Age Owner Branch data TLA Line data Source code
1 : : /*-------------------------------------------------------------------------
2 : : *
3 : : * predicate.c
4 : : * POSTGRES predicate locking
5 : : * to support full serializable transaction isolation
6 : : *
7 : : *
8 : : * The approach taken is to implement Serializable Snapshot Isolation (SSI)
9 : : * as initially described in this paper:
10 : : *
11 : : * Michael J. Cahill, Uwe Röhm, and Alan D. Fekete. 2008.
12 : : * Serializable isolation for snapshot databases.
13 : : * In SIGMOD '08: Proceedings of the 2008 ACM SIGMOD
14 : : * international conference on Management of data,
15 : : * pages 729-738, New York, NY, USA. ACM.
16 : : * http://doi.acm.org/10.1145/1376616.1376690
17 : : *
18 : : * and further elaborated in Cahill's doctoral thesis:
19 : : *
20 : : * Michael James Cahill. 2009.
21 : : * Serializable Isolation for Snapshot Databases.
22 : : * Sydney Digital Theses.
23 : : * University of Sydney, School of Information Technologies.
24 : : * http://hdl.handle.net/2123/5353
25 : : *
26 : : *
27 : : * Predicate locks for Serializable Snapshot Isolation (SSI) are SIREAD
28 : : * locks, which are so different from normal locks that a distinct set of
29 : : * structures is required to handle them. They are needed to detect
30 : : * rw-conflicts when the read happens before the write. (When the write
31 : : * occurs first, the reading transaction can check for a conflict by
32 : : * examining the MVCC data.)
33 : : *
34 : : * (1) Besides tuples actually read, they must cover ranges of tuples
35 : : * which would have been read based on the predicate. This will
36 : : * require modelling the predicates through locks against database
37 : : * objects such as pages, index ranges, or entire tables.
38 : : *
39 : : * (2) They must be kept in RAM for quick access. Because of this, it
40 : : * isn't possible to always maintain tuple-level granularity -- when
41 : : * the space allocated to store these approaches exhaustion, a
42 : : * request for a lock may need to scan for situations where a single
43 : : * transaction holds many fine-grained locks which can be coalesced
44 : : * into a single coarser-grained lock.
45 : : *
46 : : * (3) They never block anything; they are more like flags than locks
47 : : * in that regard; although they refer to database objects and are
48 : : * used to identify rw-conflicts with normal write locks.
49 : : *
50 : : * (4) While they are associated with a transaction, they must survive
51 : : * a successful COMMIT of that transaction, and remain until all
52 : : * overlapping transactions complete. This even means that they
53 : : * must survive termination of the transaction's process. If a
54 : : * top level transaction is rolled back, however, it is immediately
55 : : * flagged so that it can be ignored, and its SIREAD locks can be
56 : : * released any time after that.
57 : : *
58 : : * (5) The only transactions which create SIREAD locks or check for
59 : : * conflicts with them are serializable transactions.
60 : : *
61 : : * (6) When a write lock for a top level transaction is found to cover
62 : : * an existing SIREAD lock for the same transaction, the SIREAD lock
63 : : * can be deleted.
64 : : *
65 : : * (7) A write from a serializable transaction must ensure that an xact
66 : : * record exists for the transaction, with the same lifespan (until
67 : : * all concurrent transaction complete or the transaction is rolled
68 : : * back) so that rw-dependencies to that transaction can be
69 : : * detected.
70 : : *
71 : : * We use an optimization for read-only transactions. Under certain
72 : : * circumstances, a read-only transaction's snapshot can be shown to
73 : : * never have conflicts with other transactions. This is referred to
74 : : * as a "safe" snapshot (and one known not to be is "unsafe").
75 : : * However, it can't be determined whether a snapshot is safe until
76 : : * all concurrent read/write transactions complete.
77 : : *
78 : : * Once a read-only transaction is known to have a safe snapshot, it
79 : : * can release its predicate locks and exempt itself from further
80 : : * predicate lock tracking. READ ONLY DEFERRABLE transactions run only
81 : : * on safe snapshots, waiting as necessary for one to be available.
82 : : *
83 : : *
84 : : * Lightweight locks to manage access to the predicate locking shared
85 : : * memory objects must be taken in this order, and should be released in
86 : : * reverse order:
87 : : *
88 : : * SerializableFinishedListLock
89 : : * - Protects the list of transactions which have completed but which
90 : : * may yet matter because they overlap still-active transactions.
91 : : *
92 : : * SerializablePredicateListLock
93 : : * - Protects the linked list of locks held by a transaction. Note
94 : : * that the locks themselves are also covered by the partition
95 : : * locks of their respective lock targets; this lock only affects
96 : : * the linked list connecting the locks related to a transaction.
97 : : * - All transactions share this single lock (with no partitioning).
98 : : * - There is never a need for a process other than the one running
99 : : * an active transaction to walk the list of locks held by that
100 : : * transaction, except parallel query workers sharing the leader's
101 : : * transaction. In the parallel case, an extra per-sxact lock is
102 : : * taken; see below.
103 : : * - It is relatively infrequent that another process needs to
104 : : * modify the list for a transaction, but it does happen for such
105 : : * things as index page splits for pages with predicate locks and
106 : : * freeing of predicate locked pages by a vacuum process. When
107 : : * removing a lock in such cases, the lock itself contains the
108 : : * pointers needed to remove it from the list. When adding a
109 : : * lock in such cases, the lock can be added using the anchor in
110 : : * the transaction structure. Neither requires walking the list.
111 : : * - Cleaning up the list for a terminated transaction is sometimes
112 : : * not done on a retail basis, in which case no lock is required.
113 : : * - Due to the above, a process accessing its active transaction's
114 : : * list always uses a shared lock, regardless of whether it is
115 : : * walking or maintaining the list. This improves concurrency
116 : : * for the common access patterns.
117 : : * - A process which needs to alter the list of a transaction other
118 : : * than its own active transaction must acquire an exclusive
119 : : * lock.
120 : : *
121 : : * SERIALIZABLEXACT's member 'perXactPredicateListLock'
122 : : * - Protects the linked list of predicate locks held by a transaction.
123 : : * Only needed for parallel mode, where multiple backends share the
124 : : * same SERIALIZABLEXACT object. Not needed if
125 : : * SerializablePredicateListLock is held exclusively.
126 : : *
127 : : * PredicateLockHashPartitionLock(hashcode)
128 : : * - The same lock protects a target, all locks on that target, and
129 : : * the linked list of locks on the target.
130 : : * - When more than one is needed, acquire in ascending address order.
131 : : * - When all are needed (rare), acquire in ascending index order with
132 : : * PredicateLockHashPartitionLockByIndex(index).
133 : : *
134 : : * SerializableXactHashLock
135 : : * - Protects both PredXact and SerializableXidHash.
136 : : *
137 : : * SerialControlLock
138 : : * - Protects SerialControlData members
139 : : *
140 : : * SLRU per-bank locks
141 : : * - Protects SerialSlruCtl
142 : : *
143 : : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
144 : : * Portions Copyright (c) 1994, Regents of the University of California
145 : : *
146 : : *
147 : : * IDENTIFICATION
148 : : * src/backend/storage/lmgr/predicate.c
149 : : *
150 : : *-------------------------------------------------------------------------
151 : : */
152 : : /*
153 : : * INTERFACE ROUTINES
154 : : *
155 : : * housekeeping for setting up shared memory predicate lock structures
156 : : * PredicateLockShmemInit(void)
157 : : * PredicateLockShmemSize(void)
158 : : *
159 : : * predicate lock reporting
160 : : * GetPredicateLockStatusData(void)
161 : : * PageIsPredicateLocked(Relation relation, BlockNumber blkno)
162 : : *
163 : : * predicate lock maintenance
164 : : * GetSerializableTransactionSnapshot(Snapshot snapshot)
165 : : * SetSerializableTransactionSnapshot(Snapshot snapshot,
166 : : * VirtualTransactionId *sourcevxid)
167 : : * RegisterPredicateLockingXid(void)
168 : : * PredicateLockRelation(Relation relation, Snapshot snapshot)
169 : : * PredicateLockPage(Relation relation, BlockNumber blkno,
170 : : * Snapshot snapshot)
171 : : * PredicateLockTID(Relation relation, const ItemPointerData *tid, Snapshot snapshot,
172 : : * TransactionId tuple_xid)
173 : : * PredicateLockPageSplit(Relation relation, BlockNumber oldblkno,
174 : : * BlockNumber newblkno)
175 : : * PredicateLockPageCombine(Relation relation, BlockNumber oldblkno,
176 : : * BlockNumber newblkno)
177 : : * TransferPredicateLocksToHeapRelation(Relation relation)
178 : : * ReleasePredicateLocks(bool isCommit, bool isReadOnlySafe)
179 : : *
180 : : * conflict detection (may also trigger rollback)
181 : : * CheckForSerializableConflictOut(Relation relation, TransactionId xid,
182 : : * Snapshot snapshot)
183 : : * CheckForSerializableConflictIn(Relation relation, const ItemPointerData *tid,
184 : : * BlockNumber blkno)
185 : : * CheckTableForSerializableConflictIn(Relation relation)
186 : : *
187 : : * final rollback checking
188 : : * PreCommit_CheckForSerializationFailure(void)
189 : : *
190 : : * two-phase commit support
191 : : * AtPrepare_PredicateLocks(void);
192 : : * PostPrepare_PredicateLocks(TransactionId xid);
193 : : * PredicateLockTwoPhaseFinish(TransactionId xid, bool isCommit);
194 : : * predicatelock_twophase_recover(FullTransactionId fxid, uint16 info,
195 : : * void *recdata, uint32 len);
196 : : */
197 : :
198 : : #include "postgres.h"
199 : :
200 : : #include "access/parallel.h"
201 : : #include "access/slru.h"
202 : : #include "access/transam.h"
203 : : #include "access/twophase.h"
204 : : #include "access/twophase_rmgr.h"
205 : : #include "access/xact.h"
206 : : #include "access/xlog.h"
207 : : #include "miscadmin.h"
208 : : #include "pgstat.h"
209 : : #include "port/pg_lfind.h"
210 : : #include "storage/predicate.h"
211 : : #include "storage/predicate_internals.h"
212 : : #include "storage/proc.h"
213 : : #include "storage/procarray.h"
214 : : #include "utils/guc_hooks.h"
215 : : #include "utils/rel.h"
216 : : #include "utils/snapmgr.h"
217 : :
218 : : /* Uncomment the next line to test the graceful degradation code. */
219 : : /* #define TEST_SUMMARIZE_SERIAL */
220 : :
221 : : /*
222 : : * Test the most selective fields first, for performance.
223 : : *
224 : : * a is covered by b if all of the following hold:
225 : : * 1) a.database = b.database
226 : : * 2) a.relation = b.relation
227 : : * 3) b.offset is invalid (b is page-granularity or higher)
228 : : * 4) either of the following:
229 : : * 4a) a.offset is valid (a is tuple-granularity) and a.page = b.page
230 : : * or 4b) a.offset is invalid and b.page is invalid (a is
231 : : * page-granularity and b is relation-granularity
232 : : */
233 : : #define TargetTagIsCoveredBy(covered_target, covering_target) \
234 : : ((GET_PREDICATELOCKTARGETTAG_RELATION(covered_target) == /* (2) */ \
235 : : GET_PREDICATELOCKTARGETTAG_RELATION(covering_target)) \
236 : : && (GET_PREDICATELOCKTARGETTAG_OFFSET(covering_target) == \
237 : : InvalidOffsetNumber) /* (3) */ \
238 : : && (((GET_PREDICATELOCKTARGETTAG_OFFSET(covered_target) != \
239 : : InvalidOffsetNumber) /* (4a) */ \
240 : : && (GET_PREDICATELOCKTARGETTAG_PAGE(covering_target) == \
241 : : GET_PREDICATELOCKTARGETTAG_PAGE(covered_target))) \
242 : : || ((GET_PREDICATELOCKTARGETTAG_PAGE(covering_target) == \
243 : : InvalidBlockNumber) /* (4b) */ \
244 : : && (GET_PREDICATELOCKTARGETTAG_PAGE(covered_target) \
245 : : != InvalidBlockNumber))) \
246 : : && (GET_PREDICATELOCKTARGETTAG_DB(covered_target) == /* (1) */ \
247 : : GET_PREDICATELOCKTARGETTAG_DB(covering_target)))
248 : :
249 : : /*
250 : : * The predicate locking target and lock shared hash tables are partitioned to
251 : : * reduce contention. To determine which partition a given target belongs to,
252 : : * compute the tag's hash code with PredicateLockTargetTagHashCode(), then
253 : : * apply one of these macros.
254 : : * NB: NUM_PREDICATELOCK_PARTITIONS must be a power of 2!
255 : : */
256 : : #define PredicateLockHashPartition(hashcode) \
257 : : ((hashcode) % NUM_PREDICATELOCK_PARTITIONS)
258 : : #define PredicateLockHashPartitionLock(hashcode) \
259 : : (&MainLWLockArray[PREDICATELOCK_MANAGER_LWLOCK_OFFSET + \
260 : : PredicateLockHashPartition(hashcode)].lock)
261 : : #define PredicateLockHashPartitionLockByIndex(i) \
262 : : (&MainLWLockArray[PREDICATELOCK_MANAGER_LWLOCK_OFFSET + (i)].lock)
263 : :
264 : : #define NPREDICATELOCKTARGETENTS() \
265 : : mul_size(max_predicate_locks_per_xact, add_size(MaxBackends, max_prepared_xacts))
266 : :
267 : : #define SxactIsOnFinishedList(sxact) (!dlist_node_is_detached(&(sxact)->finishedLink))
268 : :
269 : : /*
270 : : * Note that a sxact is marked "prepared" once it has passed
271 : : * PreCommit_CheckForSerializationFailure, even if it isn't using
272 : : * 2PC. This is the point at which it can no longer be aborted.
273 : : *
274 : : * The PREPARED flag remains set after commit, so SxactIsCommitted
275 : : * implies SxactIsPrepared.
276 : : */
277 : : #define SxactIsCommitted(sxact) (((sxact)->flags & SXACT_FLAG_COMMITTED) != 0)
278 : : #define SxactIsPrepared(sxact) (((sxact)->flags & SXACT_FLAG_PREPARED) != 0)
279 : : #define SxactIsRolledBack(sxact) (((sxact)->flags & SXACT_FLAG_ROLLED_BACK) != 0)
280 : : #define SxactIsDoomed(sxact) (((sxact)->flags & SXACT_FLAG_DOOMED) != 0)
281 : : #define SxactIsReadOnly(sxact) (((sxact)->flags & SXACT_FLAG_READ_ONLY) != 0)
282 : : #define SxactHasSummaryConflictIn(sxact) (((sxact)->flags & SXACT_FLAG_SUMMARY_CONFLICT_IN) != 0)
283 : : #define SxactHasSummaryConflictOut(sxact) (((sxact)->flags & SXACT_FLAG_SUMMARY_CONFLICT_OUT) != 0)
284 : : /*
285 : : * The following macro actually means that the specified transaction has a
286 : : * conflict out *to a transaction which committed ahead of it*. It's hard
287 : : * to get that into a name of a reasonable length.
288 : : */
289 : : #define SxactHasConflictOut(sxact) (((sxact)->flags & SXACT_FLAG_CONFLICT_OUT) != 0)
290 : : #define SxactIsDeferrableWaiting(sxact) (((sxact)->flags & SXACT_FLAG_DEFERRABLE_WAITING) != 0)
291 : : #define SxactIsROSafe(sxact) (((sxact)->flags & SXACT_FLAG_RO_SAFE) != 0)
292 : : #define SxactIsROUnsafe(sxact) (((sxact)->flags & SXACT_FLAG_RO_UNSAFE) != 0)
293 : : #define SxactIsPartiallyReleased(sxact) (((sxact)->flags & SXACT_FLAG_PARTIALLY_RELEASED) != 0)
294 : :
295 : : /*
296 : : * Compute the hash code associated with a PREDICATELOCKTARGETTAG.
297 : : *
298 : : * To avoid unnecessary recomputations of the hash code, we try to do this
299 : : * just once per function, and then pass it around as needed. Aside from
300 : : * passing the hashcode to hash_search_with_hash_value(), we can extract
301 : : * the lock partition number from the hashcode.
302 : : */
303 : : #define PredicateLockTargetTagHashCode(predicatelocktargettag) \
304 : : get_hash_value(PredicateLockTargetHash, predicatelocktargettag)
305 : :
306 : : /*
307 : : * Given a predicate lock tag, and the hash for its target,
308 : : * compute the lock hash.
309 : : *
310 : : * To make the hash code also depend on the transaction, we xor the sxid
311 : : * struct's address into the hash code, left-shifted so that the
312 : : * partition-number bits don't change. Since this is only a hash, we
313 : : * don't care if we lose high-order bits of the address; use an
314 : : * intermediate variable to suppress cast-pointer-to-int warnings.
315 : : */
316 : : #define PredicateLockHashCodeFromTargetHashCode(predicatelocktag, targethash) \
317 : : ((targethash) ^ ((uint32) PointerGetDatum((predicatelocktag)->myXact)) \
318 : : << LOG2_NUM_PREDICATELOCK_PARTITIONS)
319 : :
320 : :
321 : : /*
322 : : * The SLRU buffer area through which we access the old xids.
323 : : */
324 : : static SlruCtlData SerialSlruCtlData;
325 : :
326 : : #define SerialSlruCtl (&SerialSlruCtlData)
327 : :
328 : : #define SERIAL_PAGESIZE BLCKSZ
329 : : #define SERIAL_ENTRYSIZE sizeof(SerCommitSeqNo)
330 : : #define SERIAL_ENTRIESPERPAGE (SERIAL_PAGESIZE / SERIAL_ENTRYSIZE)
331 : :
332 : : /*
333 : : * Set maximum pages based on the number needed to track all transactions.
334 : : */
335 : : #define SERIAL_MAX_PAGE (MaxTransactionId / SERIAL_ENTRIESPERPAGE)
336 : :
337 : : #define SerialNextPage(page) (((page) >= SERIAL_MAX_PAGE) ? 0 : (page) + 1)
338 : :
339 : : #define SerialValue(slotno, xid) (*((SerCommitSeqNo *) \
340 : : (SerialSlruCtl->shared->page_buffer[slotno] + \
341 : : ((((uint32) (xid)) % SERIAL_ENTRIESPERPAGE) * SERIAL_ENTRYSIZE))))
342 : :
343 : : #define SerialPage(xid) (((uint32) (xid)) / SERIAL_ENTRIESPERPAGE)
344 : :
345 : : typedef struct SerialControlData
346 : : {
347 : : int64 headPage; /* newest initialized page */
348 : : TransactionId headXid; /* newest valid Xid in the SLRU */
349 : : TransactionId tailXid; /* oldest xmin we might be interested in */
350 : : } SerialControlData;
351 : :
352 : : typedef struct SerialControlData *SerialControl;
353 : :
354 : : static SerialControl serialControl;
355 : :
356 : : /*
357 : : * When the oldest committed transaction on the "finished" list is moved to
358 : : * SLRU, its predicate locks will be moved to this "dummy" transaction,
359 : : * collapsing duplicate targets. When a duplicate is found, the later
360 : : * commitSeqNo is used.
361 : : */
362 : : static SERIALIZABLEXACT *OldCommittedSxact;
363 : :
364 : :
365 : : /*
366 : : * These configuration variables are used to set the predicate lock table size
367 : : * and to control promotion of predicate locks to coarser granularity in an
368 : : * attempt to degrade performance (mostly as false positive serialization
369 : : * failure) gracefully in the face of memory pressure.
370 : : */
371 : : int max_predicate_locks_per_xact; /* in guc_tables.c */
372 : : int max_predicate_locks_per_relation; /* in guc_tables.c */
373 : : int max_predicate_locks_per_page; /* in guc_tables.c */
374 : :
375 : : /*
376 : : * This provides a list of objects in order to track transactions
377 : : * participating in predicate locking. Entries in the list are fixed size,
378 : : * and reside in shared memory. The memory address of an entry must remain
379 : : * fixed during its lifetime. The list will be protected from concurrent
380 : : * update externally; no provision is made in this code to manage that. The
381 : : * number of entries in the list, and the size allowed for each entry is
382 : : * fixed upon creation.
383 : : */
384 : : static PredXactList PredXact;
385 : :
386 : : /*
387 : : * This provides a pool of RWConflict data elements to use in conflict lists
388 : : * between transactions.
389 : : */
390 : : static RWConflictPoolHeader RWConflictPool;
391 : :
392 : : /*
393 : : * The predicate locking hash tables are in shared memory.
394 : : * Each backend keeps pointers to them.
395 : : */
396 : : static HTAB *SerializableXidHash;
397 : : static HTAB *PredicateLockTargetHash;
398 : : static HTAB *PredicateLockHash;
399 : : static dlist_head *FinishedSerializableTransactions;
400 : :
401 : : /*
402 : : * Tag for a dummy entry in PredicateLockTargetHash. By temporarily removing
403 : : * this entry, you can ensure that there's enough scratch space available for
404 : : * inserting one entry in the hash table. This is an otherwise-invalid tag.
405 : : */
406 : : static const PREDICATELOCKTARGETTAG ScratchTargetTag = {0, 0, 0, 0};
407 : : static uint32 ScratchTargetTagHash;
408 : : static LWLock *ScratchPartitionLock;
409 : :
410 : : /*
411 : : * The local hash table used to determine when to combine multiple fine-
412 : : * grained locks into a single courser-grained lock.
413 : : */
414 : : static HTAB *LocalPredicateLockHash = NULL;
415 : :
416 : : /*
417 : : * Keep a pointer to the currently-running serializable transaction (if any)
418 : : * for quick reference. Also, remember if we have written anything that could
419 : : * cause a rw-conflict.
420 : : */
421 : : static SERIALIZABLEXACT *MySerializableXact = InvalidSerializableXact;
422 : : static bool MyXactDidWrite = false;
423 : :
424 : : /*
425 : : * The SXACT_FLAG_RO_UNSAFE optimization might lead us to release
426 : : * MySerializableXact early. If that happens in a parallel query, the leader
427 : : * needs to defer the destruction of the SERIALIZABLEXACT until end of
428 : : * transaction, because the workers still have a reference to it. In that
429 : : * case, the leader stores it here.
