include "miscadmin.h"
208 : #include "pgstat.h"
209 : #include "port/pg_lfind.h"
210 : #include "storage/predicate.h"
211 : #include "storage/predicate_internals.h"
212 : #include "storage/proc.h"
213 : #include "storage/procarray.h"
214 : #include "utils/guc_hooks.h"
215 : #include "utils/rel.h"
216 : #include "utils/snapmgr.h"
217 :
218 : /* Uncomment the next line to test the graceful degradation code. */
219 : /* #define TEST_SUMMARIZE_SERIAL */
220 :
221 : /*
222 : * Test the most selective fields first, for performance.
223 : *
224 : * a is covered by b if all of the following hold:
225 : * 1) a.database = b.database
226 : * 2) a.relation = b.relation
227 : * 3) b.offset is invalid (b is page-granularity or higher)
228 : * 4) either of the following:
229 : * 4a) a.offset is valid (a is tuple-granularity) and a.page = b.page
230 : * or 4b) a.offset is invalid and b.page is invalid (a is
231 : * page-granularity and b is relation-granularity
232 : */
233 : #define TargetTagIsCoveredBy(covered_target, covering_target) \
234 : ((GET_PREDICATELOCKTARGETTAG_RELATION(covered_target) == /* (2) */ \
235 : GET_PREDICATELOCKTARGETTAG_RELATION(covering_target)) \
236 : && (GET_PREDICATELOCKTARGETTAG_OFFSET(covering_target) == \
237 : InvalidOffsetNumber) /* (3) */ \
238 : && (((GET_PREDICATELOCKTARGETTAG_OFFSET(covered_target) != \
239 : InvalidOffsetNumber) /* (4a) */ \
240 : && (GET_PREDICATELOCKTARGETTAG_PAGE(covering_target) == \
241 : GET_PREDICATELOCKTARGETTAG_PAGE(covered_target))) \
242 : || ((GET_PREDICATELOCKTARGETTAG_PAGE(covering_target) == \
243 : InvalidBlockNumber) /* (4b) */ \
244 : && (GET_PREDICATELOCKTARGETTAG_PAGE(covered_target) \
245 : != InvalidBlockNumber))) \
246 : && (GET_PREDICATELOCKTARGETTAG_DB(covered_target) == /* (1) */ \
247 : GET_PREDICATELOCKTARGETTAG_DB(covering_target)))
248 :
249 : /*
250 : * The predicate locking target and lock shared hash tables are partitioned to
251 : * reduce contention. To determine which partition a given target belongs to,
252 : * compute the tag's hash code with PredicateLockTargetTagHashCode(), then
253 : * apply one of these macros.
254 : * NB: NUM_PREDICATELOCK_PARTITIONS must be a power of 2!
255 : */
256 : #define PredicateLockHashPartition(hashcode) \
257 : ((hashcode) % NUM_PREDICATELOCK_PARTITIONS)
258 : #define PredicateLockHashPartitionLock(hashcode) \
259 : (&MainLWLockArray[PREDICATELOCK_MANAGER_LWLOCK_OFFSET + \
260 : PredicateLockHashPartition(hashcode)].lock)
261 : #define PredicateLockHashPartitionLockByIndex(i) \
262 : (&MainLWLockArray[PREDICATELOCK_MANAGER_LWLOCK_OFFSET + (i)].lock)
263 :
264 : #define NPREDICATELOCKTARGETENTS() \
265 : mul_size(max_predicate_locks_per_xact, add_size(MaxBackends, max_prepared_xacts))
266 :
267 : #define SxactIsOnFinishedList(sxact) (!dlist_node_is_detached(&(sxact)->finishedLink))
268 :
269 : /*
270 : * Note that a sxact is marked "prepared" once it has passed
271 : * PreCommit_CheckForSerializationFailure, even if it isn't using
272 : * 2PC. This is the point at which it can no longer be aborted.
273 : *
274 : * The PREPARED flag remains set after commit, so SxactIsCommitted
275 : * implies SxactIsPrepared.
276 : */
277 : #define SxactIsCommitted(sxact) (((sxact)->flags & SXACT_FLAG_COMMITTED) != 0)
278 : #define SxactIsPrepared(sxact) (((sxact)->flags & SXACT_FLAG_PREPARED) != 0)
279 : #define SxactIsRolledBack(sxact) (((sxact)->flags & SXACT_FLAG_ROLLED_BACK) != 0)
280 : #define SxactIsDoomed(sxact) (((sxact)->flags & SXACT_FLAG_DOOMED) != 0)
281 : #define SxactIsReadOnly(sxact) (((sxact)->flags & SXACT_FLAG_READ_ONLY) != 0)
282 : #define SxactHasSummaryConflictIn(sxact) (((sxact)->flags & SXACT_FLAG_SUMMARY_CONFLICT_IN) != 0)
283 : #define SxactHasSummaryConflictOut(sxact) (((sxact)->flags & SXACT_FLAG_SUMMARY_CONFLICT_OUT) != 0)
284 : /*
285 : * The following macro actually means that the specified transaction has a
286 : * conflict out *to a transaction which committed ahead of it*. It's hard
287 : * to get that into a name of a reasonable length.
288 : */
289 : #define SxactHasConflictOut(sxact) (((sxact)->flags & SXACT_FLAG_CONFLICT_OUT) != 0)
290 : #define SxactIsDeferrableWaiting(sxact) (((sxact)->flags & SXACT_FLAG_DEFERRABLE_WAITING) != 0)
291 : #define SxactIsROSafe(sxact) (((sxact)->flags & SXACT_FLAG_RO_SAFE) != 0)
292 : #define SxactIsROUnsafe(sxact) (((sxact)->flags & SXACT_FLAG_RO_UNSAFE) != 0)
293 : #define SxactIsPartiallyReleased(sxact) (((sxact)->flags & SXACT_FLAG_PARTIALLY_RELEASED) != 0)
294 :
295 : /*
296 : * Compute the hash code associated with a PREDICATELOCKTARGETTAG.
297 : *
298 : * To avoid unnecessary recomputations of the hash code, we try to do this
299 : * just once per function, and then pass it around as needed. Aside from
300 : * passing the hashcode to hash_search_with_hash_value(), we can extract
301 : * the lock partition number from the hashcode.
302 : */
303 : #define PredicateLockTargetTagHashCode(predicatelocktargettag) \
304 : get_hash_value(PredicateLockTargetHash, predicatelocktargettag)
305 :
306 : /*
307 : * Given a predicate lock tag, and the hash for its target,
308 : * compute the lock hash.
309 : *
310 : * To make the hash code also depend on the transaction, we xor the sxid
311 : * struct's address into the hash code, left-shifted so that the
312 : * partition-number bits don't change. Since this is only a hash, we
313 : * don't care if we lose high-order bits of the address; use an
314 : * intermediate variable to suppress cast-pointer-to-int warnings.
315 : */
316 : #define PredicateLockHashCodeFromTargetHashCode(predicatelocktag, targethash) \
317 : ((targethash) ^ ((uint32) PointerGetDatum((predicatelocktag)->myXact)) \
318 : << LOG2_NUM_PREDICATELOCK_PARTITIONS)
319 :
320 :
321 : /*
322 : * The SLRU buffer area through which we access the old xids.
323 : */
324 : static SlruCtlData SerialSlruCtlData;
325 :
326 : #define SerialSlruCtl (&SerialSlruCtlData)
327 :
328 : #define SERIAL_PAGESIZE BLCKSZ
329 : #define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo)
330 : #define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE)
331 :
332 : /*
333 : * Set maximum pages based on the number needed to track all transactions.
334 : */
335 : #define SERIAL_MAX_PAGE (MaxTransactionId / SERIAL_ENTRIESPERPAGE)
336 :
337 : #define SerialNextPage(page) (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1)
338 :
339 : #define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
340 : (SerialSlruCtl->shared->page_buffer[slotno] + \
341 : ((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
342 :
343 : #define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
344 :
345 : typedef struct SerialControlData
346 : {
347 : int 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 298314 : PredicateLockingNeededForRelation(Relation relation)
499 : {
500 412616 : return !(relation->rd_id < FirstUnpinnedObjectId ||
501 114302 : 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 98619110 : SerializationNeededForRead(Relation relation, Snapshot snapshot)
517 : {
518 : /* Nothing to do if this is not a serializable transaction */
519 98619110 : if (MySerializableXact == InvalidSerializableXact)
520 98334460 : 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 284650 : if (!IsMVCCSnapshot(snapshot))
530 3302 : 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 281348 : 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 281282 : if (!PredicateLockingNeededForRelation(relation))
550 176384 : return false;
551 :
552 104898 : 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 29686872 : SerializationNeededForWrite(Relation relation)
561 : {
562 : /* Nothing to do if this is not a serializable transaction */
563 29686872 : if (MySerializableXact == InvalidSerializableXact)
564 29670070 : return false;
565 :
566 : /* Check if the relation doesn't participate in predicate locking */
567 16802 : if (!PredicateLockingNeededForRelation(relation))
568 7874 : return false;
569 :
570 8928 : 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 5074 : CreatePredXact(void)
583 : {
584 : SERIALIZABLEXACT *sxact;
585 :
586 5074 : if (dlist_is_empty(&PredXact->availableList))
587 0 : return NULL;
588 :
589 5074 : sxact = dlist_container(SERIALIZABLEXACT, xactLink,
590 : dlist_pop_head_node(&PredXact->availableList));
591 5074 : dlist_push_tail(&PredXact->activeList, &sxact->xactLink);
592 5074 : return sxact;
593 : }
594 :
595 : static void
596 3306 : ReleasePredXact(SERIALIZABLEXACT *sxact)
597 : {
598 : Assert(ShmemAddrIsValid(sxact));
599 :
600 3306 : dlist_delete(&sxact->xactLink);
601 3306 : dlist_push_tail(&PredXact->availableList, &sxact->xactLink);
602 3306 : }
603 :
604 : /*------------------------------------------------------------------------*/
605 :
606 : /*
607 : * These functions manage primitive access to the RWConflict pool and lists.
608 : */
609 : static bool
610 3744 : 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 3744 : if (SxactIsDoomed(reader)
618 3744 : || SxactIsDoomed(writer)
619 3744 : || dlist_is_empty(&reader->outConflicts)
620 1138 : || dlist_is_empty(&writer->inConflicts))
621 2690 : return false;
622 :
623 : /*
624 : * A conflict is possible; walk the list to find out.
625 : *
626 : * The unconstify is needed as we have no const version of
627 : * dlist_foreach().
628 : */
629 1086 : dlist_foreach(iter, &unconstify(SERIALIZABLEXACT *, reader)->outConflicts)
630 : {
631 1054 : RWConflict conflict =
632 1054 : dlist_container(RWConflictData, outLink, iter.cur);
633 :
634 1054 : if (conflict->sxactIn == writer)
635 1022 : return true;
636 : }
637 :
638 : /* No conflict found. */
639 32 : return false;
640 : }
641 :
642 : static void
643 1564 : SetRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
644 : {
645 : RWConflict conflict;
646 :
647 : Assert(reader != writer);
648 : Assert(!RWConflictExists(reader, writer));
649 :
650 1564 : 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 1564 : conflict = dlist_head_element(RWConflictData, outLink, &RWConflictPool->availableList);
657 1564 : dlist_delete(&conflict->outLink);
658 :
659 1564 : conflict->sxactOut = reader;
660 1564 : conflict->sxactIn = writer;
661 1564 : dlist_push_tail(&reader->outConflicts, &conflict->outLink);
662 1564 : dlist_push_tail(&writer->inConflicts, &conflict->inLink);
663 1564 : }
664 :
665 : static void
666 268 : SetPossibleUnsafeConflict(SERIALIZABLEXACT *roXact,
667 : SERIALIZABLEXACT *activeXact)
668 : {
669 : RWConflict conflict;
670 :
671 : Assert(roXact != activeXact);
672 : Assert(SxactIsReadOnly(roXact));
673 : Assert(!SxactIsReadOnly(activeXact));
674 :
675 268 : if (dlist_is_empty(&RWConflictPool->availableList))
676 0 : ereport(ERROR,
677 : (errcode(ERRCODE_OUT_OF_MEMORY),
678 : errmsg("not enough elements in RWConflictPool to record a potential read/write conflict"),
679 : errhint("You might need to run fewer transactions at a time or increase max_connections.")));
680 :
681 268 : conflict = dlist_head_element(RWConflictData, outLink, &RWConflictPool->availableList);
682 268 : dlist_delete(&conflict->outLink);
683 :
684 268 : conflict->sxactOut = activeXact;
685 268 : conflict->sxactIn = roXact;
686 268 : dlist_push_tail(&activeXact->possibleUnsafeConflicts, &conflict->outLink);
687 268 : dlist_push_tail(&roXact->possibleUnsafeConflicts, &conflict->inLink);
688 268 : }
689 :
690 : static void
691 1832 : ReleaseRWConflict(RWConflict conflict)
692 : {
693 1832 : dlist_delete(&conflict->inLink);
694 1832 : dlist_delete(&conflict->outLink);
695 1832 : dlist_push_tail(&RWConflictPool->availableList, &conflict->outLink);
696 1832 : }
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 1768 : SerialInit(void)
807 : {
808 : bool found;
809 :
810 : /*
811 : * Set up SLRU management of the pg_serial data.
812 : */
813 1768 : SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
814 1768 : 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 1768 : serialControl = (SerialControl)
827 1768 : ShmemInitStruct("SerialControlData", sizeof(SerialControlData), &found);
828 :
829 : Assert(found == IsUnderPostmaster);
830 1768 : if (!found)
831 : {
832 : /*
833 : * Set control information to reflect empty SLRU.
834 : */
835 1768 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
836 1768 : serialControl->headPage = -1;
837 1768 : serialControl->headXid = InvalidTransactionId;
838 1768 : serialControl->tailXid = InvalidTransactionId;
839 1768 : LWLockRelease(SerialControlLock);
840 : }
841 1768 : }
842 :
843 : /*
844 : * GUC check_hook for serializable_buffers
845 : */
846 : bool
847 1830 : check_serial_buffers(int *newval, void **extra, GucSource source)
848 : {
849 1830 : 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 no serializable transactions are active, there shouldn't be anything
881 : * to push out to the SLRU. Hitting this assert would mean there's
882 : * something wrong with the earlier cleanup logic.
883 : */
884 0 : tailXid = serialControl->tailXid;
885 : Assert(TransactionIdIsValid(tailXid));
886 :
887 : /*
888 : * If the SLRU is currently unused, zero out the whole active region from
889 : * tailXid to headXid before taking it into use. Otherwise zero out only
890 : * any new pages that enter the tailXid-headXid range as we advance
891 : * headXid.
892 : */
893 0 : if (serialControl->headPage < 0)
894 : {
895 0 : firstZeroPage = SerialPage(tailXid);
896 0 : isNewPage = true;
897 : }
898 : else
899 : {
900 0 : firstZeroPage = SerialNextPage(serialControl->headPage);
901 0 : isNewPage = SerialPagePrecedesLogically(serialControl->headPage,
902 : targetPage);
903 : }
904 :
905 0 : if (!TransactionIdIsValid(serialControl->headXid)
906 0 : || TransactionIdFollows(xid, serialControl->headXid))
907 0 : serialControl->headXid = xid;
908 0 : if (isNewPage)
909 0 : serialControl->headPage = targetPage;
910 :
911 0 : if (isNewPage)
912 : {
913 : /* Initialize intervening pages; might involve trading locks */
914 : for (;;)
915 : {
916 0 : lock = SimpleLruGetBankLock(SerialSlruCtl, firstZeroPage);
917 0 : LWLockAcquire(lock, LW_EXCLUSIVE);
918 0 : slotno = SimpleLruZeroPage(SerialSlruCtl, firstZeroPage);
919 0 : if (firstZeroPage == targetPage)
920 0 : break;
921 0 : firstZeroPage = SerialNextPage(firstZeroPage);
922 0 : LWLockRelease(lock);
923 : }
924 : }
925 : else
926 : {
927 0 : LWLockAcquire(lock, LW_EXCLUSIVE);
928 0 : slotno = SimpleLruReadPage(SerialSlruCtl, targetPage, true, xid);
929 : }
930 :
931 0 : SerialValue(slotno, xid) = minConflictCommitSeqNo;
932 0 : SerialSlruCtl->shared->page_dirty[slotno] = true;
933 :
934 0 : LWLockRelease(lock);
935 0 : LWLockRelease(SerialControlLock);
936 0 : }
937 :
938 : /*
939 : * Get the minimum commitSeqNo for any conflict out for the given xid. For
940 : * a transaction which exists but has no conflict out, InvalidSerCommitSeqNo
941 : * will be returned.
942 : */
943 : static SerCommitSeqNo
944 48 : SerialGetMinConflictCommitSeqNo(TransactionId xid)
945 : {
946 : TransactionId headXid;
947 : TransactionId tailXid;
948 : SerCommitSeqNo val;
949 : int slotno;
950 :
951 : Assert(TransactionIdIsValid(xid));
952 :
953 48 : LWLockAcquire(SerialControlLock, LW_SHARED);
954 48 : headXid = serialControl->headXid;
955 48 : tailXid = serialControl->tailXid;
956 48 : LWLockRelease(SerialControlLock);
957 :
958 48 : if (!TransactionIdIsValid(headXid))
959 48 : return 0;
960 :
961 : Assert(TransactionIdIsValid(tailXid));
962 :
963 0 : if (TransactionIdPrecedes(xid, tailXid)
964 0 : || TransactionIdFollows(xid, headXid))
965 0 : return 0;
966 :
967 : /*
968 : * The following function must be called without holding SLRU bank lock,
969 : * but will return with that lock held, which must then be released.
970 : */
971 0 : slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
972 : SerialPage(xid), xid);
973 0 : val = SerialValue(slotno, xid);
974 0 : LWLockRelease(SimpleLruGetBankLock(SerialSlruCtl, SerialPage(xid)));
975 0 : return val;
976 : }
977 :
978 : /*
979 : * Call this whenever there is a new xmin for active serializable
980 : * transactions. We don't need to keep information on transactions which
981 : * precede that. InvalidTransactionId means none active, so everything in
982 : * the SLRU can be discarded.
983 : */
984 : static void
985 3458 : SerialSetActiveSerXmin(TransactionId xid)
986 : {
987 3458 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
988 :
989 : /*
990 : * When no sxacts are active, nothing overlaps, set the xid values to
991 : * invalid to show that there are no valid entries. Don't clear headPage,
992 : * though. A new xmin might still land on that page, and we don't want to
993 : * repeatedly zero out the same page.
994 : */
995 3458 : if (!TransactionIdIsValid(xid))
996 : {
997 1710 : serialControl->tailXid = InvalidTransactionId;
998 1710 : serialControl->headXid = InvalidTransactionId;
999 1710 : LWLockRelease(SerialControlLock);
1000 1710 : return;
1001 : }
1002 :
1003 : /*
1004 : * When we're recovering prepared transactions, the global xmin might move
1005 : * backwards depending on the order they're recovered. Normally that's not
1006 : * OK, but during recovery no serializable transactions will commit, so
1007 : * the SLRU is empty and we can get away with it.
1008 : */
1009 1748 : if (RecoveryInProgress())
1010 : {
1011 : Assert(serialControl->headPage < 0);
1012 0 : if (!TransactionIdIsValid(serialControl->tailXid)
1013 0 : || TransactionIdPrecedes(xid, serialControl->tailXid))
1014 : {
1015 0 : serialControl->tailXid = xid;
1016 : }
1017 0 : LWLockRelease(SerialControlLock);
1018 0 : return;
1019 : }
1020 :
1021 : Assert(!TransactionIdIsValid(serialControl->tailXid)
1022 : || TransactionIdFollows(xid, serialControl->tailXid));
1023 :
1024 1748 : serialControl->tailXid = xid;
1025 :
1026 1748 : LWLockRelease(SerialControlLock);
1027 : }
1028 :
1029 : /*
1030 : * Perform a checkpoint --- either during shutdown, or on-the-fly
1031 : *
1032 : * We don't have any data that needs to survive a restart, but this is a
1033 : * convenient place to truncate the SLRU.
1034 : */
1035 : void
1036 1704 : CheckPointPredicate(void)
1037 : {
1038 : int truncateCutoffPage;
1039 :
1040 1704 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
1041 :
1042 : /* Exit quickly if the SLRU is currently not in use. */
1043 1704 : if (serialControl->headPage < 0)
1044 : {
1045 1704 : LWLockRelease(SerialControlLock);
1046 1704 : return;
1047 : }
1048 :
1049 0 : if (TransactionIdIsValid(serialControl->tailXid))
1050 : {
1051 : int tailPage;
1052 :
1053 0 : tailPage = SerialPage(serialControl->tailXid);
1054 :
1055 : /*
1056 : * It is possible for the tailXid to be ahead of the headXid. This
1057 : * occurs if we checkpoint while there are in-progress serializable
1058 : * transaction(s) advancing the tail but we are yet to summarize the
1059 : * transactions. In this case, we cutoff up to the headPage and the
1060 : * next summary will advance the headXid.
1061 : */
1062 0 : if (SerialPagePrecedesLogically(tailPage, serialControl->headPage))
1063 : {
1064 : /* We can truncate the SLRU up to the page containing tailXid */
1065 0 : truncateCutoffPage = tailPage;
1066 : }
1067 : else
1068 0 : truncateCutoffPage = serialControl->headPage;
1069 : }
1070 : else
1071 : {
1072 : /*----------
1073 : * The SLRU is no longer needed. Truncate to head before we set head
1074 : * invalid.
1075 : *
1076 : * XXX: It's possible that the SLRU is not needed again until XID
1077 : * wrap-around has happened, so that the segment containing headPage
1078 : * that we leave behind will appear to be new again. In that case it
1079 : * won't be removed until XID horizon advances enough to make it
1080 : * current again.
1081 : *
1082 : * XXX: This should happen in vac_truncate_clog(), not in checkpoints.
1083 : * Consider this scenario, starting from a system with no in-progress
1084 : * transactions and VACUUM FREEZE having maximized oldestXact:
1085 : * - Start a SERIALIZABLE transaction.
1086 : * - Start, finish, and summarize a SERIALIZABLE transaction, creating
1087 : * one SLRU page.
1088 : * - Consume XIDs to reach xidStopLimit.
1089 : * - Finish all transactions. Due to the long-running SERIALIZABLE
1090 : * transaction, earlier checkpoints did not touch headPage. The
1091 : * next checkpoint will change it, but that checkpoint happens after
1092 : * the end of the scenario.
1093 : * - VACUUM to advance XID limits.
1094 : * - Consume ~2M XIDs, crossing the former xidWrapLimit.
1095 : * - Start, finish, and summarize a SERIALIZABLE transaction.
1096 : * SerialAdd() declines to create the targetPage, because headPage
1097 : * is not regarded as in the past relative to that targetPage. The
1098 : * transaction instigating the summarize fails in
1099 : * SimpleLruReadPage().
1100 : */
1101 0 : truncateCutoffPage = serialControl->headPage;
1102 0 : serialControl->headPage = -1;
1103 : }
1104 :
1105 0 : LWLockRelease(SerialControlLock);
1106 :
1107 : /*
1108 : * Truncate away pages that are no longer required. Note that no
1109 : * additional locking is required, because this is only called as part of
1110 : * a checkpoint, and the validity limits have already been determined.