430 : : */
431 : : static SERIALIZABLEXACT *SavedSerializableXact = InvalidSerializableXact;
432 : :
433 : : /* local functions */
434 : :
435 : : static SERIALIZABLEXACT *CreatePredXact(void);
436 : : static void ReleasePredXact(SERIALIZABLEXACT *sxact);
437 : :
438 : : static bool RWConflictExists(const SERIALIZABLEXACT *reader, const SERIALIZABLEXACT *writer);
439 : : static void SetRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer);
440 : : static void SetPossibleUnsafeConflict(SERIALIZABLEXACT *roXact, SERIALIZABLEXACT *activeXact);
441 : : static void ReleaseRWConflict(RWConflict conflict);
442 : : static void FlagSxactUnsafe(SERIALIZABLEXACT *sxact);
443 : :
444 : : static bool SerialPagePrecedesLogically(int64 page1, int64 page2);
445 : : static void SerialInit(void);
446 : : static void SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo);
447 : : static SerCommitSeqNo SerialGetMinConflictCommitSeqNo(TransactionId xid);
448 : : static void SerialSetActiveSerXmin(TransactionId xid);
449 : :
450 : : static uint32 predicatelock_hash(const void *key, Size keysize);
451 : : static void SummarizeOldestCommittedSxact(void);
452 : : static Snapshot GetSafeSnapshot(Snapshot origSnapshot);
453 : : static Snapshot GetSerializableTransactionSnapshotInt(Snapshot snapshot,
454 : : VirtualTransactionId *sourcevxid,
455 : : int sourcepid);
456 : : static bool PredicateLockExists(const PREDICATELOCKTARGETTAG *targettag);
457 : : static bool GetParentPredicateLockTag(const PREDICATELOCKTARGETTAG *tag,
458 : : PREDICATELOCKTARGETTAG *parent);
459 : : static bool CoarserLockCovers(const PREDICATELOCKTARGETTAG *newtargettag);
460 : : static void RemoveScratchTarget(bool lockheld);
461 : : static void RestoreScratchTarget(bool lockheld);
462 : : static void RemoveTargetIfNoLongerUsed(PREDICATELOCKTARGET *target,
463 : : uint32 targettaghash);
464 : : static void DeleteChildTargetLocks(const PREDICATELOCKTARGETTAG *newtargettag);
465 : : static int MaxPredicateChildLocks(const PREDICATELOCKTARGETTAG *tag);
466 : : static bool CheckAndPromotePredicateLockRequest(const PREDICATELOCKTARGETTAG *reqtag);
467 : : static void DecrementParentLocks(const PREDICATELOCKTARGETTAG *targettag);
468 : : static void CreatePredicateLock(const PREDICATELOCKTARGETTAG *targettag,
469 : : uint32 targettaghash,
470 : : SERIALIZABLEXACT *sxact);
471 : : static void DeleteLockTarget(PREDICATELOCKTARGET *target, uint32 targettaghash);
472 : : static bool TransferPredicateLocksToNewTarget(PREDICATELOCKTARGETTAG oldtargettag,
473 : : PREDICATELOCKTARGETTAG newtargettag,
474 : : bool removeOld);
475 : : static void PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag);
476 : : static void DropAllPredicateLocksFromTable(Relation relation,
477 : : bool transfer);
478 : : static void SetNewSxactGlobalXmin(void);
479 : : static void ClearOldPredicateLocks(void);
480 : : static void ReleaseOneSerializableXact(SERIALIZABLEXACT *sxact, bool partial,
481 : : bool summarize);
482 : : static bool XidIsConcurrent(TransactionId xid);
483 : : static void CheckTargetForConflictsIn(PREDICATELOCKTARGETTAG *targettag);
484 : : static void FlagRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer);
485 : : static void OnConflict_CheckForSerializationFailure(const SERIALIZABLEXACT *reader,
486 : : SERIALIZABLEXACT *writer);
487 : : static void CreateLocalPredicateLockHash(void);
488 : : static void ReleasePredicateLocksLocal(void);
489 : :
490 : :
491 : : /*------------------------------------------------------------------------*/
492 : :
493 : : /*
494 : : * Does this relation participate in predicate locking? Temporary and system
495 : : * relations are exempt.
496 : : */
497 : : static inline bool
5299 heikki.linnakangas@i 498 :CBC 160727 : PredicateLockingNeededForRelation(Relation relation)
499 : : {
1616 tgl@sss.pgh.pa.us 500 [ + + ]: 217948 : return !(relation->rd_id < FirstUnpinnedObjectId ||
1112 michael@paquier.xyz 501 [ + + ]: 57221 : 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
5299 heikki.linnakangas@i 516 : 58369296 : SerializationNeededForRead(Relation relation, Snapshot snapshot)
517 : : {
518 : : /* Nothing to do if this is not a serializable transaction */
519 [ + + ]: 58369296 : if (MySerializableXact == InvalidSerializableXact)
520 : 58214986 : 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 [ + + + - ]: 154310 : if (!IsMVCCSnapshot(snapshot))
530 : 1599 : 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 [ + + ]: 152711 : if (SxactIsROSafe(MySerializableXact))
543 : : {
2469 tmunro@postgresql.or 544 : 33 : ReleasePredicateLocks(false, true);
5299 heikki.linnakangas@i 545 : 33 : return false;
546 : : }
547 : :
548 : : /* Check if the relation doesn't participate in predicate locking */
549 [ + + ]: 152678 : if (!PredicateLockingNeededForRelation(relation))
550 : 100147 : return false;
551 : :
5298 552 : 52531 : 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
5299 560 : 16650667 : SerializationNeededForWrite(Relation relation)
561 : : {
562 : : /* Nothing to do if this is not a serializable transaction */
563 [ + + ]: 16650667 : if (MySerializableXact == InvalidSerializableXact)
564 : 16642694 : return false;
565 : :
566 : : /* Check if the relation doesn't participate in predicate locking */
567 [ + + ]: 7973 : if (!PredicateLockingNeededForRelation(relation))
568 : 3487 : return false;
569 : :
5298 570 : 4486 : 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 *
5427 582 : 2738 : CreatePredXact(void)
583 : : {
584 : : SERIALIZABLEXACT *sxact;
585 : :
1063 andres@anarazel.de 586 [ - + ]: 2738 : if (dlist_is_empty(&PredXact->availableList))
5427 heikki.linnakangas@i 587 :UBC 0 : return NULL;
588 : :
1063 andres@anarazel.de 589 :CBC 2738 : sxact = dlist_container(SERIALIZABLEXACT, xactLink,
590 : : dlist_pop_head_node(&PredXact->availableList));
591 : 2738 : dlist_push_tail(&PredXact->activeList, &sxact->xactLink);
592 : 2738 : return sxact;
593 : : }
594 : :
595 : : static void
5427 heikki.linnakangas@i 596 : 1669 : ReleasePredXact(SERIALIZABLEXACT *sxact)
597 : : {
598 [ - + ]: 1669 : Assert(ShmemAddrIsValid(sxact));
599 : :
1063 andres@anarazel.de 600 : 1669 : dlist_delete(&sxact->xactLink);
601 : 1669 : dlist_push_tail(&PredXact->availableList, &sxact->xactLink);
5427 heikki.linnakangas@i 602 : 1669 : }
603 : :
604 : : /*------------------------------------------------------------------------*/
605 : :
606 : : /*
607 : : * These functions manage primitive access to the RWConflict pool and lists.
608 : : */
609 : : static bool
5292 tgl@sss.pgh.pa.us 610 : 2682 : RWConflictExists(const SERIALIZABLEXACT *reader, const SERIALIZABLEXACT *writer)
611 : : {
612 : : dlist_iter iter;
613 : :
5427 heikki.linnakangas@i 614 [ - + ]: 2682 : Assert(reader != writer);
615 : :
616 : : /* Check the ends of the purported conflict first. */
5299 617 [ + - ]: 2682 : if (SxactIsDoomed(reader)
618 [ + + ]: 2682 : || SxactIsDoomed(writer)
1063 andres@anarazel.de 619 [ + + ]: 2674 : || dlist_is_empty(&reader->outConflicts)
620 [ + + ]: 609 : || dlist_is_empty(&writer->inConflicts))
5427 heikki.linnakangas@i 621 : 2145 : 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 : : */
1063 andres@anarazel.de 629 [ + - + + ]: 561 : dlist_foreach(iter, &unconstify(SERIALIZABLEXACT *, reader)->outConflicts)
630 : : {
631 : 537 : RWConflict conflict =
943 tgl@sss.pgh.pa.us 632 :ECB (535) : dlist_container(RWConflictData, outLink, iter.cur);
633 : :
5427 heikki.linnakangas@i 634 [ + + ]:CBC 537 : if (conflict->sxactIn == writer)
635 : 513 : return true;
636 : : }
637 : :
638 : : /* No conflict found. */
639 : 24 : return false;
640 : : }
641 : :
642 : : static void
643 : 792 : SetRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
644 : : {
645 : : RWConflict conflict;
646 : :
647 [ - + ]: 792 : Assert(reader != writer);
648 [ - + ]: 792 : Assert(!RWConflictExists(reader, writer));
649 : :
1063 andres@anarazel.de 650 [ - + ]: 792 : if (dlist_is_empty(&RWConflictPool->availableList))
5427 heikki.linnakangas@i 651 [ # # ]:UBC 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 : :
1063 andres@anarazel.de 656 :CBC 792 : conflict = dlist_head_element(RWConflictData, outLink, &RWConflictPool->availableList);
657 : 792 : dlist_delete(&conflict->outLink);
658 : :
5427 heikki.linnakangas@i 659 : 792 : conflict->sxactOut = reader;
660 : 792 : conflict->sxactIn = writer;
1063 andres@anarazel.de 661 : 792 : dlist_push_tail(&reader->outConflicts, &conflict->outLink);
662 : 792 : dlist_push_tail(&writer->inConflicts, &conflict->inLink);
5427 heikki.linnakangas@i 663 : 792 : }
664 : :
665 : : static void
666 : 133 : SetPossibleUnsafeConflict(SERIALIZABLEXACT *roXact,
667 : : SERIALIZABLEXACT *activeXact)
668 : : {
669 : : RWConflict conflict;
670 : :
671 [ - + ]: 133 : Assert(roXact != activeXact);
672 [ - + ]: 133 : Assert(SxactIsReadOnly(roXact));
673 [ - + ]: 133 : Assert(!SxactIsReadOnly(activeXact));
674 : :
1063 andres@anarazel.de 675 [ - + ]: 133 : if (dlist_is_empty(&RWConflictPool->availableList))
5427 heikki.linnakangas@i 676 [ # # ]:UBC 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 : :
1063 andres@anarazel.de 681 :CBC 133 : conflict = dlist_head_element(RWConflictData, outLink, &RWConflictPool->availableList);
682 : 133 : dlist_delete(&conflict->outLink);
683 : :
5427 heikki.linnakangas@i 684 : 133 : conflict->sxactOut = activeXact;
685 : 133 : conflict->sxactIn = roXact;
1063 andres@anarazel.de 686 : 133 : dlist_push_tail(&activeXact->possibleUnsafeConflicts, &conflict->outLink);
687 : 133 : dlist_push_tail(&roXact->possibleUnsafeConflicts, &conflict->inLink);
5427 heikki.linnakangas@i 688 : 133 : }
689 : :
690 : : static void
691 : 925 : ReleaseRWConflict(RWConflict conflict)
692 : : {
1063 andres@anarazel.de 693 : 925 : dlist_delete(&conflict->inLink);
694 : 925 : dlist_delete(&conflict->outLink);
695 : 925 : dlist_push_tail(&RWConflictPool->availableList, &conflict->outLink);
5427 heikki.linnakangas@i 696 : 925 : }
697 : :
698 : : static void
699 : 3 : FlagSxactUnsafe(SERIALIZABLEXACT *sxact)
700 : : {
701 : : dlist_mutable_iter iter;
702 : :
703 [ - + ]: 3 : Assert(SxactIsReadOnly(sxact));
704 [ - + ]: 3 : Assert(!SxactIsROSafe(sxact));
705 : :
706 : 3 : 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 : : */
1063 andres@anarazel.de 712 [ + - + + ]: 6 : dlist_foreach_modify(iter, &sxact->possibleUnsafeConflicts)
713 : : {
714 : 3 : RWConflict conflict =
943 tgl@sss.pgh.pa.us 715 : 3 : dlist_container(RWConflictData, inLink, iter.cur);
716 : :
5427 heikki.linnakangas@i 717 [ - + ]: 3 : Assert(!SxactIsReadOnly(conflict->sxactOut));
718 [ - + ]: 3 : Assert(sxact == conflict->sxactIn);
719 : :
720 : 3 : ReleaseRWConflict(conflict);
721 : : }
722 : 3 : }
723 : :
724 : : /*------------------------------------------------------------------------*/
725 : :
726 : : /*
727 : : * Decide whether a Serial page number is "older" for truncation purposes.
728 : : * Analogous to CLOGPagePrecedes().
729 : : */
730 : : static bool
749 akorotkov@postgresql 731 : 43829 : SerialPagePrecedesLogically(int64 page1, int64 page2)
732 : : {
733 : : TransactionId xid1;
734 : : TransactionId xid2;
735 : :
1796 noah@leadboat.com 736 : 43829 : xid1 = ((TransactionId) page1) * SERIAL_ENTRIESPERPAGE;
737 : 43829 : xid1 += FirstNormalTransactionId + 1;
738 : 43829 : xid2 = ((TransactionId) page2) * SERIAL_ENTRIESPERPAGE;
739 : 43829 : xid2 += FirstNormalTransactionId + 1;
740 : :
741 [ + + + + ]: 73761 : return (TransactionIdPrecedes(xid1, xid2) &&
742 : 29932 : TransactionIdPrecedes(xid1, xid2 + SERIAL_ENTRIESPERPAGE - 1));
743 : : }
744 : :
745 : : #ifdef USE_ASSERT_CHECKING
746 : : static void
747 : 1069 : SerialPagePrecedesLogicallyUnitTests(void)
748 : : {
749 : 1069 : int per_page = SERIAL_ENTRIESPERPAGE,
750 : 1069 : 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 : 1069 : newestPage = 2 * SLRU_PAGES_PER_SEGMENT - 1; /* nothing special */
760 : 1069 : newestXact = newestPage * per_page + offset;
761 [ - + ]: 1069 : Assert(newestXact / per_page == newestPage);
762 : 1069 : oldestXact = newestXact + 1;
763 : 1069 : oldestXact -= 1U << 31;
764 : 1069 : 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 : 1069 : headPage = newestPage;
776 : 1069 : targetPage = oldestPage;
777 [ - + ]: 1069 : 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 : 1069 : headPage = oldestPage;
794 : 1069 : targetPage = newestPage;
795 [ - + ]: 1069 : Assert(SerialPagePrecedesLogically(headPage, targetPage - 1));
796 : : #if 0
797 : : Assert(SerialPagePrecedesLogically(headPage, targetPage));
798 : : #endif
5427 heikki.linnakangas@i 799 : 1069 : }
800 : : #endif
801 : :
802 : : /*
803 : : * Initialize for the tracking of old serializable committed xids.
804 : : */
805 : : static void
2042 tgl@sss.pgh.pa.us 806 : 1069 : SerialInit(void)
807 : : {
808 : : bool found;
809 : :
810 : : /*
811 : : * Set up SLRU management of the pg_serial data.
812 : : */
813 : 1069 : SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
658 alvherre@alvh.no-ip. 814 : 1069 : 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
1796 noah@leadboat.com 819 : 1069 : SerialPagePrecedesLogicallyUnitTests();
820 : : #endif
821 : 1069 : SlruPagePrecedesUnitTests(SerialSlruCtl, SERIAL_ENTRIESPERPAGE);
822 : :
823 : : /*
824 : : * Create or attach to the SerialControl structure.
825 : : */
2042 tgl@sss.pgh.pa.us 826 : 1069 : serialControl = (SerialControl)
827 : 1069 : ShmemInitStruct("SerialControlData", sizeof(SerialControlData), &found);
828 : :
3068 829 [ - + ]: 1069 : Assert(found == IsUnderPostmaster);
5427 heikki.linnakangas@i 830 [ + - ]: 1069 : if (!found)
831 : : {
832 : : /*
833 : : * Set control information to reflect empty SLRU.
834 : : */
687 alvherre@alvh.no-ip. 835 : 1069 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
2042 tgl@sss.pgh.pa.us 836 : 1069 : serialControl->headPage = -1;
837 : 1069 : serialControl->headXid = InvalidTransactionId;
838 : 1069 : serialControl->tailXid = InvalidTransactionId;
687 alvherre@alvh.no-ip. 839 : 1069 : LWLockRelease(SerialControlLock);
840 : : }
5427 heikki.linnakangas@i 841 : 1069 : }
842 : :
843 : : /*
844 : : * GUC check_hook for serializable_buffers
845 : : */
846 : : bool
658 alvherre@alvh.no-ip. 847 : 1107 : check_serial_buffers(int *newval, void **extra, GucSource source)
848 : : {
849 : 1107 : 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
2042 tgl@sss.pgh.pa.us 858 :UBC 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 : :
5427 heikki.linnakangas@i 867 [ # # ]: 0 : Assert(TransactionIdIsValid(xid));
868 : :
2042 tgl@sss.pgh.pa.us 869 : 0 : targetPage = SerialPage(xid);
658 alvherre@alvh.no-ip. 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 : : */
687 877 : 0 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
878 : :
879 : : /*
880 : : * If 'xid' is older than the global xmin (== tailXid), there's no need to
881 : : * store it, after all. This can happen if the oldest transaction holding
882 : : * back the global xmin just finished, making 'xid' uninteresting, but
883 : : * ClearOldPredicateLocks() has not yet run.
884 : : */
2042 tgl@sss.pgh.pa.us 885 : 0 : tailXid = serialControl->tailXid;
422 heikki.linnakangas@i 886 [ # # # # ]: 0 : if (!TransactionIdIsValid(tailXid) || TransactionIdPrecedes(xid, tailXid))
887 : : {
888 : 0 : LWLockRelease(SerialControlLock);
889 : 0 : return;
890 : : }
891 : :
892 : : /*
893 : : * If the SLRU is currently unused, zero out the whole active region from
894 : : * tailXid to headXid before taking it into use. Otherwise zero out only
895 : : * any new pages that enter the tailXid-headXid range as we advance
896 : : * headXid.
897 : : */
2042 tgl@sss.pgh.pa.us 898 [ # # ]: 0 : if (serialControl->headPage < 0)
899 : : {
900 : 0 : firstZeroPage = SerialPage(tailXid);
5427 heikki.linnakangas@i 901 : 0 : isNewPage = true;
902 : : }
903 : : else
904 : : {
2042 tgl@sss.pgh.pa.us 905 [ # # ]: 0 : firstZeroPage = SerialNextPage(serialControl->headPage);
906 : 0 : isNewPage = SerialPagePrecedesLogically(serialControl->headPage,
907 : : targetPage);
908 : : }
909 : :
910 [ # # ]: 0 : if (!TransactionIdIsValid(serialControl->headXid)
911 [ # # ]: 0 : || TransactionIdFollows(xid, serialControl->headXid))
912 : 0 : serialControl->headXid = xid;
5398 heikki.linnakangas@i 913 [ # # ]: 0 : if (isNewPage)
2042 tgl@sss.pgh.pa.us 914 : 0 : serialControl->headPage = targetPage;
915 : :
5427 heikki.linnakangas@i 916 [ # # ]: 0 : if (isNewPage)
917 : : {
918 : : /* Initialize intervening pages; might involve trading locks */
919 : : for (;;)
920 : : {
623 alvherre@alvh.no-ip. 921 : 0 : lock = SimpleLruGetBankLock(SerialSlruCtl, firstZeroPage);
922 : 0 : LWLockAcquire(lock, LW_EXCLUSIVE);
923 : 0 : slotno = SimpleLruZeroPage(SerialSlruCtl, firstZeroPage);
924 [ # # ]: 0 : if (firstZeroPage == targetPage)
925 : 0 : break;
2042 tgl@sss.pgh.pa.us 926 [ # # ]: 0 : firstZeroPage = SerialNextPage(firstZeroPage);
623 alvherre@alvh.no-ip. 927 : 0 : LWLockRelease(lock);
928 : : }
929 : : }
930 : : else
931 : : {
932 : 0 : LWLockAcquire(lock, LW_EXCLUSIVE);
2042 tgl@sss.pgh.pa.us 933 : 0 : slotno = SimpleLruReadPage(SerialSlruCtl, targetPage, true, xid);
934 : : }
935 : :
936 : 0 : SerialValue(slotno, xid) = minConflictCommitSeqNo;
937 : 0 : SerialSlruCtl->shared->page_dirty[slotno] = true;
938 : :
658 alvherre@alvh.no-ip. 939 : 0 : LWLockRelease(lock);
687 940 : 0 : LWLockRelease(SerialControlLock);
941 : : }
942 : :
943 : : /*
944 : : * Get the minimum commitSeqNo for any conflict out for the given xid. For
945 : : * a transaction which exists but has no conflict out, InvalidSerCommitSeqNo
946 : : * will be returned.
947 : : */
948 : : static SerCommitSeqNo
2042 tgl@sss.pgh.pa.us 949 :CBC 25 : SerialGetMinConflictCommitSeqNo(TransactionId xid)
950 : : {
951 : : TransactionId headXid;
952 : : TransactionId tailXid;
953 : : SerCommitSeqNo val;
954 : : int slotno;
955 : :
5427 heikki.linnakangas@i 956 [ - + ]: 25 : Assert(TransactionIdIsValid(xid));
957 : :
687 alvherre@alvh.no-ip. 958 : 25 : LWLockAcquire(SerialControlLock, LW_SHARED);
2042 tgl@sss.pgh.pa.us 959 : 25 : headXid = serialControl->headXid;
960 : 25 : tailXid = serialControl->tailXid;
687 alvherre@alvh.no-ip. 961 : 25 : LWLockRelease(SerialControlLock);
962 : :
5427 heikki.linnakangas@i 963 [ + - ]: 25 : if (!TransactionIdIsValid(headXid))
964 : 25 : return 0;
965 : :
5427 heikki.linnakangas@i 966 [ # # ]:UBC 0 : Assert(TransactionIdIsValid(tailXid));
967 : :
968 [ # # ]: 0 : if (TransactionIdPrecedes(xid, tailXid)
969 [ # # ]: 0 : || TransactionIdFollows(xid, headXid))
970 : 0 : return 0;
971 : :
972 : : /*
973 : : * The following function must be called without holding SLRU bank lock,
974 : : * but will return with that lock held, which must then be released.
975 : : */
2042 tgl@sss.pgh.pa.us 976 : 0 : slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
977 : : SerialPage(xid), xid);
978 : 0 : val = SerialValue(slotno, xid);
658 alvherre@alvh.no-ip. 979 : 0 : LWLockRelease(SimpleLruGetBankLock(SerialSlruCtl, SerialPage(xid)));
5427 heikki.linnakangas@i 980 : 0 : return val;
981 : : }
982 : :
983 : : /*
984 : : * Call this whenever there is a new xmin for active serializable
985 : : * transactions. We don't need to keep information on transactions which
986 : : * precede that. InvalidTransactionId means none active, so everything in
987 : : * the SLRU can be discarded.
988 : : */
989 : : static void
2042 tgl@sss.pgh.pa.us 990 :CBC 1717 : SerialSetActiveSerXmin(TransactionId xid)
991 : : {
687 alvherre@alvh.no-ip. 992 : 1717 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
993 : :
994 : : /*
995 : : * When no sxacts are active, nothing overlaps, set the xid values to
996 : : * invalid to show that there are no valid entries. Don't clear headPage,
997 : : * though. A new xmin might still land on that page, and we don't want to
998 : : * repeatedly zero out the same page.
999 : : */
5427 heikki.linnakangas@i 1000 [ + + ]: 1717 : if (!TransactionIdIsValid(xid))
1001 : : {
2042 tgl@sss.pgh.pa.us 1002 : 849 : serialControl->tailXid = InvalidTransactionId;
1003 : 849 : serialControl->headXid = InvalidTransactionId;
687 alvherre@alvh.no-ip. 1004 : 849 : LWLockRelease(SerialControlLock);
5427 heikki.linnakangas@i 1005 : 849 : return;
1006 : : }
1007 : :
1008 : : /*
1009 : : * When we're recovering prepared transactions, the global xmin might move
1010 : : * backwards depending on the order they're recovered. Normally that's not
1011 : : * OK, but during recovery no serializable transactions will commit, so
1012 : : * the SLRU is empty and we can get away with it.
1013 : : */
1014 [ - + ]: 868 : if (RecoveryInProgress())
1015 : : {
2042 tgl@sss.pgh.pa.us 1016 [ # # ]:UBC 0 : Assert(serialControl->headPage < 0);
1017 [ # # ]: 0 : if (!TransactionIdIsValid(serialControl->tailXid)
1018 [ # # ]: 0 : || TransactionIdPrecedes(xid, serialControl->tailXid))
1019 : : {
1020 : 0 : serialControl->tailXid = xid;
1021 : : }
687 alvherre@alvh.no-ip. 1022 : 0 : LWLockRelease(SerialControlLock);
5427 heikki.linnakangas@i 1023 : 0 : return;
1024 : : }
1025 : :
2042 tgl@sss.pgh.pa.us 1026 [ + + - + ]:CBC 868 : Assert(!TransactionIdIsValid(serialControl->tailXid)
1027 : : || TransactionIdFollows(xid, serialControl->tailXid));
1028 : :
1029 : 868 : serialControl->tailXid = xid;
1030 : :
687 alvherre@alvh.no-ip. 1031 : 868 : LWLockRelease(SerialControlLock);
1032 : : }
1033 : :
1034 : : /*
1035 : : * Perform a checkpoint --- either during shutdown, or on-the-fly
1036 : : *
1037 : : * We don't have any data that needs to survive a restart, but this is a
1038 : : * convenient place to truncate the SLRU.
1039 : : */
1040 : : void
5398 heikki.linnakangas@i 1041 : 1729 : CheckPointPredicate(void)
1042 : : {
1043 : : int64 truncateCutoffPage;
1044 : :
687 alvherre@alvh.no-ip. 1045 : 1729 : LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
1046 : :
1047 : : /* Exit quickly if the SLRU is currently not in use. */
2042 tgl@sss.pgh.pa.us 1048 [ + - ]: 1729 : if (serialControl->headPage < 0)
1049 : : {
687 alvherre@alvh.no-ip. 1050 : 1729 : LWLockRelease(SerialControlLock);
5427 heikki.linnakangas@i 1051 : 1729 : return;
1052 : : }
1053 : :
2042 tgl@sss.pgh.pa.us 1054 [ # # ]:UBC 0 : if (TransactionIdIsValid(serialControl->tailXid))
1055 : : {
1056 : : int64 tailPage;
1057 : :
1058 : 0 : tailPage = SerialPage(serialControl->tailXid);
1059 : :
1060 : : /*
1061 : : * It is possible for the tailXid to be ahead of the headXid. This
1062 : : * occurs if we checkpoint while there are in-progress serializable
1063 : : * transaction(s) advancing the tail but we are yet to summarize the
1064 : : * transactions. In this case, we cutoff up to the headPage and the
1065 : : * next summary will advance the headXid.
1066 : : */
792 michael@paquier.xyz 1067 [ # # ]: 0 : if (SerialPagePrecedesLogically(tailPage, serialControl->headPage))
1068 : : {
1069 : : /* We can truncate the SLRU up to the page containing tailXid */
1070 : 0 : truncateCutoffPage = tailPage;
1071 : : }
1072 : : else
1073 : 0 : truncateCutoffPage = serialControl->headPage;
1074 : : }
1075 : : else
1076 : : {
1077 : : /*----------
1078 : : * The SLRU is no longer needed. Truncate to head before we set head
1079 : : * invalid.
1080 : : *
1081 : : * XXX: It's possible that the SLRU is not needed again until XID
1082 : : * wrap-around has happened, so that the segment containing headPage
1083 : : * that we leave behind will appear to be new again. In that case it
1084 : : * won't be removed until XID horizon advances enough to make it
1085 : : * current again.
1086 : : *
1087 : : * XXX: This should happen in vac_truncate_clog(), not in checkpoints.
1088 : : * Consider this scenario, starting from a system with no in-progress
1089 : : * transactions and VACUUM FREEZE having maximized oldestXact:
1090 : : * - Start a SERIALIZABLE transaction.
1091 : : * - Start, finish, and summarize a SERIALIZABLE transaction, creating
1092 : : * one SLRU page.
1093 : : * - Consume XIDs to reach xidStopLimit.
1094 : : * - Finish all transactions. Due to the long-running SERIALIZABLE
1095 : : * transaction, earlier checkpoints did not touch headPage. The
1096 : : * next checkpoint will change it, but that checkpoint happens after
1097 : : * the end of the scenario.
1098 : : * - VACUUM to advance XID limits.
1099 : : * - Consume ~2M XIDs, crossing the former xidWrapLimit.
1100 : : * - Start, finish, and summarize a SERIALIZABLE transaction.
1101 : : * SerialAdd() declines to create the targetPage, because headPage
1102 : : * is not regarded as in the past relative to that targetPage. The
1103 : : * transaction instigating the summarize fails in
1104 : : * SimpleLruReadPage().
1105 : : */
1106 : 0 : truncateCutoffPage = serialControl->headPage;
2042 tgl@sss.pgh.pa.us 1107 : 0 : serialControl->headPage = -1;
1108 : : }
1109 : :
687 alvherre@alvh.no-ip. 1110 : 0 : LWLockRelease(SerialControlLock);
1111 : :
1112 : : /*
1113 : : * Truncate away pages that are no longer required. Note that no
1114 : : * additional locking is required, because this is only called as part of
1115 : : * a checkpoint, and the validity limits have already been determined.
1116 : : */
792 michael@paquier.xyz 1117 : 0 : SimpleLruTruncate(SerialSlruCtl, truncateCutoffPage);
1118 : :
1119 : : /*
1120 : : * Write dirty SLRU pages to disk
1121 : : *
1122 : : * This is not actually necessary from a correctness point of view. We do
1123 : : * it merely as a debugging aid.
1124 : : *
1125 : : * We're doing this after the truncation to avoid writing pages right
1126 : : * before deleting the file in which they sit, which would be completely
1127 : : * pointless.
1128 : : */
1909 tmunro@postgresql.or 1129 : 0 : SimpleLruWriteAll(SerialSlruCtl, true);
1130 : : }
1131 : :
1132 : : /*------------------------------------------------------------------------*/
1133 : :
1134 : : /*
1135 : : * PredicateLockShmemInit -- Initialize the predicate locking data structures.
1136 : : *
1137 : : * This is called from CreateSharedMemoryAndSemaphores(), which see for
1138 : : * more comments. In the normal postmaster case, the shared hash tables
1139 : : * are created here. Backends inherit the pointers
1140 : : * to the shared tables via fork(). In the EXEC_BACKEND case, each
1141 : : * backend re-executes this code to obtain pointers to the already existing
1142 : : * shared hash tables.
1143 : : */
1144 : : void
475 heikki.linnakangas@i 1145 :CBC 1069 : PredicateLockShmemInit(void)
1146 : : {
1147 : : HASHCTL info;
1148 : : int64 max_table_size;
1149 : : Size requestSize;
1150 : : bool found;
1151 : :
1152 : : #ifndef EXEC_BACKEND
3068 tgl@sss.pgh.pa.us 1153 [ - + ]: 1069 : Assert(!IsUnderPostmaster);
1154 : : #endif
1155 : :
1156 : : /*
1157 : : * Compute size of predicate lock target hashtable. Note these
1158 : : * calculations must agree with PredicateLockShmemSize!
1159 : : */
5427 heikki.linnakangas@i 1160 : 1069 : max_table_size = NPREDICATELOCKTARGETENTS();
1161 : :
1162 : : /*
1163 : : * Allocate hash table for PREDICATELOCKTARGET structs. This stores
1164 : : * per-predicate-lock-target information.
1165 : : */
1166 : 1069 : info.keysize = sizeof(PREDICATELOCKTARGETTAG);
1167 : 1069 : info.entrysize = sizeof(PREDICATELOCKTARGET);
1168 : 1069 : info.num_partitions = NUM_PREDICATELOCK_PARTITIONS;
1169 : :
1170 : 1069 : PredicateLockTargetHash = ShmemInitHash("PREDICATELOCKTARGET hash",
1171 : : max_table_size,
1172 : : max_table_size,
1173 : : &info,
1174 : : HASH_ELEM | HASH_BLOBS |
1175 : : HASH_PARTITION | HASH_FIXED_SIZE);
1176 : :
1177 : : /*
1178 : : * Reserve a dummy entry in the hash table; we use it to make sure there's
1179 : : * always one entry available when we need to split or combine a page,
1180 : : * because running out of space there could mean aborting a
1181 : : * non-serializable transaction.
1182 : : */
3068 tgl@sss.pgh.pa.us 1183 [ + - ]: 1069 : if (!IsUnderPostmaster)
1184 : : {
1185 : 1069 : (void) hash_search(PredicateLockTargetHash, &ScratchTargetTag,
1186 : : HASH_ENTER, &found);
1187 [ - + ]: 1069 : Assert(!found);
1188 : : }
1189 : :
1190 : : /* Pre-calculate the hash and partition lock of the scratch entry */
1191 : 1069 : ScratchTargetTagHash = PredicateLockTargetTagHashCode(&ScratchTargetTag);
1192 : 1069 : ScratchPartitionLock = PredicateLockHashPartitionLock(ScratchTargetTagHash);
1193 : :
1194 : : /*
1195 : : * Allocate hash table for PREDICATELOCK structs. This stores per
1196 : : * xact-lock-of-a-target information.
1197 : : */
5427 heikki.linnakangas@i 1198 : 1069 : info.keysize = sizeof(PREDICATELOCKTAG);
1199 : 1069 : info.entrysize = sizeof(PREDICATELOCK);
1200 : 1069 : info.hash = predicatelock_hash;
1201 : 1069 : info.num_partitions = NUM_PREDICATELOCK_PARTITIONS;
1202 : :
1203 : : /* Assume an average of 2 xacts per target */
3068 tgl@sss.pgh.pa.us 1204 : 1069 : max_table_size *= 2;
1205 : :
5427 heikki.linnakangas@i 1206 : 1069 : PredicateLockHash = ShmemInitHash("PREDICATELOCK hash",
1207 : : max_table_size,
1208 : : max_table_size,
1209 : : &info,
1210 : : HASH_ELEM | HASH_FUNCTION |
1211 : : HASH_PARTITION | HASH_FIXED_SIZE);
1212 : :
1213 : : /*
1214 : : * Compute size for serializable transaction hashtable. Note these
1215 : : * calculations must agree with PredicateLockShmemSize!
1216 : : */
1345 rhaas@postgresql.org 1217 : 1069 : max_table_size = (MaxBackends + max_prepared_xacts);
1218 : :
1219 : : /*
1220 : : * Allocate a list to hold information on transactions participating in
1221 : : * predicate locking.
1222 : : *
1223 : : * Assume an average of 10 predicate locking transactions per backend.
1224 : : * This allows aggressive cleanup while detail is present before data must
1225 : : * be summarized for storage in SLRU and the "dummy" transaction.
1226 : : */
5427 heikki.linnakangas@i 1227 : 1069 : max_table_size *= 10;
1228 : :
259 tomas.vondra@postgre 1229 : 1069 : requestSize = add_size(PredXactListDataSize,
1230 : : (mul_size((Size) max_table_size,
1231 : : sizeof(SERIALIZABLEXACT))));
1232 : :
5427 heikki.linnakangas@i 1233 : 1069 : PredXact = ShmemInitStruct("PredXactList",
1234 : : requestSize,
1235 : : &found);
3068 tgl@sss.pgh.pa.us 1236 [ - + ]: 1069 : Assert(found == IsUnderPostmaster);
5427 heikki.linnakangas@i 1237 [ + - ]: 1069 : if (!found)
1238 : : {
1239 : : int i;
1240 : :
1241 : : /* clean everything, both the header and the element */
259 tomas.vondra@postgre 1242 : 1069 : memset(PredXact, 0, requestSize);
1243 : :
1063 andres@anarazel.de 1244 : 1069 : dlist_init(&PredXact->availableList);
1245 : 1069 : dlist_init(&PredXact->activeList);
5427 heikki.linnakangas@i 1246 : 1069 : PredXact->SxactGlobalXmin = InvalidTransactionId;
1247 : 1069 : PredXact->SxactGlobalXminCount = 0;
1248 : 1069 : PredXact->WritableSxactCount = 0;
1249 : 1069 : PredXact->LastSxactCommitSeqNo = FirstNormalSerCommitSeqNo - 1;
1250 : 1069 : PredXact->CanPartialClearThrough = 0;
1251 : 1069 : PredXact->HavePartialClearedThrough = 0;
259 tomas.vondra@postgre 1252 : 1069 : PredXact->element
1253 : 1069 : = (SERIALIZABLEXACT *) ((char *) PredXact + PredXactListDataSize);
1254 : : /* Add all elements to available list, clean. */
5427 heikki.linnakangas@i 1255 [ + + ]: 998609 : for (i = 0; i < max_table_size; i++)
1256 : : {
1063 andres@anarazel.de 1257 : 997540 : LWLockInitialize(&PredXact->element[i].perXactPredicateListLock,
1258 : : LWTRANCHE_PER_XACT_PREDICATE_LIST);
1259 : 997540 : dlist_push_tail(&PredXact->availableList, &PredXact->element[i].xactLink);
1260 : : }
5427 heikki.linnakangas@i 1261 : 1069 : PredXact->OldCommittedSxact = CreatePredXact();
1262 : 1069 : SetInvalidVirtualTransactionId(PredXact->OldCommittedSxact->vxid);
5277 1263 : 1069 : PredXact->OldCommittedSxact->prepareSeqNo = 0;
5427 1264 : 1069 : PredXact->OldCommittedSxact->commitSeqNo = 0;
1265 : 1069 : PredXact->OldCommittedSxact->SeqNo.lastCommitBeforeSnapshot = 0;
1063 andres@anarazel.de 1266 : 1069 : dlist_init(&PredXact->OldCommittedSxact->outConflicts);
1267 : 1069 : dlist_init(&PredXact->OldCommittedSxact->inConflicts);
1268 : 1069 : dlist_init(&PredXact->OldCommittedSxact->predicateLocks);
1269 : 1069 : dlist_node_init(&PredXact->OldCommittedSxact->finishedLink);
1270 : 1069 : dlist_init(&PredXact->OldCommittedSxact->possibleUnsafeConflicts);
5427 heikki.linnakangas@i 1271 : 1069 : PredXact->OldCommittedSxact->topXid = InvalidTransactionId;
1272 : 1069 : PredXact->OldCommittedSxact->finishedBefore = InvalidTransactionId;
1273 : 1069 : PredXact->OldCommittedSxact->xmin = InvalidTransactionId;
1274 : 1069 : PredXact->OldCommittedSxact->flags = SXACT_FLAG_COMMITTED;
1275 : 1069 : PredXact->OldCommittedSxact->pid = 0;
654 1276 : 1069 : PredXact->OldCommittedSxact->pgprocno = INVALID_PROC_NUMBER;
1277 : : }
1278 : : /* This never changes, so let's keep a local copy. */
5427 1279 : 1069 : OldCommittedSxact = PredXact->OldCommittedSxact;
1280 : :
1281 : : /*
1282 : : * Allocate hash table for SERIALIZABLEXID structs. This stores per-xid
1283 : : * information for serializable transactions which have accessed data.
1284 : : */
1285 : 1069 : info.keysize = sizeof(SERIALIZABLEXIDTAG);
1286 : 1069 : info.entrysize = sizeof(SERIALIZABLEXID);
1287 : :
1288 : 1069 : SerializableXidHash = ShmemInitHash("SERIALIZABLEXID hash",
1289 : : max_table_size,
1290 : : max_table_size,
1291 : : &info,
1292 : : HASH_ELEM | HASH_BLOBS |
1293 : : HASH_FIXED_SIZE);
1294 : :
1295 : : /*
1296 : : * Allocate space for tracking rw-conflicts in lists attached to the
1297 : : * transactions.
1298 : : *
1299 : : * Assume an average of 5 conflicts per transaction. Calculations suggest
1300 : : * that this will prevent resource exhaustion in even the most pessimal
1301 : : * loads up to max_connections = 200 with all 200 connections pounding the
1302 : : * database with serializable transactions. Beyond that, there may be
1303 : : * occasional transactions canceled when trying to flag conflicts. That's
1304 : : * probably OK.
1305 : : */
1306 : 1069 : max_table_size *= 5;
1307 : :
259 tomas.vondra@postgre 1308 : 1069 : requestSize = RWConflictPoolHeaderDataSize +
1309 : 1069 : mul_size((Size) max_table_size,
1310 : : RWConflictDataSize);
1311 : :
5427 heikki.linnakangas@i 1312 : 1069 : RWConflictPool = ShmemInitStruct("RWConflictPool",
1313 : : requestSize,
1314 : : &found);
3068 tgl@sss.pgh.pa.us 1315 [ - + ]: 1069 : Assert(found == IsUnderPostmaster);
5427 heikki.linnakangas@i 1316 [ + - ]: 1069 : if (!found)
1317 : : {
1318 : : int i;
1319 : :
1320 : : /* clean everything, including the elements */
259 tomas.vondra@postgre 1321 : 1069 : memset(RWConflictPool, 0, requestSize);
1322 : :
1063 andres@anarazel.de 1323 : 1069 : dlist_init(&RWConflictPool->availableList);
259 tomas.vondra@postgre 1324 : 1069 : RWConflictPool->element = (RWConflict) ((char *) RWConflictPool +
1325 : : RWConflictPoolHeaderDataSize);
1326 : : /* Add all elements to available list, clean. */
5427 heikki.linnakangas@i 1327 [ + + ]: 4988769 : for (i = 0; i < max_table_size; i++)
1328 : : {
1063 andres@anarazel.de 1329 : 4987700 : dlist_push_tail(&RWConflictPool->availableList,
1330 : 4987700 : &RWConflictPool->element[i].outLink);
1331 : : }
1332 : : }
1333 : :
1334 : : /*
1335 : : * Create or attach to the header for the list of finished serializable
1336 : : * transactions.
1337 : : */
1338 : 1069 : FinishedSerializableTransactions = (dlist_head *)
5427 heikki.linnakangas@i 1339 : 1069 : ShmemInitStruct("FinishedSerializableTransactions",
1340 : : sizeof(dlist_head),
1341 : : &found);
3068 tgl@sss.pgh.pa.us 1342 [ - + ]: 1069 : Assert(found == IsUnderPostmaster);
5427 heikki.linnakangas@i 1343 [ + - ]: 1069 : if (!found)
1063 andres@anarazel.de 1344 : 1069 : dlist_init(FinishedSerializableTransactions);
1345 : :
1346 : : /*
1347 : : * Initialize the SLRU storage for old committed serializable
1348 : : * transactions.
1349 : : */
2042 tgl@sss.pgh.pa.us 1350 : 1069 : SerialInit();
5427 heikki.linnakangas@i 1351 : 1069 : }
1352 : :
1353 : : /*
1354 : : * Estimate shared-memory space used for predicate lock table
1355 : : */
1356 : : Size
1357 : 1986 : PredicateLockShmemSize(void)
1358 : : {
1359 : 1986 : Size size = 0;
1360 : : long max_table_size;
1361 : :
1362 : : /* predicate lock target hash table */
1363 : 1986 : max_table_size = NPREDICATELOCKTARGETENTS();
1364 : 1986 : size = add_size(size, hash_estimate_size(max_table_size,
1365 : : sizeof(PREDICATELOCKTARGET)));
1366 : :
1367 : : /* predicate lock hash table */
1368 : 1986 : max_table_size *= 2;
1369 : 1986 : size = add_size(size, hash_estimate_size(max_table_size,
1370 : : sizeof(PREDICATELOCK)));
1371 : :
1372 : : /*
1373 : : * Since NPREDICATELOCKTARGETENTS is only an estimate, add 10% safety
1374 : : * margin.
1375 : : */
1376 : 1986 : size = add_size(size, size / 10);
1377 : :
1378 : : /* transaction list */
1345 rhaas@postgresql.org 1379 : 1986 : max_table_size = MaxBackends + max_prepared_xacts;
5427 heikki.linnakangas@i 1380 : 1986 : max_table_size *= 10;
1381 : 1986 : size = add_size(size, PredXactListDataSize);
1382 : 1986 : size = add_size(size, mul_size((Size) max_table_size,
1383 : : sizeof(SERIALIZABLEXACT)));
1384 : :
1385 : : /* transaction xid table */
1386 : 1986 : size = add_size(size, hash_estimate_size(max_table_size,
1387 : : sizeof(SERIALIZABLEXID)));
1388 : :
1389 : : /* rw-conflict pool */
5425 1390 : 1986 : max_table_size *= 5;
1391 : 1986 : size = add_size(size, RWConflictPoolHeaderDataSize);
1392 : 1986 : size = add_size(size, mul_size((Size) max_table_size,
1393 : : RWConflictDataSize));
1394 : :
1395 : : /* Head for list of finished serializable transactions. */
1063 andres@anarazel.de 1396 : 1986 : size = add_size(size, sizeof(dlist_head));
1397 : :
1398 : : /* Shared memory structures for SLRU tracking of old committed xids. */
2042 tgl@sss.pgh.pa.us 1399 : 1986 : size = add_size(size, sizeof(SerialControlData));
658 alvherre@alvh.no-ip. 1400 : 1986 : size = add_size(size, SimpleLruShmemSize(serializable_buffers, 0));
1401 : :
5427 heikki.linnakangas@i 1402 : 1986 : return size;
1403 : : }
1404 : :
1405 : :
1406 : : /*
1407 : : * Compute the hash code associated with a PREDICATELOCKTAG.
1408 : : *
1409 : : * Because we want to use just one set of partition locks for both the
1410 : : * PREDICATELOCKTARGET and PREDICATELOCK hash tables, we have to make sure
1411 : : * that PREDICATELOCKs fall into the same partition number as their
1412 : : * associated PREDICATELOCKTARGETs. dynahash.c expects the partition number
1413 : : * to be the low-order bits of the hash code, and therefore a
1414 : : * PREDICATELOCKTAG's hash code must have the same low-order bits as the
1415 : : * associated PREDICATELOCKTARGETTAG's hash code. We achieve this with this
1416 : : * specialized hash function.
1417 : : */
1418 : : static uint32
5427 heikki.linnakangas@i 1419 :UBC 0 : predicatelock_hash(const void *key, Size keysize)
1420 : : {
5292 tgl@sss.pgh.pa.us 1421 : 0 : const PREDICATELOCKTAG *predicatelocktag = (const PREDICATELOCKTAG *) key;
1422 : : uint32 targethash;
1423 : :
5427 heikki.linnakangas@i 1424 [ # # ]: 0 : Assert(keysize == sizeof(PREDICATELOCKTAG));
1425 : :
1426 : : /* Look into the associated target object, and compute its hash code */
1427 : 0 : targethash = PredicateLockTargetTagHashCode(&predicatelocktag->myTarget->tag);
1428 : :
1429 : 0 : return PredicateLockHashCodeFromTargetHashCode(predicatelocktag, targethash);
1430 : : }
1431 : :
1432 : :
1433 : : /*
1434 : : * GetPredicateLockStatusData
1435 : : * Return a table containing the internal state of the predicate
1436 : : * lock manager for use in pg_lock_status.
1437 : : *
1438 : : * Like GetLockStatusData, this function tries to hold the partition LWLocks
1439 : : * for as short a time as possible by returning two arrays that simply
1440 : : * contain the PREDICATELOCKTARGETTAG and SERIALIZABLEXACT for each lock
1441 : : * table entry. Multiple copies of the same PREDICATELOCKTARGETTAG and
1442 : : * SERIALIZABLEXACT will likely appear.
1443 : : */
1444 : : PredicateLockData *
5427 heikki.linnakangas@i 1445 :CBC 219 : GetPredicateLockStatusData(void)
1446 : : {
1447 : : PredicateLockData *data;
1448 : : int i;
1449 : : int els,
1450 : : el;
1451 : : HASH_SEQ_STATUS seqstat;
1452 : : PREDICATELOCK *predlock;
1453 : :
7 michael@paquier.xyz 1454 :GNC 219 : data = palloc_object(PredicateLockData);
1455 : :
1456 : : /*
1457 : : * To ensure consistency, take simultaneous locks on all partition locks
1458 : : * in ascending order, then SerializableXactHashLock.