1111 : */
1112 0 : SimpleLruTruncate(SerialSlruCtl, truncateCutoffPage);
1113 :
1114 : /*
1115 : * Write dirty SLRU pages to disk
1116 : *
1117 : * This is not actually necessary from a correctness point of view. We do
1118 : * it merely as a debugging aid.
1119 : *
1120 : * We're doing this after the truncation to avoid writing pages right
1121 : * before deleting the file in which they sit, which would be completely
1122 : * pointless.
1123 : */
1124 0 : SimpleLruWriteAll(SerialSlruCtl, true);
1125 : }
1126 :
1127 : /*------------------------------------------------------------------------*/
1128 :
1129 : /*
1130 : * InitPredicateLocks -- Initialize the predicate locking data structures.
1131 : *
1132 : * This is called from CreateSharedMemoryAndSemaphores(), which see for
1133 : * more comments. In the normal postmaster case, the shared hash tables
1134 : * are created here. Backends inherit the pointers
1135 : * to the shared tables via fork(). In the EXEC_BACKEND case, each
1136 : * backend re-executes this code to obtain pointers to the already existing
1137 : * shared hash tables.
1138 : */
1139 : void
1140 1768 : InitPredicateLocks(void)
1141 : {
1142 : HASHCTL info;
1143 : long max_table_size;
1144 : Size requestSize;
1145 : bool found;
1146 :
1147 : #ifndef EXEC_BACKEND
1148 : Assert(!IsUnderPostmaster);
1149 : #endif
1150 :
1151 : /*
1152 : * Compute size of predicate lock target hashtable. Note these
1153 : * calculations must agree with PredicateLockShmemSize!
1154 : */
1155 1768 : max_table_size = NPREDICATELOCKTARGETENTS();
1156 :
1157 : /*
1158 : * Allocate hash table for PREDICATELOCKTARGET structs. This stores
1159 : * per-predicate-lock-target information.
1160 : */
1161 1768 : info.keysize = sizeof(PREDICATELOCKTARGETTAG);
1162 1768 : info.entrysize = sizeof(PREDICATELOCKTARGET);
1163 1768 : info.num_partitions = NUM_PREDICATELOCK_PARTITIONS;
1164 :
1165 1768 : PredicateLockTargetHash = ShmemInitHash("PREDICATELOCKTARGET hash",
1166 : max_table_size,
1167 : max_table_size,
1168 : &info,
1169 : HASH_ELEM | HASH_BLOBS |
1170 : HASH_PARTITION | HASH_FIXED_SIZE);
1171 :
1172 : /*
1173 : * Reserve a dummy entry in the hash table; we use it to make sure there's
1174 : * always one entry available when we need to split or combine a page,
1175 : * because running out of space there could mean aborting a
1176 : * non-serializable transaction.
1177 : */
1178 1768 : if (!IsUnderPostmaster)
1179 : {
1180 1768 : (void) hash_search(PredicateLockTargetHash, &ScratchTargetTag,
1181 : HASH_ENTER, &found);
1182 : Assert(!found);
1183 : }
1184 :
1185 : /* Pre-calculate the hash and partition lock of the scratch entry */
1186 1768 : ScratchTargetTagHash = PredicateLockTargetTagHashCode(&ScratchTargetTag);
1187 1768 : ScratchPartitionLock = PredicateLockHashPartitionLock(ScratchTargetTagHash);
1188 :
1189 : /*
1190 : * Allocate hash table for PREDICATELOCK structs. This stores per
1191 : * xact-lock-of-a-target information.
1192 : */
1193 1768 : info.keysize = sizeof(PREDICATELOCKTAG);
1194 1768 : info.entrysize = sizeof(PREDICATELOCK);
1195 1768 : info.hash = predicatelock_hash;
1196 1768 : info.num_partitions = NUM_PREDICATELOCK_PARTITIONS;
1197 :
1198 : /* Assume an average of 2 xacts per target */
1199 1768 : max_table_size *= 2;
1200 :
1201 1768 : PredicateLockHash = ShmemInitHash("PREDICATELOCK hash",
1202 : max_table_size,
1203 : max_table_size,
1204 : &info,
1205 : HASH_ELEM | HASH_FUNCTION |
1206 : HASH_PARTITION | HASH_FIXED_SIZE);
1207 :
1208 : /*
1209 : * Compute size for serializable transaction hashtable. Note these
1210 : * calculations must agree with PredicateLockShmemSize!
1211 : */
1212 1768 : max_table_size = (MaxBackends + max_prepared_xacts);
1213 :
1214 : /*
1215 : * Allocate a list to hold information on transactions participating in
1216 : * predicate locking.
1217 : *
1218 : * Assume an average of 10 predicate locking transactions per backend.
1219 : * This allows aggressive cleanup while detail is present before data must
1220 : * be summarized for storage in SLRU and the "dummy" transaction.
1221 : */
1222 1768 : max_table_size *= 10;
1223 :
1224 1768 : PredXact = ShmemInitStruct("PredXactList",
1225 : PredXactListDataSize,
1226 : &found);
1227 : Assert(found == IsUnderPostmaster);
1228 1768 : if (!found)
1229 : {
1230 : int i;
1231 :
1232 1768 : dlist_init(&PredXact->availableList);
1233 1768 : dlist_init(&PredXact->activeList);
1234 1768 : PredXact->SxactGlobalXmin = InvalidTransactionId;
1235 1768 : PredXact->SxactGlobalXminCount = 0;
1236 1768 : PredXact->WritableSxactCount = 0;
1237 1768 : PredXact->LastSxactCommitSeqNo = FirstNormalSerCommitSeqNo - 1;
1238 1768 : PredXact->CanPartialClearThrough = 0;
1239 1768 : PredXact->HavePartialClearedThrough = 0;
1240 1768 : requestSize = mul_size((Size) max_table_size,
1241 : sizeof(SERIALIZABLEXACT));
1242 1768 : PredXact->element = ShmemAlloc(requestSize);
1243 : /* Add all elements to available list, clean. */
1244 1768 : memset(PredXact->element, 0, requestSize);
1245 1480068 : for (i = 0; i < max_table_size; i++)
1246 : {
1247 1478300 : LWLockInitialize(&PredXact->element[i].perXactPredicateListLock,
1248 : LWTRANCHE_PER_XACT_PREDICATE_LIST);
1249 1478300 : dlist_push_tail(&PredXact->availableList, &PredXact->element[i].xactLink);
1250 : }
1251 1768 : PredXact->OldCommittedSxact = CreatePredXact();
1252 1768 : SetInvalidVirtualTransactionId(PredXact->OldCommittedSxact->vxid);
1253 1768 : PredXact->OldCommittedSxact->prepareSeqNo = 0;
1254 1768 : PredXact->OldCommittedSxact->commitSeqNo = 0;
1255 1768 : PredXact->OldCommittedSxact->SeqNo.lastCommitBeforeSnapshot = 0;
1256 1768 : dlist_init(&PredXact->OldCommittedSxact->outConflicts);
1257 1768 : dlist_init(&PredXact->OldCommittedSxact->inConflicts);
1258 1768 : dlist_init(&PredXact->OldCommittedSxact->predicateLocks);
1259 1768 : dlist_node_init(&PredXact->OldCommittedSxact->finishedLink);
1260 1768 : dlist_init(&PredXact->OldCommittedSxact->possibleUnsafeConflicts);
1261 1768 : PredXact->OldCommittedSxact->topXid = InvalidTransactionId;
1262 1768 : PredXact->OldCommittedSxact->finishedBefore = InvalidTransactionId;
1263 1768 : PredXact->OldCommittedSxact->xmin = InvalidTransactionId;
1264 1768 : PredXact->OldCommittedSxact->flags = SXACT_FLAG_COMMITTED;
1265 1768 : PredXact->OldCommittedSxact->pid = 0;
1266 1768 : PredXact->OldCommittedSxact->pgprocno = INVALID_PROC_NUMBER;
1267 : }
1268 : /* This never changes, so let's keep a local copy. */
1269 1768 : OldCommittedSxact = PredXact->OldCommittedSxact;
1270 :
1271 : /*
1272 : * Allocate hash table for SERIALIZABLEXID structs. This stores per-xid
1273 : * information for serializable transactions which have accessed data.
1274 : */
1275 1768 : info.keysize = sizeof(SERIALIZABLEXIDTAG);
1276 1768 : info.entrysize = sizeof(SERIALIZABLEXID);
1277 :
1278 1768 : SerializableXidHash = ShmemInitHash("SERIALIZABLEXID hash",
1279 : max_table_size,
1280 : max_table_size,
1281 : &info,
1282 : HASH_ELEM | HASH_BLOBS |
1283 : HASH_FIXED_SIZE);
1284 :
1285 : /*
1286 : * Allocate space for tracking rw-conflicts in lists attached to the
1287 : * transactions.
1288 : *
1289 : * Assume an average of 5 conflicts per transaction. Calculations suggest
1290 : * that this will prevent resource exhaustion in even the most pessimal
1291 : * loads up to max_connections = 200 with all 200 connections pounding the
1292 : * database with serializable transactions. Beyond that, there may be
1293 : * occasional transactions canceled when trying to flag conflicts. That's
1294 : * probably OK.
1295 : */
1296 1768 : max_table_size *= 5;
1297 :
1298 1768 : RWConflictPool = ShmemInitStruct("RWConflictPool",
1299 : RWConflictPoolHeaderDataSize,
1300 : &found);
1301 : Assert(found == IsUnderPostmaster);
1302 1768 : if (!found)
1303 : {
1304 : int i;
1305 :
1306 1768 : dlist_init(&RWConflictPool->availableList);
1307 1768 : requestSize = mul_size((Size) max_table_size,
1308 : RWConflictDataSize);
1309 1768 : RWConflictPool->element = ShmemAlloc(requestSize);
1310 : /* Add all elements to available list, clean. */
1311 1768 : memset(RWConflictPool->element, 0, requestSize);
1312 7393268 : for (i = 0; i < max_table_size; i++)
1313 : {
1314 7391500 : dlist_push_tail(&RWConflictPool->availableList,
1315 7391500 : &RWConflictPool->element[i].outLink);
1316 : }
1317 : }
1318 :
1319 : /*
1320 : * Create or attach to the header for the list of finished serializable
1321 : * transactions.
1322 : */
1323 1768 : FinishedSerializableTransactions = (dlist_head *)
1324 1768 : ShmemInitStruct("FinishedSerializableTransactions",
1325 : sizeof(dlist_head),
1326 : &found);
1327 : Assert(found == IsUnderPostmaster);
1328 1768 : if (!found)
1329 1768 : dlist_init(FinishedSerializableTransactions);
1330 :
1331 : /*
1332 : * Initialize the SLRU storage for old committed serializable
1333 : * transactions.
1334 : */
1335 1768 : SerialInit();
1336 1768 : }
1337 :
1338 : /*
1339 : * Estimate shared-memory space used for predicate lock table
1340 : */
1341 : Size
1342 3298 : PredicateLockShmemSize(void)
1343 : {
1344 3298 : Size size = 0;
1345 : long max_table_size;
1346 :
1347 : /* predicate lock target hash table */
1348 3298 : max_table_size = NPREDICATELOCKTARGETENTS();
1349 3298 : size = add_size(size, hash_estimate_size(max_table_size,
1350 : sizeof(PREDICATELOCKTARGET)));
1351 :
1352 : /* predicate lock hash table */
1353 3298 : max_table_size *= 2;
1354 3298 : size = add_size(size, hash_estimate_size(max_table_size,
1355 : sizeof(PREDICATELOCK)));
1356 :
1357 : /*
1358 : * Since NPREDICATELOCKTARGETENTS is only an estimate, add 10% safety
1359 : * margin.
1360 : */
1361 3298 : size = add_size(size, size / 10);
1362 :
1363 : /* transaction list */
1364 3298 : max_table_size = MaxBackends + max_prepared_xacts;
1365 3298 : max_table_size *= 10;
1366 3298 : size = add_size(size, PredXactListDataSize);
1367 3298 : size = add_size(size, mul_size((Size) max_table_size,
1368 : sizeof(SERIALIZABLEXACT)));
1369 :
1370 : /* transaction xid table */
1371 3298 : size = add_size(size, hash_estimate_size(max_table_size,
1372 : sizeof(SERIALIZABLEXID)));
1373 :
1374 : /* rw-conflict pool */
1375 3298 : max_table_size *= 5;
1376 3298 : size = add_size(size, RWConflictPoolHeaderDataSize);
1377 3298 : size = add_size(size, mul_size((Size) max_table_size,
1378 : RWConflictDataSize));
1379 :
1380 : /* Head for list of finished serializable transactions. */
1381 3298 : size = add_size(size, sizeof(dlist_head));
1382 :
1383 : /* Shared memory structures for SLRU tracking of old committed xids. */
1384 3298 : size = add_size(size, sizeof(SerialControlData));
1385 3298 : size = add_size(size, SimpleLruShmemSize(serializable_buffers, 0));
1386 :
1387 3298 : return size;
1388 : }
1389 :
1390 :
1391 : /*
1392 : * Compute the hash code associated with a PREDICATELOCKTAG.
1393 : *
1394 : * Because we want to use just one set of partition locks for both the
1395 : * PREDICATELOCKTARGET and PREDICATELOCK hash tables, we have to make sure
1396 : * that PREDICATELOCKs fall into the same partition number as their
1397 : * associated PREDICATELOCKTARGETs. dynahash.c expects the partition number
1398 : * to be the low-order bits of the hash code, and therefore a
1399 : * PREDICATELOCKTAG's hash code must have the same low-order bits as the
1400 : * associated PREDICATELOCKTARGETTAG's hash code. We achieve this with this
1401 : * specialized hash function.
1402 : */
1403 : static uint32
1404 0 : predicatelock_hash(const void *key, Size keysize)
1405 : {
1406 0 : const PREDICATELOCKTAG *predicatelocktag = (const PREDICATELOCKTAG *) key;
1407 : uint32 targethash;
1408 :
1409 : Assert(keysize == sizeof(PREDICATELOCKTAG));
1410 :
1411 : /* Look into the associated target object, and compute its hash code */
1412 0 : targethash = PredicateLockTargetTagHashCode(&predicatelocktag->myTarget->tag);
1413 :
1414 0 : return PredicateLockHashCodeFromTargetHashCode(predicatelocktag, targethash);
1415 : }
1416 :
1417 :
1418 : /*
1419 : * GetPredicateLockStatusData
1420 : * Return a table containing the internal state of the predicate
1421 : * lock manager for use in pg_lock_status.
1422 : *
1423 : * Like GetLockStatusData, this function tries to hold the partition LWLocks
1424 : * for as short a time as possible by returning two arrays that simply
1425 : * contain the PREDICATELOCKTARGETTAG and SERIALIZABLEXACT for each lock
1426 : * table entry. Multiple copies of the same PREDICATELOCKTARGETTAG and
1427 : * SERIALIZABLEXACT will likely appear.
1428 : */
1429 : PredicateLockData *
1430 524 : GetPredicateLockStatusData(void)
1431 : {
1432 : PredicateLockData *data;
1433 : int i;
1434 : int els,
1435 : el;
1436 : HASH_SEQ_STATUS seqstat;
1437 : PREDICATELOCK *predlock;
1438 :
1439 524 : data = (PredicateLockData *) palloc(sizeof(PredicateLockData));
1440 :
1441 : /*
1442 : * To ensure consistency, take simultaneous locks on all partition locks
1443 : * in ascending order, then SerializableXactHashLock.
1444 : */
1445 8908 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
1446 8384 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_SHARED);
1447 524 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1448 :
1449 : /* Get number of locks and allocate appropriately-sized arrays. */
1450 524 : els = hash_get_num_entries(PredicateLockHash);
1451 524 : data->nelements = els;
1452 524 : data->locktags = (PREDICATELOCKTARGETTAG *)
1453 524 : palloc(sizeof(PREDICATELOCKTARGETTAG) * els);
1454 524 : data->xacts = (SERIALIZABLEXACT *)
1455 524 : palloc(sizeof(SERIALIZABLEXACT) * els);
1456 :
1457 :
1458 : /* Scan through PredicateLockHash and copy contents */
1459 524 : hash_seq_init(&seqstat, PredicateLockHash);
1460 :
1461 524 : el = 0;
1462 :
1463 530 : while ((predlock = (PREDICATELOCK *) hash_seq_search(&seqstat)))
1464 : {
1465 6 : data->locktags[el] = predlock->tag.myTarget->tag;
1466 6 : data->xacts[el] = *predlock->tag.myXact;
1467 6 : el++;
1468 : }
1469 :
1470 : Assert(el == els);
1471 :
1472 : /* Release locks in reverse order */
1473 524 : LWLockRelease(SerializableXactHashLock);
1474 8908 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
1475 8384 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
1476 :
1477 524 : return data;
1478 : }
1479 :
1480 : /*
1481 : * Free up shared memory structures by pushing the oldest sxact (the one at
1482 : * the front of the SummarizeOldestCommittedSxact queue) into summary form.
1483 : * Each call will free exactly one SERIALIZABLEXACT structure and may also
1484 : * free one or more of these structures: SERIALIZABLEXID, PREDICATELOCK,
1485 : * PREDICATELOCKTARGET, RWConflictData.
1486 : */
1487 : static void
1488 0 : SummarizeOldestCommittedSxact(void)
1489 : {
1490 : SERIALIZABLEXACT *sxact;
1491 :
1492 0 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
1493 :
1494 : /*
1495 : * This function is only called if there are no sxact slots available.
1496 : * Some of them must belong to old, already-finished transactions, so
1497 : * there should be something in FinishedSerializableTransactions list that
1498 : * we can summarize. However, there's a race condition: while we were not
1499 : * holding any locks, a transaction might have ended and cleaned up all
1500 : * the finished sxact entries already, freeing up their sxact slots. In
1501 : * that case, we have nothing to do here. The caller will find one of the
1502 : * slots released by the other backend when it retries.
1503 : */
1504 0 : if (dlist_is_empty(FinishedSerializableTransactions))
1505 : {
1506 0 : LWLockRelease(SerializableFinishedListLock);
1507 0 : return;
1508 : }
1509 :
1510 : /*
1511 : * Grab the first sxact off the finished list -- this will be the earliest
1512 : * commit. Remove it from the list.
1513 : */
1514 0 : sxact = dlist_head_element(SERIALIZABLEXACT, finishedLink,
1515 : FinishedSerializableTransactions);
1516 0 : dlist_delete_thoroughly(&sxact->finishedLink);
1517 :
1518 : /* Add to SLRU summary information. */
1519 0 : if (TransactionIdIsValid(sxact->topXid) && !SxactIsReadOnly(sxact))
1520 0 : SerialAdd(sxact->topXid, SxactHasConflictOut(sxact)
1521 : ? sxact->SeqNo.earliestOutConflictCommit : InvalidSerCommitSeqNo);
1522 :
1523 : /* Summarize and release the detail. */
1524 0 : ReleaseOneSerializableXact(sxact, false, true);
1525 :
1526 0 : LWLockRelease(SerializableFinishedListLock);
1527 : }
1528 :
1529 : /*
1530 : * GetSafeSnapshot
1531 : * Obtain and register a snapshot for a READ ONLY DEFERRABLE
1532 : * transaction. Ensures that the snapshot is "safe", i.e. a
1533 : * read-only transaction running on it can execute serializably
1534 : * without further checks. This requires waiting for concurrent
1535 : * transactions to complete, and retrying with a new snapshot if
1536 : * one of them could possibly create a conflict.
1537 : *
1538 : * As with GetSerializableTransactionSnapshot (which this is a subroutine
1539 : * for), the passed-in Snapshot pointer should reference a static data
1540 : * area that can safely be passed to GetSnapshotData.
1541 : */
1542 : static Snapshot
1543 10 : GetSafeSnapshot(Snapshot origSnapshot)
1544 : {
1545 : Snapshot snapshot;
1546 :
1547 : Assert(XactReadOnly && XactDeferrable);
1548 :
1549 : while (true)
1550 : {
1551 : /*
1552 : * GetSerializableTransactionSnapshotInt is going to call
1553 : * GetSnapshotData, so we need to provide it the static snapshot area
1554 : * our caller passed to us. The pointer returned is actually the same
1555 : * one passed to it, but we avoid assuming that here.
1556 : */
1557 10 : snapshot = GetSerializableTransactionSnapshotInt(origSnapshot,
1558 : NULL, InvalidPid);
1559 :
1560 10 : if (MySerializableXact == InvalidSerializableXact)
1561 6 : return snapshot; /* no concurrent r/w xacts; it's safe */
1562 :
1563 4 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1564 :
1565 : /*
1566 : * Wait for concurrent transactions to finish. Stop early if one of
1567 : * them marked us as conflicted.
1568 : */
1569 4 : MySerializableXact->flags |= SXACT_FLAG_DEFERRABLE_WAITING;
1570 8 : while (!(dlist_is_empty(&MySerializableXact->possibleUnsafeConflicts) ||
1571 4 : SxactIsROUnsafe(MySerializableXact)))
1572 : {
1573 4 : LWLockRelease(SerializableXactHashLock);
1574 4 : ProcWaitForSignal(WAIT_EVENT_SAFE_SNAPSHOT);
1575 4 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1576 : }
1577 4 : MySerializableXact->flags &= ~SXACT_FLAG_DEFERRABLE_WAITING;
1578 :
1579 4 : if (!SxactIsROUnsafe(MySerializableXact))
1580 : {
1581 2 : LWLockRelease(SerializableXactHashLock);
1582 2 : break; /* success */
1583 : }
1584 :
1585 2 : LWLockRelease(SerializableXactHashLock);
1586 :
1587 : /* else, need to retry... */
1588 2 : ereport(DEBUG2,
1589 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
1590 : errmsg_internal("deferrable snapshot was unsafe; trying a new one")));
1591 2 : ReleasePredicateLocks(false, false);
1592 : }
1593 :
1594 : /*
1595 : * Now we have a safe snapshot, so we don't need to do any further checks.
1596 : */
1597 : Assert(SxactIsROSafe(MySerializableXact));
1598 2 : ReleasePredicateLocks(false, true);
1599 :
1600 2 : return snapshot;
1601 : }
1602 :
1603 : /*
1604 : * GetSafeSnapshotBlockingPids
1605 : * If the specified process is currently blocked in GetSafeSnapshot,
1606 : * write the process IDs of all processes that it is blocked by
1607 : * into the caller-supplied buffer output[]. The list is truncated at
1608 : * output_size, and the number of PIDs written into the buffer is
1609 : * returned. Returns zero if the given PID is not currently blocked
1610 : * in GetSafeSnapshot.
1611 : */
1612 : int
1613 904 : GetSafeSnapshotBlockingPids(int blocked_pid, int *output, int output_size)
1614 : {
1615 904 : int num_written = 0;
1616 : dlist_iter iter;
1617 904 : SERIALIZABLEXACT *blocking_sxact = NULL;
1618 :
1619 904 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1620 :
1621 : /* Find blocked_pid's SERIALIZABLEXACT by linear search. */
1622 2462 : dlist_foreach(iter, &PredXact->activeList)
1623 : {
1624 1980 : SERIALIZABLEXACT *sxact =
1625 1980 : dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
1626 :
1627 1980 : if (sxact->pid == blocked_pid)
1628 : {
1629 422 : blocking_sxact = sxact;
1630 422 : break;
1631 : }
1632 : }
1633 :
1634 : /* Did we find it, and is it currently waiting in GetSafeSnapshot? */
1635 904 : if (blocking_sxact != NULL && SxactIsDeferrableWaiting(blocking_sxact))
1636 : {
1637 : /* Traverse the list of possible unsafe conflicts collecting PIDs. */
1638 4 : dlist_foreach(iter, &blocking_sxact->possibleUnsafeConflicts)
1639 : {
1640 4 : RWConflict possibleUnsafeConflict =
1641 4 : dlist_container(RWConflictData, inLink, iter.cur);
1642 :
1643 4 : output[num_written++] = possibleUnsafeConflict->sxactOut->pid;
1644 :
1645 4 : if (num_written >= output_size)
1646 4 : break;
1647 : }
1648 : }
1649 :
1650 904 : LWLockRelease(SerializableXactHashLock);
1651 :
1652 904 : return num_written;
1653 : }
1654 :
1655 : /*
1656 : * Acquire a snapshot that can be used for the current transaction.