1459 : : */
5427 heikki.linnakangas@i 1460 [ + + ]:CBC 3723 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
4342 rhaas@postgresql.org 1461 : 3504 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_SHARED);
5427 heikki.linnakangas@i 1462 : 219 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1463 : :
1464 : : /* Get number of locks and allocate appropriately-sized arrays. */
1465 : 219 : els = hash_get_num_entries(PredicateLockHash);
1466 : 219 : data->nelements = els;
7 michael@paquier.xyz 1467 :GNC 219 : data->locktags = palloc_array(PREDICATELOCKTARGETTAG, els);
1468 : 219 : data->xacts = palloc_array(SERIALIZABLEXACT, els);
1469 : :
1470 : :
1471 : : /* Scan through PredicateLockHash and copy contents */
5427 heikki.linnakangas@i 1472 :CBC 219 : hash_seq_init(&seqstat, PredicateLockHash);
1473 : :
1474 : 219 : el = 0;
1475 : :
1476 [ + + ]: 222 : while ((predlock = (PREDICATELOCK *) hash_seq_search(&seqstat)))
1477 : : {
1478 : 3 : data->locktags[el] = predlock->tag.myTarget->tag;
1479 : 3 : data->xacts[el] = *predlock->tag.myXact;
1480 : 3 : el++;
1481 : : }
1482 : :
1483 [ - + ]: 219 : Assert(el == els);
1484 : :
1485 : : /* Release locks in reverse order */
1486 : 219 : LWLockRelease(SerializableXactHashLock);
1487 [ + + ]: 3723 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
4342 rhaas@postgresql.org 1488 : 3504 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
1489 : :
5427 heikki.linnakangas@i 1490 : 219 : return data;
1491 : : }
1492 : :
1493 : : /*
1494 : : * Free up shared memory structures by pushing the oldest sxact (the one at
1495 : : * the front of the SummarizeOldestCommittedSxact queue) into summary form.
1496 : : * Each call will free exactly one SERIALIZABLEXACT structure and may also
1497 : : * free one or more of these structures: SERIALIZABLEXID, PREDICATELOCK,
1498 : : * PREDICATELOCKTARGET, RWConflictData.
1499 : : */
1500 : : static void
5427 heikki.linnakangas@i 1501 :UBC 0 : SummarizeOldestCommittedSxact(void)
1502 : : {
1503 : : SERIALIZABLEXACT *sxact;
1504 : :
1505 : 0 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
1506 : :
1507 : : /*
1508 : : * This function is only called if there are no sxact slots available.
1509 : : * Some of them must belong to old, already-finished transactions, so
1510 : : * there should be something in FinishedSerializableTransactions list that
1511 : : * we can summarize. However, there's a race condition: while we were not
1512 : : * holding any locks, a transaction might have ended and cleaned up all
1513 : : * the finished sxact entries already, freeing up their sxact slots. In
1514 : : * that case, we have nothing to do here. The caller will find one of the
1515 : : * slots released by the other backend when it retries.
1516 : : */
1063 andres@anarazel.de 1517 [ # # ]: 0 : if (dlist_is_empty(FinishedSerializableTransactions))
1518 : : {
5427 heikki.linnakangas@i 1519 : 0 : LWLockRelease(SerializableFinishedListLock);
1520 : 0 : return;
1521 : : }
1522 : :
1523 : : /*
1524 : : * Grab the first sxact off the finished list -- this will be the earliest
1525 : : * commit. Remove it from the list.
1526 : : */
1063 andres@anarazel.de 1527 : 0 : sxact = dlist_head_element(SERIALIZABLEXACT, finishedLink,
1528 : : FinishedSerializableTransactions);
1529 : 0 : dlist_delete_thoroughly(&sxact->finishedLink);
1530 : :
1531 : : /* Add to SLRU summary information. */
5427 heikki.linnakangas@i 1532 [ # # # # ]: 0 : if (TransactionIdIsValid(sxact->topXid) && !SxactIsReadOnly(sxact))
2042 tgl@sss.pgh.pa.us 1533 [ # # ]: 0 : SerialAdd(sxact->topXid, SxactHasConflictOut(sxact)
1534 : : ? sxact->SeqNo.earliestOutConflictCommit : InvalidSerCommitSeqNo);
1535 : :
1536 : : /* Summarize and release the detail. */
5427 heikki.linnakangas@i 1537 : 0 : ReleaseOneSerializableXact(sxact, false, true);
1538 : :
1539 : 0 : LWLockRelease(SerializableFinishedListLock);
1540 : : }
1541 : :
1542 : : /*
1543 : : * GetSafeSnapshot
1544 : : * Obtain and register a snapshot for a READ ONLY DEFERRABLE
1545 : : * transaction. Ensures that the snapshot is "safe", i.e. a
1546 : : * read-only transaction running on it can execute serializably
1547 : : * without further checks. This requires waiting for concurrent
1548 : : * transactions to complete, and retrying with a new snapshot if
1549 : : * one of them could possibly create a conflict.
1550 : : *
1551 : : * As with GetSerializableTransactionSnapshot (which this is a subroutine
1552 : : * for), the passed-in Snapshot pointer should reference a static data
1553 : : * area that can safely be passed to GetSnapshotData.
1554 : : */
1555 : : static Snapshot
5427 heikki.linnakangas@i 1556 :CBC 4 : GetSafeSnapshot(Snapshot origSnapshot)
1557 : : {
1558 : : Snapshot snapshot;
1559 : :
1560 [ + - + - ]: 4 : Assert(XactReadOnly && XactDeferrable);
1561 : :
1562 : : while (true)
1563 : : {
1564 : : /*
1565 : : * GetSerializableTransactionSnapshotInt is going to call
1566 : : * GetSnapshotData, so we need to provide it the static snapshot area
1567 : : * our caller passed to us. The pointer returned is actually the same
1568 : : * one passed to it, but we avoid assuming that here.
1569 : : */
5170 tgl@sss.pgh.pa.us 1570 : 5 : snapshot = GetSerializableTransactionSnapshotInt(origSnapshot,
1571 : : NULL, InvalidPid);
1572 : :
5427 heikki.linnakangas@i 1573 [ + + ]: 5 : if (MySerializableXact == InvalidSerializableXact)
1574 : 3 : return snapshot; /* no concurrent r/w xacts; it's safe */
1575 : :
5304 1576 : 2 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1577 : :
1578 : : /*
1579 : : * Wait for concurrent transactions to finish. Stop early if one of
1580 : : * them marked us as conflicted.
1581 : : */
1582 : 2 : MySerializableXact->flags |= SXACT_FLAG_DEFERRABLE_WAITING;
1063 andres@anarazel.de 1583 [ + + ]: 4 : while (!(dlist_is_empty(&MySerializableXact->possibleUnsafeConflicts) ||
5427 heikki.linnakangas@i 1584 [ + - ]: 2 : SxactIsROUnsafe(MySerializableXact)))
1585 : : {
5304 1586 : 2 : LWLockRelease(SerializableXactHashLock);
3361 rhaas@postgresql.org 1587 : 2 : ProcWaitForSignal(WAIT_EVENT_SAFE_SNAPSHOT);
5304 heikki.linnakangas@i 1588 : 2 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1589 : : }
5427 1590 : 2 : MySerializableXact->flags &= ~SXACT_FLAG_DEFERRABLE_WAITING;
1591 : :
1592 [ + + ]: 2 : if (!SxactIsROUnsafe(MySerializableXact))
1593 : : {
5304 1594 : 1 : LWLockRelease(SerializableXactHashLock);
5427 1595 : 1 : break; /* success */
1596 : : }
1597 : :
5304 1598 : 1 : LWLockRelease(SerializableXactHashLock);
1599 : :
1600 : : /* else, need to retry... */
5427 1601 [ - + ]: 1 : ereport(DEBUG2,
1602 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
1603 : : errmsg_internal("deferrable snapshot was unsafe; trying a new one")));
2469 tmunro@postgresql.or 1604 : 1 : ReleasePredicateLocks(false, false);
1605 : : }
1606 : :
1607 : : /*
1608 : : * Now we have a safe snapshot, so we don't need to do any further checks.
1609 : : */
5427 heikki.linnakangas@i 1610 [ - + ]: 1 : Assert(SxactIsROSafe(MySerializableXact));
2469 tmunro@postgresql.or 1611 : 1 : ReleasePredicateLocks(false, true);
1612 : :
5427 heikki.linnakangas@i 1613 : 1 : return snapshot;
1614 : : }
1615 : :
1616 : : /*
1617 : : * GetSafeSnapshotBlockingPids
1618 : : * If the specified process is currently blocked in GetSafeSnapshot,
1619 : : * write the process IDs of all processes that it is blocked by
1620 : : * into the caller-supplied buffer output[]. The list is truncated at
1621 : : * output_size, and the number of PIDs written into the buffer is
1622 : : * returned. Returns zero if the given PID is not currently blocked
1623 : : * in GetSafeSnapshot.
1624 : : */
1625 : : int
3173 tgl@sss.pgh.pa.us 1626 : 706 : GetSafeSnapshotBlockingPids(int blocked_pid, int *output, int output_size)
1627 : : {
1628 : 706 : int num_written = 0;
1629 : : dlist_iter iter;
1043 andres@anarazel.de 1630 : 706 : SERIALIZABLEXACT *blocking_sxact = NULL;
1631 : :
3173 tgl@sss.pgh.pa.us 1632 : 706 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1633 : :
1634 : : /* Find blocked_pid's SERIALIZABLEXACT by linear search. */
1063 andres@anarazel.de 1635 [ + - + + ]: 1563 : dlist_foreach(iter, &PredXact->activeList)
1636 : : {
1043 1637 : 954 : SERIALIZABLEXACT *sxact =
1638 : 954 : dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
1639 : :
3173 tgl@sss.pgh.pa.us 1640 [ + + ]: 954 : if (sxact->pid == blocked_pid)
1641 : : {
1043 andres@anarazel.de 1642 : 97 : blocking_sxact = sxact;
3173 tgl@sss.pgh.pa.us 1643 : 97 : break;
1644 : : }
1645 : : }
1646 : :
1647 : : /* Did we find it, and is it currently waiting in GetSafeSnapshot? */
1043 andres@anarazel.de 1648 [ + + + + ]: 706 : if (blocking_sxact != NULL && SxactIsDeferrableWaiting(blocking_sxact))
1649 : : {
1650 : : /* Traverse the list of possible unsafe conflicts collecting PIDs. */
1651 [ + - + - ]: 2 : dlist_foreach(iter, &blocking_sxact->possibleUnsafeConflicts)
1652 : : {
1063 1653 : 2 : RWConflict possibleUnsafeConflict =
943 tgl@sss.pgh.pa.us 1654 : 2 : dlist_container(RWConflictData, inLink, iter.cur);
1655 : :
3173 1656 : 2 : output[num_written++] = possibleUnsafeConflict->sxactOut->pid;
1657 : :
1043 andres@anarazel.de 1658 [ + - ]: 2 : if (num_written >= output_size)
1659 : 2 : break;
1660 : : }
1661 : : }
1662 : :
3173 tgl@sss.pgh.pa.us 1663 : 706 : LWLockRelease(SerializableXactHashLock);
1664 : :
1665 : 706 : return num_written;
1666 : : }
1667 : :
1668 : : /*
1669 : : * Acquire a snapshot that can be used for the current transaction.
1670 : : *
1671 : : * Make sure we have a SERIALIZABLEXACT reference in MySerializableXact.
1672 : : * It should be current for this process and be contained in PredXact.
1673 : : *
1674 : : * The passed-in Snapshot pointer should reference a static data area that
1675 : : * can safely be passed to GetSnapshotData. The return value is actually
1676 : : * always this same pointer; no new snapshot data structure is allocated
1677 : : * within this function.
1678 : : */
1679 : : Snapshot
5196 1680 : 1668 : GetSerializableTransactionSnapshot(Snapshot snapshot)
1681 : : {
5427 heikki.linnakangas@i 1682 [ - + ]: 1668 : Assert(IsolationIsSerializable());
1683 : :
1684 : : /*
1685 : : * Can't use serializable mode while recovery is still active, as it is,
1686 : : * for example, on a hot standby. We could get here despite the check in
1687 : : * check_transaction_isolation() if default_transaction_isolation is set
1688 : : * to serializable, so phrase the hint accordingly.
1689 : : */
4863 tgl@sss.pgh.pa.us 1690 [ - + ]: 1668 : if (RecoveryInProgress())
4863 tgl@sss.pgh.pa.us 1691 [ # # ]:UBC 0 : ereport(ERROR,
1692 : : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1693 : : errmsg("cannot use serializable mode in a hot standby"),
1694 : : errdetail("\"default_transaction_isolation\" is set to \"serializable\"."),
1695 : : errhint("You can use \"SET default_transaction_isolation = 'repeatable read'\" to change the default.")));
1696 : :
1697 : : /*
1698 : : * A special optimization is available for SERIALIZABLE READ ONLY
1699 : : * DEFERRABLE transactions -- we can wait for a suitable snapshot and
1700 : : * thereby avoid all SSI overhead once it's running.
1701 : : */
5427 heikki.linnakangas@i 1702 [ + + + + ]:CBC 1668 : if (XactReadOnly && XactDeferrable)
1703 : 4 : return GetSafeSnapshot(snapshot);
1704 : :
5170 tgl@sss.pgh.pa.us 1705 : 1664 : return GetSerializableTransactionSnapshotInt(snapshot,
1706 : : NULL, InvalidPid);
1707 : : }
1708 : :
1709 : : /*
1710 : : * Import a snapshot to be used for the current transaction.
1711 : : *
1712 : : * This is nearly the same as GetSerializableTransactionSnapshot, except that
1713 : : * we don't take a new snapshot, but rather use the data we're handed.
1714 : : *
1715 : : * The caller must have verified that the snapshot came from a serializable
1716 : : * transaction; and if we're read-write, the source transaction must not be
1717 : : * read-only.
1718 : : */
1719 : : void
1720 : 13 : SetSerializableTransactionSnapshot(Snapshot snapshot,
1721 : : VirtualTransactionId *sourcevxid,
1722 : : int sourcepid)
1723 : : {
1724 [ - + ]: 13 : Assert(IsolationIsSerializable());
1725 : :
1726 : : /*
1727 : : * If this is called by parallel.c in a parallel worker, we don't want to
1728 : : * create a SERIALIZABLEXACT just yet because the leader's
1729 : : * SERIALIZABLEXACT will be installed with AttachSerializableXact(). We
1730 : : * also don't want to reject SERIALIZABLE READ ONLY DEFERRABLE in this
1731 : : * case, because the leader has already determined that the snapshot it
1732 : : * has passed us is safe. So there is nothing for us to do.
1733 : : */
2469 tmunro@postgresql.or 1734 [ + - ]: 13 : if (IsParallelWorker())
1735 : 13 : return;
1736 : :
1737 : : /*
1738 : : * We do not allow SERIALIZABLE READ ONLY DEFERRABLE transactions to
1739 : : * import snapshots, since there's no way to wait for a safe snapshot when
1740 : : * we're using the snap we're told to. (XXX instead of throwing an error,
1741 : : * we could just ignore the XactDeferrable flag?)
1742 : : */
5170 tgl@sss.pgh.pa.us 1743 [ # # # # ]:UBC 0 : if (XactReadOnly && XactDeferrable)
1744 [ # # ]: 0 : ereport(ERROR,
1745 : : (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1746 : : errmsg("a snapshot-importing transaction must not be READ ONLY DEFERRABLE")));
1747 : :
3108 andres@anarazel.de 1748 : 0 : (void) GetSerializableTransactionSnapshotInt(snapshot, sourcevxid,
1749 : : sourcepid);
1750 : : }
1751 : :
1752 : : /*
1753 : : * Guts of GetSerializableTransactionSnapshot
1754 : : *
1755 : : * If sourcevxid is valid, this is actually an import operation and we should
1756 : : * skip calling GetSnapshotData, because the snapshot contents are already
1757 : : * loaded up. HOWEVER: to avoid race conditions, we must check that the
1758 : : * source xact is still running after we acquire SerializableXactHashLock.
1759 : : * We do that by calling ProcArrayInstallImportedXmin.
1760 : : */
1761 : : static Snapshot
5170 tgl@sss.pgh.pa.us 1762 :CBC 1669 : GetSerializableTransactionSnapshotInt(Snapshot snapshot,
1763 : : VirtualTransactionId *sourcevxid,
1764 : : int sourcepid)
1765 : : {
1766 : : PGPROC *proc;
1767 : : VirtualTransactionId vxid;
1768 : : SERIALIZABLEXACT *sxact,
1769 : : *othersxact;
1770 : :
1771 : : /* We only do this for serializable transactions. Once. */
5427 heikki.linnakangas@i 1772 [ - + ]: 1669 : Assert(MySerializableXact == InvalidSerializableXact);
1773 : :
1774 [ - + ]: 1669 : Assert(!RecoveryInProgress());
1775 : :
1776 : : /*
1777 : : * Since all parts of a serializable transaction must use the same
1778 : : * snapshot, it is too late to establish one after a parallel operation
1779 : : * has begun.
1780 : : */
3884 rhaas@postgresql.org 1781 [ - + ]: 1669 : if (IsInParallelMode())
3884 rhaas@postgresql.org 1782 [ # # ]:UBC 0 : elog(ERROR, "cannot establish serializable snapshot during a parallel operation");
1783 : :
5427 heikki.linnakangas@i 1784 :CBC 1669 : proc = MyProc;
1785 [ - + ]: 1669 : Assert(proc != NULL);
1786 : 1669 : GET_VXID_FROM_PGPROC(vxid, *proc);
1787 : :
1788 : : /*
1789 : : * First we get the sxact structure, which may involve looping and access
1790 : : * to the "finished" list to free a structure for use.
1791 : : *
1792 : : * We must hold SerializableXactHashLock when taking/checking the snapshot
1793 : : * to avoid race conditions, for much the same reasons that
1794 : : * GetSnapshotData takes the ProcArrayLock. Since we might have to
1795 : : * release SerializableXactHashLock to call SummarizeOldestCommittedSxact,
1796 : : * this means we have to create the sxact first, which is a bit annoying
1797 : : * (in particular, an elog(ERROR) in procarray.c would cause us to leak
1798 : : * the sxact). Consider refactoring to avoid this.
1799 : : */
1800 : : #ifdef TEST_SUMMARIZE_SERIAL
1801 : : SummarizeOldestCommittedSxact();
1802 : : #endif
1803 : 1669 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1804 : : do
1805 : : {
1806 : 1669 : sxact = CreatePredXact();
1807 : : /* If null, push out committed sxact to SLRU summary & retry. */
1808 [ - + ]: 1669 : if (!sxact)
1809 : : {
5427 heikki.linnakangas@i 1810 :UBC 0 : LWLockRelease(SerializableXactHashLock);
1811 : 0 : SummarizeOldestCommittedSxact();
1812 : 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1813 : : }
5427 heikki.linnakangas@i 1814 [ - + ]:CBC 1669 : } while (!sxact);
1815 : :
1816 : : /* Get the snapshot, or check that it's safe to use */
3108 andres@anarazel.de 1817 [ + - ]: 1669 : if (!sourcevxid)
5170 tgl@sss.pgh.pa.us 1818 : 1669 : snapshot = GetSnapshotData(snapshot);
3108 andres@anarazel.de 1819 [ # # ]:UBC 0 : else if (!ProcArrayInstallImportedXmin(snapshot->xmin, sourcevxid))
1820 : : {
5170 tgl@sss.pgh.pa.us 1821 : 0 : ReleasePredXact(sxact);
1822 : 0 : LWLockRelease(SerializableXactHashLock);
1823 [ # # ]: 0 : ereport(ERROR,
1824 : : (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
1825 : : errmsg("could not import the requested snapshot"),
1826 : : errdetail("The source process with PID %d is not running anymore.",
1827 : : sourcepid)));
1828 : : }
1829 : :
1830 : : /*
1831 : : * If there are no serializable transactions which are not read-only, we
1832 : : * can "opt out" of predicate locking and conflict checking for a
1833 : : * read-only transaction.
1834 : : *
1835 : : * The reason this is safe is that a read-only transaction can only become
1836 : : * part of a dangerous structure if it overlaps a writable transaction
1837 : : * which in turn overlaps a writable transaction which committed before
1838 : : * the read-only transaction started. A new writable transaction can
1839 : : * overlap this one, but it can't meet the other condition of overlapping
1840 : : * a transaction which committed before this one started.
1841 : : */
5427 heikki.linnakangas@i 1842 [ + + + + ]:CBC 1669 : if (XactReadOnly && PredXact->WritableSxactCount == 0)
1843 : : {
1844 : 111 : ReleasePredXact(sxact);
1845 : 111 : LWLockRelease(SerializableXactHashLock);
1846 : 111 : return snapshot;
1847 : : }
1848 : :
1849 : : /* Initialize the structure. */
1850 : 1558 : sxact->vxid = vxid;
1851 : 1558 : sxact->SeqNo.lastCommitBeforeSnapshot = PredXact->LastSxactCommitSeqNo;
5277 1852 : 1558 : sxact->prepareSeqNo = InvalidSerCommitSeqNo;
5427 1853 : 1558 : sxact->commitSeqNo = InvalidSerCommitSeqNo;
1063 andres@anarazel.de 1854 : 1558 : dlist_init(&(sxact->outConflicts));
1855 : 1558 : dlist_init(&(sxact->inConflicts));
1856 : 1558 : dlist_init(&(sxact->possibleUnsafeConflicts));
5427 heikki.linnakangas@i 1857 : 1558 : sxact->topXid = GetTopTransactionIdIfAny();
1858 : 1558 : sxact->finishedBefore = InvalidTransactionId;
1859 : 1558 : sxact->xmin = snapshot->xmin;
1860 : 1558 : sxact->pid = MyProcPid;
664 1861 : 1558 : sxact->pgprocno = MyProcNumber;
1063 andres@anarazel.de 1862 : 1558 : dlist_init(&sxact->predicateLocks);
1863 : 1558 : dlist_node_init(&sxact->finishedLink);
5427 heikki.linnakangas@i 1864 : 1558 : sxact->flags = 0;
1865 [ + + ]: 1558 : if (XactReadOnly)
1866 : : {
1867 : : dlist_iter iter;
1868 : :
1869 : 107 : sxact->flags |= SXACT_FLAG_READ_ONLY;
1870 : :
1871 : : /*
1872 : : * Register all concurrent r/w transactions as possible conflicts; if
1873 : : * all of them commit without any outgoing conflicts to earlier
1874 : : * transactions then this snapshot can be deemed safe (and we can run
1875 : : * without tracking predicate locks).
1876 : : */
1063 andres@anarazel.de 1877 [ + - + + ]: 468 : dlist_foreach(iter, &PredXact->activeList)
1878 : : {
1879 : 361 : othersxact = dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
1880 : :
5276 heikki.linnakangas@i 1881 [ + + ]: 361 : if (!SxactIsCommitted(othersxact)
1882 [ + - ]: 241 : && !SxactIsDoomed(othersxact)
1883 [ + + ]: 241 : && !SxactIsReadOnly(othersxact))
1884 : : {
5427 1885 : 133 : SetPossibleUnsafeConflict(sxact, othersxact);
1886 : : }
1887 : : }
1888 : :
1889 : : /*
1890 : : * If we didn't find any possibly unsafe conflicts because every
1891 : : * uncommitted writable transaction turned out to be doomed, then we
1892 : : * can "opt out" immediately. See comments above the earlier check
1893 : : * for PredXact->WritableSxactCount == 0.