1657 : *
1658 : * Make sure we have a SERIALIZABLEXACT reference in MySerializableXact.
1659 : * It should be current for this process and be contained in PredXact.
1660 : *
1661 : * The passed-in Snapshot pointer should reference a static data area that
1662 : * can safely be passed to GetSnapshotData. The return value is actually
1663 : * always this same pointer; no new snapshot data structure is allocated
1664 : * within this function.
1665 : */
1666 : Snapshot
1667 3304 : GetSerializableTransactionSnapshot(Snapshot snapshot)
1668 : {
1669 : Assert(IsolationIsSerializable());
1670 :
1671 : /*
1672 : * Can't use serializable mode while recovery is still active, as it is,
1673 : * for example, on a hot standby. We could get here despite the check in
1674 : * check_transaction_isolation() if default_transaction_isolation is set
1675 : * to serializable, so phrase the hint accordingly.
1676 : */
1677 3304 : if (RecoveryInProgress())
1678 0 : ereport(ERROR,
1679 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1680 : errmsg("cannot use serializable mode in a hot standby"),
1681 : errdetail("default_transaction_isolation is set to \"serializable\"."),
1682 : errhint("You can use \"SET default_transaction_isolation = 'repeatable read'\" to change the default.")));
1683 :
1684 : /*
1685 : * A special optimization is available for SERIALIZABLE READ ONLY
1686 : * DEFERRABLE transactions -- we can wait for a suitable snapshot and
1687 : * thereby avoid all SSI overhead once it's running.
1688 : */
1689 3304 : if (XactReadOnly && XactDeferrable)
1690 8 : return GetSafeSnapshot(snapshot);
1691 :
1692 3296 : return GetSerializableTransactionSnapshotInt(snapshot,
1693 : NULL, InvalidPid);
1694 : }
1695 :
1696 : /*
1697 : * Import a snapshot to be used for the current transaction.
1698 : *
1699 : * This is nearly the same as GetSerializableTransactionSnapshot, except that
1700 : * we don't take a new snapshot, but rather use the data we're handed.
1701 : *
1702 : * The caller must have verified that the snapshot came from a serializable
1703 : * transaction; and if we're read-write, the source transaction must not be
1704 : * read-only.
1705 : */
1706 : void
1707 26 : SetSerializableTransactionSnapshot(Snapshot snapshot,
1708 : VirtualTransactionId *sourcevxid,
1709 : int sourcepid)
1710 : {
1711 : Assert(IsolationIsSerializable());
1712 :
1713 : /*
1714 : * If this is called by parallel.c in a parallel worker, we don't want to
1715 : * create a SERIALIZABLEXACT just yet because the leader's
1716 : * SERIALIZABLEXACT will be installed with AttachSerializableXact(). We
1717 : * also don't want to reject SERIALIZABLE READ ONLY DEFERRABLE in this
1718 : * case, because the leader has already determined that the snapshot it
1719 : * has passed us is safe. So there is nothing for us to do.
1720 : */
1721 26 : if (IsParallelWorker())
1722 26 : return;
1723 :
1724 : /*
1725 : * We do not allow SERIALIZABLE READ ONLY DEFERRABLE transactions to
1726 : * import snapshots, since there's no way to wait for a safe snapshot when
1727 : * we're using the snap we're told to. (XXX instead of throwing an error,
1728 : * we could just ignore the XactDeferrable flag?)
1729 : */
1730 0 : if (XactReadOnly && XactDeferrable)
1731 0 : ereport(ERROR,
1732 : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1733 : errmsg("a snapshot-importing transaction must not be READ ONLY DEFERRABLE")));
1734 :
1735 0 : (void) GetSerializableTransactionSnapshotInt(snapshot, sourcevxid,
1736 : sourcepid);
1737 : }
1738 :
1739 : /*
1740 : * Guts of GetSerializableTransactionSnapshot
1741 : *
1742 : * If sourcevxid is valid, this is actually an import operation and we should
1743 : * skip calling GetSnapshotData, because the snapshot contents are already
1744 : * loaded up. HOWEVER: to avoid race conditions, we must check that the
1745 : * source xact is still running after we acquire SerializableXactHashLock.
1746 : * We do that by calling ProcArrayInstallImportedXmin.
1747 : */
1748 : static Snapshot
1749 3306 : GetSerializableTransactionSnapshotInt(Snapshot snapshot,
1750 : VirtualTransactionId *sourcevxid,
1751 : int sourcepid)
1752 : {
1753 : PGPROC *proc;
1754 : VirtualTransactionId vxid;
1755 : SERIALIZABLEXACT *sxact,
1756 : *othersxact;
1757 :
1758 : /* We only do this for serializable transactions. Once. */
1759 : Assert(MySerializableXact == InvalidSerializableXact);
1760 :
1761 : Assert(!RecoveryInProgress());
1762 :
1763 : /*
1764 : * Since all parts of a serializable transaction must use the same
1765 : * snapshot, it is too late to establish one after a parallel operation
1766 : * has begun.
1767 : */
1768 3306 : if (IsInParallelMode())
1769 0 : elog(ERROR, "cannot establish serializable snapshot during a parallel operation");
1770 :
1771 3306 : proc = MyProc;
1772 : Assert(proc != NULL);
1773 3306 : GET_VXID_FROM_PGPROC(vxid, *proc);
1774 :
1775 : /*
1776 : * First we get the sxact structure, which may involve looping and access
1777 : * to the "finished" list to free a structure for use.
1778 : *
1779 : * We must hold SerializableXactHashLock when taking/checking the snapshot
1780 : * to avoid race conditions, for much the same reasons that
1781 : * GetSnapshotData takes the ProcArrayLock. Since we might have to
1782 : * release SerializableXactHashLock to call SummarizeOldestCommittedSxact,
1783 : * this means we have to create the sxact first, which is a bit annoying
1784 : * (in particular, an elog(ERROR) in procarray.c would cause us to leak
1785 : * the sxact). Consider refactoring to avoid this.
1786 : */
1787 : #ifdef TEST_SUMMARIZE_SERIAL
1788 : SummarizeOldestCommittedSxact();
1789 : #endif
1790 3306 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1791 : do
1792 : {
1793 3306 : sxact = CreatePredXact();
1794 : /* If null, push out committed sxact to SLRU summary & retry. */
1795 3306 : if (!sxact)
1796 : {
1797 0 : LWLockRelease(SerializableXactHashLock);
1798 0 : SummarizeOldestCommittedSxact();
1799 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1800 : }
1801 3306 : } while (!sxact);
1802 :
1803 : /* Get the snapshot, or check that it's safe to use */
1804 3306 : if (!sourcevxid)
1805 3306 : snapshot = GetSnapshotData(snapshot);
1806 0 : else if (!ProcArrayInstallImportedXmin(snapshot->xmin, sourcevxid))
1807 : {
1808 0 : ReleasePredXact(sxact);
1809 0 : LWLockRelease(SerializableXactHashLock);
1810 0 : ereport(ERROR,
1811 : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1812 : errmsg("could not import the requested snapshot"),
1813 : errdetail("The source process with PID %d is not running anymore.",
1814 : sourcepid)));
1815 : }
1816 :
1817 : /*
1818 : * If there are no serializable transactions which are not read-only, we
1819 : * can "opt out" of predicate locking and conflict checking for a
1820 : * read-only transaction.
1821 : *
1822 : * The reason this is safe is that a read-only transaction can only become
1823 : * part of a dangerous structure if it overlaps a writable transaction
1824 : * which in turn overlaps a writable transaction which committed before
1825 : * the read-only transaction started. A new writable transaction can
1826 : * overlap this one, but it can't meet the other condition of overlapping
1827 : * a transaction which committed before this one started.
1828 : */
1829 3306 : if (XactReadOnly && PredXact->WritableSxactCount == 0)
1830 : {
1831 222 : ReleasePredXact(sxact);
1832 222 : LWLockRelease(SerializableXactHashLock);
1833 222 : return snapshot;
1834 : }
1835 :
1836 : /* Initialize the structure. */
1837 3084 : sxact->vxid = vxid;
1838 3084 : sxact->SeqNo.lastCommitBeforeSnapshot = PredXact->LastSxactCommitSeqNo;
1839 3084 : sxact->prepareSeqNo = InvalidSerCommitSeqNo;
1840 3084 : sxact->commitSeqNo = InvalidSerCommitSeqNo;
1841 3084 : dlist_init(&(sxact->outConflicts));
1842 3084 : dlist_init(&(sxact->inConflicts));
1843 3084 : dlist_init(&(sxact->possibleUnsafeConflicts));
1844 3084 : sxact->topXid = GetTopTransactionIdIfAny();
1845 3084 : sxact->finishedBefore = InvalidTransactionId;
1846 3084 : sxact->xmin = snapshot->xmin;
1847 3084 : sxact->pid = MyProcPid;
1848 3084 : sxact->pgprocno = MyProcNumber;
1849 3084 : dlist_init(&sxact->predicateLocks);
1850 3084 : dlist_node_init(&sxact->finishedLink);
1851 3084 : sxact->flags = 0;
1852 3084 : if (XactReadOnly)
1853 : {
1854 : dlist_iter iter;
1855 :
1856 214 : sxact->flags |= SXACT_FLAG_READ_ONLY;
1857 :
1858 : /*
1859 : * Register all concurrent r/w transactions as possible conflicts; if
1860 : * all of them commit without any outgoing conflicts to earlier
1861 : * transactions then this snapshot can be deemed safe (and we can run
1862 : * without tracking predicate locks).
1863 : */
1864 938 : dlist_foreach(iter, &PredXact->activeList)
1865 : {
1866 724 : othersxact = dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
1867 :
1868 724 : if (!SxactIsCommitted(othersxact)
1869 484 : && !SxactIsDoomed(othersxact)
1870 484 : && !SxactIsReadOnly(othersxact))
1871 : {
1872 268 : SetPossibleUnsafeConflict(sxact, othersxact);
1873 : }
1874 : }
1875 :
1876 : /*
1877 : * If we didn't find any possibly unsafe conflicts because every
1878 : * uncommitted writable transaction turned out to be doomed, then we
1879 : * can "opt out" immediately. See comments above the earlier check
1880 : * for PredXact->WritableSxactCount == 0.
1881 : */
1882 214 : if (dlist_is_empty(&sxact->possibleUnsafeConflicts))
1883 : {
1884 0 : ReleasePredXact(sxact);
1885 0 : LWLockRelease(SerializableXactHashLock);
1886 0 : return snapshot;
1887 : }
1888 : }
1889 : else
1890 : {
1891 2870 : ++(PredXact->WritableSxactCount);
1892 : Assert(PredXact->WritableSxactCount <=
1893 : (MaxBackends + max_prepared_xacts));
1894 : }
1895 :
1896 : /* Maintain serializable global xmin info. */
1897 3084 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
1898 : {
1899 : Assert(PredXact->SxactGlobalXminCount == 0);
1900 1710 : PredXact->SxactGlobalXmin = snapshot->xmin;
1901 1710 : PredXact->SxactGlobalXminCount = 1;
1902 1710 : SerialSetActiveSerXmin(snapshot->xmin);
1903 : }
1904 1374 : else if (TransactionIdEquals(snapshot->xmin, PredXact->SxactGlobalXmin))
1905 : {
1906 : Assert(PredXact->SxactGlobalXminCount > 0);
1907 1300 : PredXact->SxactGlobalXminCount++;
1908 : }
1909 : else
1910 : {
1911 : Assert(TransactionIdFollows(snapshot->xmin, PredXact->SxactGlobalXmin));
1912 : }
1913 :
1914 3084 : MySerializableXact = sxact;
1915 3084 : MyXactDidWrite = false; /* haven't written anything yet */
1916 :
1917 3084 : LWLockRelease(SerializableXactHashLock);
1918 :
1919 3084 : CreateLocalPredicateLockHash();
1920 :
1921 3084 : return snapshot;
1922 : }
1923 :
1924 : static void
1925 3110 : CreateLocalPredicateLockHash(void)
1926 : {
1927 : HASHCTL hash_ctl;
1928 :
1929 : /* Initialize the backend-local hash table of parent locks */
1930 : Assert(LocalPredicateLockHash == NULL);
1931 3110 : hash_ctl.keysize = sizeof(PREDICATELOCKTARGETTAG);
1932 3110 : hash_ctl.entrysize = sizeof(LOCALPREDICATELOCK);
1933 3110 : LocalPredicateLockHash = hash_create("Local predicate lock",
1934 : max_predicate_locks_per_xact,
1935 : &hash_ctl,
1936 : HASH_ELEM | HASH_BLOBS);
1937 3110 : }
1938 :
1939 : /*
1940 : * Register the top level XID in SerializableXidHash.
1941 : * Also store it for easy reference in MySerializableXact.
1942 : */
1943 : void
1944 232016 : RegisterPredicateLockingXid(TransactionId xid)
1945 : {
1946 : SERIALIZABLEXIDTAG sxidtag;
1947 : SERIALIZABLEXID *sxid;
1948 : bool found;
1949 :
1950 : /*
1951 : * If we're not tracking predicate lock data for this transaction, we
1952 : * should ignore the request and return quickly.
1953 : */
1954 232016 : if (MySerializableXact == InvalidSerializableXact)
1955 229466 : return;
1956 :
1957 : /* We should have a valid XID and be at the top level. */
1958 : Assert(TransactionIdIsValid(xid));
1959 :
1960 2550 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1961 :
1962 : /* This should only be done once per transaction. */
1963 : Assert(MySerializableXact->topXid == InvalidTransactionId);
1964 :
1965 2550 : MySerializableXact->topXid = xid;
1966 :
1967 2550 : sxidtag.xid = xid;
1968 2550 : sxid = (SERIALIZABLEXID *) hash_search(SerializableXidHash,
1969 : &sxidtag,
1970 : HASH_ENTER, &found);
1971 : Assert(!found);
1972 :
1973 : /* Initialize the structure. */
1974 2550 : sxid->myXact = MySerializableXact;
1975 2550 : LWLockRelease(SerializableXactHashLock);
1976 : }
1977 :
1978 :
1979 : /*
1980 : * Check whether there are any predicate locks held by any transaction
1981 : * for the page at the given block number.
1982 : *
1983 : * Note that the transaction may be completed but not yet subject to
1984 : * cleanup due to overlapping serializable transactions. This must
1985 : * return valid information regardless of transaction isolation level.
1986 : *
1987 : * Also note that this doesn't check for a conflicting relation lock,
1988 : * just a lock specifically on the given page.
1989 : *
1990 : * One use is to support proper behavior during GiST index vacuum.
1991 : */
1992 : bool
1993 0 : PageIsPredicateLocked(Relation relation, BlockNumber blkno)
1994 : {
1995 : PREDICATELOCKTARGETTAG targettag;
1996 : uint32 targettaghash;
1997 : LWLock *partitionLock;
1998 : PREDICATELOCKTARGET *target;
1999 :
2000 0 : SET_PREDICATELOCKTARGETTAG_PAGE(targettag,
2001 : relation->rd_locator.dbOid,
2002 : relation->rd_id,
2003 : blkno);
2004 :
2005 0 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
2006 0 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
2007 0 : LWLockAcquire(partitionLock, LW_SHARED);
2008 : target = (PREDICATELOCKTARGET *)
2009 0 : hash_search_with_hash_value(PredicateLockTargetHash,
2010 : &targettag, targettaghash,
2011 : HASH_FIND, NULL);
2012 0 : LWLockRelease(partitionLock);
2013 :
2014 0 : return (target != NULL);
2015 : }
2016 :
2017 :
2018 : /*
2019 : * Check whether a particular lock is held by this transaction.
2020 : *
2021 : * Important note: this function may return false even if the lock is
2022 : * being held, because it uses the local lock table which is not
2023 : * updated if another transaction modifies our lock list (e.g. to
2024 : * split an index page). It can also return true when a coarser
2025 : * granularity lock that covers this target is being held. Be careful
2026 : * to only use this function in circumstances where such errors are
2027 : * acceptable!
2028 : */
2029 : static bool
2030 154398 : PredicateLockExists(const PREDICATELOCKTARGETTAG *targettag)
2031 : {
2032 : LOCALPREDICATELOCK *lock;
2033 :
2034 : /* check local hash table */
2035 154398 : lock = (LOCALPREDICATELOCK *) hash_search(LocalPredicateLockHash,
2036 : targettag,
2037 : HASH_FIND, NULL);
2038 :
2039 154398 : if (!lock)
2040 60196 : return false;
2041 :
2042 : /*
2043 : * Found entry in the table, but still need to check whether it's actually
2044 : * held -- it could just be a parent of some held lock.
2045 : */
2046 94202 : return lock->held;
2047 : }
2048 :
2049 : /*
2050 : * Return the parent lock tag in the lock hierarchy: the next coarser
2051 : * lock that covers the provided tag.
2052 : *
2053 : * Returns true and sets *parent to the parent tag if one exists,
2054 : * returns false if none exists.
2055 : */
2056 : static bool
2057 90334 : GetParentPredicateLockTag(const PREDICATELOCKTARGETTAG *tag,
2058 : PREDICATELOCKTARGETTAG *parent)
2059 : {
2060 90334 : switch (GET_PREDICATELOCKTARGETTAG_TYPE(*tag))
2061 : {
2062 19636 : case PREDLOCKTAG_RELATION:
2063 : /* relation locks have no parent lock */
2064 19636 : return false;
2065 :
2066 16832 : case PREDLOCKTAG_PAGE:
2067 : /* parent lock is relation lock */
2068 16832 : SET_PREDICATELOCKTARGETTAG_RELATION(*parent,
2069 : GET_PREDICATELOCKTARGETTAG_DB(*tag),
2070 : GET_PREDICATELOCKTARGETTAG_RELATION(*tag));
2071 :
2072 16832 : return true;
2073 :
2074 53866 : case PREDLOCKTAG_TUPLE:
2075 : /* parent lock is page lock */
2076 53866 : SET_PREDICATELOCKTARGETTAG_PAGE(*parent,
2077 : GET_PREDICATELOCKTARGETTAG_DB(*tag),
2078 : GET_PREDICATELOCKTARGETTAG_RELATION(*tag),
2079 : GET_PREDICATELOCKTARGETTAG_PAGE(*tag));
2080 53866 : return true;
2081 : }
2082 :
2083 : /* not reachable */
2084 : Assert(false);
2085 0 : return false;
2086 : }
2087 :
2088 : /*
2089 : * Check whether the lock we are considering is already covered by a
2090 : * coarser lock for our transaction.
2091 : *
2092 : * Like PredicateLockExists, this function might return a false
2093 : * negative, but it will never return a false positive.
2094 : */
2095 : static bool
2096 52080 : CoarserLockCovers(const PREDICATELOCKTARGETTAG *newtargettag)
2097 : {
2098 : PREDICATELOCKTARGETTAG targettag,
2099 : parenttag;
2100 :
2101 52080 : targettag = *newtargettag;
2102 :
2103 : /* check parents iteratively until no more */
2104 62848 : while (GetParentPredicateLockTag(&targettag, &parenttag))
2105 : {
2106 54410 : targettag = parenttag;
2107 54410 : if (PredicateLockExists(&targettag))
2108 43642 : return true;
2109 : }
2110 :
2111 : /* no more parents to check; lock is not covered */
2112 8438 : return false;
2113 : }
2114 :
2115 : /*
2116 : * Remove the dummy entry from the predicate lock target hash, to free up some
2117 : * scratch space. The caller must be holding SerializablePredicateListLock,
2118 : * and must restore the entry with RestoreScratchTarget() before releasing the
2119 : * lock.
2120 : *
2121 : * If lockheld is true, the caller is already holding the partition lock
2122 : * of the partition containing the scratch entry.
2123 : */
2124 : static void
2125 112 : RemoveScratchTarget(bool lockheld)
2126 : {
2127 : bool found;
2128 :
2129 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2130 :
2131 112 : if (!lockheld)
2132 0 : LWLockAcquire(ScratchPartitionLock, LW_EXCLUSIVE);
2133 112 : hash_search_with_hash_value(PredicateLockTargetHash,
2134 : &ScratchTargetTag,
2135 : ScratchTargetTagHash,
2136 : HASH_REMOVE, &found);
2137 : Assert(found);
2138 112 : if (!lockheld)
2139 0 : LWLockRelease(ScratchPartitionLock);
2140 112 : }
2141 :
2142 : /*
2143 : * Re-insert the dummy entry in predicate lock target hash.
2144 : */
2145 : static void
2146 112 : RestoreScratchTarget(bool lockheld)
2147 : {
2148 : bool found;
2149 :
2150 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2151 :
2152 112 : if (!lockheld)
2153 0 : LWLockAcquire(ScratchPartitionLock, LW_EXCLUSIVE);
2154 112 : hash_search_with_hash_value(PredicateLockTargetHash,
2155 : &ScratchTargetTag,
2156 : ScratchTargetTagHash,
2157 : HASH_ENTER, &found);
2158 : Assert(!found);
2159 112 : if (!lockheld)
2160 0 : LWLockRelease(ScratchPartitionLock);
2161 112 : }
2162 :
2163 : /*
2164 : * Check whether the list of related predicate locks is empty for a
2165 : * predicate lock target, and remove the target if it is.
2166 : */
2167 : static void
2168 8426 : RemoveTargetIfNoLongerUsed(PREDICATELOCKTARGET *target, uint32 targettaghash)
2169 : {
2170 : PREDICATELOCKTARGET *rmtarget PG_USED_FOR_ASSERTS_ONLY;
2171 :
2172 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2173 :
2174 : /* Can't remove it until no locks at this target. */
2175 8426 : if (!dlist_is_empty(&target->predicateLocks))
2176 1924 : return;
2177 :
2178 : /* Actually remove the target. */
2179 6502 : rmtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2180 6502 : &target->tag,
2181 : targettaghash,
2182 : HASH_REMOVE, NULL);
2183 : Assert(rmtarget == target);
2184 : }
2185 :
2186 : /*
2187 : * Delete child target locks owned by this process.
2188 : * This implementation is assuming that the usage of each target tag field
2189 : * is uniform. No need to make this hard if we don't have to.
2190 : *
2191 : * We acquire an LWLock in the case of parallel mode, because worker
2192 : * backends have access to the leader's SERIALIZABLEXACT. Otherwise,
2193 : * we aren't acquiring LWLocks for the predicate lock or lock
2194 : * target structures associated with this transaction unless we're going
2195 : * to modify them, because no other process is permitted to modify our
2196 : * locks.
2197 : */
2198 : static void
2199 4706 : DeleteChildTargetLocks(const PREDICATELOCKTARGETTAG *newtargettag)
2200 : {
2201 : SERIALIZABLEXACT *sxact;
2202 : PREDICATELOCK *predlock;
2203 : dlist_mutable_iter iter;
2204 :
2205 4706 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2206 4706 : sxact = MySerializableXact;
2207 4706 : if (IsInParallelMode())
2208 22 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
2209 :
2210 15574 : dlist_foreach_modify(iter, &sxact->predicateLocks)
2211 : {
2212 : PREDICATELOCKTAG oldlocktag;
2213 : PREDICATELOCKTARGET *oldtarget;
2214 : PREDICATELOCKTARGETTAG oldtargettag;
2215 :
2216 10868 : predlock = dlist_container(PREDICATELOCK, xactLink, iter.cur);
2217 :
2218 10868 : oldlocktag = predlock->tag;
2219 : Assert(oldlocktag.myXact == sxact);
2220 10868 : oldtarget = oldlocktag.myTarget;
2221 10868 : oldtargettag = oldtarget->tag;
2222 :
2223 10868 : if (TargetTagIsCoveredBy(oldtargettag, *newtargettag))
2224 : {
2225 : uint32 oldtargettaghash;
2226 : LWLock *partitionLock;
2227 : PREDICATELOCK *rmpredlock PG_USED_FOR_ASSERTS_ONLY;
2228 :
2229 1998 : oldtargettaghash = PredicateLockTargetTagHashCode(&oldtargettag);
2230 1998 : partitionLock = PredicateLockHashPartitionLock(oldtargettaghash);
2231 :
2232 1998 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
2233 :
2234 1998 : dlist_delete(&predlock->xactLink);
2235 1998 : dlist_delete(&predlock->targetLink);
2236 1998 : rmpredlock = hash_search_with_hash_value
2237 : (PredicateLockHash,
2238 : &oldlocktag,
2239 1998 : PredicateLockHashCodeFromTargetHashCode(&oldlocktag,
2240 : oldtargettaghash),
2241 : HASH_REMOVE, NULL);
2242 : Assert(rmpredlock == predlock);
2243 :
2244 1998 : RemoveTargetIfNoLongerUsed(oldtarget, oldtargettaghash);
2245 :
2246 1998 : LWLockRelease(partitionLock);
2247 :
2248 1998 : DecrementParentLocks(&oldtargettag);
2249 : }
2250 : }
2251 4706 : if (IsInParallelMode())
2252 22 : LWLockRelease(&sxact->perXactPredicateListLock);
2253 4706 : LWLockRelease(SerializablePredicateListLock);
2254 4706 : }
2255 :
2256 : /*
2257 : * Returns the promotion limit for a given predicate lock target. This is the
2258 : * max number of descendant locks allowed before promoting to the specified
2259 : * tag. Note that the limit includes non-direct descendants (e.g., both tuples
2260 : * and pages for a relation lock).
2261 : *
2262 : * Currently the default limit is 2 for a page lock, and half of the value of
2263 : * max_pred_locks_per_transaction - 1 for a relation lock, to match behavior
2264 : * of earlier releases when upgrading.
2265 : *
2266 : * TODO SSI: We should probably add additional GUCs to allow a maximum ratio
2267 : * of page and tuple locks based on the pages in a relation, and the maximum
2268 : * ratio of tuple locks to tuples in a page. This would provide more
2269 : * generally "balanced" allocation of locks to where they are most useful,
2270 : * while still allowing the absolute numbers to prevent one relation from
2271 : * tying up all predicate lock resources.
2272 : */
2273 : static int
2274 10768 : MaxPredicateChildLocks(const PREDICATELOCKTARGETTAG *tag)
2275 : {
2276 10768 : switch (GET_PREDICATELOCKTARGETTAG_TYPE(*tag))
2277 : {
2278 7036 : case PREDLOCKTAG_RELATION:
2279 7036 : return max_predicate_locks_per_relation < 0
2280 : ? (max_predicate_locks_per_xact
2281 7036 : / (-max_predicate_locks_per_relation)) - 1
2282 7036 : : max_predicate_locks_per_relation;
2283 :
2284 3732 : case PREDLOCKTAG_PAGE:
2285 3732 : return max_predicate_locks_per_page;
2286 :
2287 0 : case PREDLOCKTAG_TUPLE:
2288 :
2289 : /*
2290 : * not reachable: nothing is finer-granularity than a tuple, so we
2291 : * should never try to promote to it.
2292 : */
2293 : Assert(false);
2294 0 : return 0;
2295 : }
2296 :
2297 : /* not reachable */
2298 : Assert(false);
2299 0 : return 0;
2300 : }
2301 :
2302 : /*
2303 : * For all ancestors of a newly-acquired predicate lock, increment
2304 : * their child count in the parent hash table. If any of them have
2305 : * more descendants than their promotion threshold, acquire the
2306 : * coarsest such lock.
2307 : *
2308 : * Returns true if a parent lock was acquired and false otherwise.
2309 : */
2310 : static bool
2311 8438 : CheckAndPromotePredicateLockRequest(const PREDICATELOCKTARGETTAG *reqtag)
2312 : {
2313 : PREDICATELOCKTARGETTAG targettag,
2314 : nexttag,
2315 : promotiontag;
2316 : LOCALPREDICATELOCK *parentlock;
2317 : bool found,
2318 : promote;
2319 :
2320 8438 : promote = false;
2321 :
2322 8438 : targettag = *reqtag;
2323 :
2324 : /* check parents iteratively */
2325 19206 : while (GetParentPredicateLockTag(&targettag, &nexttag))
2326 : {
2327 10768 : targettag = nexttag;
2328 10768 : parentlock = (LOCALPREDICATELOCK *) hash_search(LocalPredicateLockHash,
2329 : &targettag,
2330 : HASH_ENTER,
2331 : &found);
2332 10768 : if (!found)
2333 : {
2334 6646 : parentlock->held = false;
2335 6646 : parentlock->childLocks = 1;
2336 : }
2337 : else
2338 4122 : parentlock->childLocks++;
2339 :
2340 10768 : if (parentlock->childLocks >
2341 10768 : MaxPredicateChildLocks(&targettag))
2342 : {
2343 : /*
2344 : * We should promote to this parent lock. Continue to check its
2345 : * ancestors, however, both to get their child counts right and to
2346 : * check whether we should just go ahead and promote to one of
2347 : * them.
2348 : */
2349 666 : promotiontag = targettag;
2350 666 : promote = true;
2351 : }
2352 : }
2353 :
2354 8438 : if (promote)
2355 : {
2356 : /* acquire coarsest ancestor eligible for promotion */
2357 666 : PredicateLockAcquire(&promotiontag);
2358 666 : return true;
2359 : }
2360 : else
2361 7772 : return false;
2362 : }
2363 :
2364 : /*
2365 : * When releasing a lock, decrement the child count on all ancestor
2366 : * locks.
2367 : *
2368 : * This is called only when releasing a lock via
2369 : * DeleteChildTargetLocks (i.e. when a lock becomes redundant because
2370 : * we've acquired its parent, possibly due to promotion) or when a new
2371 : * MVCC write lock makes the predicate lock unnecessary. There's no
2372 : * point in calling it when locks are released at transaction end, as
2373 : * this information is no longer needed.
2374 : */
2375 : static void
2376 2760 : DecrementParentLocks(const PREDICATELOCKTARGETTAG *targettag)
2377 : {
2378 : PREDICATELOCKTARGETTAG parenttag,
2379 : nexttag;
2380 :
2381 2760 : parenttag = *targettag;
2382 :
2383 8280 : while (GetParentPredicateLockTag(&parenttag, &nexttag))
2384 : {
2385 : uint32 targettaghash;
2386 : LOCALPREDICATELOCK *parentlock,
2387 : *rmlock PG_USED_FOR_ASSERTS_ONLY;
2388 :
2389 5520 : parenttag = nexttag;
2390 5520 : targettaghash = PredicateLockTargetTagHashCode(&parenttag);
2391 : parentlock = (LOCALPREDICATELOCK *)
2392 5520 : hash_search_with_hash_value(LocalPredicateLockHash,
2393 : &parenttag, targettaghash,
2394 : HASH_FIND, NULL);
2395 :
2396 : /*
2397 : * There's a small chance the parent lock doesn't exist in the lock
2398 : * table. This can happen if we prematurely removed it because an
2399 : * index split caused the child refcount to be off.
2400 : */
2401 5520 : if (parentlock == NULL)
2402 0 : continue;
2403 :
2404 5520 : parentlock->childLocks--;
2405 :
2406 : /*
2407 : * Under similar circumstances the parent lock's refcount might be
2408 : * zero. This only happens if we're holding that lock (otherwise we
2409 : * would have removed the entry).
2410 : */
2411 5520 : if (parentlock->childLocks < 0)
2412 : {
2413 : Assert(parentlock->held);
2414 0 : parentlock->childLocks = 0;
2415 : }
2416 :
2417 5520 : if ((parentlock->childLocks == 0) && (!parentlock->held))
2418 : {
2419 : rmlock = (LOCALPREDICATELOCK *)
2420 1500 : hash_search_with_hash_value(LocalPredicateLockHash,
2421 : &parenttag, targettaghash,
2422 : HASH_REMOVE, NULL);
2423 : Assert(rmlock == parentlock);
2424 : }
2425 : }
2426 2760 : }
2427 :
2428 : /*
2429 : * Indicate that a predicate lock on the given target is held by the
2430 : * specified transaction. Has no effect if the lock is already held.
2431 : *
2432 : * This updates the lock table and the sxact's lock list, and creates
2433 : * the lock target if necessary, but does *not* do anything related to
2434 : * granularity promotion or the local lock table. See
2435 : * PredicateLockAcquire for that.
2436 : */
2437 : static void
2438 8438 : CreatePredicateLock(const PREDICATELOCKTARGETTAG *targettag,
2439 : uint32 targettaghash,
2440 : SERIALIZABLEXACT *sxact)
2441 : {
2442 : PREDICATELOCKTARGET *target;
2443 : PREDICATELOCKTAG locktag;
2444 : PREDICATELOCK *lock;
2445 : LWLock *partitionLock;
2446 : bool found;
2447 :
2448 8438 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
2449 :
2450 8438 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2451 8438 : if (IsInParallelMode())
2452 32 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
2453 8438 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
2454 :
2455 : /* Make sure that the target is represented. */
2456 : target = (PREDICATELOCKTARGET *)
2457 8438 : hash_search_with_hash_value(PredicateLockTargetHash,
2458 : targettag, targettaghash,
2459 : HASH_ENTER_NULL, &found);
2460 8438 : if (!target)
2461 0 : ereport(ERROR,
2462 : (errcode(ERRCODE_OUT_OF_MEMORY),
2463 : errmsg("out of shared memory"),
2464 : errhint("You might need to increase %s.", "max_pred_locks_per_transaction")));
2465 8438 : if (!found)
2466 6502 : dlist_init(&target->predicateLocks);
2467 :
2468 : /* We've got the sxact and target, make sure they're joined. */
2469 8438 : locktag.myTarget = target;
2470 8438 : locktag.myXact = sxact;
2471 : lock = (PREDICATELOCK *)
2472 8438 : hash_search_with_hash_value(PredicateLockHash, &locktag,
2473 8438 : PredicateLockHashCodeFromTargetHashCode(&locktag, targettaghash),
2474 : HASH_ENTER_NULL, &found);
2475 8438 : if (!lock)
2476 0 : ereport(ERROR,
2477 : (errcode(ERRCODE_OUT_OF_MEMORY),
2478 : errmsg("out of shared memory"),
2479 : errhint("You might need to increase %s.", "max_pred_locks_per_transaction")));
2480 :
2481 8438 : if (!found)
2482 : {
2483 8426 : dlist_push_tail(&target->predicateLocks, &lock->targetLink);
2484 8426 : dlist_push_tail(&sxact->predicateLocks, &lock->xactLink);
2485 8426 : lock->commitSeqNo = InvalidSerCommitSeqNo;
2486 : }
2487 :
2488 8438 : LWLockRelease(partitionLock);
2489 8438 : if (IsInParallelMode())
2490 32 : LWLockRelease(&sxact->perXactPredicateListLock);
2491 8438 : LWLockRelease(SerializablePredicateListLock);
2492 8438 : }
2493 :
2494 : /*
2495 : * Acquire a predicate lock on the specified target for the current
2496 : * connection if not already held. This updates the local lock table
2497 : * and uses it to implement granularity promotion. It will consolidate
2498 : * multiple locks into a coarser lock if warranted, and will release
2499 : * any finer-grained locks covered by the new one.
2500 : */
2501 : static void
2502 52480 : PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag)
2503 : {
2504 : uint32 targettaghash;
2505 : bool found;
2506 : LOCALPREDICATELOCK *locallock;
2507 :
2508 : /* Do we have the lock already, or a covering lock? */
2509 52480 : if (PredicateLockExists(targettag))
2510 44042 : return;
2511 :
2512 52080 : if (CoarserLockCovers(targettag))
2513 43642 : return;
2514 :
2515 : /* the same hash and LW lock apply to the lock target and the local lock. */
2516 8438 : targettaghash = PredicateLockTargetTagHashCode(targettag);
2517 :
2518 : /* Acquire lock in local table */
2519 : locallock = (LOCALPREDICATELOCK *)
2520 8438 : hash_search_with_hash_value(LocalPredicateLockHash,
2521 : targettag, targettaghash,
2522 : HASH_ENTER, &found);
2523 8438 : locallock->held = true;
2524 8438 : if (!found)
2525 7772 : locallock->childLocks = 0;
2526 :
2527 : /* Actually create the lock */
2528 8438 : CreatePredicateLock(targettag, targettaghash, MySerializableXact);
2529 :
2530 : /*
2531 : * Lock has been acquired. Check whether it should be promoted to a
2532 : * coarser granularity, or whether there are finer-granularity locks to
2533 : * clean up.
2534 : */
2535 8438 : if (CheckAndPromotePredicateLockRequest(targettag))
2536 : {
2537 : /*
2538 : * Lock request was promoted to a coarser-granularity lock, and that
2539 : * lock was acquired. It will delete this lock and any of its
2540 : * children, so we're done.
2541 : */
2542 : }
2543 : else
2544 : {
2545 : /* Clean up any finer-granularity locks */
2546 7772 : if (GET_PREDICATELOCKTARGETTAG_TYPE(*targettag) != PREDLOCKTAG_TUPLE)
2547 4706 : DeleteChildTargetLocks(targettag);
2548 : }
2549 : }
2550 :
2551 :
2552 : /*
2553 : * PredicateLockRelation
2554 : *
2555 : * Gets a predicate lock at the relation level.
2556 : * Skip if not in full serializable transaction isolation level.
2557 : * Skip if this is a temporary table.
2558 : * Clear any finer-grained predicate locks this session has on the relation.
2559 : */
2560 : void
2561 591704 : PredicateLockRelation(Relation relation, Snapshot snapshot)
2562 : {
2563 : PREDICATELOCKTARGETTAG tag;
2564 :
2565 591704 : if (!SerializationNeededForRead(relation, snapshot))
2566 590260 : return;
2567 :
2568 1444 : SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2569 : relation->rd_locator.dbOid,
2570 : relation->rd_id);
2571 1444 : PredicateLockAcquire(&tag);
2572 : }
2573 :
2574 : /*
2575 : * PredicateLockPage
2576 : *
2577 : * Gets a predicate lock at the page level.
2578 : * Skip if not in full serializable transaction isolation level.
2579 : * Skip if this is a temporary table.
2580 : * Skip if a coarser predicate lock already covers this page.
2581 : * Clear any finer-grained predicate locks this session has on the relation.
2582 : */
2583 : void
2584 14342450 : PredicateLockPage(Relation relation, BlockNumber blkno, Snapshot snapshot)
2585 : {
2586 : PREDICATELOCKTARGETTAG tag;
2587 :
2588 14342450 : if (!SerializationNeededForRead(relation, snapshot))
2589 14339588 : return;
2590 :
2591 2862 : SET_PREDICATELOCKTARGETTAG_PAGE(tag,
2592 : relation->rd_locator.dbOid,
2593 : relation->rd_id,
2594 : blkno);
2595 2862 : PredicateLockAcquire(&tag);
2596 : }
2597 :
2598 : /*
2599 : * PredicateLockTID
2600 : *
2601 : * Gets a predicate lock at the tuple level.
2602 : * Skip if not in full serializable transaction isolation level.
2603 : * Skip if this is a temporary table.
2604 : */
2605 : void
2606 25582438 : PredicateLockTID(Relation relation, ItemPointer tid, Snapshot snapshot,
2607 : TransactionId tuple_xid)
2608 : {
2609 : PREDICATELOCKTARGETTAG tag;
2610 :
2611 25582438 : if (!SerializationNeededForRead(relation, snapshot))
2612 25534930 : return;
2613 :
2614 : /*
2615 : * Return if this xact wrote it.
2616 : */
2617 47508 : if (relation->rd_index == NULL)
2618 : {
2619 : /* If we wrote it; we already have a write lock. */
2620 47508 : if (TransactionIdIsCurrentTransactionId(tuple_xid))
2621 0 : return;
2622 : }
2623 :
2624 : /*
2625 : * Do quick-but-not-definitive test for a relation lock first. This will
2626 : * never cause a return when the relation is *not* locked, but will
2627 : * occasionally let the check continue when there really *is* a relation
2628 : * level lock.
2629 : */
2630 47508 : SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2631 : relation->rd_locator.dbOid,
2632 : relation->rd_id);
2633 47508 : if (PredicateLockExists(&tag))
2634 0 : return;
2635 :
2636 47508 : SET_PREDICATELOCKTARGETTAG_TUPLE(tag,
2637 : relation->rd_locator.dbOid,
2638 : relation->rd_id,
2639 : ItemPointerGetBlockNumber(tid),
2640 : ItemPointerGetOffsetNumber(tid));
2641 47508 : PredicateLockAcquire(&tag);
2642 : }
2643 :
2644 :
2645 : /*
2646 : * DeleteLockTarget
2647 : *
2648 : * Remove a predicate lock target along with any locks held for it.
2649 : *
2650 : * Caller must hold SerializablePredicateListLock and the
2651 : * appropriate hash partition lock for the target.
2652 : */
2653 : static void
2654 0 : DeleteLockTarget(PREDICATELOCKTARGET *target, uint32 targettaghash)
2655 : {
2656 : dlist_mutable_iter iter;
2657 :
2658 : Assert(LWLockHeldByMeInMode(SerializablePredicateListLock,
2659 : LW_EXCLUSIVE));
2660 : Assert(LWLockHeldByMe(PredicateLockHashPartitionLock(targettaghash)));
2661 :
2662 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2663 :
2664 0 : dlist_foreach_modify(iter, &target->predicateLocks)
2665 : {
2666 0 : PREDICATELOCK *predlock =
2667 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
2668 : bool found;
2669 :
2670 0 : dlist_delete(&(predlock->xactLink));
2671 0 : dlist_delete(&(predlock->targetLink));
2672 :
2673 0 : hash_search_with_hash_value
2674 : (PredicateLockHash,
2675 0 : &predlock->tag,
2676 0 : PredicateLockHashCodeFromTargetHashCode(&predlock->tag,
2677 : targettaghash),
2678 : HASH_REMOVE, &found);
2679 : Assert(found);
2680 : }
2681 0 : LWLockRelease(SerializableXactHashLock);
2682 :
2683 : /* Remove the target itself, if possible. */
2684 0 : RemoveTargetIfNoLongerUsed(target, targettaghash);
2685 0 : }
2686 :
2687 :
2688 : /*
2689 : * TransferPredicateLocksToNewTarget
2690 : *
2691 : * Move or copy all the predicate locks for a lock target, for use by
2692 : * index page splits/combines and other things that create or replace
2693 : * lock targets. If 'removeOld' is true, the old locks and the target
2694 : * will be removed.
2695 : *
2696 : * Returns true on success, or false if we ran out of shared memory to
2697 : * allocate the new target or locks. Guaranteed to always succeed if
2698 : * removeOld is set (by using the scratch entry in PredicateLockTargetHash
2699 : * for scratch space).
2700 : *
2701 : * Warning: the "removeOld" option should be used only with care,
2702 : * because this function does not (indeed, can not) update other
2703 : * backends' LocalPredicateLockHash. If we are only adding new
2704 : * entries, this is not a problem: the local lock table is used only
2705 : * as a hint, so missing entries for locks that are held are
2706 : * OK. Having entries for locks that are no longer held, as can happen
2707 : * when using "removeOld", is not in general OK. We can only use it
2708 : * safely when replacing a lock with a coarser-granularity lock that
2709 : * covers it, or if we are absolutely certain that no one will need to
2710 : * refer to that lock in the future.
2711 : *
2712 : * Caller must hold SerializablePredicateListLock exclusively.
2713 : */
2714 : static bool
2715 20 : TransferPredicateLocksToNewTarget(PREDICATELOCKTARGETTAG oldtargettag,
2716 : PREDICATELOCKTARGETTAG newtargettag,
2717 : bool removeOld)
2718 : {
2719 : uint32 oldtargettaghash;
2720 : LWLock *oldpartitionLock;
2721 : PREDICATELOCKTARGET *oldtarget;
2722 : uint32 newtargettaghash;
2723 : LWLock *newpartitionLock;
2724 : bool found;
2725 20 : bool outOfShmem = false;
2726 :
2727 : Assert(LWLockHeldByMeInMode(SerializablePredicateListLock,
2728 : LW_EXCLUSIVE));
2729 :
2730 20 : oldtargettaghash = PredicateLockTargetTagHashCode(&oldtargettag);
2731 20 : newtargettaghash = PredicateLockTargetTagHashCode(&newtargettag);
2732 20 : oldpartitionLock = PredicateLockHashPartitionLock(oldtargettaghash);
2733 20 : newpartitionLock = PredicateLockHashPartitionLock(newtargettaghash);
2734 :
2735 20 : if (removeOld)
2736 : {
2737 : /*
2738 : * Remove the dummy entry to give us scratch space, so we know we'll
2739 : * be able to create the new lock target.
2740 : */
2741 0 : RemoveScratchTarget(false);
2742 : }
2743 :
2744 : /*
2745 : * We must get the partition locks in ascending sequence to avoid
2746 : * deadlocks. If old and new partitions are the same, we must request the
2747 : * lock only once.
2748 : */
2749 20 : if (oldpartitionLock < newpartitionLock)
2750 : {
2751 14 : LWLockAcquire(oldpartitionLock,
2752 14 : (removeOld ? LW_EXCLUSIVE : LW_SHARED));
2753 14 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2754 : }
2755 6 : else if (oldpartitionLock > newpartitionLock)
2756 : {
2757 6 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2758 6 : LWLockAcquire(oldpartitionLock,
2759 6 : (removeOld ? LW_EXCLUSIVE : LW_SHARED));
2760 : }
2761 : else
2762 0 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2763 :
2764 : /*
2765 : * Look for the old target. If not found, that's OK; no predicate locks
2766 : * are affected, so we can just clean up and return. If it does exist,
2767 : * walk its list of predicate locks and move or copy them to the new
2768 : * target.