1894 : : */
1014 tmunro@postgresql.or 1895 [ - + ]: 107 : if (dlist_is_empty(&sxact->possibleUnsafeConflicts))
1896 : : {
1014 tmunro@postgresql.or 1897 :UBC 0 : ReleasePredXact(sxact);
1898 : 0 : LWLockRelease(SerializableXactHashLock);
1899 : 0 : return snapshot;
1900 : : }
1901 : : }
1902 : : else
1903 : : {
5427 heikki.linnakangas@i 1904 :CBC 1451 : ++(PredXact->WritableSxactCount);
1905 [ - + ]: 1451 : Assert(PredXact->WritableSxactCount <=
1906 : : (MaxBackends + max_prepared_xacts));
1907 : : }
1908 : :
1909 : : /* Maintain serializable global xmin info. */
1014 tmunro@postgresql.or 1910 [ + + ]: 1558 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
1911 : : {
1912 [ - + ]: 849 : Assert(PredXact->SxactGlobalXminCount == 0);
1913 : 849 : PredXact->SxactGlobalXmin = snapshot->xmin;
1914 : 849 : PredXact->SxactGlobalXminCount = 1;
1915 : 849 : SerialSetActiveSerXmin(snapshot->xmin);
1916 : : }
1917 [ + + ]: 709 : else if (TransactionIdEquals(snapshot->xmin, PredXact->SxactGlobalXmin))
1918 : : {
1919 [ - + ]: 673 : Assert(PredXact->SxactGlobalXminCount > 0);
1920 : 673 : PredXact->SxactGlobalXminCount++;
1921 : : }
1922 : : else
1923 : : {
1924 [ - + ]: 36 : Assert(TransactionIdFollows(snapshot->xmin, PredXact->SxactGlobalXmin));
1925 : : }
1926 : :
5427 heikki.linnakangas@i 1927 : 1558 : MySerializableXact = sxact;
5304 1928 : 1558 : MyXactDidWrite = false; /* haven't written anything yet */
1929 : :
5427 1930 : 1558 : LWLockRelease(SerializableXactHashLock);
1931 : :
2469 tmunro@postgresql.or 1932 : 1558 : CreateLocalPredicateLockHash();
1933 : :
1934 : 1558 : return snapshot;
1935 : : }
1936 : :
1937 : : static void
1938 : 1571 : CreateLocalPredicateLockHash(void)
1939 : : {
1940 : : HASHCTL hash_ctl;
1941 : :
1942 : : /* Initialize the backend-local hash table of parent locks */
5427 heikki.linnakangas@i 1943 [ - + ]: 1571 : Assert(LocalPredicateLockHash == NULL);
1944 : 1571 : hash_ctl.keysize = sizeof(PREDICATELOCKTARGETTAG);
1945 : 1571 : hash_ctl.entrysize = sizeof(LOCALPREDICATELOCK);
1946 : 1571 : LocalPredicateLockHash = hash_create("Local predicate lock",
1947 : : max_predicate_locks_per_xact,
1948 : : &hash_ctl,
1949 : : HASH_ELEM | HASH_BLOBS);
1950 : 1571 : }
1951 : :
1952 : : /*
1953 : : * Register the top level XID in SerializableXidHash.
1954 : : * Also store it for easy reference in MySerializableXact.
1955 : : */
1956 : : void
5292 1957 : 135799 : RegisterPredicateLockingXid(TransactionId xid)
1958 : : {
1959 : : SERIALIZABLEXIDTAG sxidtag;
1960 : : SERIALIZABLEXID *sxid;
1961 : : bool found;
1962 : :
1963 : : /*
1964 : : * If we're not tracking predicate lock data for this transaction, we
1965 : : * should ignore the request and return quickly.
1966 : : */
5427 1967 [ + + ]: 135799 : if (MySerializableXact == InvalidSerializableXact)
1968 : 134508 : return;
1969 : :
1970 : : /* We should have a valid XID and be at the top level. */
1971 [ - + ]: 1291 : Assert(TransactionIdIsValid(xid));
1972 : :
5304 1973 : 1291 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
1974 : :
1975 : : /* This should only be done once per transaction. */
1976 [ - + ]: 1291 : Assert(MySerializableXact->topXid == InvalidTransactionId);
1977 : :
5427 1978 : 1291 : MySerializableXact->topXid = xid;
1979 : :
1980 : 1291 : sxidtag.xid = xid;
1981 : 1291 : sxid = (SERIALIZABLEXID *) hash_search(SerializableXidHash,
1982 : : &sxidtag,
1983 : : HASH_ENTER, &found);
1984 [ - + ]: 1291 : Assert(!found);
1985 : :
1986 : : /* Initialize the structure. */
5304 1987 : 1291 : sxid->myXact = MySerializableXact;
5427 1988 : 1291 : LWLockRelease(SerializableXactHashLock);
1989 : : }
1990 : :
1991 : :
1992 : : /*
1993 : : * Check whether there are any predicate locks held by any transaction
1994 : : * for the page at the given block number.
1995 : : *
1996 : : * Note that the transaction may be completed but not yet subject to
1997 : : * cleanup due to overlapping serializable transactions. This must
1998 : : * return valid information regardless of transaction isolation level.
1999 : : *
2000 : : * Also note that this doesn't check for a conflicting relation lock,
2001 : : * just a lock specifically on the given page.
2002 : : *
2003 : : * One use is to support proper behavior during GiST index vacuum.
2004 : : */
2005 : : bool
5292 heikki.linnakangas@i 2006 :UBC 0 : PageIsPredicateLocked(Relation relation, BlockNumber blkno)
2007 : : {
2008 : : PREDICATELOCKTARGETTAG targettag;
2009 : : uint32 targettaghash;
2010 : : LWLock *partitionLock;
2011 : : PREDICATELOCKTARGET *target;
2012 : :
5427 2013 : 0 : SET_PREDICATELOCKTARGETTAG_PAGE(targettag,
2014 : : relation->rd_locator.dbOid,
2015 : : relation->rd_id,
2016 : : blkno);
2017 : :
2018 : 0 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
2019 : 0 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
2020 : 0 : LWLockAcquire(partitionLock, LW_SHARED);
2021 : : target = (PREDICATELOCKTARGET *)
2022 : 0 : hash_search_with_hash_value(PredicateLockTargetHash,
2023 : : &targettag, targettaghash,
2024 : : HASH_FIND, NULL);
2025 : 0 : LWLockRelease(partitionLock);
2026 : :
2027 : 0 : return (target != NULL);
2028 : : }
2029 : :
2030 : :
2031 : : /*
2032 : : * Check whether a particular lock is held by this transaction.
2033 : : *
2034 : : * Important note: this function may return false even if the lock is
2035 : : * being held, because it uses the local lock table which is not
2036 : : * updated if another transaction modifies our lock list (e.g. to
2037 : : * split an index page). It can also return true when a coarser
2038 : : * granularity lock that covers this target is being held. Be careful
2039 : : * to only use this function in circumstances where such errors are
2040 : : * acceptable!
2041 : : */
2042 : : static bool
5292 tgl@sss.pgh.pa.us 2043 :CBC 77308 : PredicateLockExists(const PREDICATELOCKTARGETTAG *targettag)
2044 : : {
2045 : : LOCALPREDICATELOCK *lock;
2046 : :
2047 : : /* check local hash table */
5427 heikki.linnakangas@i 2048 : 77308 : lock = (LOCALPREDICATELOCK *) hash_search(LocalPredicateLockHash,
2049 : : targettag,
2050 : : HASH_FIND, NULL);
2051 : :
2052 [ + + ]: 77308 : if (!lock)
2053 : 30202 : return false;
2054 : :
2055 : : /*
2056 : : * Found entry in the table, but still need to check whether it's actually
2057 : : * held -- it could just be a parent of some held lock.
2058 : : */
2059 : 47106 : return lock->held;
2060 : : }
2061 : :
2062 : : /*
2063 : : * Return the parent lock tag in the lock hierarchy: the next coarser
2064 : : * lock that covers the provided tag.
2065 : : *
2066 : : * Returns true and sets *parent to the parent tag if one exists,
2067 : : * returns false if none exists.
2068 : : */
2069 : : static bool
5292 tgl@sss.pgh.pa.us 2070 : 45367 : GetParentPredicateLockTag(const PREDICATELOCKTARGETTAG *tag,
2071 : : PREDICATELOCKTARGETTAG *parent)
2072 : : {
5427 heikki.linnakangas@i 2073 [ + + + + : 45367 : switch (GET_PREDICATELOCKTARGETTAG_TYPE(*tag))
+ - ]
2074 : : {
2075 : 9906 : case PREDLOCKTAG_RELATION:
2076 : : /* relation locks have no parent lock */
2077 : 9906 : return false;
2078 : :
2079 : 8488 : case PREDLOCKTAG_PAGE:
2080 : : /* parent lock is relation lock */
2081 : 8488 : SET_PREDICATELOCKTARGETTAG_RELATION(*parent,
2082 : : GET_PREDICATELOCKTARGETTAG_DB(*tag),
2083 : : GET_PREDICATELOCKTARGETTAG_RELATION(*tag));
2084 : :
2085 : 8488 : return true;
2086 : :
2087 : 26973 : case PREDLOCKTAG_TUPLE:
2088 : : /* parent lock is page lock */
2089 : 26973 : SET_PREDICATELOCKTARGETTAG_PAGE(*parent,
2090 : : GET_PREDICATELOCKTARGETTAG_DB(*tag),
2091 : : GET_PREDICATELOCKTARGETTAG_RELATION(*tag),
2092 : : GET_PREDICATELOCKTARGETTAG_PAGE(*tag));
2093 : 26973 : return true;
2094 : : }
2095 : :
2096 : : /* not reachable */
5427 heikki.linnakangas@i 2097 :UBC 0 : Assert(false);
2098 : : return false;
2099 : : }
2100 : :
2101 : : /*
2102 : : * Check whether the lock we are considering is already covered by a
2103 : : * coarser lock for our transaction.
2104 : : *
2105 : : * Like PredicateLockExists, this function might return a false
2106 : : * negative, but it will never return a false positive.
2107 : : */
2108 : : static bool
5292 tgl@sss.pgh.pa.us 2109 :CBC 26080 : CoarserLockCovers(const PREDICATELOCKTARGETTAG *newtargettag)
2110 : : {
2111 : : PREDICATELOCKTARGETTAG targettag,
2112 : : parenttag;
2113 : :
5427 heikki.linnakangas@i 2114 : 26080 : targettag = *newtargettag;
2115 : :
2116 : : /* check parents iteratively until no more */
2117 [ + + ]: 31512 : while (GetParentPredicateLockTag(&targettag, &parenttag))
2118 : : {
2119 : 27253 : targettag = parenttag;
2120 [ + + ]: 27253 : if (PredicateLockExists(&targettag))
2121 : 21821 : return true;
2122 : : }
2123 : :
2124 : : /* no more parents to check; lock is not covered */
2125 : 4259 : return false;
2126 : : }
2127 : :
2128 : : /*
2129 : : * Remove the dummy entry from the predicate lock target hash, to free up some
2130 : : * scratch space. The caller must be holding SerializablePredicateListLock,
2131 : : * and must restore the entry with RestoreScratchTarget() before releasing the
2132 : : * lock.
2133 : : *
2134 : : * If lockheld is true, the caller is already holding the partition lock
2135 : : * of the partition containing the scratch entry.
2136 : : */
2137 : : static void
5306 2138 : 41 : RemoveScratchTarget(bool lockheld)
2139 : : {
2140 : : bool found;
2141 : :
2042 tgl@sss.pgh.pa.us 2142 [ - + ]: 41 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2143 : :
5306 heikki.linnakangas@i 2144 [ - + ]: 41 : if (!lockheld)
5306 heikki.linnakangas@i 2145 :UBC 0 : LWLockAcquire(ScratchPartitionLock, LW_EXCLUSIVE);
5306 heikki.linnakangas@i 2146 :CBC 41 : hash_search_with_hash_value(PredicateLockTargetHash,
2147 : : &ScratchTargetTag,
2148 : : ScratchTargetTagHash,
2149 : : HASH_REMOVE, &found);
2150 [ - + ]: 41 : Assert(found);
2151 [ - + ]: 41 : if (!lockheld)
5306 heikki.linnakangas@i 2152 :UBC 0 : LWLockRelease(ScratchPartitionLock);
5306 heikki.linnakangas@i 2153 :CBC 41 : }
2154 : :
2155 : : /*
2156 : : * Re-insert the dummy entry in predicate lock target hash.
2157 : : */
2158 : : static void
2159 : 41 : RestoreScratchTarget(bool lockheld)
2160 : : {
2161 : : bool found;
2162 : :
2042 tgl@sss.pgh.pa.us 2163 [ - + ]: 41 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2164 : :
5306 heikki.linnakangas@i 2165 [ - + ]: 41 : if (!lockheld)
5306 heikki.linnakangas@i 2166 :UBC 0 : LWLockAcquire(ScratchPartitionLock, LW_EXCLUSIVE);
5306 heikki.linnakangas@i 2167 :CBC 41 : hash_search_with_hash_value(PredicateLockTargetHash,
2168 : : &ScratchTargetTag,
2169 : : ScratchTargetTagHash,
2170 : : HASH_ENTER, &found);
2171 [ - + ]: 41 : Assert(!found);
2172 [ - + ]: 41 : if (!lockheld)
5306 heikki.linnakangas@i 2173 :UBC 0 : LWLockRelease(ScratchPartitionLock);
5306 heikki.linnakangas@i 2174 :CBC 41 : }
2175 : :
2176 : : /*
2177 : : * Check whether the list of related predicate locks is empty for a
2178 : : * predicate lock target, and remove the target if it is.
2179 : : */
2180 : : static void
5427 2181 : 4253 : RemoveTargetIfNoLongerUsed(PREDICATELOCKTARGET *target, uint32 targettaghash)
2182 : : {
2183 : : PREDICATELOCKTARGET *rmtarget PG_USED_FOR_ASSERTS_ONLY;
2184 : :
2042 tgl@sss.pgh.pa.us 2185 [ - + ]: 4253 : Assert(LWLockHeldByMe(SerializablePredicateListLock));
2186 : :
2187 : : /* Can't remove it until no locks at this target. */
1063 andres@anarazel.de 2188 [ + + ]: 4253 : if (!dlist_is_empty(&target->predicateLocks))
5427 heikki.linnakangas@i 2189 : 973 : return;
2190 : :
2191 : : /* Actually remove the target. */
2192 : 3280 : rmtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2193 : 3280 : &target->tag,
2194 : : targettaghash,
2195 : : HASH_REMOVE, NULL);
2196 [ - + ]: 3280 : Assert(rmtarget == target);
2197 : : }
2198 : :
2199 : : /*
2200 : : * Delete child target locks owned by this process.
2201 : : * This implementation is assuming that the usage of each target tag field
2202 : : * is uniform. No need to make this hard if we don't have to.
2203 : : *
2204 : : * We acquire an LWLock in the case of parallel mode, because worker
2205 : : * backends have access to the leader's SERIALIZABLEXACT. Otherwise,
2206 : : * we aren't acquiring LWLocks for the predicate lock or lock
2207 : : * target structures associated with this transaction unless we're going
2208 : : * to modify them, because no other process is permitted to modify our
2209 : : * locks.
2210 : : */
2211 : : static void
5292 tgl@sss.pgh.pa.us 2212 : 2377 : DeleteChildTargetLocks(const PREDICATELOCKTARGETTAG *newtargettag)
2213 : : {
2214 : : SERIALIZABLEXACT *sxact;
2215 : : PREDICATELOCK *predlock;
2216 : : dlist_mutable_iter iter;
2217 : :
2042 2218 : 2377 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
5304 heikki.linnakangas@i 2219 : 2377 : sxact = MySerializableXact;
2469 tmunro@postgresql.or 2220 [ + + ]: 2377 : if (IsInParallelMode())
2042 tgl@sss.pgh.pa.us 2221 : 11 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
2222 : :
1063 andres@anarazel.de 2223 [ + - + + ]: 7843 : dlist_foreach_modify(iter, &sxact->predicateLocks)
2224 : : {
2225 : : PREDICATELOCKTAG oldlocktag;
2226 : : PREDICATELOCKTARGET *oldtarget;
2227 : : PREDICATELOCKTARGETTAG oldtargettag;
2228 : :
2229 : 5466 : predlock = dlist_container(PREDICATELOCK, xactLink, iter.cur);
2230 : :
5427 heikki.linnakangas@i 2231 : 5466 : oldlocktag = predlock->tag;
2232 [ - + ]: 5466 : Assert(oldlocktag.myXact == sxact);
2233 : 5466 : oldtarget = oldlocktag.myTarget;
2234 : 5466 : oldtargettag = oldtarget->tag;
2235 : :
2236 [ + + + - : 5466 : if (TargetTagIsCoveredBy(oldtargettag, *newtargettag))
+ + + + +
+ - + +
- ]
2237 : : {
2238 : : uint32 oldtargettaghash;
2239 : : LWLock *partitionLock;
2240 : : PREDICATELOCK *rmpredlock PG_USED_FOR_ASSERTS_ONLY;
2241 : :
2242 : 999 : oldtargettaghash = PredicateLockTargetTagHashCode(&oldtargettag);
2243 : 999 : partitionLock = PredicateLockHashPartitionLock(oldtargettaghash);
2244 : :
2245 : 999 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
2246 : :
1063 andres@anarazel.de 2247 : 999 : dlist_delete(&predlock->xactLink);
2248 : 999 : dlist_delete(&predlock->targetLink);
5427 heikki.linnakangas@i 2249 : 999 : rmpredlock = hash_search_with_hash_value
2250 : : (PredicateLockHash,
2251 : : &oldlocktag,
2252 : 999 : PredicateLockHashCodeFromTargetHashCode(&oldlocktag,
2253 : : oldtargettaghash),
2254 : : HASH_REMOVE, NULL);
2255 [ - + ]: 999 : Assert(rmpredlock == predlock);
2256 : :
2257 : 999 : RemoveTargetIfNoLongerUsed(oldtarget, oldtargettaghash);
2258 : :
2259 : 999 : LWLockRelease(partitionLock);
2260 : :
2261 : 999 : DecrementParentLocks(&oldtargettag);
2262 : : }
2263 : : }
2469 tmunro@postgresql.or 2264 [ + + ]: 2377 : if (IsInParallelMode())
2042 tgl@sss.pgh.pa.us 2265 : 11 : LWLockRelease(&sxact->perXactPredicateListLock);
2266 : 2377 : LWLockRelease(SerializablePredicateListLock);
5427 heikki.linnakangas@i 2267 : 2377 : }
2268 : :
2269 : : /*
2270 : : * Returns the promotion limit for a given predicate lock target. This is the
2271 : : * max number of descendant locks allowed before promoting to the specified
2272 : : * tag. Note that the limit includes non-direct descendants (e.g., both tuples
2273 : : * and pages for a relation lock).
2274 : : *
2275 : : * Currently the default limit is 2 for a page lock, and half of the value of
2276 : : * max_pred_locks_per_transaction - 1 for a relation lock, to match behavior
2277 : : * of earlier releases when upgrading.
2278 : : *
2279 : : * TODO SSI: We should probably add additional GUCs to allow a maximum ratio
2280 : : * of page and tuple locks based on the pages in a relation, and the maximum
2281 : : * ratio of tuple locks to tuples in a page. This would provide more
2282 : : * generally "balanced" allocation of locks to where they are most useful,
2283 : : * while still allowing the absolute numbers to prevent one relation from
2284 : : * tying up all predicate lock resources.
2285 : : */
2286 : : static int
3176 kgrittn@postgresql.o 2287 : 5432 : MaxPredicateChildLocks(const PREDICATELOCKTARGETTAG *tag)
2288 : : {
5427 heikki.linnakangas@i 2289 [ + - + + : 5432 : switch (GET_PREDICATELOCKTARGETTAG_TYPE(*tag))
- - ]
2290 : : {
2291 : 3550 : case PREDLOCKTAG_RELATION:
3176 kgrittn@postgresql.o 2292 : 3550 : return max_predicate_locks_per_relation < 0
2293 : : ? (max_predicate_locks_per_xact
2294 : 3550 : / (-max_predicate_locks_per_relation)) - 1
2295 [ + - ]: 3550 : : max_predicate_locks_per_relation;
2296 : :
5427 heikki.linnakangas@i 2297 : 1882 : case PREDLOCKTAG_PAGE:
3176 kgrittn@postgresql.o 2298 : 1882 : return max_predicate_locks_per_page;
2299 : :
5427 heikki.linnakangas@i 2300 :UBC 0 : case PREDLOCKTAG_TUPLE:
2301 : :
2302 : : /*
2303 : : * not reachable: nothing is finer-granularity than a tuple, so we
2304 : : * should never try to promote to it.
2305 : : */
2306 : 0 : Assert(false);
2307 : : return 0;
2308 : : }
2309 : :
2310 : : /* not reachable */
2311 : 0 : Assert(false);
2312 : : return 0;
2313 : : }
2314 : :
2315 : : /*
2316 : : * For all ancestors of a newly-acquired predicate lock, increment
2317 : : * their child count in the parent hash table. If any of them have
2318 : : * more descendants than their promotion threshold, acquire the
2319 : : * coarsest such lock.
2320 : : *
2321 : : * Returns true if a parent lock was acquired and false otherwise.
2322 : : */
2323 : : static bool
5292 tgl@sss.pgh.pa.us 2324 :CBC 4259 : CheckAndPromotePredicateLockRequest(const PREDICATELOCKTARGETTAG *reqtag)
2325 : : {
2326 : : PREDICATELOCKTARGETTAG targettag,
2327 : : nexttag,
2328 : : promotiontag;
2329 : : LOCALPREDICATELOCK *parentlock;
2330 : : bool found,
2331 : : promote;
2332 : :
5427 heikki.linnakangas@i 2333 : 4259 : promote = false;
2334 : :
2335 : 4259 : targettag = *reqtag;
2336 : :
2337 : : /* check parents iteratively */
2338 [ + + ]: 13950 : while (GetParentPredicateLockTag(&targettag, &nexttag))
2339 : : {
2340 : 5432 : targettag = nexttag;
2341 : 5432 : parentlock = (LOCALPREDICATELOCK *) hash_search(LocalPredicateLockHash,
2342 : : &targettag,
2343 : : HASH_ENTER,
2344 : : &found);
2345 [ + + ]: 5432 : if (!found)
2346 : : {
2347 : 3371 : parentlock->held = false;
2348 : 3371 : parentlock->childLocks = 1;
2349 : : }
2350 : : else
2351 : 2061 : parentlock->childLocks++;
2352 : :
3176 kgrittn@postgresql.o 2353 [ + + ]: 5432 : if (parentlock->childLocks >
2354 : 5432 : MaxPredicateChildLocks(&targettag))
2355 : : {
2356 : : /*
2357 : : * We should promote to this parent lock. Continue to check its
2358 : : * ancestors, however, both to get their child counts right and to
2359 : : * check whether we should just go ahead and promote to one of
2360 : : * them.
2361 : : */
5427 heikki.linnakangas@i 2362 : 333 : promotiontag = targettag;
2363 : 333 : promote = true;
2364 : : }
2365 : : }
2366 : :
2367 [ + + ]: 4259 : if (promote)
2368 : : {
2369 : : /* acquire coarsest ancestor eligible for promotion */
2370 : 333 : PredicateLockAcquire(&promotiontag);
2371 : 333 : return true;
2372 : : }
2373 : : else
2374 : 3926 : return false;
2375 : : }
2376 : :
2377 : : /*
2378 : : * When releasing a lock, decrement the child count on all ancestor
2379 : : * locks.