2769 : */
2770 20 : oldtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2771 : &oldtargettag,
2772 : oldtargettaghash,
2773 : HASH_FIND, NULL);
2774 :
2775 20 : if (oldtarget)
2776 : {
2777 : PREDICATELOCKTARGET *newtarget;
2778 : PREDICATELOCKTAG newpredlocktag;
2779 : dlist_mutable_iter iter;
2780 :
2781 0 : newtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2782 : &newtargettag,
2783 : newtargettaghash,
2784 : HASH_ENTER_NULL, &found);
2785 :
2786 0 : if (!newtarget)
2787 : {
2788 : /* Failed to allocate due to insufficient shmem */
2789 0 : outOfShmem = true;
2790 0 : goto exit;
2791 : }
2792 :
2793 : /* If we created a new entry, initialize it */
2794 0 : if (!found)
2795 0 : dlist_init(&newtarget->predicateLocks);
2796 :
2797 0 : newpredlocktag.myTarget = newtarget;
2798 :
2799 : /*
2800 : * Loop through all the locks on the old target, replacing them with
2801 : * locks on the new target.
2802 : */
2803 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2804 :
2805 0 : dlist_foreach_modify(iter, &oldtarget->predicateLocks)
2806 : {
2807 0 : PREDICATELOCK *oldpredlock =
2808 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
2809 : PREDICATELOCK *newpredlock;
2810 0 : SerCommitSeqNo oldCommitSeqNo = oldpredlock->commitSeqNo;
2811 :
2812 0 : newpredlocktag.myXact = oldpredlock->tag.myXact;
2813 :
2814 0 : if (removeOld)
2815 : {
2816 0 : dlist_delete(&(oldpredlock->xactLink));
2817 0 : dlist_delete(&(oldpredlock->targetLink));
2818 :
2819 0 : hash_search_with_hash_value
2820 : (PredicateLockHash,
2821 0 : &oldpredlock->tag,
2822 0 : PredicateLockHashCodeFromTargetHashCode(&oldpredlock->tag,
2823 : oldtargettaghash),
2824 : HASH_REMOVE, &found);
2825 : Assert(found);
2826 : }
2827 :
2828 : newpredlock = (PREDICATELOCK *)
2829 0 : hash_search_with_hash_value(PredicateLockHash,
2830 : &newpredlocktag,
2831 0 : PredicateLockHashCodeFromTargetHashCode(&newpredlocktag,
2832 : newtargettaghash),
2833 : HASH_ENTER_NULL,
2834 : &found);
2835 0 : if (!newpredlock)
2836 : {
2837 : /* Out of shared memory. Undo what we've done so far. */
2838 0 : LWLockRelease(SerializableXactHashLock);
2839 0 : DeleteLockTarget(newtarget, newtargettaghash);
2840 0 : outOfShmem = true;
2841 0 : goto exit;
2842 : }
2843 0 : if (!found)
2844 : {
2845 0 : dlist_push_tail(&(newtarget->predicateLocks),
2846 : &(newpredlock->targetLink));
2847 0 : dlist_push_tail(&(newpredlocktag.myXact->predicateLocks),
2848 : &(newpredlock->xactLink));
2849 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
2850 : }
2851 : else
2852 : {
2853 0 : if (newpredlock->commitSeqNo < oldCommitSeqNo)
2854 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
2855 : }
2856 :
2857 : Assert(newpredlock->commitSeqNo != 0);
2858 : Assert((newpredlock->commitSeqNo == InvalidSerCommitSeqNo)
2859 : || (newpredlock->tag.myXact == OldCommittedSxact));
2860 : }
2861 0 : LWLockRelease(SerializableXactHashLock);
2862 :
2863 0 : if (removeOld)
2864 : {
2865 : Assert(dlist_is_empty(&oldtarget->predicateLocks));
2866 0 : RemoveTargetIfNoLongerUsed(oldtarget, oldtargettaghash);
2867 : }
2868 : }
2869 :
2870 :
2871 20 : exit:
2872 : /* Release partition locks in reverse order of acquisition. */
2873 20 : if (oldpartitionLock < newpartitionLock)
2874 : {
2875 14 : LWLockRelease(newpartitionLock);
2876 14 : LWLockRelease(oldpartitionLock);
2877 : }
2878 6 : else if (oldpartitionLock > newpartitionLock)
2879 : {
2880 6 : LWLockRelease(oldpartitionLock);
2881 6 : LWLockRelease(newpartitionLock);
2882 : }
2883 : else
2884 0 : LWLockRelease(newpartitionLock);
2885 :
2886 20 : if (removeOld)
2887 : {
2888 : /* We shouldn't run out of memory if we're moving locks */
2889 : Assert(!outOfShmem);
2890 :
2891 : /* Put the scratch entry back */
2892 0 : RestoreScratchTarget(false);
2893 : }
2894 :
2895 20 : return !outOfShmem;
2896 : }
2897 :
2898 : /*
2899 : * Drop all predicate locks of any granularity from the specified relation,
2900 : * which can be a heap relation or an index relation. If 'transfer' is true,
2901 : * acquire a relation lock on the heap for any transactions with any lock(s)
2902 : * on the specified relation.
2903 : *
2904 : * This requires grabbing a lot of LW locks and scanning the entire lock
2905 : * target table for matches. That makes this more expensive than most
2906 : * predicate lock management functions, but it will only be called for DDL
2907 : * type commands that are expensive anyway, and there are fast returns when
2908 : * no serializable transactions are active or the relation is temporary.
2909 : *
2910 : * We don't use the TransferPredicateLocksToNewTarget function because it
2911 : * acquires its own locks on the partitions of the two targets involved,
2912 : * and we'll already be holding all partition locks.
2913 : *
2914 : * We can't throw an error from here, because the call could be from a
2915 : * transaction which is not serializable.
2916 : *
2917 : * NOTE: This is currently only called with transfer set to true, but that may
2918 : * change. If we decide to clean up the locks from a table on commit of a
2919 : * transaction which executed DROP TABLE, the false condition will be useful.
2920 : */
2921 : static void
2922 32214 : DropAllPredicateLocksFromTable(Relation relation, bool transfer)
2923 : {
2924 : HASH_SEQ_STATUS seqstat;
2925 : PREDICATELOCKTARGET *oldtarget;
2926 : PREDICATELOCKTARGET *heaptarget;
2927 : Oid dbId;
2928 : Oid relId;
2929 : Oid heapId;
2930 : int i;
2931 : bool isIndex;
2932 : bool found;
2933 : uint32 heaptargettaghash;
2934 :
2935 : /*
2936 : * Bail out quickly if there are no serializable transactions running.
2937 : * It's safe to check this without taking locks because the caller is
2938 : * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
2939 : * would matter here can be acquired while that is held.
2940 : */
2941 32214 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
2942 32102 : return;
2943 :
2944 168 : if (!PredicateLockingNeededForRelation(relation))
2945 56 : return;
2946 :
2947 112 : dbId = relation->rd_locator.dbOid;
2948 112 : relId = relation->rd_id;
2949 112 : if (relation->rd_index == NULL)
2950 : {
2951 2 : isIndex = false;
2952 2 : heapId = relId;
2953 : }
2954 : else
2955 : {
2956 110 : isIndex = true;
2957 110 : heapId = relation->rd_index->indrelid;
2958 : }
2959 : Assert(heapId != InvalidOid);
2960 : Assert(transfer || !isIndex); /* index OID only makes sense with
2961 : * transfer */
2962 :
2963 : /* Retrieve first time needed, then keep. */
2964 112 : heaptargettaghash = 0;
2965 112 : heaptarget = NULL;
2966 :
2967 : /* Acquire locks on all lock partitions */
2968 112 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
2969 1904 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
2970 1792 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_EXCLUSIVE);
2971 112 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2972 :
2973 : /*
2974 : * Remove the dummy entry to give us scratch space, so we know we'll be
2975 : * able to create the new lock target.
2976 : */
2977 112 : if (transfer)
2978 112 : RemoveScratchTarget(true);
2979 :
2980 : /* Scan through target map */
2981 112 : hash_seq_init(&seqstat, PredicateLockTargetHash);
2982 :
2983 234 : while ((oldtarget = (PREDICATELOCKTARGET *) hash_seq_search(&seqstat)))
2984 : {
2985 : dlist_mutable_iter iter;
2986 :
2987 : /*
2988 : * Check whether this is a target which needs attention.
2989 : */
2990 122 : if (GET_PREDICATELOCKTARGETTAG_RELATION(oldtarget->tag) != relId)
2991 122 : continue; /* wrong relation id */
2992 0 : if (GET_PREDICATELOCKTARGETTAG_DB(oldtarget->tag) != dbId)
2993 0 : continue; /* wrong database id */
2994 0 : if (transfer && !isIndex
2995 0 : && GET_PREDICATELOCKTARGETTAG_TYPE(oldtarget->tag) == PREDLOCKTAG_RELATION)
2996 0 : continue; /* already the right lock */
2997 :
2998 : /*
2999 : * If we made it here, we have work to do. We make sure the heap
3000 : * relation lock exists, then we walk the list of predicate locks for
3001 : * the old target we found, moving all locks to the heap relation lock
3002 : * -- unless they already hold that.
3003 : */
3004 :
3005 : /*
3006 : * First make sure we have the heap relation target. We only need to
3007 : * do this once.
3008 : */
3009 0 : if (transfer && heaptarget == NULL)
3010 : {
3011 : PREDICATELOCKTARGETTAG heaptargettag;
3012 :
3013 0 : SET_PREDICATELOCKTARGETTAG_RELATION(heaptargettag, dbId, heapId);
3014 0 : heaptargettaghash = PredicateLockTargetTagHashCode(&heaptargettag);
3015 0 : heaptarget = hash_search_with_hash_value(PredicateLockTargetHash,
3016 : &heaptargettag,
3017 : heaptargettaghash,
3018 : HASH_ENTER, &found);
3019 0 : if (!found)
3020 0 : dlist_init(&heaptarget->predicateLocks);
3021 : }
3022 :
3023 : /*
3024 : * Loop through all the locks on the old target, replacing them with
3025 : * locks on the new target.
3026 : */
3027 0 : dlist_foreach_modify(iter, &oldtarget->predicateLocks)
3028 : {
3029 0 : PREDICATELOCK *oldpredlock =
3030 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
3031 : PREDICATELOCK *newpredlock;
3032 : SerCommitSeqNo oldCommitSeqNo;
3033 : SERIALIZABLEXACT *oldXact;
3034 :
3035 : /*
3036 : * Remove the old lock first. This avoids the chance of running
3037 : * out of lock structure entries for the hash table.
3038 : */
3039 0 : oldCommitSeqNo = oldpredlock->commitSeqNo;
3040 0 : oldXact = oldpredlock->tag.myXact;
3041 :
3042 0 : dlist_delete(&(oldpredlock->xactLink));
3043 :
3044 : /*
3045 : * No need for retail delete from oldtarget list, we're removing
3046 : * the whole target anyway.
3047 : */
3048 0 : hash_search(PredicateLockHash,
3049 0 : &oldpredlock->tag,
3050 : HASH_REMOVE, &found);
3051 : Assert(found);
3052 :
3053 0 : if (transfer)
3054 : {
3055 : PREDICATELOCKTAG newpredlocktag;
3056 :
3057 0 : newpredlocktag.myTarget = heaptarget;
3058 0 : newpredlocktag.myXact = oldXact;
3059 : newpredlock = (PREDICATELOCK *)
3060 0 : hash_search_with_hash_value(PredicateLockHash,
3061 : &newpredlocktag,
3062 0 : PredicateLockHashCodeFromTargetHashCode(&newpredlocktag,
3063 : heaptargettaghash),
3064 : HASH_ENTER,
3065 : &found);
3066 0 : if (!found)
3067 : {
3068 0 : dlist_push_tail(&(heaptarget->predicateLocks),
3069 : &(newpredlock->targetLink));
3070 0 : dlist_push_tail(&(newpredlocktag.myXact->predicateLocks),
3071 : &(newpredlock->xactLink));
3072 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
3073 : }
3074 : else
3075 : {
3076 0 : if (newpredlock->commitSeqNo < oldCommitSeqNo)
3077 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
3078 : }
3079 :
3080 : Assert(newpredlock->commitSeqNo != 0);
3081 : Assert((newpredlock->commitSeqNo == InvalidSerCommitSeqNo)
3082 : || (newpredlock->tag.myXact == OldCommittedSxact));
3083 : }
3084 : }
3085 :
3086 0 : hash_search(PredicateLockTargetHash, &oldtarget->tag, HASH_REMOVE,
3087 : &found);
3088 : Assert(found);
3089 : }
3090 :
3091 : /* Put the scratch entry back */
3092 112 : if (transfer)
3093 112 : RestoreScratchTarget(true);
3094 :
3095 : /* Release locks in reverse order */
3096 112 : LWLockRelease(SerializableXactHashLock);
3097 1904 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
3098 1792 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
3099 112 : LWLockRelease(SerializablePredicateListLock);
3100 : }
3101 :
3102 : /*
3103 : * TransferPredicateLocksToHeapRelation
3104 : * For all transactions, transfer all predicate locks for the given
3105 : * relation to a single relation lock on the heap.
3106 : */
3107 : void
3108 32214 : TransferPredicateLocksToHeapRelation(Relation relation)
3109 : {
3110 32214 : DropAllPredicateLocksFromTable(relation, true);
3111 32214 : }
3112 :
3113 :
3114 : /*
3115 : * PredicateLockPageSplit
3116 : *
3117 : * Copies any predicate locks for the old page to the new page.
3118 : * Skip if this is a temporary table or toast table.
3119 : *
3120 : * NOTE: A page split (or overflow) affects all serializable transactions,
3121 : * even if it occurs in the context of another transaction isolation level.
3122 : *
3123 : * NOTE: This currently leaves the local copy of the locks without
3124 : * information on the new lock which is in shared memory. This could cause
3125 : * problems if enough page splits occur on locked pages without the processes
3126 : * which hold the locks getting in and noticing.
3127 : */
3128 : void
3129 55304 : PredicateLockPageSplit(Relation relation, BlockNumber oldblkno,
3130 : BlockNumber newblkno)
3131 : {
3132 : PREDICATELOCKTARGETTAG oldtargettag;
3133 : PREDICATELOCKTARGETTAG newtargettag;
3134 : bool success;
3135 :
3136 : /*
3137 : * Bail out quickly if there are no serializable transactions running.
3138 : *
3139 : * It's safe to do this check without taking any additional locks. Even if
3140 : * a serializable transaction starts concurrently, we know it can't take
3141 : * any SIREAD locks on the page being split because the caller is holding
3142 : * the associated buffer page lock. Memory reordering isn't an issue; the
3143 : * memory barrier in the LWLock acquisition guarantees that this read
3144 : * occurs while the buffer page lock is held.
3145 : */
3146 55304 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
3147 55284 : return;
3148 :
3149 62 : if (!PredicateLockingNeededForRelation(relation))
3150 42 : return;
3151 :
3152 : Assert(oldblkno != newblkno);
3153 : Assert(BlockNumberIsValid(oldblkno));
3154 : Assert(BlockNumberIsValid(newblkno));
3155 :
3156 20 : SET_PREDICATELOCKTARGETTAG_PAGE(oldtargettag,
3157 : relation->rd_locator.dbOid,
3158 : relation->rd_id,
3159 : oldblkno);
3160 20 : SET_PREDICATELOCKTARGETTAG_PAGE(newtargettag,
3161 : relation->rd_locator.dbOid,
3162 : relation->rd_id,
3163 : newblkno);
3164 :
3165 20 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
3166 :
3167 : /*
3168 : * Try copying the locks over to the new page's tag, creating it if
3169 : * necessary.
3170 : */
3171 20 : success = TransferPredicateLocksToNewTarget(oldtargettag,
3172 : newtargettag,
3173 : false);
3174 :
3175 20 : if (!success)
3176 : {
3177 : /*
3178 : * No more predicate lock entries are available. Failure isn't an
3179 : * option here, so promote the page lock to a relation lock.
3180 : */
3181 :
3182 : /* Get the parent relation lock's lock tag */
3183 0 : success = GetParentPredicateLockTag(&oldtargettag,
3184 : &newtargettag);
3185 : Assert(success);
3186 :
3187 : /*
3188 : * Move the locks to the parent. This shouldn't fail.
3189 : *
3190 : * Note that here we are removing locks held by other backends,
3191 : * leading to a possible inconsistency in their local lock hash table.
3192 : * This is OK because we're replacing it with a lock that covers the
3193 : * old one.
3194 : */
3195 0 : success = TransferPredicateLocksToNewTarget(oldtargettag,
3196 : newtargettag,
3197 : true);
3198 : Assert(success);
3199 : }
3200 :
3201 20 : LWLockRelease(SerializablePredicateListLock);
3202 : }
3203 :
3204 : /*
3205 : * PredicateLockPageCombine
3206 : *
3207 : * Combines predicate locks for two existing pages.
3208 : * Skip if this is a temporary table or toast table.
3209 : *
3210 : * NOTE: A page combine affects all serializable transactions, even if it
3211 : * occurs in the context of another transaction isolation level.
3212 : */
3213 : void
3214 4700 : PredicateLockPageCombine(Relation relation, BlockNumber oldblkno,
3215 : BlockNumber newblkno)
3216 : {
3217 : /*
3218 : * Page combines differ from page splits in that we ought to be able to
3219 : * remove the locks on the old page after transferring them to the new
3220 : * page, instead of duplicating them. However, because we can't edit other
3221 : * backends' local lock tables, removing the old lock would leave them
3222 : * with an entry in their LocalPredicateLockHash for a lock they're not
3223 : * holding, which isn't acceptable. So we wind up having to do the same
3224 : * work as a page split, acquiring a lock on the new page and keeping the
3225 : * old page locked too. That can lead to some false positives, but should
3226 : * be rare in practice.
3227 : */
3228 4700 : PredicateLockPageSplit(relation, oldblkno, newblkno);
3229 4700 : }
3230 :
3231 : /*
3232 : * Walk the list of in-progress serializable transactions and find the new
3233 : * xmin.
3234 : */
3235 : static void
3236 1748 : SetNewSxactGlobalXmin(void)
3237 : {
3238 : dlist_iter iter;
3239 :
3240 : Assert(LWLockHeldByMe(SerializableXactHashLock));
3241 :
3242 1748 : PredXact->SxactGlobalXmin = InvalidTransactionId;
3243 1748 : PredXact->SxactGlobalXminCount = 0;
3244 :
3245 6616 : dlist_foreach(iter, &PredXact->activeList)
3246 : {
3247 4868 : SERIALIZABLEXACT *sxact =
3248 4868 : dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
3249 :
3250 4868 : if (!SxactIsRolledBack(sxact)
3251 4272 : && !SxactIsCommitted(sxact)
3252 40 : && sxact != OldCommittedSxact)
3253 : {
3254 : Assert(sxact->xmin != InvalidTransactionId);
3255 40 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)
3256 2 : || TransactionIdPrecedes(sxact->xmin,
3257 2 : PredXact->SxactGlobalXmin))
3258 : {
3259 38 : PredXact->SxactGlobalXmin = sxact->xmin;
3260 38 : PredXact->SxactGlobalXminCount = 1;
3261 : }
3262 2 : else if (TransactionIdEquals(sxact->xmin,
3263 : PredXact->SxactGlobalXmin))
3264 0 : PredXact->SxactGlobalXminCount++;
3265 : }
3266 : }
3267 :
3268 1748 : SerialSetActiveSerXmin(PredXact->SxactGlobalXmin);
3269 1748 : }
3270 :
3271 : /*
3272 : * ReleasePredicateLocks
3273 : *
3274 : * Releases predicate locks based on completion of the current transaction,
3275 : * whether committed or rolled back. It can also be called for a read only
3276 : * transaction when it becomes impossible for the transaction to become
3277 : * part of a dangerous structure.
3278 : *
3279 : * We do nothing unless this is a serializable transaction.
3280 : *
3281 : * This method must ensure that shared memory hash tables are cleaned
3282 : * up in some relatively timely fashion.
3283 : *
3284 : * If this transaction is committing and is holding any predicate locks,
3285 : * it must be added to a list of completed serializable transactions still
3286 : * holding locks.
3287 : *
3288 : * If isReadOnlySafe is true, then predicate locks are being released before
3289 : * the end of the transaction because MySerializableXact has been determined
3290 : * to be RO_SAFE. In non-parallel mode we can release it completely, but it
3291 : * in parallel mode we partially release the SERIALIZABLEXACT and keep it
3292 : * around until the end of the transaction, allowing each backend to clear its
3293 : * MySerializableXact variable and benefit from the optimization in its own
3294 : * time.
3295 : */
3296 : void
3297 565928 : ReleasePredicateLocks(bool isCommit, bool isReadOnlySafe)
3298 : {
3299 565928 : bool partiallyReleasing = false;
3300 : bool needToClear;
3301 : SERIALIZABLEXACT *roXact;
3302 : dlist_mutable_iter iter;
3303 :
3304 : /*
3305 : * We can't trust XactReadOnly here, because a transaction which started
3306 : * as READ WRITE can show as READ ONLY later, e.g., within
3307 : * subtransactions. We want to flag a transaction as READ ONLY if it
3308 : * commits without writing so that de facto READ ONLY transactions get the
3309 : * benefit of some RO optimizations, so we will use this local variable to
3310 : * get some cleanup logic right which is based on whether the transaction
3311 : * was declared READ ONLY at the top level.
3312 : */
3313 : bool topLevelIsDeclaredReadOnly;
3314 :
3315 : /* We can't be both committing and releasing early due to RO_SAFE. */
3316 : Assert(!(isCommit && isReadOnlySafe));
3317 :
3318 : /* Are we at the end of a transaction, that is, a commit or abort? */
3319 565928 : if (!isReadOnlySafe)
3320 : {
3321 : /*
3322 : * Parallel workers mustn't release predicate locks at the end of
3323 : * their transaction. The leader will do that at the end of its
3324 : * transaction.
3325 : */
3326 565860 : if (IsParallelWorker())
3327 : {
3328 7926 : ReleasePredicateLocksLocal();
3329 562840 : return;
3330 : }
3331 :
3332 : /*
3333 : * By the time the leader in a parallel query reaches end of
3334 : * transaction, it has waited for all workers to exit.
3335 : */
3336 : Assert(!ParallelContextActive());
3337 :
3338 : /*
3339 : * If the leader in a parallel query earlier stashed a partially
3340 : * released SERIALIZABLEXACT for final clean-up at end of transaction
3341 : * (because workers might still have been accessing it), then it's
3342 : * time to restore it.
3343 : */
3344 557934 : if (SavedSerializableXact != InvalidSerializableXact)
3345 : {
3346 : Assert(MySerializableXact == InvalidSerializableXact);
3347 2 : MySerializableXact = SavedSerializableXact;
3348 2 : SavedSerializableXact = InvalidSerializableXact;
3349 : Assert(SxactIsPartiallyReleased(MySerializableXact));
3350 : }
3351 : }
3352 :
3353 558002 : if (MySerializableXact == InvalidSerializableXact)
3354 : {
3355 : Assert(LocalPredicateLockHash == NULL);
3356 554908 : return;
3357 : }
3358 :
3359 3094 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
3360 :
3361 : /*
3362 : * If the transaction is committing, but it has been partially released
3363 : * already, then treat this as a roll back. It was marked as rolled back.
3364 : */
3365 3094 : if (isCommit && SxactIsPartiallyReleased(MySerializableXact))
3366 4 : isCommit = false;
3367 :
3368 : /*
3369 : * If we're called in the middle of a transaction because we discovered
3370 : * that the SXACT_FLAG_RO_SAFE flag was set, then we'll partially release
3371 : * it (that is, release the predicate locks and conflicts, but not the
3372 : * SERIALIZABLEXACT itself) if we're the first backend to have noticed.
3373 : */
3374 3094 : if (isReadOnlySafe && IsInParallelMode())
3375 : {
3376 : /*
3377 : * The leader needs to stash a pointer to it, so that it can
3378 : * completely release it at end-of-transaction.
3379 : */
3380 10 : if (!IsParallelWorker())
3381 2 : SavedSerializableXact = MySerializableXact;
3382 :
3383 : /*
3384 : * The first backend to reach this condition will partially release
3385 : * the SERIALIZABLEXACT. All others will just clear their
3386 : * backend-local state so that they stop doing SSI checks for the rest
3387 : * of the transaction.
3388 : */
3389 10 : if (SxactIsPartiallyReleased(MySerializableXact))
3390 : {
3391 6 : LWLockRelease(SerializableXactHashLock);
3392 6 : ReleasePredicateLocksLocal();
3393 6 : return;
3394 : }
3395 : else
3396 : {
3397 4 : MySerializableXact->flags |= SXACT_FLAG_PARTIALLY_RELEASED;
3398 4 : partiallyReleasing = true;
3399 : /* ... and proceed to perform the partial release below. */
3400 : }
3401 : }
3402 : Assert(!isCommit || SxactIsPrepared(MySerializableXact));
3403 : Assert(!isCommit || !SxactIsDoomed(MySerializableXact));
3404 : Assert(!SxactIsCommitted(MySerializableXact));
3405 : Assert(SxactIsPartiallyReleased(MySerializableXact)
3406 : || !SxactIsRolledBack(MySerializableXact));
3407 :
3408 : /* may not be serializable during COMMIT/ROLLBACK PREPARED */
3409 : Assert(MySerializableXact->pid == 0 || IsolationIsSerializable());
3410 :
3411 : /* We'd better not already be on the cleanup list. */
3412 : Assert(!SxactIsOnFinishedList(MySerializableXact));
3413 :
3414 3088 : topLevelIsDeclaredReadOnly = SxactIsReadOnly(MySerializableXact);
3415 :
3416 : /*
3417 : * We don't hold XidGenLock lock here, assuming that TransactionId is
3418 : * atomic!
3419 : *
3420 : * If this value is changing, we don't care that much whether we get the
3421 : * old or new value -- it is just used to determine how far
3422 : * SxactGlobalXmin must advance before this transaction can be fully
3423 : * cleaned up. The worst that could happen is we wait for one more
3424 : * transaction to complete before freeing some RAM; correctness of visible
3425 : * behavior is not affected.
3426 : */
3427 3088 : MySerializableXact->finishedBefore = XidFromFullTransactionId(TransamVariables->nextXid);
3428 :
3429 : /*
3430 : * If it's not a commit it's either a rollback or a read-only transaction
3431 : * flagged SXACT_FLAG_RO_SAFE, and we can clear our locks immediately.
3432 : */
3433 3088 : if (isCommit)
3434 : {
3435 2444 : MySerializableXact->flags |= SXACT_FLAG_COMMITTED;
3436 2444 : MySerializableXact->commitSeqNo = ++(PredXact->LastSxactCommitSeqNo);
3437 : /* Recognize implicit read-only transaction (commit without write). */
3438 2444 : if (!MyXactDidWrite)
3439 466 : MySerializableXact->flags |= SXACT_FLAG_READ_ONLY;
3440 : }
3441 : else
3442 : {
3443 : /*
3444 : * The DOOMED flag indicates that we intend to roll back this
3445 : * transaction and so it should not cause serialization failures for
3446 : * other transactions that conflict with it. Note that this flag might
3447 : * already be set, if another backend marked this transaction for
3448 : * abort.
3449 : *
3450 : * The ROLLED_BACK flag further indicates that ReleasePredicateLocks
3451 : * has been called, and so the SerializableXact is eligible for
3452 : * cleanup. This means it should not be considered when calculating
3453 : * SxactGlobalXmin.
3454 : */
3455 644 : MySerializableXact->flags |= SXACT_FLAG_DOOMED;
3456 644 : MySerializableXact->flags |= SXACT_FLAG_ROLLED_BACK;
3457 :
3458 : /*
3459 : * If the transaction was previously prepared, but is now failing due
3460 : * to a ROLLBACK PREPARED or (hopefully very rare) error after the
3461 : * prepare, clear the prepared flag. This simplifies conflict
3462 : * checking.
3463 : */
3464 644 : MySerializableXact->flags &= ~SXACT_FLAG_PREPARED;
3465 : }
3466 :
3467 3088 : if (!topLevelIsDeclaredReadOnly)
3468 : {
3469 : Assert(PredXact->WritableSxactCount > 0);
3470 2870 : if (--(PredXact->WritableSxactCount) == 0)
3471 : {
3472 : /*
3473 : * Release predicate locks and rw-conflicts in for all committed
3474 : * transactions. There are no longer any transactions which might
3475 : * conflict with the locks and no chance for new transactions to
3476 : * overlap. Similarly, existing conflicts in can't cause pivots,
3477 : * and any conflicts in which could have completed a dangerous
3478 : * structure would already have caused a rollback, so any
3479 : * remaining ones must be benign.
3480 : */
3481 1730 : PredXact->CanPartialClearThrough = PredXact->LastSxactCommitSeqNo;
3482 : }
3483 : }
3484 : else
3485 : {
3486 : /*
3487 : * Read-only transactions: clear the list of transactions that might
3488 : * make us unsafe. Note that we use 'inLink' for the iteration as
3489 : * opposed to 'outLink' for the r/w xacts.
3490 : */
3491 302 : dlist_foreach_modify(iter, &MySerializableXact->possibleUnsafeConflicts)
3492 : {
3493 84 : RWConflict possibleUnsafeConflict =
3494 84 : dlist_container(RWConflictData, inLink, iter.cur);
3495 :
3496 : Assert(!SxactIsReadOnly(possibleUnsafeConflict->sxactOut));
3497 : Assert(MySerializableXact == possibleUnsafeConflict->sxactIn);
3498 :
3499 84 : ReleaseRWConflict(possibleUnsafeConflict);
3500 : }
3501 : }
3502 :
3503 : /* Check for conflict out to old committed transactions. */
3504 3088 : if (isCommit
3505 2444 : && !SxactIsReadOnly(MySerializableXact)
3506 1978 : && SxactHasSummaryConflictOut(MySerializableXact))
3507 : {
3508 : /*
3509 : * we don't know which old committed transaction we conflicted with,
3510 : * so be conservative and use FirstNormalSerCommitSeqNo here
3511 : */
3512 0 : MySerializableXact->SeqNo.earliestOutConflictCommit =
3513 : FirstNormalSerCommitSeqNo;
3514 0 : MySerializableXact->flags |= SXACT_FLAG_CONFLICT_OUT;
3515 : }
3516 :
3517 : /*
3518 : * Release all outConflicts to committed transactions. If we're rolling
3519 : * back clear them all. Set SXACT_FLAG_CONFLICT_OUT if any point to
3520 : * previously committed transactions.
3521 : */
3522 4446 : dlist_foreach_modify(iter, &MySerializableXact->outConflicts)
3523 : {
3524 1358 : RWConflict conflict =
3525 1358 : dlist_container(RWConflictData, outLink, iter.cur);
3526 :
3527 1358 : if (isCommit
3528 902 : && !SxactIsReadOnly(MySerializableXact)
3529 686 : && SxactIsCommitted(conflict->sxactIn))
3530 : {
3531 192 : if ((MySerializableXact->flags & SXACT_FLAG_CONFLICT_OUT) == 0
3532 0 : || conflict->sxactIn->prepareSeqNo < MySerializableXact->SeqNo.earliestOutConflictCommit)
3533 192 : MySerializableXact->SeqNo.earliestOutConflictCommit = conflict->sxactIn->prepareSeqNo;
3534 192 : MySerializableXact->flags |= SXACT_FLAG_CONFLICT_OUT;
3535 : }
3536 :
3537 1358 : if (!isCommit
3538 902 : || SxactIsCommitted(conflict->sxactIn)
3539 666 : || (conflict->sxactIn->SeqNo.lastCommitBeforeSnapshot >= PredXact->LastSxactCommitSeqNo))
3540 692 : ReleaseRWConflict(conflict);
3541 : }
3542 :
3543 : /*
3544 : * Release all inConflicts from committed and read-only transactions. If
3545 : * we're rolling back, clear them all.
3546 : */
3547 4630 : dlist_foreach_modify(iter, &MySerializableXact->inConflicts)
3548 : {
3549 1542 : RWConflict conflict =
3550 1542 : dlist_container(RWConflictData, inLink, iter.cur);
3551 :
3552 1542 : if (!isCommit
3553 1198 : || SxactIsCommitted(conflict->sxactOut)
3554 830 : || SxactIsReadOnly(conflict->sxactOut))
3555 872 : ReleaseRWConflict(conflict);
3556 : }
3557 :
3558 3088 : if (!topLevelIsDeclaredReadOnly)
3559 : {
3560 : /*
3561 : * Remove ourselves from the list of possible conflicts for concurrent
3562 : * READ ONLY transactions, flagging them as unsafe if we have a
3563 : * conflict out. If any are waiting DEFERRABLE transactions, wake them
3564 : * up if they are known safe or known unsafe.
3565 : */
3566 3054 : dlist_foreach_modify(iter, &MySerializableXact->possibleUnsafeConflicts)
3567 : {
3568 184 : RWConflict possibleUnsafeConflict =
3569 184 : dlist_container(RWConflictData, outLink, iter.cur);
3570 :
3571 184 : roXact = possibleUnsafeConflict->sxactIn;
3572 : Assert(MySerializableXact == possibleUnsafeConflict->sxactOut);
3573 : Assert(SxactIsReadOnly(roXact));
3574 :
3575 : /* Mark conflicted if necessary. */
3576 184 : if (isCommit
3577 180 : && MyXactDidWrite
3578 170 : && SxactHasConflictOut(MySerializableXact)
3579 26 : && (MySerializableXact->SeqNo.earliestOutConflictCommit
3580 26 : <= roXact->SeqNo.lastCommitBeforeSnapshot))
3581 : {
3582 : /*
3583 : * This releases possibleUnsafeConflict (as well as all other
3584 : * possible conflicts for roXact)
3585 : */
3586 6 : FlagSxactUnsafe(roXact);
3587 : }
3588 : else
3589 : {
3590 178 : ReleaseRWConflict(possibleUnsafeConflict);
3591 :
3592 : /*
3593 : * If we were the last possible conflict, flag it safe. The
3594 : * transaction can now safely release its predicate locks (but
3595 : * that transaction's backend has to do that itself).
3596 : */
3597 178 : if (dlist_is_empty(&roXact->possibleUnsafeConflicts))
3598 132 : roXact->flags |= SXACT_FLAG_RO_SAFE;
3599 : }
3600 :
3601 : /*
3602 : * Wake up the process for a waiting DEFERRABLE transaction if we
3603 : * now know it's either safe or conflicted.
3604 : */
3605 184 : if (SxactIsDeferrableWaiting(roXact) &&
3606 6 : (SxactIsROUnsafe(roXact) || SxactIsROSafe(roXact)))
3607 4 : ProcSendSignal(roXact->pgprocno);
3608 : }
3609 : }
3610 :
3611 : /*
3612 : * Check whether it's time to clean up old transactions. This can only be
3613 : * done when the last serializable transaction with the oldest xmin among
3614 : * serializable transactions completes. We then find the "new oldest"
3615 : * xmin and purge any transactions which finished before this transaction
3616 : * was launched.
3617 : *
3618 : * For parallel queries in read-only transactions, it might run twice. We
3619 : * only release the reference on the first call.
3620 : */
3621 3088 : needToClear = false;
3622 3088 : if ((partiallyReleasing ||
3623 3084 : !SxactIsPartiallyReleased(MySerializableXact)) &&
3624 3084 : TransactionIdEquals(MySerializableXact->xmin,
3625 : PredXact->SxactGlobalXmin))
3626 : {
3627 : Assert(PredXact->SxactGlobalXminCount > 0);
3628 3048 : if (--(PredXact->SxactGlobalXminCount) == 0)
3629 : {
3630 1748 : SetNewSxactGlobalXmin();
3631 1748 : needToClear = true;
3632 : }
3633 : }
3634 :
3635 3088 : LWLockRelease(SerializableXactHashLock);
3636 :
3637 3088 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
3638 :
3639 : /* Add this to the list of transactions to check for later cleanup. */
3640 3088 : if (isCommit)
3641 2444 : dlist_push_tail(FinishedSerializableTransactions,
3642 2444 : &MySerializableXact->finishedLink);
3643 :
3644 : /*
3645 : * If we're releasing a RO_SAFE transaction in parallel mode, we'll only
3646 : * partially release it. That's necessary because other backends may have
3647 : * a reference to it. The leader will release the SERIALIZABLEXACT itself
3648 : * at the end of the transaction after workers have stopped running.
3649 : */
3650 3088 : if (!isCommit)
3651 644 : ReleaseOneSerializableXact(MySerializableXact,
3652 644 : isReadOnlySafe && IsInParallelMode(),
3653 : false);
3654 :
3655 3088 : LWLockRelease(SerializableFinishedListLock);
3656 :
3657 3088 : if (needToClear)
3658 1748 : ClearOldPredicateLocks();
3659 :
3660 3088 : ReleasePredicateLocksLocal();
3661 : }
3662 :
3663 : static void
3664 11020 : ReleasePredicateLocksLocal(void)
3665 : {
3666 11020 : MySerializableXact = InvalidSerializableXact;
3667 11020 : MyXactDidWrite = false;
3668 :
3669 : /* Delete per-transaction lock table */
3670 11020 : if (LocalPredicateLockHash != NULL)
3671 : {
3672 3086 : hash_destroy(LocalPredicateLockHash);
3673 3086 : LocalPredicateLockHash = NULL;
3674 : }
3675 11020 : }
3676 :
3677 : /*
3678 : * Clear old predicate locks, belonging to committed transactions that are no
3679 : * longer interesting to any in-progress transaction.
3680 : */
3681 : static void
3682 1748 : ClearOldPredicateLocks(void)
3683 : {
3684 : dlist_mutable_iter iter;
3685 :
3686 : /*
3687 : * Loop through finished transactions. They are in commit order, so we can
3688 : * stop as soon as we find one that's still interesting.
3689 : */
3690 1748 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
3691 1748 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3692 4212 : dlist_foreach_modify(iter, FinishedSerializableTransactions)
3693 : {
3694 2482 : SERIALIZABLEXACT *finishedSxact =
3695 2482 : dlist_container(SERIALIZABLEXACT, finishedLink, iter.cur);
3696 :
3697 2482 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)
3698 60 : || TransactionIdPrecedesOrEquals(finishedSxact->finishedBefore,
3699 60 : PredXact->SxactGlobalXmin))
3700 : {
3701 : /*
3702 : * This transaction committed before any in-progress transaction
3703 : * took its snapshot. It's no longer interesting.
3704 : */
3705 2444 : LWLockRelease(SerializableXactHashLock);
3706 2444 : dlist_delete_thoroughly(&finishedSxact->finishedLink);
3707 2444 : ReleaseOneSerializableXact(finishedSxact, false, false);
3708 2444 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3709 : }
3710 38 : else if (finishedSxact->commitSeqNo > PredXact->HavePartialClearedThrough
3711 38 : && finishedSxact->commitSeqNo <= PredXact->CanPartialClearThrough)
3712 : {
3713 : /*
3714 : * Any active transactions that took their snapshot before this
3715 : * transaction committed are read-only, so we can clear part of
3716 : * its state.
3717 : */
3718 20 : LWLockRelease(SerializableXactHashLock);
3719 :
3720 20 : if (SxactIsReadOnly(finishedSxact))
3721 : {
3722 : /* A read-only transaction can be removed entirely */
3723 0 : dlist_delete_thoroughly(&(finishedSxact->finishedLink));
3724 0 : ReleaseOneSerializableXact(finishedSxact, false, false);
3725 : }
3726 : else
3727 : {
3728 : /*
3729 : * A read-write transaction can only be partially cleared. We
3730 : * need to keep the SERIALIZABLEXACT but can release the
3731 : * SIREAD locks and conflicts in.
3732 : */
3733 20 : ReleaseOneSerializableXact(finishedSxact, true, false);
3734 : }
3735 :
3736 20 : PredXact->HavePartialClearedThrough = finishedSxact->commitSeqNo;
3737 20 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3738 : }
3739 : else
3740 : {
3741 : /* Still interesting. */
3742 : break;
3743 : }
3744 : }
3745 1748 : LWLockRelease(SerializableXactHashLock);
3746 :
3747 : /*
3748 : * Loop through predicate locks on dummy transaction for summarized data.
3749 : */
3750 1748 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
3751 1748 : dlist_foreach_modify(iter, &OldCommittedSxact->predicateLocks)
3752 : {
3753 0 : PREDICATELOCK *predlock =
3754 0 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
3755 : bool canDoPartialCleanup;
3756 :
3757 0 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3758 : Assert(predlock->commitSeqNo != 0);
3759 : Assert(predlock->commitSeqNo != InvalidSerCommitSeqNo);
3760 0 : canDoPartialCleanup = (predlock->commitSeqNo <= PredXact->CanPartialClearThrough);
3761 0 : LWLockRelease(SerializableXactHashLock);
3762 :
3763 : /*
3764 : * If this lock originally belonged to an old enough transaction, we
3765 : * can release it.
3766 : */
3767 0 : if (canDoPartialCleanup)
3768 : {
3769 : PREDICATELOCKTAG tag;
3770 : PREDICATELOCKTARGET *target;
3771 : PREDICATELOCKTARGETTAG targettag;
3772 : uint32 targettaghash;
3773 : LWLock *partitionLock;
3774 :
3775 0 : tag = predlock->tag;
3776 0 : target = tag.myTarget;
3777 0 : targettag = target->tag;
3778 0 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
3779 0 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
3780 :
3781 0 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
3782 :
3783 0 : dlist_delete(&(predlock->targetLink));
3784 0 : dlist_delete(&(predlock->xactLink));
3785 :
3786 0 : hash_search_with_hash_value(PredicateLockHash, &tag,
3787 0 : PredicateLockHashCodeFromTargetHashCode(&tag,
3788 : targettaghash),
3789 : HASH_REMOVE, NULL);
3790 0 : RemoveTargetIfNoLongerUsed(target, targettaghash);
3791 :
3792 0 : LWLockRelease(partitionLock);
3793 : }
3794 : }
3795 :
3796 1748 : LWLockRelease(SerializablePredicateListLock);
3797 1748 : LWLockRelease(SerializableFinishedListLock);
3798 1748 : }
3799 :
3800 : /*
3801 : * This is the normal way to delete anything from any of the predicate
3802 : * locking hash tables. Given a transaction which we know can be deleted:
3803 : * delete all predicate locks held by that transaction and any predicate
3804 : * lock targets which are now unreferenced by a lock; delete all conflicts
3805 : * for the transaction; delete all xid values for the transaction; then
3806 : * delete the transaction.
3807 : *
3808 : * When the partial flag is set, we can release all predicate locks and
3809 : * in-conflict information -- we've established that there are no longer
3810 : * any overlapping read write transactions for which this transaction could
3811 : * matter -- but keep the transaction entry itself and any outConflicts.
3812 : *
3813 : * When the summarize flag is set, we've run short of room for sxact data
3814 : * and must summarize to the SLRU. Predicate locks are transferred to a
3815 : * dummy "old" transaction, with duplicate locks on a single target
3816 : * collapsing to a single lock with the "latest" commitSeqNo from among
3817 : * the conflicting locks..
3818 : */
3819 : static void
3820 3108 : ReleaseOneSerializableXact(SERIALIZABLEXACT *sxact, bool partial,
3821 : bool summarize)
3822 : {
3823 : SERIALIZABLEXIDTAG sxidtag;
3824 : dlist_mutable_iter iter;
3825 :
3826 : Assert(sxact != NULL);
3827 : Assert(SxactIsRolledBack(sxact) || SxactIsCommitted(sxact));
3828 : Assert(partial || !SxactIsOnFinishedList(sxact));
3829 : Assert(LWLockHeldByMe(SerializableFinishedListLock));
3830 :
3831 : /*
3832 : * First release all the predicate locks held by this xact (or transfer
3833 : * them to OldCommittedSxact if summarize is true)
3834 : */
3835 3108 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
3836 3108 : if (IsInParallelMode())
3837 6 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
3838 8774 : dlist_foreach_modify(iter, &sxact->predicateLocks)
3839 : {
3840 5666 : PREDICATELOCK *predlock =
3841 5666 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
3842 : PREDICATELOCKTAG tag;
3843 : PREDICATELOCKTARGET *target;
3844 : PREDICATELOCKTARGETTAG targettag;
3845 : uint32 targettaghash;
3846 : LWLock *partitionLock;
3847 :
3848 5666 : tag = predlock->tag;
3849 5666 : target = tag.myTarget;
3850 5666 : targettag = target->tag;
3851 5666 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
3852 5666 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
3853 :
3854 5666 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
3855 :
3856 5666 : dlist_delete(&predlock->targetLink);
3857 :
3858 5666 : hash_search_with_hash_value(PredicateLockHash, &tag,
3859 5666 : PredicateLockHashCodeFromTargetHashCode(&tag,
3860 : targettaghash),
3861 : HASH_REMOVE, NULL);
3862 5666 : if (summarize)
3863 : {
3864 : bool found;
3865 :
3866 : /* Fold into dummy transaction list. */
3867 0 : tag.myXact = OldCommittedSxact;
3868 0 : predlock = hash_search_with_hash_value(PredicateLockHash, &tag,
3869 0 : PredicateLockHashCodeFromTargetHashCode(&tag,
3870 : targettaghash),
3871 : HASH_ENTER_NULL, &found);
3872 0 : if (!predlock)
3873 0 : ereport(ERROR,
3874 : (errcode(ERRCODE_OUT_OF_MEMORY),
3875 : errmsg("out of shared memory"),
3876 : errhint("You might need to increase %s.", "max_pred_locks_per_transaction")));
3877 0 : if (found)
3878 : {
3879 : Assert(predlock->commitSeqNo != 0);
3880 : Assert(predlock->commitSeqNo != InvalidSerCommitSeqNo);
3881 0 : if (predlock->commitSeqNo < sxact->commitSeqNo)
3882 0 : predlock->commitSeqNo = sxact->commitSeqNo;
3883 : }
3884 : else
3885 : {
3886 0 : dlist_push_tail(&target->predicateLocks,
3887 : &predlock->targetLink);
3888 0 : dlist_push_tail(&OldCommittedSxact->predicateLocks,
3889 : &predlock->xactLink);
3890 0 : predlock->commitSeqNo = sxact->commitSeqNo;
3891 : }
3892 : }
3893 : else
3894 5666 : RemoveTargetIfNoLongerUsed(target, targettaghash);
3895 :
3896 5666 : LWLockRelease(partitionLock);
3897 : }
3898 :
3899 : /*
3900 : * Rather than retail removal, just re-init the head after we've run
3901 : * through the list.