2380 : : *
2381 : : * This is called only when releasing a lock via
2382 : : * DeleteChildTargetLocks (i.e. when a lock becomes redundant because
2383 : : * we've acquired its parent, possibly due to promotion) or when a new
2384 : : * MVCC write lock makes the predicate lock unnecessary. There's no
2385 : : * point in calling it when locks are released at transaction end, as
2386 : : * this information is no longer needed.
2387 : : */
2388 : : static void
5292 tgl@sss.pgh.pa.us 2389 : 1388 : DecrementParentLocks(const PREDICATELOCKTARGETTAG *targettag)
2390 : : {
2391 : : PREDICATELOCKTARGETTAG parenttag,
2392 : : nexttag;
2393 : :
5427 heikki.linnakangas@i 2394 : 1388 : parenttag = *targettag;
2395 : :
2396 [ + + ]: 4164 : while (GetParentPredicateLockTag(&parenttag, &nexttag))
2397 : : {
2398 : : uint32 targettaghash;
2399 : : LOCALPREDICATELOCK *parentlock,
2400 : : *rmlock PG_USED_FOR_ASSERTS_ONLY;
2401 : :
2402 : 2776 : parenttag = nexttag;
2403 : 2776 : targettaghash = PredicateLockTargetTagHashCode(&parenttag);
2404 : : parentlock = (LOCALPREDICATELOCK *)
2405 : 2776 : hash_search_with_hash_value(LocalPredicateLockHash,
2406 : : &parenttag, targettaghash,
2407 : : HASH_FIND, NULL);
2408 : :
2409 : : /*
2410 : : * There's a small chance the parent lock doesn't exist in the lock
2411 : : * table. This can happen if we prematurely removed it because an
2412 : : * index split caused the child refcount to be off.
2413 : : */
2414 [ - + ]: 2776 : if (parentlock == NULL)
5427 heikki.linnakangas@i 2415 :UBC 0 : continue;
2416 : :
5427 heikki.linnakangas@i 2417 :CBC 2776 : parentlock->childLocks--;
2418 : :
2419 : : /*
2420 : : * Under similar circumstances the parent lock's refcount might be
2421 : : * zero. This only happens if we're holding that lock (otherwise we
2422 : : * would have removed the entry).
2423 : : */
2424 [ - + ]: 2776 : if (parentlock->childLocks < 0)
2425 : : {
5427 heikki.linnakangas@i 2426 [ # # ]:UBC 0 : Assert(parentlock->held);
2427 : 0 : parentlock->childLocks = 0;
2428 : : }
2429 : :
5427 heikki.linnakangas@i 2430 [ + + + + ]:CBC 2776 : if ((parentlock->childLocks == 0) && (!parentlock->held))
2431 : : {
2432 : : rmlock = (LOCALPREDICATELOCK *)
2433 : 766 : hash_search_with_hash_value(LocalPredicateLockHash,
2434 : : &parenttag, targettaghash,
2435 : : HASH_REMOVE, NULL);
2436 [ - + ]: 766 : Assert(rmlock == parentlock);
2437 : : }
2438 : : }
2439 : 1388 : }
2440 : :
2441 : : /*
2442 : : * Indicate that a predicate lock on the given target is held by the
2443 : : * specified transaction. Has no effect if the lock is already held.
2444 : : *
2445 : : * This updates the lock table and the sxact's lock list, and creates
2446 : : * the lock target if necessary, but does *not* do anything related to
2447 : : * granularity promotion or the local lock table. See
2448 : : * PredicateLockAcquire for that.
2449 : : */
2450 : : static void
5292 tgl@sss.pgh.pa.us 2451 : 4259 : CreatePredicateLock(const PREDICATELOCKTARGETTAG *targettag,
2452 : : uint32 targettaghash,
2453 : : SERIALIZABLEXACT *sxact)
2454 : : {
2455 : : PREDICATELOCKTARGET *target;
2456 : : PREDICATELOCKTAG locktag;
2457 : : PREDICATELOCK *lock;
2458 : : LWLock *partitionLock;
2459 : : bool found;
2460 : :
5427 heikki.linnakangas@i 2461 : 4259 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
2462 : :
2042 tgl@sss.pgh.pa.us 2463 : 4259 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2469 tmunro@postgresql.or 2464 [ + + ]: 4259 : if (IsInParallelMode())
2042 tgl@sss.pgh.pa.us 2465 : 16 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
5427 heikki.linnakangas@i 2466 : 4259 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
2467 : :
2468 : : /* Make sure that the target is represented. */
2469 : : target = (PREDICATELOCKTARGET *)
2470 : 4259 : hash_search_with_hash_value(PredicateLockTargetHash,
2471 : : targettag, targettaghash,
2472 : : HASH_ENTER_NULL, &found);
2473 [ - + ]: 4259 : if (!target)
5427 heikki.linnakangas@i 2474 [ # # ]:UBC 0 : ereport(ERROR,
2475 : : (errcode(ERRCODE_OUT_OF_MEMORY),
2476 : : errmsg("out of shared memory"),
2477 : : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
5427 heikki.linnakangas@i 2478 [ + + ]:CBC 4259 : if (!found)
1063 andres@anarazel.de 2479 : 3280 : dlist_init(&target->predicateLocks);
2480 : :
2481 : : /* We've got the sxact and target, make sure they're joined. */
5427 heikki.linnakangas@i 2482 : 4259 : locktag.myTarget = target;
2483 : 4259 : locktag.myXact = sxact;
2484 : : lock = (PREDICATELOCK *)
2485 : 4259 : hash_search_with_hash_value(PredicateLockHash, &locktag,
3101 tgl@sss.pgh.pa.us 2486 : 4259 : PredicateLockHashCodeFromTargetHashCode(&locktag, targettaghash),
2487 : : HASH_ENTER_NULL, &found);
5427 heikki.linnakangas@i 2488 [ - + ]: 4259 : if (!lock)
5427 heikki.linnakangas@i 2489 [ # # ]:UBC 0 : ereport(ERROR,
2490 : : (errcode(ERRCODE_OUT_OF_MEMORY),
2491 : : errmsg("out of shared memory"),
2492 : : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
2493 : :
5427 heikki.linnakangas@i 2494 [ + + ]:CBC 4259 : if (!found)
2495 : : {
1063 andres@anarazel.de 2496 : 4253 : dlist_push_tail(&target->predicateLocks, &lock->targetLink);
2497 : 4253 : dlist_push_tail(&sxact->predicateLocks, &lock->xactLink);
5364 heikki.linnakangas@i 2498 : 4253 : lock->commitSeqNo = InvalidSerCommitSeqNo;
2499 : : }
2500 : :
5427 2501 : 4259 : LWLockRelease(partitionLock);
2469 tmunro@postgresql.or 2502 [ + + ]: 4259 : if (IsInParallelMode())
2042 tgl@sss.pgh.pa.us 2503 : 16 : LWLockRelease(&sxact->perXactPredicateListLock);
2504 : 4259 : LWLockRelease(SerializablePredicateListLock);
5427 heikki.linnakangas@i 2505 : 4259 : }
2506 : :
2507 : : /*
2508 : : * Acquire a predicate lock on the specified target for the current
2509 : : * connection if not already held. This updates the local lock table
2510 : : * and uses it to implement granularity promotion. It will consolidate
2511 : : * multiple locks into a coarser lock if warranted, and will release
2512 : : * any finer-grained locks covered by the new one.
2513 : : */
2514 : : static void
5292 tgl@sss.pgh.pa.us 2515 : 26285 : PredicateLockAcquire(const PREDICATELOCKTARGETTAG *targettag)
2516 : : {
2517 : : uint32 targettaghash;
2518 : : bool found;
2519 : : LOCALPREDICATELOCK *locallock;
2520 : :
2521 : : /* Do we have the lock already, or a covering lock? */
5427 heikki.linnakangas@i 2522 [ + + ]: 26285 : if (PredicateLockExists(targettag))
2523 : 22026 : return;
2524 : :
2525 [ + + ]: 26080 : if (CoarserLockCovers(targettag))
2526 : 21821 : return;
2527 : :
2528 : : /* the same hash and LW lock apply to the lock target and the local lock. */
2529 : 4259 : targettaghash = PredicateLockTargetTagHashCode(targettag);
2530 : :
2531 : : /* Acquire lock in local table */
2532 : : locallock = (LOCALPREDICATELOCK *)
2533 : 4259 : hash_search_with_hash_value(LocalPredicateLockHash,
2534 : : targettag, targettaghash,
2535 : : HASH_ENTER, &found);
2536 : 4259 : locallock->held = true;
2537 [ + + ]: 4259 : if (!found)
2538 : 3926 : locallock->childLocks = 0;
2539 : :
2540 : : /* Actually create the lock */
5304 2541 : 4259 : CreatePredicateLock(targettag, targettaghash, MySerializableXact);
2542 : :
2543 : : /*
2544 : : * Lock has been acquired. Check whether it should be promoted to a
2545 : : * coarser granularity, or whether there are finer-granularity locks to
2546 : : * clean up.
2547 : : */
5427 2548 [ + + ]: 4259 : if (CheckAndPromotePredicateLockRequest(targettag))
2549 : : {
2550 : : /*
2551 : : * Lock request was promoted to a coarser-granularity lock, and that
2552 : : * lock was acquired. It will delete this lock and any of its
2553 : : * children, so we're done.
2554 : : */
2555 : : }
2556 : : else
2557 : : {
2558 : : /* Clean up any finer-granularity locks */
2559 [ + + ]: 3926 : if (GET_PREDICATELOCKTARGETTAG_TYPE(*targettag) != PREDLOCKTAG_TUPLE)
2560 : 2377 : DeleteChildTargetLocks(targettag);
2561 : : }
2562 : : }
2563 : :
2564 : :
2565 : : /*
2566 : : * PredicateLockRelation
2567 : : *
2568 : : * Gets a predicate lock at the relation level.
2569 : : * Skip if not in full serializable transaction isolation level.
2570 : : * Skip if this is a temporary table.
2571 : : * Clear any finer-grained predicate locks this session has on the relation.
2572 : : */
2573 : : void
5292 2574 : 364976 : PredicateLockRelation(Relation relation, Snapshot snapshot)
2575 : : {
2576 : : PREDICATELOCKTARGETTAG tag;
2577 : :
5299 2578 [ + + ]: 364976 : if (!SerializationNeededForRead(relation, snapshot))
5427 2579 : 364246 : return;
2580 : :
2581 : 730 : SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2582 : : relation->rd_locator.dbOid,
2583 : : relation->rd_id);
2584 : 730 : PredicateLockAcquire(&tag);
2585 : : }
2586 : :
2587 : : /*
2588 : : * PredicateLockPage
2589 : : *
2590 : : * Gets a predicate lock at the page level.
2591 : : * Skip if not in full serializable transaction isolation level.
2592 : : * Skip if this is a temporary table.
2593 : : * Skip if a coarser predicate lock already covers this page.
2594 : : * Clear any finer-grained predicate locks this session has on the relation.
2595 : : */
2596 : : void
5292 2597 : 9054932 : PredicateLockPage(Relation relation, BlockNumber blkno, Snapshot snapshot)
2598 : : {
2599 : : PREDICATELOCKTARGETTAG tag;
2600 : :
5299 2601 [ + + ]: 9054932 : if (!SerializationNeededForRead(relation, snapshot))
5427 2602 : 9053480 : return;
2603 : :
2604 : 1452 : SET_PREDICATELOCKTARGETTAG_PAGE(tag,
2605 : : relation->rd_locator.dbOid,
2606 : : relation->rd_id,
2607 : : blkno);
2608 : 1452 : PredicateLockAcquire(&tag);
2609 : : }
2610 : :
2611 : : /*
2612 : : * PredicateLockTID
2613 : : *
2614 : : * Gets a predicate lock at the tuple level.
2615 : : * Skip if not in full serializable transaction isolation level.
2616 : : * Skip if this is a temporary table.
2617 : : */
2618 : : void
48 peter@eisentraut.org 2619 :GNC 15728074 : PredicateLockTID(Relation relation, const ItemPointerData *tid, Snapshot snapshot,
2620 : : TransactionId tuple_xid)
2621 : : {
2622 : : PREDICATELOCKTARGETTAG tag;
2623 : :
5299 heikki.linnakangas@i 2624 [ + + ]:CBC 15728074 : if (!SerializationNeededForRead(relation, snapshot))
5427 2625 : 15704304 : return;
2626 : :
2627 : : /*
2628 : : * Return if this xact wrote it.
2629 : : */
2630 [ + - ]: 23772 : if (relation->rd_index == NULL)
2631 : : {
2632 : : /* If we wrote it; we already have a write lock. */
2150 tmunro@postgresql.or 2633 [ + + ]: 23772 : if (TransactionIdIsCurrentTransactionId(tuple_xid))
2228 tmunro@postgresql.or 2634 :GBC 2 : return;
2635 : : }
2636 : :
2637 : : /*
2638 : : * Do quick-but-not-definitive test for a relation lock first. This will
2639 : : * never cause a return when the relation is *not* locked, but will
2640 : : * occasionally let the check continue when there really *is* a relation
2641 : : * level lock.
2642 : : */
5427 heikki.linnakangas@i 2643 :CBC 23770 : SET_PREDICATELOCKTARGETTAG_RELATION(tag,
2644 : : relation->rd_locator.dbOid,
2645 : : relation->rd_id);
2646 [ - + ]: 23770 : if (PredicateLockExists(&tag))
5427 heikki.linnakangas@i 2647 :UBC 0 : return;
2648 : :
5427 heikki.linnakangas@i 2649 :CBC 23770 : SET_PREDICATELOCKTARGETTAG_TUPLE(tag,
2650 : : relation->rd_locator.dbOid,
2651 : : relation->rd_id,
2652 : : ItemPointerGetBlockNumber(tid),
2653 : : ItemPointerGetOffsetNumber(tid));
2654 : 23770 : PredicateLockAcquire(&tag);
2655 : : }
2656 : :
2657 : :
2658 : : /*
2659 : : * DeleteLockTarget
2660 : : *
2661 : : * Remove a predicate lock target along with any locks held for it.
2662 : : *
2663 : : * Caller must hold SerializablePredicateListLock and the
2664 : : * appropriate hash partition lock for the target.
2665 : : */
2666 : : static void
5427 heikki.linnakangas@i 2667 :UBC 0 : DeleteLockTarget(PREDICATELOCKTARGET *target, uint32 targettaghash)
2668 : : {
2669 : : dlist_mutable_iter iter;
2670 : :
2042 tgl@sss.pgh.pa.us 2671 [ # # ]: 0 : Assert(LWLockHeldByMeInMode(SerializablePredicateListLock,
2672 : : LW_EXCLUSIVE));
5427 heikki.linnakangas@i 2673 [ # # ]: 0 : Assert(LWLockHeldByMe(PredicateLockHashPartitionLock(targettaghash)));
2674 : :
2675 : 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2676 : :
1063 andres@anarazel.de 2677 [ # # # # ]: 0 : dlist_foreach_modify(iter, &target->predicateLocks)
2678 : : {
2679 : 0 : PREDICATELOCK *predlock =
943 tgl@sss.pgh.pa.us 2680 : 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
2681 : : bool found;
2682 : :
1063 andres@anarazel.de 2683 : 0 : dlist_delete(&(predlock->xactLink));
2684 : 0 : dlist_delete(&(predlock->targetLink));
2685 : :
5427 heikki.linnakangas@i 2686 : 0 : hash_search_with_hash_value
2687 : : (PredicateLockHash,
2688 : 0 : &predlock->tag,
2689 : 0 : PredicateLockHashCodeFromTargetHashCode(&predlock->tag,
2690 : : targettaghash),
2691 : : HASH_REMOVE, &found);
2692 [ # # ]: 0 : Assert(found);
2693 : : }
2694 : 0 : LWLockRelease(SerializableXactHashLock);
2695 : :
2696 : : /* Remove the target itself, if possible. */
2697 : 0 : RemoveTargetIfNoLongerUsed(target, targettaghash);
2698 : 0 : }
2699 : :
2700 : :
2701 : : /*
2702 : : * TransferPredicateLocksToNewTarget
2703 : : *
2704 : : * Move or copy all the predicate locks for a lock target, for use by
2705 : : * index page splits/combines and other things that create or replace
2706 : : * lock targets. If 'removeOld' is true, the old locks and the target
2707 : : * will be removed.
2708 : : *
2709 : : * Returns true on success, or false if we ran out of shared memory to
2710 : : * allocate the new target or locks. Guaranteed to always succeed if
2711 : : * removeOld is set (by using the scratch entry in PredicateLockTargetHash
2712 : : * for scratch space).
2713 : : *
2714 : : * Warning: the "removeOld" option should be used only with care,
2715 : : * because this function does not (indeed, can not) update other
2716 : : * backends' LocalPredicateLockHash. If we are only adding new
2717 : : * entries, this is not a problem: the local lock table is used only
2718 : : * as a hint, so missing entries for locks that are held are
2719 : : * OK. Having entries for locks that are no longer held, as can happen
2720 : : * when using "removeOld", is not in general OK. We can only use it
2721 : : * safely when replacing a lock with a coarser-granularity lock that
2722 : : * covers it, or if we are absolutely certain that no one will need to
2723 : : * refer to that lock in the future.
2724 : : *
2725 : : * Caller must hold SerializablePredicateListLock exclusively.
2726 : : */
2727 : : static bool
5292 2728 : 0 : TransferPredicateLocksToNewTarget(PREDICATELOCKTARGETTAG oldtargettag,
2729 : : PREDICATELOCKTARGETTAG newtargettag,
2730 : : bool removeOld)
2731 : : {
2732 : : uint32 oldtargettaghash;
2733 : : LWLock *oldpartitionLock;
2734 : : PREDICATELOCKTARGET *oldtarget;
2735 : : uint32 newtargettaghash;
2736 : : LWLock *newpartitionLock;
2737 : : bool found;
5427 2738 : 0 : bool outOfShmem = false;
2739 : :
2042 tgl@sss.pgh.pa.us 2740 [ # # ]: 0 : Assert(LWLockHeldByMeInMode(SerializablePredicateListLock,
2741 : : LW_EXCLUSIVE));
2742 : :
5427 heikki.linnakangas@i 2743 : 0 : oldtargettaghash = PredicateLockTargetTagHashCode(&oldtargettag);
2744 : 0 : newtargettaghash = PredicateLockTargetTagHashCode(&newtargettag);
2745 : 0 : oldpartitionLock = PredicateLockHashPartitionLock(oldtargettaghash);
2746 : 0 : newpartitionLock = PredicateLockHashPartitionLock(newtargettaghash);
2747 : :
2748 [ # # ]: 0 : if (removeOld)
2749 : : {
2750 : : /*
2751 : : * Remove the dummy entry to give us scratch space, so we know we'll
2752 : : * be able to create the new lock target.
2753 : : */
5306 2754 : 0 : RemoveScratchTarget(false);
2755 : : }
2756 : :
2757 : : /*
2758 : : * We must get the partition locks in ascending sequence to avoid
2759 : : * deadlocks. If old and new partitions are the same, we must request the
2760 : : * lock only once.
2761 : : */
5427 2762 [ # # ]: 0 : if (oldpartitionLock < newpartitionLock)
2763 : : {
2764 : 0 : LWLockAcquire(oldpartitionLock,
2765 : 0 : (removeOld ? LW_EXCLUSIVE : LW_SHARED));
2766 : 0 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2767 : : }
2768 [ # # ]: 0 : else if (oldpartitionLock > newpartitionLock)
2769 : : {
2770 : 0 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2771 : 0 : LWLockAcquire(oldpartitionLock,
2772 : 0 : (removeOld ? LW_EXCLUSIVE : LW_SHARED));
2773 : : }
2774 : : else
2775 : 0 : LWLockAcquire(newpartitionLock, LW_EXCLUSIVE);
2776 : :
2777 : : /*
2778 : : * Look for the old target. If not found, that's OK; no predicate locks
2779 : : * are affected, so we can just clean up and return. If it does exist,
2780 : : * walk its list of predicate locks and move or copy them to the new
2781 : : * target.
2782 : : */
2783 : 0 : oldtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2784 : : &oldtargettag,
2785 : : oldtargettaghash,
2786 : : HASH_FIND, NULL);
2787 : :
2788 [ # # ]: 0 : if (oldtarget)
2789 : : {
2790 : : PREDICATELOCKTARGET *newtarget;
2791 : : PREDICATELOCKTAG newpredlocktag;
2792 : : dlist_mutable_iter iter;
2793 : :
2794 : 0 : newtarget = hash_search_with_hash_value(PredicateLockTargetHash,
2795 : : &newtargettag,
2796 : : newtargettaghash,
2797 : : HASH_ENTER_NULL, &found);
2798 : :
2799 [ # # ]: 0 : if (!newtarget)
2800 : : {
2801 : : /* Failed to allocate due to insufficient shmem */
2802 : 0 : outOfShmem = true;
2803 : 0 : goto exit;
2804 : : }
2805 : :
2806 : : /* If we created a new entry, initialize it */
2807 [ # # ]: 0 : if (!found)
1063 andres@anarazel.de 2808 : 0 : dlist_init(&newtarget->predicateLocks);
2809 : :
5416 magnus@hagander.net 2810 : 0 : newpredlocktag.myTarget = newtarget;
2811 : :
2812 : : /*
2813 : : * Loop through all the locks on the old target, replacing them with
2814 : : * locks on the new target.