3902 : */
3903 3108 : dlist_init(&sxact->predicateLocks);
3904 :
3905 3108 : if (IsInParallelMode())
3906 6 : LWLockRelease(&sxact->perXactPredicateListLock);
3907 3108 : LWLockRelease(SerializablePredicateListLock);
3908 :
3909 3108 : sxidtag.xid = sxact->topXid;
3910 3108 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
3911 :
3912 : /* Release all outConflicts (unless 'partial' is true) */
3913 3108 : if (!partial)
3914 : {
3915 3084 : dlist_foreach_modify(iter, &sxact->outConflicts)
3916 : {
3917 0 : RWConflict conflict =
3918 0 : dlist_container(RWConflictData, outLink, iter.cur);
3919 :
3920 0 : if (summarize)
3921 0 : conflict->sxactIn->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
3922 0 : ReleaseRWConflict(conflict);
3923 : }
3924 : }
3925 :
3926 : /* Release all inConflicts. */
3927 3108 : dlist_foreach_modify(iter, &sxact->inConflicts)
3928 : {
3929 0 : RWConflict conflict =
3930 0 : dlist_container(RWConflictData, inLink, iter.cur);
3931 :
3932 0 : if (summarize)
3933 0 : conflict->sxactOut->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
3934 0 : ReleaseRWConflict(conflict);
3935 : }
3936 :
3937 : /* Finally, get rid of the xid and the record of the transaction itself. */
3938 3108 : if (!partial)
3939 : {
3940 3084 : if (sxidtag.xid != InvalidTransactionId)
3941 2550 : hash_search(SerializableXidHash, &sxidtag, HASH_REMOVE, NULL);
3942 3084 : ReleasePredXact(sxact);
3943 : }
3944 :
3945 3108 : LWLockRelease(SerializableXactHashLock);
3946 3108 : }
3947 :
3948 : /*
3949 : * Tests whether the given top level transaction is concurrent with
3950 : * (overlaps) our current transaction.
3951 : *
3952 : * We need to identify the top level transaction for SSI, anyway, so pass
3953 : * that to this function to save the overhead of checking the snapshot's
3954 : * subxip array.
3955 : */
3956 : static bool
3957 1066 : XidIsConcurrent(TransactionId xid)
3958 : {
3959 : Snapshot snap;
3960 :
3961 : Assert(TransactionIdIsValid(xid));
3962 : Assert(!TransactionIdEquals(xid, GetTopTransactionIdIfAny()));
3963 :
3964 1066 : snap = GetTransactionSnapshot();
3965 :
3966 1066 : if (TransactionIdPrecedes(xid, snap->xmin))
3967 0 : return false;
3968 :
3969 1066 : if (TransactionIdFollowsOrEquals(xid, snap->xmax))
3970 1040 : return true;
3971 :
3972 26 : return pg_lfind32(xid, snap->xip, snap->xcnt);
3973 : }
3974 :
3975 : bool
3976 58101384 : CheckForSerializableConflictOutNeeded(Relation relation, Snapshot snapshot)
3977 : {
3978 58101384 : if (!SerializationNeededForRead(relation, snapshot))
3979 58049434 : return false;
3980 :
3981 : /* Check if someone else has already decided that we need to die */
3982 51950 : if (SxactIsDoomed(MySerializableXact))
3983 : {
3984 0 : ereport(ERROR,
3985 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
3986 : errmsg("could not serialize access due to read/write dependencies among transactions"),
3987 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
3988 : errhint("The transaction might succeed if retried.")));
3989 : }
3990 :
3991 51950 : return true;
3992 : }
3993 :
3994 : /*
3995 : * CheckForSerializableConflictOut
3996 : * A table AM is reading a tuple that has been modified. If it determines
3997 : * that the tuple version it is reading is not visible to us, it should
3998 : * pass in the top level xid of the transaction that created it.
3999 : * Otherwise, if it determines that it is visible to us but it has been
4000 : * deleted or there is a newer version available due to an update, it
4001 : * should pass in the top level xid of the modifying transaction.
4002 : *
4003 : * This function will check for overlap with our own transaction. If the given
4004 : * xid is also serializable and the transactions overlap (i.e., they cannot see
4005 : * each other's writes), then we have a conflict out.
4006 : */
4007 : void
4008 1134 : CheckForSerializableConflictOut(Relation relation, TransactionId xid, Snapshot snapshot)
4009 : {
4010 : SERIALIZABLEXIDTAG sxidtag;
4011 : SERIALIZABLEXID *sxid;
4012 : SERIALIZABLEXACT *sxact;
4013 :
4014 1134 : if (!SerializationNeededForRead(relation, snapshot))
4015 408 : return;
4016 :
4017 : /* Check if someone else has already decided that we need to die */
4018 1134 : if (SxactIsDoomed(MySerializableXact))
4019 : {
4020 0 : ereport(ERROR,
4021 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4022 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4023 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4024 : errhint("The transaction might succeed if retried.")));
4025 : }
4026 : Assert(TransactionIdIsValid(xid));
4027 :
4028 1134 : if (TransactionIdEquals(xid, GetTopTransactionIdIfAny()))
4029 0 : return;
4030 :
4031 : /*
4032 : * Find sxact or summarized info for the top level xid.
4033 : */
4034 1134 : sxidtag.xid = xid;
4035 1134 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4036 : sxid = (SERIALIZABLEXID *)
4037 1134 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
4038 1134 : if (!sxid)
4039 : {
4040 : /*
4041 : * Transaction not found in "normal" SSI structures. Check whether it
4042 : * got pushed out to SLRU storage for "old committed" transactions.
4043 : */
4044 : SerCommitSeqNo conflictCommitSeqNo;
4045 :
4046 48 : conflictCommitSeqNo = SerialGetMinConflictCommitSeqNo(xid);
4047 48 : if (conflictCommitSeqNo != 0)
4048 : {
4049 0 : if (conflictCommitSeqNo != InvalidSerCommitSeqNo
4050 0 : && (!SxactIsReadOnly(MySerializableXact)
4051 0 : || conflictCommitSeqNo
4052 0 : <= MySerializableXact->SeqNo.lastCommitBeforeSnapshot))
4053 0 : ereport(ERROR,
4054 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4055 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4056 : errdetail_internal("Reason code: Canceled on conflict out to old pivot %u.", xid),
4057 : errhint("The transaction might succeed if retried.")));
4058 :
4059 0 : if (SxactHasSummaryConflictIn(MySerializableXact)
4060 0 : || !dlist_is_empty(&MySerializableXact->inConflicts))
4061 0 : ereport(ERROR,
4062 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4063 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4064 : errdetail_internal("Reason code: Canceled on identification as a pivot, with conflict out to old committed transaction %u.", xid),
4065 : errhint("The transaction might succeed if retried.")));
4066 :
4067 0 : MySerializableXact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4068 : }
4069 :
4070 : /* It's not serializable or otherwise not important. */
4071 48 : LWLockRelease(SerializableXactHashLock);
4072 48 : return;
4073 : }
4074 1086 : sxact = sxid->myXact;
4075 : Assert(TransactionIdEquals(sxact->topXid, xid));
4076 1086 : if (sxact == MySerializableXact || SxactIsDoomed(sxact))
4077 : {
4078 : /* Can't conflict with ourself or a transaction that will roll back. */
4079 8 : LWLockRelease(SerializableXactHashLock);
4080 8 : return;
4081 : }
4082 :
4083 : /*
4084 : * We have a conflict out to a transaction which has a conflict out to a
4085 : * summarized transaction. That summarized transaction must have
4086 : * committed first, and we can't tell when it committed in relation to our
4087 : * snapshot acquisition, so something needs to be canceled.
4088 : */
4089 1078 : if (SxactHasSummaryConflictOut(sxact))
4090 : {
4091 0 : if (!SxactIsPrepared(sxact))
4092 : {
4093 0 : sxact->flags |= SXACT_FLAG_DOOMED;
4094 0 : LWLockRelease(SerializableXactHashLock);
4095 0 : return;
4096 : }
4097 : else
4098 : {
4099 0 : LWLockRelease(SerializableXactHashLock);
4100 0 : ereport(ERROR,
4101 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4102 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4103 : errdetail_internal("Reason code: Canceled on conflict out to old pivot."),
4104 : errhint("The transaction might succeed if retried.")));
4105 : }
4106 : }
4107 :
4108 : /*
4109 : * If this is a read-only transaction and the writing transaction has
4110 : * committed, and it doesn't have a rw-conflict to a transaction which
4111 : * committed before it, no conflict.
4112 : */
4113 1078 : if (SxactIsReadOnly(MySerializableXact)
4114 238 : && SxactIsCommitted(sxact)
4115 16 : && !SxactHasSummaryConflictOut(sxact)
4116 16 : && (!SxactHasConflictOut(sxact)
4117 4 : || MySerializableXact->SeqNo.lastCommitBeforeSnapshot < sxact->SeqNo.earliestOutConflictCommit))
4118 : {
4119 : /* Read-only transaction will appear to run first. No conflict. */
4120 12 : LWLockRelease(SerializableXactHashLock);
4121 12 : return;
4122 : }
4123 :
4124 1066 : if (!XidIsConcurrent(xid))
4125 : {
4126 : /* This write was already in our snapshot; no conflict. */
4127 2 : LWLockRelease(SerializableXactHashLock);
4128 2 : return;
4129 : }
4130 :
4131 1064 : if (RWConflictExists(MySerializableXact, sxact))
4132 : {
4133 : /* We don't want duplicate conflict records in the list. */
4134 338 : LWLockRelease(SerializableXactHashLock);
4135 338 : return;
4136 : }
4137 :
4138 : /*
4139 : * Flag the conflict. But first, if this conflict creates a dangerous
4140 : * structure, ereport an error.
4141 : */
4142 726 : FlagRWConflict(MySerializableXact, sxact);
4143 700 : LWLockRelease(SerializableXactHashLock);
4144 : }
4145 :
4146 : /*
4147 : * Check a particular target for rw-dependency conflict in. A subroutine of
4148 : * CheckForSerializableConflictIn().
4149 : */
4150 : static void
4151 14992 : CheckTargetForConflictsIn(PREDICATELOCKTARGETTAG *targettag)
4152 : {
4153 : uint32 targettaghash;
4154 : LWLock *partitionLock;
4155 : PREDICATELOCKTARGET *target;
4156 14992 : PREDICATELOCK *mypredlock = NULL;
4157 : PREDICATELOCKTAG mypredlocktag;
4158 : dlist_mutable_iter iter;
4159 :
4160 : Assert(MySerializableXact != InvalidSerializableXact);
4161 :
4162 : /*
4163 : * The same hash and LW lock apply to the lock target and the lock itself.
4164 : */
4165 14992 : targettaghash = PredicateLockTargetTagHashCode(targettag);
4166 14992 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
4167 14992 : LWLockAcquire(partitionLock, LW_SHARED);
4168 : target = (PREDICATELOCKTARGET *)
4169 14992 : hash_search_with_hash_value(PredicateLockTargetHash,
4170 : targettag, targettaghash,
4171 : HASH_FIND, NULL);
4172 14992 : if (!target)
4173 : {
4174 : /* Nothing has this target locked; we're done here. */
4175 11250 : LWLockRelease(partitionLock);
4176 11250 : return;
4177 : }
4178 :
4179 : /*
4180 : * Each lock for an overlapping transaction represents a conflict: a
4181 : * rw-dependency in to this transaction.
4182 : */
4183 3742 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4184 :
4185 8436 : dlist_foreach_modify(iter, &target->predicateLocks)
4186 : {
4187 4828 : PREDICATELOCK *predlock =
4188 4828 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
4189 4828 : SERIALIZABLEXACT *sxact = predlock->tag.myXact;
4190 :
4191 4828 : if (sxact == MySerializableXact)
4192 : {
4193 : /*
4194 : * If we're getting a write lock on a tuple, we don't need a
4195 : * predicate (SIREAD) lock on the same tuple. We can safely remove
4196 : * our SIREAD lock, but we'll defer doing so until after the loop
4197 : * because that requires upgrading to an exclusive partition lock.
4198 : *
4199 : * We can't use this optimization within a subtransaction because
4200 : * the subtransaction could roll back, and we would be left
4201 : * without any lock at the top level.
4202 : */
4203 3128 : if (!IsSubTransaction()
4204 3128 : && GET_PREDICATELOCKTARGETTAG_OFFSET(*targettag))
4205 : {
4206 776 : mypredlock = predlock;
4207 776 : mypredlocktag = predlock->tag;
4208 : }
4209 : }
4210 1700 : else if (!SxactIsDoomed(sxact)
4211 1700 : && (!SxactIsCommitted(sxact)
4212 170 : || TransactionIdPrecedes(GetTransactionSnapshot()->xmin,
4213 : sxact->finishedBefore))
4214 1682 : && !RWConflictExists(sxact, MySerializableXact))
4215 : {
4216 998 : LWLockRelease(SerializableXactHashLock);
4217 998 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4218 :
4219 : /*
4220 : * Re-check after getting exclusive lock because the other
4221 : * transaction may have flagged a conflict.
4222 : */
4223 998 : if (!SxactIsDoomed(sxact)
4224 998 : && (!SxactIsCommitted(sxact)
4225 152 : || TransactionIdPrecedes(GetTransactionSnapshot()->xmin,
4226 : sxact->finishedBefore))
4227 998 : && !RWConflictExists(sxact, MySerializableXact))
4228 : {
4229 998 : FlagRWConflict(sxact, MySerializableXact);
4230 : }
4231 :
4232 864 : LWLockRelease(SerializableXactHashLock);
4233 864 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4234 : }
4235 : }
4236 3608 : LWLockRelease(SerializableXactHashLock);
4237 3608 : LWLockRelease(partitionLock);
4238 :
4239 : /*
4240 : * If we found one of our own SIREAD locks to remove, remove it now.
4241 : *
4242 : * At this point our transaction already has a RowExclusiveLock on the
4243 : * relation, so we are OK to drop the predicate lock on the tuple, if
4244 : * found, without fearing that another write against the tuple will occur
4245 : * before the MVCC information makes it to the buffer.
4246 : */
4247 3608 : if (mypredlock != NULL)
4248 : {
4249 : uint32 predlockhashcode;
4250 : PREDICATELOCK *rmpredlock;
4251 :
4252 762 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
4253 762 : if (IsInParallelMode())
4254 0 : LWLockAcquire(&MySerializableXact->perXactPredicateListLock, LW_EXCLUSIVE);
4255 762 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
4256 762 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4257 :
4258 : /*
4259 : * Remove the predicate lock from shared memory, if it wasn't removed
4260 : * while the locks were released. One way that could happen is from
4261 : * autovacuum cleaning up an index.
4262 : */
4263 762 : predlockhashcode = PredicateLockHashCodeFromTargetHashCode
4264 : (&mypredlocktag, targettaghash);
4265 : rmpredlock = (PREDICATELOCK *)
4266 762 : hash_search_with_hash_value(PredicateLockHash,
4267 : &mypredlocktag,
4268 : predlockhashcode,
4269 : HASH_FIND, NULL);
4270 762 : if (rmpredlock != NULL)
4271 : {
4272 : Assert(rmpredlock == mypredlock);
4273 :
4274 762 : dlist_delete(&(mypredlock->targetLink));
4275 762 : dlist_delete(&(mypredlock->xactLink));
4276 :
4277 : rmpredlock = (PREDICATELOCK *)
4278 762 : hash_search_with_hash_value(PredicateLockHash,
4279 : &mypredlocktag,
4280 : predlockhashcode,
4281 : HASH_REMOVE, NULL);
4282 : Assert(rmpredlock == mypredlock);
4283 :
4284 762 : RemoveTargetIfNoLongerUsed(target, targettaghash);
4285 : }
4286 :
4287 762 : LWLockRelease(SerializableXactHashLock);
4288 762 : LWLockRelease(partitionLock);
4289 762 : if (IsInParallelMode())
4290 0 : LWLockRelease(&MySerializableXact->perXactPredicateListLock);
4291 762 : LWLockRelease(SerializablePredicateListLock);
4292 :
4293 762 : if (rmpredlock != NULL)
4294 : {
4295 : /*
4296 : * Remove entry in local lock table if it exists. It's OK if it
4297 : * doesn't exist; that means the lock was transferred to a new
4298 : * target by a different backend.
4299 : */
4300 762 : hash_search_with_hash_value(LocalPredicateLockHash,
4301 : targettag, targettaghash,
4302 : HASH_REMOVE, NULL);
4303 :
4304 762 : DecrementParentLocks(targettag);
4305 : }
4306 : }
4307 : }
4308 :
4309 : /*
4310 : * CheckForSerializableConflictIn
4311 : * We are writing the given tuple. If that indicates a rw-conflict
4312 : * in from another serializable transaction, take appropriate action.
4313 : *
4314 : * Skip checking for any granularity for which a parameter is missing.
4315 : *
4316 : * A tuple update or delete is in conflict if we have a predicate lock
4317 : * against the relation or page in which the tuple exists, or against the
4318 : * tuple itself.
4319 : */
4320 : void
4321 29686558 : CheckForSerializableConflictIn(Relation relation, ItemPointer tid, BlockNumber blkno)
4322 : {
4323 : PREDICATELOCKTARGETTAG targettag;
4324 :
4325 29686558 : if (!SerializationNeededForWrite(relation))
4326 29677672 : return;
4327 :
4328 : /* Check if someone else has already decided that we need to die */
4329 8886 : if (SxactIsDoomed(MySerializableXact))
4330 2 : ereport(ERROR,
4331 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4332 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4333 : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict in checking."),
4334 : errhint("The transaction might succeed if retried.")));
4335 :
4336 : /*
4337 : * We're doing a write which might cause rw-conflicts now or later.
4338 : * Memorize that fact.
4339 : */
4340 8884 : MyXactDidWrite = true;
4341 :
4342 : /*
4343 : * It is important that we check for locks from the finest granularity to
4344 : * the coarsest granularity, so that granularity promotion doesn't cause
4345 : * us to miss a lock. The new (coarser) lock will be acquired before the
4346 : * old (finer) locks are released.
4347 : *
4348 : * It is not possible to take and hold a lock across the checks for all
4349 : * granularities because each target could be in a separate partition.
4350 : */
4351 8884 : if (tid != NULL)
4352 : {
4353 1286 : SET_PREDICATELOCKTARGETTAG_TUPLE(targettag,
4354 : relation->rd_locator.dbOid,
4355 : relation->rd_id,
4356 : ItemPointerGetBlockNumber(tid),
4357 : ItemPointerGetOffsetNumber(tid));
4358 1286 : CheckTargetForConflictsIn(&targettag);
4359 : }
4360 :
4361 8838 : if (blkno != InvalidBlockNumber)
4362 : {
4363 4928 : SET_PREDICATELOCKTARGETTAG_PAGE(targettag,
4364 : relation->rd_locator.dbOid,
4365 : relation->rd_id,
4366 : blkno);
4367 4928 : CheckTargetForConflictsIn(&targettag);
4368 : }
4369 :
4370 8778 : SET_PREDICATELOCKTARGETTAG_RELATION(targettag,
4371 : relation->rd_locator.dbOid,
4372 : relation->rd_id);
4373 8778 : CheckTargetForConflictsIn(&targettag);
4374 : }
4375 :
4376 : /*
4377 : * CheckTableForSerializableConflictIn
4378 : * The entire table is going through a DDL-style logical mass delete
4379 : * like TRUNCATE or DROP TABLE. If that causes a rw-conflict in from
4380 : * another serializable transaction, take appropriate action.
4381 : *
4382 : * While these operations do not operate entirely within the bounds of
4383 : * snapshot isolation, they can occur inside a serializable transaction, and
4384 : * will logically occur after any reads which saw rows which were destroyed
4385 : * by these operations, so we do what we can to serialize properly under
4386 : * SSI.
4387 : *
4388 : * The relation passed in must be a heap relation. Any predicate lock of any
4389 : * granularity on the heap will cause a rw-conflict in to this transaction.
4390 : * Predicate locks on indexes do not matter because they only exist to guard
4391 : * against conflicting inserts into the index, and this is a mass *delete*.
4392 : * When a table is truncated or dropped, the index will also be truncated
4393 : * or dropped, and we'll deal with locks on the index when that happens.
4394 : *
4395 : * Dropping or truncating a table also needs to drop any existing predicate
4396 : * locks on heap tuples or pages, because they're about to go away. This
4397 : * should be done before altering the predicate locks because the transaction
4398 : * could be rolled back because of a conflict, in which case the lock changes
4399 : * are not needed. (At the moment, we don't actually bother to drop the
4400 : * existing locks on a dropped or truncated table at the moment. That might
4401 : * lead to some false positives, but it doesn't seem worth the trouble.)
4402 : */
4403 : void
4404 44930 : CheckTableForSerializableConflictIn(Relation relation)
4405 : {
4406 : HASH_SEQ_STATUS seqstat;
4407 : PREDICATELOCKTARGET *target;
4408 : Oid dbId;
4409 : Oid heapId;
4410 : int i;
4411 :
4412 : /*
4413 : * Bail out quickly if there are no serializable transactions running.
4414 : * It's safe to check this without taking locks because the caller is
4415 : * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
4416 : * would matter here can be acquired while that is held.
4417 : */
4418 44930 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
4419 44888 : return;
4420 :
4421 314 : if (!SerializationNeededForWrite(relation))
4422 272 : return;
4423 :
4424 : /*
4425 : * We're doing a write which might cause rw-conflicts now or later.
4426 : * Memorize that fact.
4427 : */
4428 42 : MyXactDidWrite = true;
4429 :
4430 : Assert(relation->rd_index == NULL); /* not an index relation */
4431 :
4432 42 : dbId = relation->rd_locator.dbOid;
4433 42 : heapId = relation->rd_id;
4434 :
4435 42 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
4436 714 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
4437 672 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_SHARED);
4438 42 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4439 :
4440 : /* Scan through target list */
4441 42 : hash_seq_init(&seqstat, PredicateLockTargetHash);
4442 :
4443 144 : while ((target = (PREDICATELOCKTARGET *) hash_seq_search(&seqstat)))
4444 : {
4445 : dlist_mutable_iter iter;
4446 :
4447 : /*
4448 : * Check whether this is a target which needs attention.
4449 : */
4450 102 : if (GET_PREDICATELOCKTARGETTAG_RELATION(target->tag) != heapId)
4451 84 : continue; /* wrong relation id */
4452 18 : if (GET_PREDICATELOCKTARGETTAG_DB(target->tag) != dbId)
4453 0 : continue; /* wrong database id */
4454 :
4455 : /*
4456 : * Loop through locks for this target and flag conflicts.
4457 : */
4458 36 : dlist_foreach_modify(iter, &target->predicateLocks)
4459 : {
4460 18 : PREDICATELOCK *predlock =
4461 18 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
4462 :
4463 18 : if (predlock->tag.myXact != MySerializableXact
4464 0 : && !RWConflictExists(predlock->tag.myXact, MySerializableXact))
4465 : {
4466 0 : FlagRWConflict(predlock->tag.myXact, MySerializableXact);
4467 : }
4468 : }
4469 : }
4470 :
4471 : /* Release locks in reverse order */
4472 42 : LWLockRelease(SerializableXactHashLock);
4473 714 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
4474 672 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
4475 42 : LWLockRelease(SerializablePredicateListLock);
4476 : }
4477 :
4478 :
4479 : /*
4480 : * Flag a rw-dependency between two serializable transactions.
4481 : *
4482 : * The caller is responsible for ensuring that we have a LW lock on
4483 : * the transaction hash table.