2815 : : */
5427 heikki.linnakangas@i 2816 : 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2817 : :
1063 andres@anarazel.de 2818 [ # # # # ]: 0 : dlist_foreach_modify(iter, &oldtarget->predicateLocks)
2819 : : {
2820 : 0 : PREDICATELOCK *oldpredlock =
943 tgl@sss.pgh.pa.us 2821 : 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
2822 : : PREDICATELOCK *newpredlock;
5364 heikki.linnakangas@i 2823 : 0 : SerCommitSeqNo oldCommitSeqNo = oldpredlock->commitSeqNo;
2824 : :
5427 2825 : 0 : newpredlocktag.myXact = oldpredlock->tag.myXact;
2826 : :
2827 [ # # ]: 0 : if (removeOld)
2828 : : {
1063 andres@anarazel.de 2829 : 0 : dlist_delete(&(oldpredlock->xactLink));
2830 : 0 : dlist_delete(&(oldpredlock->targetLink));
2831 : :
5427 heikki.linnakangas@i 2832 : 0 : hash_search_with_hash_value
2833 : : (PredicateLockHash,
2834 : 0 : &oldpredlock->tag,
3101 tgl@sss.pgh.pa.us 2835 : 0 : PredicateLockHashCodeFromTargetHashCode(&oldpredlock->tag,
2836 : : oldtargettaghash),
2837 : : HASH_REMOVE, &found);
5427 heikki.linnakangas@i 2838 [ # # ]: 0 : Assert(found);
2839 : : }
2840 : :
2841 : : newpredlock = (PREDICATELOCK *)
5292 tgl@sss.pgh.pa.us 2842 : 0 : hash_search_with_hash_value(PredicateLockHash,
2843 : : &newpredlocktag,
3101 2844 : 0 : PredicateLockHashCodeFromTargetHashCode(&newpredlocktag,
2845 : : newtargettaghash),
2846 : : HASH_ENTER_NULL,
2847 : : &found);
5427 heikki.linnakangas@i 2848 [ # # ]: 0 : if (!newpredlock)
2849 : : {
2850 : : /* Out of shared memory. Undo what we've done so far. */
2851 : 0 : LWLockRelease(SerializableXactHashLock);
2852 : 0 : DeleteLockTarget(newtarget, newtargettaghash);
2853 : 0 : outOfShmem = true;
2854 : 0 : goto exit;
2855 : : }
5416 magnus@hagander.net 2856 [ # # ]: 0 : if (!found)
2857 : : {
1063 andres@anarazel.de 2858 : 0 : dlist_push_tail(&(newtarget->predicateLocks),
2859 : : &(newpredlock->targetLink));
2860 : 0 : dlist_push_tail(&(newpredlocktag.myXact->predicateLocks),
2861 : : &(newpredlock->xactLink));
5364 heikki.linnakangas@i 2862 : 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
2863 : : }
2864 : : else
2865 : : {
2866 [ # # ]: 0 : if (newpredlock->commitSeqNo < oldCommitSeqNo)
2867 : 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
2868 : : }
2869 : :
2870 [ # # ]: 0 : Assert(newpredlock->commitSeqNo != 0);
2871 [ # # # # ]: 0 : Assert((newpredlock->commitSeqNo == InvalidSerCommitSeqNo)
2872 : : || (newpredlock->tag.myXact == OldCommittedSxact));
2873 : : }
5427 2874 : 0 : LWLockRelease(SerializableXactHashLock);
2875 : :
2876 [ # # ]: 0 : if (removeOld)
2877 : : {
1063 andres@anarazel.de 2878 [ # # ]: 0 : Assert(dlist_is_empty(&oldtarget->predicateLocks));
5427 heikki.linnakangas@i 2879 : 0 : RemoveTargetIfNoLongerUsed(oldtarget, oldtargettaghash);
2880 : : }
2881 : : }
2882 : :
2883 : :
2884 : 0 : exit:
2885 : : /* Release partition locks in reverse order of acquisition. */
2886 [ # # ]: 0 : if (oldpartitionLock < newpartitionLock)
2887 : : {
2888 : 0 : LWLockRelease(newpartitionLock);
2889 : 0 : LWLockRelease(oldpartitionLock);
2890 : : }
2891 [ # # ]: 0 : else if (oldpartitionLock > newpartitionLock)
2892 : : {
2893 : 0 : LWLockRelease(oldpartitionLock);
2894 : 0 : LWLockRelease(newpartitionLock);
2895 : : }
2896 : : else
2897 : 0 : LWLockRelease(newpartitionLock);
2898 : :
2899 [ # # ]: 0 : if (removeOld)
2900 : : {
2901 : : /* We shouldn't run out of memory if we're moving locks */
2902 [ # # ]: 0 : Assert(!outOfShmem);
2903 : :
2904 : : /* Put the scratch entry back */
5306 2905 : 0 : RestoreScratchTarget(false);
2906 : : }
2907 : :
5427 2908 : 0 : return !outOfShmem;
2909 : : }
2910 : :
2911 : : /*
2912 : : * Drop all predicate locks of any granularity from the specified relation,
2913 : : * which can be a heap relation or an index relation. If 'transfer' is true,
2914 : : * acquire a relation lock on the heap for any transactions with any lock(s)
2915 : : * on the specified relation.
2916 : : *
2917 : : * This requires grabbing a lot of LW locks and scanning the entire lock
2918 : : * target table for matches. That makes this more expensive than most
2919 : : * predicate lock management functions, but it will only be called for DDL
2920 : : * type commands that are expensive anyway, and there are fast returns when
2921 : : * no serializable transactions are active or the relation is temporary.
2922 : : *
2923 : : * We don't use the TransferPredicateLocksToNewTarget function because it
2924 : : * acquires its own locks on the partitions of the two targets involved,
2925 : : * and we'll already be holding all partition locks.
2926 : : *
2927 : : * We can't throw an error from here, because the call could be from a
2928 : : * transaction which is not serializable.
2929 : : *
2930 : : * NOTE: This is currently only called with transfer set to true, but that may
2931 : : * change. If we decide to clean up the locks from a table on commit of a
2932 : : * transaction which executed DROP TABLE, the false condition will be useful.
2933 : : */
2934 : : static void
5292 heikki.linnakangas@i 2935 :CBC 17508 : DropAllPredicateLocksFromTable(Relation relation, bool transfer)
2936 : : {
2937 : : HASH_SEQ_STATUS seqstat;
2938 : : PREDICATELOCKTARGET *oldtarget;
2939 : : PREDICATELOCKTARGET *heaptarget;
2940 : : Oid dbId;
2941 : : Oid relId;
2942 : : Oid heapId;
2943 : : int i;
2944 : : bool isIndex;
2945 : : bool found;
2946 : : uint32 heaptargettaghash;
2947 : :
2948 : : /*
2949 : : * Bail out quickly if there are no serializable transactions running.
2950 : : * It's safe to check this without taking locks because the caller is
2951 : : * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
2952 : : * would matter here can be acquired while that is held.
2953 : : */
5306 2954 [ + + ]: 17508 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
2955 : 17467 : return;
2956 : :
5299 2957 [ + + ]: 60 : if (!PredicateLockingNeededForRelation(relation))
5306 2958 : 19 : return;
2959 : :
1260 rhaas@postgresql.org 2960 : 41 : dbId = relation->rd_locator.dbOid;
5306 heikki.linnakangas@i 2961 : 41 : relId = relation->rd_id;
2962 [ + + ]: 41 : if (relation->rd_index == NULL)
2963 : : {
2964 : 2 : isIndex = false;
2965 : 2 : heapId = relId;
2966 : : }
2967 : : else
2968 : : {
2969 : 39 : isIndex = true;
2970 : 39 : heapId = relation->rd_index->indrelid;
2971 : : }
2972 [ - + ]: 41 : Assert(heapId != InvalidOid);
3101 tgl@sss.pgh.pa.us 2973 [ - + - - ]: 41 : Assert(transfer || !isIndex); /* index OID only makes sense with
2974 : : * transfer */
2975 : :
2976 : : /* Retrieve first time needed, then keep. */
5306 heikki.linnakangas@i 2977 : 41 : heaptargettaghash = 0;
2978 : 41 : heaptarget = NULL;
2979 : :
2980 : : /* Acquire locks on all lock partitions */
2042 tgl@sss.pgh.pa.us 2981 : 41 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
5306 heikki.linnakangas@i 2982 [ + + ]: 697 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
4342 rhaas@postgresql.org 2983 : 656 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_EXCLUSIVE);
5306 heikki.linnakangas@i 2984 : 41 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
2985 : :
2986 : : /*
2987 : : * Remove the dummy entry to give us scratch space, so we know we'll be
2988 : : * able to create the new lock target.
2989 : : */
2990 [ + - ]: 41 : if (transfer)
2991 : 41 : RemoveScratchTarget(true);
2992 : :
2993 : : /* Scan through target map */
2994 : 41 : hash_seq_init(&seqstat, PredicateLockTargetHash);
2995 : :
2996 [ + + ]: 82 : while ((oldtarget = (PREDICATELOCKTARGET *) hash_seq_search(&seqstat)))
2997 : : {
2998 : : dlist_mutable_iter iter;
2999 : :
3000 : : /*
3001 : : * Check whether this is a target which needs attention.
3002 : : */
3003 [ + - ]: 41 : if (GET_PREDICATELOCKTARGETTAG_RELATION(oldtarget->tag) != relId)
3004 : 41 : continue; /* wrong relation id */
5306 heikki.linnakangas@i 3005 [ # # ]:UBC 0 : if (GET_PREDICATELOCKTARGETTAG_DB(oldtarget->tag) != dbId)
3006 : 0 : continue; /* wrong database id */
3007 [ # # # # ]: 0 : if (transfer && !isIndex
3008 [ # # # # ]: 0 : && GET_PREDICATELOCKTARGETTAG_TYPE(oldtarget->tag) == PREDLOCKTAG_RELATION)
3009 : 0 : continue; /* already the right lock */
3010 : :
3011 : : /*
3012 : : * If we made it here, we have work to do. We make sure the heap
3013 : : * relation lock exists, then we walk the list of predicate locks for
3014 : : * the old target we found, moving all locks to the heap relation lock
3015 : : * -- unless they already hold that.
3016 : : */
3017 : :
3018 : : /*
3019 : : * First make sure we have the heap relation target. We only need to
3020 : : * do this once.
3021 : : */
3022 [ # # # # ]: 0 : if (transfer && heaptarget == NULL)
3023 : : {
3024 : : PREDICATELOCKTARGETTAG heaptargettag;
3025 : :
3026 : 0 : SET_PREDICATELOCKTARGETTAG_RELATION(heaptargettag, dbId, heapId);
3027 : 0 : heaptargettaghash = PredicateLockTargetTagHashCode(&heaptargettag);
3028 : 0 : heaptarget = hash_search_with_hash_value(PredicateLockTargetHash,
3029 : : &heaptargettag,
3030 : : heaptargettaghash,
3031 : : HASH_ENTER, &found);
3032 [ # # ]: 0 : if (!found)
1063 andres@anarazel.de 3033 : 0 : dlist_init(&heaptarget->predicateLocks);
3034 : : }
3035 : :
3036 : : /*
3037 : : * Loop through all the locks on the old target, replacing them with
3038 : : * locks on the new target.
3039 : : */
3040 [ # # # # ]: 0 : dlist_foreach_modify(iter, &oldtarget->predicateLocks)
3041 : : {
3042 : 0 : PREDICATELOCK *oldpredlock =
943 tgl@sss.pgh.pa.us 3043 : 0 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
3044 : : PREDICATELOCK *newpredlock;
3045 : : SerCommitSeqNo oldCommitSeqNo;
3046 : : SERIALIZABLEXACT *oldXact;
3047 : :
3048 : : /*
3049 : : * Remove the old lock first. This avoids the chance of running
3050 : : * out of lock structure entries for the hash table.
3051 : : */
5306 heikki.linnakangas@i 3052 : 0 : oldCommitSeqNo = oldpredlock->commitSeqNo;
3053 : 0 : oldXact = oldpredlock->tag.myXact;
3054 : :
1063 andres@anarazel.de 3055 : 0 : dlist_delete(&(oldpredlock->xactLink));
3056 : :
3057 : : /*
3058 : : * No need for retail delete from oldtarget list, we're removing
3059 : : * the whole target anyway.
3060 : : */
5306 heikki.linnakangas@i 3061 : 0 : hash_search(PredicateLockHash,
3062 : 0 : &oldpredlock->tag,
3063 : : HASH_REMOVE, &found);
3064 [ # # ]: 0 : Assert(found);
3065 : :
3066 [ # # ]: 0 : if (transfer)
3067 : : {
3068 : : PREDICATELOCKTAG newpredlocktag;
3069 : :
3070 : 0 : newpredlocktag.myTarget = heaptarget;
3071 : 0 : newpredlocktag.myXact = oldXact;
3072 : : newpredlock = (PREDICATELOCK *)
5292 tgl@sss.pgh.pa.us 3073 : 0 : hash_search_with_hash_value(PredicateLockHash,
3074 : : &newpredlocktag,
3101 3075 : 0 : PredicateLockHashCodeFromTargetHashCode(&newpredlocktag,
3076 : : heaptargettaghash),
3077 : : HASH_ENTER,
3078 : : &found);
5306 heikki.linnakangas@i 3079 [ # # ]: 0 : if (!found)
3080 : : {
1063 andres@anarazel.de 3081 : 0 : dlist_push_tail(&(heaptarget->predicateLocks),
3082 : : &(newpredlock->targetLink));
3083 : 0 : dlist_push_tail(&(newpredlocktag.myXact->predicateLocks),
3084 : : &(newpredlock->xactLink));
5306 heikki.linnakangas@i 3085 : 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
3086 : : }
3087 : : else
3088 : : {
3089 [ # # ]: 0 : if (newpredlock->commitSeqNo < oldCommitSeqNo)
3090 : 0 : newpredlock->commitSeqNo = oldCommitSeqNo;
3091 : : }
3092 : :
3093 [ # # ]: 0 : Assert(newpredlock->commitSeqNo != 0);
3094 [ # # # # ]: 0 : Assert((newpredlock->commitSeqNo == InvalidSerCommitSeqNo)
3095 : : || (newpredlock->tag.myXact == OldCommittedSxact));
3096 : : }
3097 : : }
3098 : :
3099 : 0 : hash_search(PredicateLockTargetHash, &oldtarget->tag, HASH_REMOVE,
3100 : : &found);
3101 [ # # ]: 0 : Assert(found);
3102 : : }
3103 : :
3104 : : /* Put the scratch entry back */
5306 heikki.linnakangas@i 3105 [ + - ]:CBC 41 : if (transfer)
3106 : 41 : RestoreScratchTarget(true);
3107 : :
3108 : : /* Release locks in reverse order */
3109 : 41 : LWLockRelease(SerializableXactHashLock);
3110 [ + + ]: 697 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
4342 rhaas@postgresql.org 3111 : 656 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
2042 tgl@sss.pgh.pa.us 3112 : 41 : LWLockRelease(SerializablePredicateListLock);
3113 : : }
3114 : :
3115 : : /*
3116 : : * TransferPredicateLocksToHeapRelation
3117 : : * For all transactions, transfer all predicate locks for the given
3118 : : * relation to a single relation lock on the heap.
3119 : : */
3120 : : void
5292 heikki.linnakangas@i 3121 : 17508 : TransferPredicateLocksToHeapRelation(Relation relation)
3122 : : {
5306 3123 : 17508 : DropAllPredicateLocksFromTable(relation, true);
3124 : 17508 : }
3125 : :
3126 : :
3127 : : /*
3128 : : * PredicateLockPageSplit
3129 : : *
3130 : : * Copies any predicate locks for the old page to the new page.
3131 : : * Skip if this is a temporary table or toast table.
3132 : : *
3133 : : * NOTE: A page split (or overflow) affects all serializable transactions,
3134 : : * even if it occurs in the context of another transaction isolation level.
3135 : : *
3136 : : * NOTE: This currently leaves the local copy of the locks without
3137 : : * information on the new lock which is in shared memory. This could cause
3138 : : * problems if enough page splits occur on locked pages without the processes
3139 : : * which hold the locks getting in and noticing.
3140 : : */
3141 : : void
5292 3142 : 30422 : PredicateLockPageSplit(Relation relation, BlockNumber oldblkno,
3143 : : BlockNumber newblkno)
3144 : : {
3145 : : PREDICATELOCKTARGETTAG oldtargettag;
3146 : : PREDICATELOCKTARGETTAG newtargettag;
3147 : : bool success;
3148 : :
3149 : : /*
3150 : : * Bail out quickly if there are no serializable transactions running.
3151 : : *
3152 : : * It's safe to do this check without taking any additional locks. Even if
3153 : : * a serializable transaction starts concurrently, we know it can't take
3154 : : * any SIREAD locks on the page being split because the caller is holding
3155 : : * the associated buffer page lock. Memory reordering isn't an issue; the
3156 : : * memory barrier in the LWLock acquisition guarantees that this read
3157 : : * occurs while the buffer page lock is held.
3158 : : */
5350 rhaas@postgresql.org 3159 [ + + ]: 30422 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
3160 : 30422 : return;
3161 : :
5299 heikki.linnakangas@i 3162 [ + - ]: 16 : if (!PredicateLockingNeededForRelation(relation))
5427 3163 : 16 : return;
3164 : :
5427 heikki.linnakangas@i 3165 [ # # ]:UBC 0 : Assert(oldblkno != newblkno);
3166 [ # # ]: 0 : Assert(BlockNumberIsValid(oldblkno));
3167 [ # # ]: 0 : Assert(BlockNumberIsValid(newblkno));
3168 : :
3169 : 0 : SET_PREDICATELOCKTARGETTAG_PAGE(oldtargettag,
3170 : : relation->rd_locator.dbOid,
3171 : : relation->rd_id,
3172 : : oldblkno);
3173 : 0 : SET_PREDICATELOCKTARGETTAG_PAGE(newtargettag,
3174 : : relation->rd_locator.dbOid,
3175 : : relation->rd_id,
3176 : : newblkno);
3177 : :
2042 tgl@sss.pgh.pa.us 3178 : 0 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
3179 : :
3180 : : /*
3181 : : * Try copying the locks over to the new page's tag, creating it if
3182 : : * necessary.
3183 : : */
5427 heikki.linnakangas@i 3184 : 0 : success = TransferPredicateLocksToNewTarget(oldtargettag,
3185 : : newtargettag,
3186 : : false);
3187 : :
3188 [ # # ]: 0 : if (!success)
3189 : : {
3190 : : /*
3191 : : * No more predicate lock entries are available. Failure isn't an
3192 : : * option here, so promote the page lock to a relation lock.
3193 : : */
3194 : :
3195 : : /* Get the parent relation lock's lock tag */
3196 : 0 : success = GetParentPredicateLockTag(&oldtargettag,
3197 : : &newtargettag);
3198 [ # # ]: 0 : Assert(success);
3199 : :
3200 : : /*
3201 : : * Move the locks to the parent. This shouldn't fail.
3202 : : *
3203 : : * Note that here we are removing locks held by other backends,
3204 : : * leading to a possible inconsistency in their local lock hash table.
3205 : : * This is OK because we're replacing it with a lock that covers the
3206 : : * old one.
3207 : : */
3208 : 0 : success = TransferPredicateLocksToNewTarget(oldtargettag,
3209 : : newtargettag,
3210 : : true);
3211 [ # # ]: 0 : Assert(success);
3212 : : }
3213 : :
2042 tgl@sss.pgh.pa.us 3214 : 0 : LWLockRelease(SerializablePredicateListLock);
3215 : : }
3216 : :
3217 : : /*
3218 : : * PredicateLockPageCombine
3219 : : *
3220 : : * Combines predicate locks for two existing pages.
3221 : : * Skip if this is a temporary table or toast table.
3222 : : *
3223 : : * NOTE: A page combine affects all serializable transactions, even if it
3224 : : * occurs in the context of another transaction isolation level.
3225 : : */
3226 : : void
5292 heikki.linnakangas@i 3227 :CBC 3002 : PredicateLockPageCombine(Relation relation, BlockNumber oldblkno,
3228 : : BlockNumber newblkno)
3229 : : {
3230 : : /*
3231 : : * Page combines differ from page splits in that we ought to be able to
3232 : : * remove the locks on the old page after transferring them to the new
3233 : : * page, instead of duplicating them. However, because we can't edit other
3234 : : * backends' local lock tables, removing the old lock would leave them
3235 : : * with an entry in their LocalPredicateLockHash for a lock they're not
3236 : : * holding, which isn't acceptable. So we wind up having to do the same
3237 : : * work as a page split, acquiring a lock on the new page and keeping the
3238 : : * old page locked too. That can lead to some false positives, but should
3239 : : * be rare in practice.
3240 : : */
5405 3241 : 3002 : PredicateLockPageSplit(relation, oldblkno, newblkno);
5427 3242 : 3002 : }
3243 : :
3244 : : /*
3245 : : * Walk the list of in-progress serializable transactions and find the new
3246 : : * xmin.
3247 : : */
3248 : : static void
3249 : 868 : SetNewSxactGlobalXmin(void)
3250 : : {
3251 : : dlist_iter iter;
3252 : :
3253 [ - + ]: 868 : Assert(LWLockHeldByMe(SerializableXactHashLock));
3254 : :
3255 : 868 : PredXact->SxactGlobalXmin = InvalidTransactionId;
3256 : 868 : PredXact->SxactGlobalXminCount = 0;
3257 : :
1063 andres@anarazel.de 3258 [ + - + + ]: 3307 : dlist_foreach(iter, &PredXact->activeList)
3259 : : {
3260 : 2439 : SERIALIZABLEXACT *sxact =
943 tgl@sss.pgh.pa.us 3261 : 2439 : dlist_container(SERIALIZABLEXACT, xactLink, iter.cur);
3262 : :
5293 heikki.linnakangas@i 3263 [ + + ]: 2439 : if (!SxactIsRolledBack(sxact)
5427 3264 [ + + ]: 2138 : && !SxactIsCommitted(sxact)
3265 [ + - ]: 19 : && sxact != OldCommittedSxact)
3266 : : {
3267 [ - + ]: 19 : Assert(sxact->xmin != InvalidTransactionId);
3268 [ - + ]: 19 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)
5427 heikki.linnakangas@i 3269 [ # # ]:UBC 0 : || TransactionIdPrecedes(sxact->xmin,
3270 : 0 : PredXact->SxactGlobalXmin))
3271 : : {
5427 heikki.linnakangas@i 3272 :CBC 19 : PredXact->SxactGlobalXmin = sxact->xmin;
3273 : 19 : PredXact->SxactGlobalXminCount = 1;
3274 : : }
5427 heikki.linnakangas@i 3275 [ # # ]:UBC 0 : else if (TransactionIdEquals(sxact->xmin,
3276 : : PredXact->SxactGlobalXmin))
3277 : 0 : PredXact->SxactGlobalXminCount++;
3278 : : }
3279 : : }
3280 : :
2042 tgl@sss.pgh.pa.us 3281 :CBC 868 : SerialSetActiveSerXmin(PredXact->SxactGlobalXmin);
5427 heikki.linnakangas@i 3282 : 868 : }
3283 : :
3284 : : /*
3285 : : * ReleasePredicateLocks
3286 : : *
3287 : : * Releases predicate locks based on completion of the current transaction,
3288 : : * whether committed or rolled back. It can also be called for a read only
3289 : : * transaction when it becomes impossible for the transaction to become
3290 : : * part of a dangerous structure.
3291 : : *
3292 : : * We do nothing unless this is a serializable transaction.
3293 : : *
3294 : : * This method must ensure that shared memory hash tables are cleaned
3295 : : * up in some relatively timely fashion.
3296 : : *
3297 : : * If this transaction is committing and is holding any predicate locks,
3298 : : * it must be added to a list of completed serializable transactions still
3299 : : * holding locks.
3300 : : *
3301 : : * If isReadOnlySafe is true, then predicate locks are being released before
3302 : : * the end of the transaction because MySerializableXact has been determined
3303 : : * to be RO_SAFE. In non-parallel mode we can release it completely, but it
3304 : : * in parallel mode we partially release the SERIALIZABLEXACT and keep it
3305 : : * around until the end of the transaction, allowing each backend to clear its
3306 : : * MySerializableXact variable and benefit from the optimization in its own
3307 : : * time.
3308 : : */
3309 : : void
2469 tmunro@postgresql.or 3310 : 331934 : ReleasePredicateLocks(bool isCommit, bool isReadOnlySafe)
3311 : : {
1017 3312 : 331934 : bool partiallyReleasing = false;
3313 : : bool needToClear;
3314 : : SERIALIZABLEXACT *roXact;
3315 : : dlist_mutable_iter iter;
3316 : :
3317 : : /*
3318 : : * We can't trust XactReadOnly here, because a transaction which started
3319 : : * as READ WRITE can show as READ ONLY later, e.g., within
3320 : : * subtransactions. We want to flag a transaction as READ ONLY if it
3321 : : * commits without writing so that de facto READ ONLY transactions get the
3322 : : * benefit of some RO optimizations, so we will use this local variable to
3323 : : * get some cleanup logic right which is based on whether the transaction
3324 : : * was declared READ ONLY at the top level.
3325 : : */
3326 : : bool topLevelIsDeclaredReadOnly;
3327 : :
3328 : : /* We can't be both committing and releasing early due to RO_SAFE. */
2469 3329 [ + + - + ]: 331934 : Assert(!(isCommit && isReadOnlySafe));
3330 : :
3331 : : /* Are we at the end of a transaction, that is, a commit or abort? */
3332 [ + + ]: 331934 : if (!isReadOnlySafe)
3333 : : {
3334 : : /*
3335 : : * Parallel workers mustn't release predicate locks at the end of
3336 : : * their transaction. The leader will do that at the end of its
3337 : : * transaction.