4484 : */
4485 : static void
4486 1724 : FlagRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
4487 : {
4488 : Assert(reader != writer);
4489 :
4490 : /* First, see if this conflict causes failure. */
4491 1724 : OnConflict_CheckForSerializationFailure(reader, writer);
4492 :
4493 : /* Actually do the conflict flagging. */
4494 1564 : if (reader == OldCommittedSxact)
4495 0 : writer->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
4496 1564 : else if (writer == OldCommittedSxact)
4497 0 : reader->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4498 : else
4499 1564 : SetRWConflict(reader, writer);
4500 1564 : }
4501 :
4502 : /*----------------------------------------------------------------------------
4503 : * We are about to add a RW-edge to the dependency graph - check that we don't
4504 : * introduce a dangerous structure by doing so, and abort one of the
4505 : * transactions if so.
4506 : *
4507 : * A serialization failure can only occur if there is a dangerous structure
4508 : * in the dependency graph:
4509 : *
4510 : * Tin ------> Tpivot ------> Tout
4511 : * rw rw
4512 : *
4513 : * Furthermore, Tout must commit first.
4514 : *
4515 : * One more optimization is that if Tin is declared READ ONLY (or commits
4516 : * without writing), we can only have a problem if Tout committed before Tin
4517 : * acquired its snapshot.
4518 : *----------------------------------------------------------------------------
4519 : */
4520 : static void
4521 1724 : OnConflict_CheckForSerializationFailure(const SERIALIZABLEXACT *reader,
4522 : SERIALIZABLEXACT *writer)
4523 : {
4524 : bool failure;
4525 :
4526 : Assert(LWLockHeldByMe(SerializableXactHashLock));
4527 :
4528 1724 : failure = false;
4529 :
4530 : /*------------------------------------------------------------------------
4531 : * Check for already-committed writer with rw-conflict out flagged
4532 : * (conflict-flag on W means that T2 committed before W):
4533 : *
4534 : * R ------> W ------> T2
4535 : * rw rw
4536 : *
4537 : * That is a dangerous structure, so we must abort. (Since the writer
4538 : * has already committed, we must be the reader)
4539 : *------------------------------------------------------------------------
4540 : */
4541 1724 : if (SxactIsCommitted(writer)
4542 36 : && (SxactHasConflictOut(writer) || SxactHasSummaryConflictOut(writer)))
4543 4 : failure = true;
4544 :
4545 : /*------------------------------------------------------------------------
4546 : * Check whether the writer has become a pivot with an out-conflict
4547 : * committed transaction (T2), and T2 committed first:
4548 : *
4549 : * R ------> W ------> T2
4550 : * rw rw
4551 : *
4552 : * Because T2 must've committed first, there is no anomaly if:
4553 : * - the reader committed before T2
4554 : * - the writer committed before T2
4555 : * - the reader is a READ ONLY transaction and the reader was concurrent
4556 : * with T2 (= reader acquired its snapshot before T2 committed)
4557 : *
4558 : * We also handle the case that T2 is prepared but not yet committed
4559 : * here. In that case T2 has already checked for conflicts, so if it
4560 : * commits first, making the above conflict real, it's too late for it
4561 : * to abort.
4562 : *------------------------------------------------------------------------
4563 : */
4564 1724 : if (!failure && SxactHasSummaryConflictOut(writer))
4565 0 : failure = true;
4566 1724 : else if (!failure)
4567 : {
4568 : dlist_iter iter;
4569 :
4570 2146 : dlist_foreach(iter, &writer->outConflicts)
4571 : {
4572 576 : RWConflict conflict =
4573 576 : dlist_container(RWConflictData, outLink, iter.cur);
4574 576 : SERIALIZABLEXACT *t2 = conflict->sxactIn;
4575 :
4576 576 : if (SxactIsPrepared(t2)
4577 162 : && (!SxactIsCommitted(reader)
4578 130 : || t2->prepareSeqNo <= reader->commitSeqNo)
4579 162 : && (!SxactIsCommitted(writer)
4580 0 : || t2->prepareSeqNo <= writer->commitSeqNo)
4581 162 : && (!SxactIsReadOnly(reader)
4582 24 : || t2->prepareSeqNo <= reader->SeqNo.lastCommitBeforeSnapshot))
4583 : {
4584 150 : failure = true;
4585 150 : break;
4586 : }
4587 : }
4588 : }
4589 :
4590 : /*------------------------------------------------------------------------
4591 : * Check whether the reader has become a pivot with a writer
4592 : * that's committed (or prepared):
4593 : *
4594 : * T0 ------> R ------> W
4595 : * rw rw
4596 : *
4597 : * Because W must've committed first for an anomaly to occur, there is no
4598 : * anomaly if:
4599 : * - T0 committed before the writer
4600 : * - T0 is READ ONLY, and overlaps the writer
4601 : *------------------------------------------------------------------------
4602 : */
4603 1724 : if (!failure && SxactIsPrepared(writer) && !SxactIsReadOnly(reader))
4604 : {
4605 36 : if (SxactHasSummaryConflictIn(reader))
4606 : {
4607 0 : failure = true;
4608 : }
4609 : else
4610 : {
4611 : dlist_iter iter;
4612 :
4613 : /*
4614 : * The unconstify is needed as we have no const version of
4615 : * dlist_foreach().
4616 : */
4617 36 : dlist_foreach(iter, &unconstify(SERIALIZABLEXACT *, reader)->inConflicts)
4618 : {
4619 22 : const RWConflict conflict =
4620 22 : dlist_container(RWConflictData, inLink, iter.cur);
4621 22 : const SERIALIZABLEXACT *t0 = conflict->sxactOut;
4622 :
4623 22 : if (!SxactIsDoomed(t0)
4624 22 : && (!SxactIsCommitted(t0)
4625 22 : || t0->commitSeqNo >= writer->prepareSeqNo)
4626 22 : && (!SxactIsReadOnly(t0)
4627 0 : || t0->SeqNo.lastCommitBeforeSnapshot >= writer->prepareSeqNo))
4628 : {
4629 22 : failure = true;
4630 22 : break;
4631 : }
4632 : }
4633 : }
4634 : }
4635 :
4636 1724 : if (failure)
4637 : {
4638 : /*
4639 : * We have to kill a transaction to avoid a possible anomaly from
4640 : * occurring. If the writer is us, we can just ereport() to cause a
4641 : * transaction abort. Otherwise we flag the writer for termination,
4642 : * causing it to abort when it tries to commit. However, if the writer
4643 : * is a prepared transaction, already prepared, we can't abort it
4644 : * anymore, so we have to kill the reader instead.
4645 : */
4646 176 : if (MySerializableXact == writer)
4647 : {
4648 134 : LWLockRelease(SerializableXactHashLock);
4649 134 : ereport(ERROR,
4650 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4651 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4652 : errdetail_internal("Reason code: Canceled on identification as a pivot, during write."),
4653 : errhint("The transaction might succeed if retried.")));
4654 : }
4655 42 : else if (SxactIsPrepared(writer))
4656 : {
4657 26 : LWLockRelease(SerializableXactHashLock);
4658 :
4659 : /* if we're not the writer, we have to be the reader */
4660 : Assert(MySerializableXact == reader);
4661 26 : ereport(ERROR,
4662 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4663 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4664 : errdetail_internal("Reason code: Canceled on conflict out to pivot %u, during read.", writer->topXid),
4665 : errhint("The transaction might succeed if retried.")));
4666 : }
4667 16 : writer->flags |= SXACT_FLAG_DOOMED;
4668 : }
4669 1564 : }
4670 :
4671 : /*
4672 : * PreCommit_CheckForSerializationFailure
4673 : * Check for dangerous structures in a serializable transaction
4674 : * at commit.
4675 : *
4676 : * We're checking for a dangerous structure as each conflict is recorded.
4677 : * The only way we could have a problem at commit is if this is the "out"
4678 : * side of a pivot, and neither the "in" side nor the pivot has yet
4679 : * committed.
4680 : *
4681 : * If a dangerous structure is found, the pivot (the near conflict) is
4682 : * marked for death, because rolling back another transaction might mean
4683 : * that we fail without ever making progress. This transaction is
4684 : * committing writes, so letting it commit ensures progress. If we
4685 : * canceled the far conflict, it might immediately fail again on retry.
4686 : */
4687 : void
4688 518452 : PreCommit_CheckForSerializationFailure(void)
4689 : {
4690 : dlist_iter near_iter;
4691 :
4692 518452 : if (MySerializableXact == InvalidSerializableXact)
4693 515664 : return;
4694 :
4695 : Assert(IsolationIsSerializable());
4696 :
4697 2788 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4698 :
4699 : /*
4700 : * Check if someone else has already decided that we need to die. Since
4701 : * we set our own DOOMED flag when partially releasing, ignore in that
4702 : * case.
4703 : */
4704 2788 : if (SxactIsDoomed(MySerializableXact) &&
4705 312 : !SxactIsPartiallyReleased(MySerializableXact))
4706 : {
4707 310 : LWLockRelease(SerializableXactHashLock);
4708 310 : ereport(ERROR,
4709 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4710 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4711 : errdetail_internal("Reason code: Canceled on identification as a pivot, during commit attempt."),
4712 : errhint("The transaction might succeed if retried.")));
4713 : }
4714 :
4715 3684 : dlist_foreach(near_iter, &MySerializableXact->inConflicts)
4716 : {
4717 1206 : RWConflict nearConflict =
4718 1206 : dlist_container(RWConflictData, inLink, near_iter.cur);
4719 :
4720 1206 : if (!SxactIsCommitted(nearConflict->sxactOut)
4721 834 : && !SxactIsDoomed(nearConflict->sxactOut))
4722 : {
4723 : dlist_iter far_iter;
4724 :
4725 896 : dlist_foreach(far_iter, &nearConflict->sxactOut->inConflicts)
4726 : {
4727 358 : RWConflict farConflict =
4728 358 : dlist_container(RWConflictData, inLink, far_iter.cur);
4729 :
4730 358 : if (farConflict->sxactOut == MySerializableXact
4731 86 : || (!SxactIsCommitted(farConflict->sxactOut)
4732 48 : && !SxactIsReadOnly(farConflict->sxactOut)
4733 24 : && !SxactIsDoomed(farConflict->sxactOut)))
4734 : {
4735 : /*
4736 : * Normally, we kill the pivot transaction to make sure we
4737 : * make progress if the failing transaction is retried.
4738 : * However, we can't kill it if it's already prepared, so
4739 : * in that case we commit suicide instead.
4740 : */
4741 296 : if (SxactIsPrepared(nearConflict->sxactOut))
4742 : {
4743 0 : LWLockRelease(SerializableXactHashLock);
4744 0 : ereport(ERROR,
4745 : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4746 : errmsg("could not serialize access due to read/write dependencies among transactions"),
4747 : errdetail_internal("Reason code: Canceled on commit attempt with conflict in from prepared pivot."),
4748 : errhint("The transaction might succeed if retried.")));
4749 : }
4750 296 : nearConflict->sxactOut->flags |= SXACT_FLAG_DOOMED;
4751 296 : break;
4752 : }
4753 : }
4754 : }
4755 : }
4756 :
4757 2478 : MySerializableXact->prepareSeqNo = ++(PredXact->LastSxactCommitSeqNo);
4758 2478 : MySerializableXact->flags |= SXACT_FLAG_PREPARED;
4759 :
4760 2478 : LWLockRelease(SerializableXactHashLock);
4761 : }
4762 :
4763 : /*------------------------------------------------------------------------*/
4764 :
4765 : /*
4766 : * Two-phase commit support
4767 : */
4768 :
4769 : /*
4770 : * AtPrepare_Locks
4771 : * Do the preparatory work for a PREPARE: make 2PC state file
4772 : * records for all predicate locks currently held.
4773 : */
4774 : void
4775 750 : AtPrepare_PredicateLocks(void)
4776 : {
4777 : SERIALIZABLEXACT *sxact;
4778 : TwoPhasePredicateRecord record;
4779 : TwoPhasePredicateXactRecord *xactRecord;
4780 : TwoPhasePredicateLockRecord *lockRecord;
4781 : dlist_iter iter;
4782 :
4783 750 : sxact = MySerializableXact;
4784 750 : xactRecord = &(record.data.xactRecord);
4785 750 : lockRecord = &(record.data.lockRecord);
4786 :
4787 750 : if (MySerializableXact == InvalidSerializableXact)
4788 726 : return;
4789 :
4790 : /* Generate an xact record for our SERIALIZABLEXACT */
4791 24 : record.type = TWOPHASEPREDICATERECORD_XACT;
4792 24 : xactRecord->xmin = MySerializableXact->xmin;
4793 24 : xactRecord->flags = MySerializableXact->flags;
4794 :
4795 : /*
4796 : * Note that we don't include the list of conflicts in our out in the
4797 : * statefile, because new conflicts can be added even after the
4798 : * transaction prepares. We'll just make a conservative assumption during
4799 : * recovery instead.
4800 : */
4801 :
4802 24 : RegisterTwoPhaseRecord(TWOPHASE_RM_PREDICATELOCK_ID, 0,
4803 : &record, sizeof(record));
4804 :
4805 : /*
4806 : * Generate a lock record for each lock.
4807 : *
4808 : * To do this, we need to walk the predicate lock list in our sxact rather
4809 : * than using the local predicate lock table because the latter is not
4810 : * guaranteed to be accurate.
4811 : */
4812 24 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
4813 :
4814 : /*
4815 : * No need to take sxact->perXactPredicateListLock in parallel mode
4816 : * because there cannot be any parallel workers running while we are
4817 : * preparing a transaction.
4818 : */
4819 : Assert(!IsParallelWorker() && !ParallelContextActive());
4820 :
4821 44 : dlist_foreach(iter, &sxact->predicateLocks)
4822 : {
4823 20 : PREDICATELOCK *predlock =
4824 20 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
4825 :
4826 20 : record.type = TWOPHASEPREDICATERECORD_LOCK;
4827 20 : lockRecord->target = predlock->tag.myTarget->tag;
4828 :
4829 20 : RegisterTwoPhaseRecord(TWOPHASE_RM_PREDICATELOCK_ID, 0,
4830 : &record, sizeof(record));
4831 : }
4832 :
4833 24 : LWLockRelease(SerializablePredicateListLock);
4834 : }
4835 :
4836 : /*
4837 : * PostPrepare_Locks
4838 : * Clean up after successful PREPARE. Unlike the non-predicate
4839 : * lock manager, we do not need to transfer locks to a dummy
4840 : * PGPROC because our SERIALIZABLEXACT will stay around
4841 : * anyway. We only need to clean up our local state.
4842 : */
4843 : void
4844 750 : PostPrepare_PredicateLocks(TransactionId xid)
4845 : {
4846 750 : if (MySerializableXact == InvalidSerializableXact)
4847 726 : return;
4848 :
4849 : Assert(SxactIsPrepared(MySerializableXact));
4850 :
4851 24 : MySerializableXact->pid = 0;
4852 24 : MySerializableXact->pgprocno = INVALID_PROC_NUMBER;
4853 :
4854 24 : hash_destroy(LocalPredicateLockHash);
4855 24 : LocalPredicateLockHash = NULL;
4856 :
4857 24 : MySerializableXact = InvalidSerializableXact;
4858 24 : MyXactDidWrite = false;
4859 : }
4860 :
4861 : /*
4862 : * PredicateLockTwoPhaseFinish
4863 : * Release a prepared transaction's predicate locks once it
4864 : * commits or aborts.
4865 : */
4866 : void
4867 764 : PredicateLockTwoPhaseFinish(TransactionId xid, bool isCommit)
4868 : {
4869 : SERIALIZABLEXID *sxid;
4870 : SERIALIZABLEXIDTAG sxidtag;
4871 :
4872 764 : sxidtag.xid = xid;
4873 :
4874 764 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4875 : sxid = (SERIALIZABLEXID *)
4876 764 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
4877 764 : LWLockRelease(SerializableXactHashLock);
4878 :
4879 : /* xid will not be found if it wasn't a serializable transaction */
4880 764 : if (sxid == NULL)
4881 740 : return;
4882 :
4883 : /* Release its locks */
4884 24 : MySerializableXact = sxid->myXact;
4885 24 : MyXactDidWrite = true; /* conservatively assume that we wrote
4886 : * something */
4887 24 : ReleasePredicateLocks(isCommit, false);
4888 : }
4889 :
4890 : /*
4891 : * Re-acquire a predicate lock belonging to a transaction that was prepared.
4892 : */
4893 : void
4894 0 : predicatelock_twophase_recover(TransactionId xid, uint16 info,
4895 : void *recdata, uint32 len)
4896 : {
4897 : TwoPhasePredicateRecord *record;
4898 :
4899 : Assert(len == sizeof(TwoPhasePredicateRecord));
4900 :
4901 0 : record = (TwoPhasePredicateRecord *) recdata;
4902 :
4903 : Assert((record->type == TWOPHASEPREDICATERECORD_XACT) ||
4904 : (record->type == TWOPHASEPREDICATERECORD_LOCK));
4905 :
4906 0 : if (record->type == TWOPHASEPREDICATERECORD_XACT)
4907 : {
4908 : /* Per-transaction record. Set up a SERIALIZABLEXACT. */
4909 : TwoPhasePredicateXactRecord *xactRecord;
4910 : SERIALIZABLEXACT *sxact;
4911 : SERIALIZABLEXID *sxid;
4912 : SERIALIZABLEXIDTAG sxidtag;
4913 : bool found;
4914 :
4915 0 : xactRecord = (TwoPhasePredicateXactRecord *) &record->data.xactRecord;
4916 :
4917 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4918 0 : sxact = CreatePredXact();
4919 0 : if (!sxact)
4920 0 : ereport(ERROR,
4921 : (errcode(ERRCODE_OUT_OF_MEMORY),
4922 : errmsg("out of shared memory")));
4923 :
4924 : /* vxid for a prepared xact is INVALID_PROC_NUMBER/xid; no pid */
4925 0 : sxact->vxid.procNumber = INVALID_PROC_NUMBER;
4926 0 : sxact->vxid.localTransactionId = (LocalTransactionId) xid;
4927 0 : sxact->pid = 0;
4928 0 : sxact->pgprocno = INVALID_PROC_NUMBER;
4929 :
4930 : /* a prepared xact hasn't committed yet */
4931 0 : sxact->prepareSeqNo = RecoverySerCommitSeqNo;
4932 0 : sxact->commitSeqNo = InvalidSerCommitSeqNo;
4933 0 : sxact->finishedBefore = InvalidTransactionId;
4934 :
4935 0 : sxact->SeqNo.lastCommitBeforeSnapshot = RecoverySerCommitSeqNo;
4936 :
4937 : /*
4938 : * Don't need to track this; no transactions running at the time the
4939 : * recovered xact started are still active, except possibly other
4940 : * prepared xacts and we don't care whether those are RO_SAFE or not.
4941 : */
4942 0 : dlist_init(&(sxact->possibleUnsafeConflicts));
4943 :
4944 0 : dlist_init(&(sxact->predicateLocks));
4945 0 : dlist_node_init(&sxact->finishedLink);
4946 :
4947 0 : sxact->topXid = xid;
4948 0 : sxact->xmin = xactRecord->xmin;
4949 0 : sxact->flags = xactRecord->flags;
4950 : Assert(SxactIsPrepared(sxact));
4951 0 : if (!SxactIsReadOnly(sxact))
4952 : {
4953 0 : ++(PredXact->WritableSxactCount);
4954 : Assert(PredXact->WritableSxactCount <=
4955 : (MaxBackends + max_prepared_xacts));
4956 : }
4957 :
4958 : /*
4959 : * We don't know whether the transaction had any conflicts or not, so
4960 : * we'll conservatively assume that it had both a conflict in and a
4961 : * conflict out, and represent that with the summary conflict flags.
4962 : */
4963 0 : dlist_init(&(sxact->outConflicts));
4964 0 : dlist_init(&(sxact->inConflicts));
4965 0 : sxact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
4966 0 : sxact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4967 :
4968 : /* Register the transaction's xid */
4969 0 : sxidtag.xid = xid;
4970 0 : sxid = (SERIALIZABLEXID *) hash_search(SerializableXidHash,
4971 : &sxidtag,
4972 : HASH_ENTER, &found);
4973 : Assert(sxid != NULL);
4974 : Assert(!found);
4975 0 : sxid->myXact = (SERIALIZABLEXACT *) sxact;
4976 :
4977 : /*
4978 : * Update global xmin. Note that this is a special case compared to
4979 : * registering a normal transaction, because the global xmin might go
4980 : * backwards. That's OK, because until recovery is over we're not
4981 : * going to complete any transactions or create any non-prepared
4982 : * transactions, so there's no danger of throwing away.
4983 : */
4984 0 : if ((!TransactionIdIsValid(PredXact->SxactGlobalXmin)) ||
4985 0 : (TransactionIdFollows(PredXact->SxactGlobalXmin, sxact->xmin)))
4986 : {
4987 0 : PredXact->SxactGlobalXmin = sxact->xmin;
4988 0 : PredXact->SxactGlobalXminCount = 1;
4989 0 : SerialSetActiveSerXmin(sxact->xmin);
4990 : }
4991 0 : else if (TransactionIdEquals(sxact->xmin, PredXact->SxactGlobalXmin))
4992 : {
4993 : Assert(PredXact->SxactGlobalXminCount > 0);
4994 0 : PredXact->SxactGlobalXminCount++;
4995 : }
4996 :
4997 0 : LWLockRelease(SerializableXactHashLock);
4998 : }
4999 0 : else if (record->type == TWOPHASEPREDICATERECORD_LOCK)
5000 : {
5001 : /* Lock record. Recreate the PREDICATELOCK */
5002 : TwoPhasePredicateLockRecord *lockRecord;
5003 : SERIALIZABLEXID *sxid;
5004 : SERIALIZABLEXACT *sxact;
5005 : SERIALIZABLEXIDTAG sxidtag;
5006 : uint32 targettaghash;
5007 :
5008 0 : lockRecord = (TwoPhasePredicateLockRecord *) &record->data.lockRecord;
5009 0 : targettaghash = PredicateLockTargetTagHashCode(&lockRecord->target);
5010 :
5011 0 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
5012 0 : sxidtag.xid = xid;
5013 : sxid = (SERIALIZABLEXID *)
5014 0 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
5015 0 : LWLockRelease(SerializableXactHashLock);
5016 :
5017 : Assert(sxid != NULL);
5018 0 : sxact = sxid->myXact;
5019 : Assert(sxact != InvalidSerializableXact);
5020 :
5021 0 : CreatePredicateLock(&lockRecord->target, targettaghash, sxact);
5022 : }
5023 0 : }
5024 :
5025 : /*
5026 : * Prepare to share the current SERIALIZABLEXACT with parallel workers.
5027 : * Return a handle object that can be used by AttachSerializableXact() in a
5028 : * parallel worker.
5029 : */
5030 : SerializableXactHandle
5031 828 : ShareSerializableXact(void)
5032 : {
5033 828 : return MySerializableXact;
5034 : }
5035 :
5036 : /*
5037 : * Allow parallel workers to import the leader's SERIALIZABLEXACT.
5038 : */
5039 : void
5040 2642 : AttachSerializableXact(SerializableXactHandle handle)
5041 : {
5042 :
5043 : Assert(MySerializableXact == InvalidSerializableXact);
5044 :
5045 2642 : MySerializableXact = (SERIALIZABLEXACT *) handle;
5046 2642 : if (MySerializableXact != InvalidSerializableXact)
5047 26 : CreateLocalPredicateLockHash();
5048 2642 : }
|