3338 : : */
3339 [ + + ]: 331900 : if (IsParallelWorker())
3340 : : {
3341 : 4320 : ReleasePredicateLocksLocal();
3342 : 330374 : return;
3343 : : }
3344 : :
3345 : : /*
3346 : : * By the time the leader in a parallel query reaches end of
3347 : : * transaction, it has waited for all workers to exit.
3348 : : */
3349 [ - + ]: 327580 : Assert(!ParallelContextActive());
3350 : :
3351 : : /*
3352 : : * If the leader in a parallel query earlier stashed a partially
3353 : : * released SERIALIZABLEXACT for final clean-up at end of transaction
3354 : : * (because workers might still have been accessing it), then it's
3355 : : * time to restore it.
3356 : : */
3357 [ + + ]: 327580 : if (SavedSerializableXact != InvalidSerializableXact)
3358 : : {
3359 [ - + ]: 1 : Assert(MySerializableXact == InvalidSerializableXact);
3360 : 1 : MySerializableXact = SavedSerializableXact;
3361 : 1 : SavedSerializableXact = InvalidSerializableXact;
3362 [ - + ]: 1 : Assert(SxactIsPartiallyReleased(MySerializableXact));
3363 : : }
3364 : : }
3365 : :
5427 heikki.linnakangas@i 3366 [ + + ]: 327614 : if (MySerializableXact == InvalidSerializableXact)
3367 : : {
3368 [ - + ]: 326051 : Assert(LocalPredicateLockHash == NULL);
3369 : 326051 : return;
3370 : : }
3371 : :
3700 kgrittn@postgresql.o 3372 : 1563 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
3373 : :
3374 : : /*
3375 : : * If the transaction is committing, but it has been partially released
3376 : : * already, then treat this as a roll back. It was marked as rolled back.
3377 : : */
2469 tmunro@postgresql.or 3378 [ + + + + ]: 1563 : if (isCommit && SxactIsPartiallyReleased(MySerializableXact))
3379 : 2 : isCommit = false;
3380 : :
3381 : : /*
3382 : : * If we're called in the middle of a transaction because we discovered
3383 : : * that the SXACT_FLAG_RO_SAFE flag was set, then we'll partially release
3384 : : * it (that is, release the predicate locks and conflicts, but not the
3385 : : * SERIALIZABLEXACT itself) if we're the first backend to have noticed.
3386 : : */
3387 [ + + + + ]: 1563 : if (isReadOnlySafe && IsInParallelMode())
3388 : : {
3389 : : /*
3390 : : * The leader needs to stash a pointer to it, so that it can
3391 : : * completely release it at end-of-transaction.
3392 : : */
3393 [ + + ]: 5 : if (!IsParallelWorker())
3394 : 1 : SavedSerializableXact = MySerializableXact;
3395 : :
3396 : : /*
3397 : : * The first backend to reach this condition will partially release
3398 : : * the SERIALIZABLEXACT. All others will just clear their
3399 : : * backend-local state so that they stop doing SSI checks for the rest
3400 : : * of the transaction.
3401 : : */
3402 [ + + ]: 5 : if (SxactIsPartiallyReleased(MySerializableXact))
3403 : : {
3404 : 3 : LWLockRelease(SerializableXactHashLock);
3405 : 3 : ReleasePredicateLocksLocal();
3406 : 3 : return;
3407 : : }
3408 : : else
3409 : : {
3410 : 2 : MySerializableXact->flags |= SXACT_FLAG_PARTIALLY_RELEASED;
1017 3411 : 2 : partiallyReleasing = true;
3412 : : /* ... and proceed to perform the partial release below. */
3413 : : }
3414 : : }
5427 heikki.linnakangas@i 3415 [ + + - + ]: 1560 : Assert(!isCommit || SxactIsPrepared(MySerializableXact));
5299 3416 [ + + - + ]: 1560 : Assert(!isCommit || !SxactIsDoomed(MySerializableXact));
5427 3417 [ - + ]: 1560 : Assert(!SxactIsCommitted(MySerializableXact));
2469 tmunro@postgresql.or 3418 [ + + - + ]: 1560 : Assert(SxactIsPartiallyReleased(MySerializableXact)
3419 : : || !SxactIsRolledBack(MySerializableXact));
3420 : :
3421 : : /* may not be serializable during COMMIT/ROLLBACK PREPARED */
3700 kgrittn@postgresql.o 3422 [ + + - + ]: 1560 : Assert(MySerializableXact->pid == 0 || IsolationIsSerializable());
3423 : :
3424 : : /* We'd better not already be on the cleanup list. */
5304 heikki.linnakangas@i 3425 [ - + ]: 1560 : Assert(!SxactIsOnFinishedList(MySerializableXact));
3426 : :
5427 3427 : 1560 : topLevelIsDeclaredReadOnly = SxactIsReadOnly(MySerializableXact);
3428 : :
3429 : : /*
3430 : : * We don't hold XidGenLock lock here, assuming that TransactionId is
3431 : : * atomic!
3432 : : *
3433 : : * If this value is changing, we don't care that much whether we get the
3434 : : * old or new value -- it is just used to determine how far
3435 : : * SxactGlobalXmin must advance before this transaction can be fully
3436 : : * cleaned up. The worst that could happen is we wait for one more
3437 : : * transaction to complete before freeing some RAM; correctness of visible
3438 : : * behavior is not affected.
3439 : : */
740 3440 : 1560 : MySerializableXact->finishedBefore = XidFromFullTransactionId(TransamVariables->nextXid);
3441 : :
3442 : : /*
3443 : : * If it's not a commit it's either a rollback or a read-only transaction
3444 : : * flagged SXACT_FLAG_RO_SAFE, and we can clear our locks immediately.
3445 : : */
5427 3446 [ + + ]: 1560 : if (isCommit)
3447 : : {
3448 : 1231 : MySerializableXact->flags |= SXACT_FLAG_COMMITTED;
3449 : 1231 : MySerializableXact->commitSeqNo = ++(PredXact->LastSxactCommitSeqNo);
3450 : : /* Recognize implicit read-only transaction (commit without write). */
5304 3451 [ + + ]: 1231 : if (!MyXactDidWrite)
5427 3452 : 233 : MySerializableXact->flags |= SXACT_FLAG_READ_ONLY;
3453 : : }
3454 : : else
3455 : : {
3456 : : /*
3457 : : * The DOOMED flag indicates that we intend to roll back this
3458 : : * transaction and so it should not cause serialization failures for
3459 : : * other transactions that conflict with it. Note that this flag might
3460 : : * already be set, if another backend marked this transaction for
3461 : : * abort.
3462 : : *
3463 : : * The ROLLED_BACK flag further indicates that ReleasePredicateLocks
3464 : : * has been called, and so the SerializableXact is eligible for
3465 : : * cleanup. This means it should not be considered when calculating
3466 : : * SxactGlobalXmin.
3467 : : */
2414 tmunro@postgresql.or 3468 : 329 : MySerializableXact->flags |= SXACT_FLAG_DOOMED;
5293 heikki.linnakangas@i 3469 : 329 : MySerializableXact->flags |= SXACT_FLAG_ROLLED_BACK;
3470 : :
3471 : : /*
3472 : : * If the transaction was previously prepared, but is now failing due
3473 : : * to a ROLLBACK PREPARED or (hopefully very rare) error after the
3474 : : * prepare, clear the prepared flag. This simplifies conflict
3475 : : * checking.
3476 : : */
5277 3477 : 329 : MySerializableXact->flags &= ~SXACT_FLAG_PREPARED;
3478 : : }
3479 : :
5427 3480 [ + + ]: 1560 : if (!topLevelIsDeclaredReadOnly)
3481 : : {
3482 [ - + ]: 1451 : Assert(PredXact->WritableSxactCount > 0);
3483 [ + + ]: 1451 : if (--(PredXact->WritableSxactCount) == 0)
3484 : : {
3485 : : /*
3486 : : * Release predicate locks and rw-conflicts in for all committed
3487 : : * transactions. There are no longer any transactions which might
3488 : : * conflict with the locks and no chance for new transactions to
3489 : : * overlap. Similarly, existing conflicts in can't cause pivots,
3490 : : * and any conflicts in which could have completed a dangerous
3491 : : * structure would already have caused a rollback, so any
3492 : : * remaining ones must be benign.
3493 : : */
3494 : 859 : PredXact->CanPartialClearThrough = PredXact->LastSxactCommitSeqNo;
3495 : : }
3496 : : }
3497 : : else
3498 : : {
3499 : : /*
3500 : : * Read-only transactions: clear the list of transactions that might
3501 : : * make us unsafe. Note that we use 'inLink' for the iteration as
3502 : : * opposed to 'outLink' for the r/w xacts.
3503 : : */
1063 andres@anarazel.de 3504 [ + - + + ]: 151 : dlist_foreach_modify(iter, &MySerializableXact->possibleUnsafeConflicts)
3505 : : {
3506 : 42 : RWConflict possibleUnsafeConflict =
943 tgl@sss.pgh.pa.us 3507 : 42 : dlist_container(RWConflictData, inLink, iter.cur);
3508 : :
5427 heikki.linnakangas@i 3509 [ - + ]: 42 : Assert(!SxactIsReadOnly(possibleUnsafeConflict->sxactOut));
3510 [ - + ]: 42 : Assert(MySerializableXact == possibleUnsafeConflict->sxactIn);
3511 : :
3512 : 42 : ReleaseRWConflict(possibleUnsafeConflict);
3513 : : }
3514 : : }
3515 : :
3516 : : /* Check for conflict out to old committed transactions. */
3517 [ + + ]: 1560 : if (isCommit
3518 [ + + ]: 1231 : && !SxactIsReadOnly(MySerializableXact)
3519 [ - + ]: 998 : && SxactHasSummaryConflictOut(MySerializableXact))
3520 : : {
3521 : : /*
3522 : : * we don't know which old committed transaction we conflicted with,
3523 : : * so be conservative and use FirstNormalSerCommitSeqNo here
3524 : : */
5427 heikki.linnakangas@i 3525 :UBC 0 : MySerializableXact->SeqNo.earliestOutConflictCommit =
3526 : : FirstNormalSerCommitSeqNo;
3527 : 0 : MySerializableXact->flags |= SXACT_FLAG_CONFLICT_OUT;
3528 : : }
3529 : :
3530 : : /*
3531 : : * Release all outConflicts to committed transactions. If we're rolling
3532 : : * back clear them all. Set SXACT_FLAG_CONFLICT_OUT if any point to
3533 : : * previously committed transactions.
3534 : : */
1063 andres@anarazel.de 3535 [ + - + + ]:CBC 2247 : dlist_foreach_modify(iter, &MySerializableXact->outConflicts)
3536 : : {
3537 : 687 : RWConflict conflict =
943 tgl@sss.pgh.pa.us 3538 :ECB (679) : dlist_container(RWConflictData, outLink, iter.cur);
3539 : :
5427 heikki.linnakangas@i 3540 [ + + ]:CBC 687 : if (isCommit
3541 [ + + ]: 455 : && !SxactIsReadOnly(MySerializableXact)
3542 [ + + ]: 347 : && SxactIsCommitted(conflict->sxactIn))
3543 : : {
3544 [ - + ]: 96 : if ((MySerializableXact->flags & SXACT_FLAG_CONFLICT_OUT) == 0
5277 heikki.linnakangas@i 3545 [ # # ]:UBC 0 : || conflict->sxactIn->prepareSeqNo < MySerializableXact->SeqNo.earliestOutConflictCommit)
5277 heikki.linnakangas@i 3546 :CBC 96 : MySerializableXact->SeqNo.earliestOutConflictCommit = conflict->sxactIn->prepareSeqNo;
5427 3547 : 96 : MySerializableXact->flags |= SXACT_FLAG_CONFLICT_OUT;
3548 : : }
3549 : :
3550 [ + + ]: 687 : if (!isCommit
3551 [ + + ]: 455 : || SxactIsCommitted(conflict->sxactIn)
3552 [ - + ]: 337 : || (conflict->sxactIn->SeqNo.lastCommitBeforeSnapshot >= PredXact->LastSxactCommitSeqNo))
3553 : 350 : ReleaseRWConflict(conflict);
3554 : : }
3555 : :
3556 : : /*
3557 : : * Release all inConflicts from committed and read-only transactions. If
3558 : : * we're rolling back, clear them all.
3559 : : */
1063 andres@anarazel.de 3560 [ + - + + ]: 2339 : dlist_foreach_modify(iter, &MySerializableXact->inConflicts)
3561 : : {
3562 : 779 : RWConflict conflict =
943 tgl@sss.pgh.pa.us 3563 : 779 : dlist_container(RWConflictData, inLink, iter.cur);
3564 : :
5427 heikki.linnakangas@i 3565 [ + + ]: 779 : if (!isCommit
3566 [ + + ]: 602 : || SxactIsCommitted(conflict->sxactOut)
3567 [ + + ]: 417 : || SxactIsReadOnly(conflict->sxactOut))
3568 : 442 : ReleaseRWConflict(conflict);
3569 : : }
3570 : :
3571 [ + + ]: 1560 : if (!topLevelIsDeclaredReadOnly)
3572 : : {
3573 : : /*
3574 : : * Remove ourselves from the list of possible conflicts for concurrent
3575 : : * READ ONLY transactions, flagging them as unsafe if we have a
3576 : : * conflict out. If any are waiting DEFERRABLE transactions, wake them
3577 : : * up if they are known safe or known unsafe.
3578 : : */
1063 andres@anarazel.de 3579 [ + - + + ]: 1542 : dlist_foreach_modify(iter, &MySerializableXact->possibleUnsafeConflicts)
3580 : : {
3581 : 91 : RWConflict possibleUnsafeConflict =
943 tgl@sss.pgh.pa.us 3582 :ECB (91) : dlist_container(RWConflictData, outLink, iter.cur);
3583 : :
5427 heikki.linnakangas@i 3584 :CBC 91 : roXact = possibleUnsafeConflict->sxactIn;
3585 [ - + ]: 91 : Assert(MySerializableXact == possibleUnsafeConflict->sxactOut);
3586 [ - + ]: 91 : Assert(SxactIsReadOnly(roXact));
3587 : :
3588 : : /* Mark conflicted if necessary. */
3589 [ + + ]: 91 : if (isCommit
5304 3590 [ + + ]: 88 : && MyXactDidWrite
5427 3591 [ + + ]: 83 : && SxactHasConflictOut(MySerializableXact)
3592 : 13 : && (MySerializableXact->SeqNo.earliestOutConflictCommit
3593 [ + + ]: 13 : <= roXact->SeqNo.lastCommitBeforeSnapshot))
3594 : : {
3595 : : /*
3596 : : * This releases possibleUnsafeConflict (as well as all other
3597 : : * possible conflicts for roXact)
3598 : : */
3599 : 3 : FlagSxactUnsafe(roXact);
3600 : : }
3601 : : else
3602 : : {
3603 : 88 : ReleaseRWConflict(possibleUnsafeConflict);
3604 : :
3605 : : /*
3606 : : * If we were the last possible conflict, flag it safe. The
3607 : : * transaction can now safely release its predicate locks (but
3608 : : * that transaction's backend has to do that itself).
3609 : : */
1063 andres@anarazel.de 3610 [ + + ]: 88 : if (dlist_is_empty(&roXact->possibleUnsafeConflicts))
5427 heikki.linnakangas@i 3611 : 66 : roXact->flags |= SXACT_FLAG_RO_SAFE;
3612 : : }
3613 : :
3614 : : /*
3615 : : * Wake up the process for a waiting DEFERRABLE transaction if we
3616 : : * now know it's either safe or conflicted.
3617 : : */
3618 [ + + ]: 91 : if (SxactIsDeferrableWaiting(roXact) &&
3619 [ + + + - ]: 2 : (SxactIsROUnsafe(roXact) || SxactIsROSafe(roXact)))
1462 tmunro@postgresql.or 3620 : 2 : ProcSendSignal(roXact->pgprocno);
3621 : : }
3622 : : }
3623 : :
3624 : : /*
3625 : : * Check whether it's time to clean up old transactions. This can only be
3626 : : * done when the last serializable transaction with the oldest xmin among
3627 : : * serializable transactions completes. We then find the "new oldest"
3628 : : * xmin and purge any transactions which finished before this transaction
3629 : : * was launched.
3630 : : *
3631 : : * For parallel queries in read-only transactions, it might run twice. We
3632 : : * only release the reference on the first call.
3633 : : */
5427 heikki.linnakangas@i 3634 : 1560 : needToClear = false;
1017 tmunro@postgresql.or 3635 [ + + ]: 1560 : if ((partiallyReleasing ||
3636 [ + + ]: 1558 : !SxactIsPartiallyReleased(MySerializableXact)) &&
3637 [ + + ]: 1558 : TransactionIdEquals(MySerializableXact->xmin,
3638 : : PredXact->SxactGlobalXmin))
3639 : : {
5427 heikki.linnakangas@i 3640 [ - + ]: 1541 : Assert(PredXact->SxactGlobalXminCount > 0);
3641 [ + + ]: 1541 : if (--(PredXact->SxactGlobalXminCount) == 0)
3642 : : {
3643 : 868 : SetNewSxactGlobalXmin();
3644 : 868 : needToClear = true;
3645 : : }
3646 : : }
3647 : :
3648 : 1560 : LWLockRelease(SerializableXactHashLock);
3649 : :
3650 : 1560 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
3651 : :
3652 : : /* Add this to the list of transactions to check for later cleanup. */
3653 [ + + ]: 1560 : if (isCommit)
1063 andres@anarazel.de 3654 : 1231 : dlist_push_tail(FinishedSerializableTransactions,
3655 : 1231 : &MySerializableXact->finishedLink);
3656 : :
3657 : : /*
3658 : : * If we're releasing a RO_SAFE transaction in parallel mode, we'll only
3659 : : * partially release it. That's necessary because other backends may have
3660 : : * a reference to it. The leader will release the SERIALIZABLEXACT itself
3661 : : * at the end of the transaction after workers have stopped running.
3662 : : */
5427 heikki.linnakangas@i 3663 [ + + ]: 1560 : if (!isCommit)
2469 tmunro@postgresql.or 3664 : 329 : ReleaseOneSerializableXact(MySerializableXact,
3665 [ + + + + ]: 329 : isReadOnlySafe && IsInParallelMode(),
3666 : : false);
3667 : :
5427 heikki.linnakangas@i 3668 : 1560 : LWLockRelease(SerializableFinishedListLock);
3669 : :
3670 [ + + ]: 1560 : if (needToClear)
3671 : 868 : ClearOldPredicateLocks();
3672 : :
2469 tmunro@postgresql.or 3673 : 1560 : ReleasePredicateLocksLocal();
3674 : : }
3675 : :
3676 : : static void
3677 : 5883 : ReleasePredicateLocksLocal(void)
3678 : : {
5427 heikki.linnakangas@i 3679 : 5883 : MySerializableXact = InvalidSerializableXact;
5304 3680 : 5883 : MyXactDidWrite = false;
3681 : :
3682 : : /* Delete per-transaction lock table */
5427 3683 [ + + ]: 5883 : if (LocalPredicateLockHash != NULL)
3684 : : {
3685 : 1559 : hash_destroy(LocalPredicateLockHash);
3686 : 1559 : LocalPredicateLockHash = NULL;
3687 : : }
3688 : 5883 : }
3689 : :
3690 : : /*
3691 : : * Clear old predicate locks, belonging to committed transactions that are no
3692 : : * longer interesting to any in-progress transaction.
3693 : : */
3694 : : static void
3695 : 868 : ClearOldPredicateLocks(void)
3696 : : {
3697 : : dlist_mutable_iter iter;
3698 : :
3699 : : /*
3700 : : * Loop through finished transactions. They are in commit order, so we can
3701 : : * stop as soon as we find one that's still interesting.
3702 : : */
3703 : 868 : LWLockAcquire(SerializableFinishedListLock, LW_EXCLUSIVE);
3704 : 868 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
1063 andres@anarazel.de 3705 [ + - + + ]: 2106 : dlist_foreach_modify(iter, FinishedSerializableTransactions)
3706 : : {
3707 : 1248 : SERIALIZABLEXACT *finishedSxact =
943 tgl@sss.pgh.pa.us 3708 : 1248 : dlist_container(SERIALIZABLEXACT, finishedLink, iter.cur);
3709 : :
5427 heikki.linnakangas@i 3710 [ + + ]: 1248 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin)
3711 [ + + ]: 27 : || TransactionIdPrecedesOrEquals(finishedSxact->finishedBefore,
3712 : 27 : PredXact->SxactGlobalXmin))
3713 : : {
3714 : : /*
3715 : : * This transaction committed before any in-progress transaction
3716 : : * took its snapshot. It's no longer interesting.
3717 : : */
3718 : 1231 : LWLockRelease(SerializableXactHashLock);
1063 andres@anarazel.de 3719 : 1231 : dlist_delete_thoroughly(&finishedSxact->finishedLink);
5427 heikki.linnakangas@i 3720 : 1231 : ReleaseOneSerializableXact(finishedSxact, false, false);
3721 : 1231 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3722 : : }
3723 [ + - ]: 17 : else if (finishedSxact->commitSeqNo > PredXact->HavePartialClearedThrough
3101 tgl@sss.pgh.pa.us 3724 [ + + ]: 17 : && finishedSxact->commitSeqNo <= PredXact->CanPartialClearThrough)
3725 : : {
3726 : : /*
3727 : : * Any active transactions that took their snapshot before this
3728 : : * transaction committed are read-only, so we can clear part of
3729 : : * its state.
3730 : : */
5427 heikki.linnakangas@i 3731 : 7 : LWLockRelease(SerializableXactHashLock);
3732 : :
5082 3733 [ - + ]: 7 : if (SxactIsReadOnly(finishedSxact))
3734 : : {
3735 : : /* A read-only transaction can be removed entirely */
1063 andres@anarazel.de 3736 :UBC 0 : dlist_delete_thoroughly(&(finishedSxact->finishedLink));
5082 heikki.linnakangas@i 3737 : 0 : ReleaseOneSerializableXact(finishedSxact, false, false);
3738 : : }
3739 : : else
3740 : : {
3741 : : /*
3742 : : * A read-write transaction can only be partially cleared. We
3743 : : * need to keep the SERIALIZABLEXACT but can release the
3744 : : * SIREAD locks and conflicts in.
3745 : : */
5082 heikki.linnakangas@i 3746 :CBC 7 : ReleaseOneSerializableXact(finishedSxact, true, false);
3747 : : }
3748 : :
5427 3749 : 7 : PredXact->HavePartialClearedThrough = finishedSxact->commitSeqNo;
3750 : 7 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
3751 : : }
3752 : : else
3753 : : {
3754 : : /* Still interesting. */
3755 : : break;
3756 : : }
3757 : : }
3758 : 868 : LWLockRelease(SerializableXactHashLock);
3759 : :
3760 : : /*
3761 : : * Loop through predicate locks on dummy transaction for summarized data.
3762 : : */
2042 tgl@sss.pgh.pa.us 3763 : 868 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
1063 andres@anarazel.de 3764 [ + - - + ]: 868 : dlist_foreach_modify(iter, &OldCommittedSxact->predicateLocks)
3765 : : {
1063 andres@anarazel.de 3766 :UBC 0 : PREDICATELOCK *predlock =
943 tgl@sss.pgh.pa.us 3767 : 0 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
3768 : : bool canDoPartialCleanup;
3769 : :
5427 heikki.linnakangas@i 3770 : 0 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
5364 3771 [ # # ]: 0 : Assert(predlock->commitSeqNo != 0);
3772 [ # # ]: 0 : Assert(predlock->commitSeqNo != InvalidSerCommitSeqNo);
5427 3773 : 0 : canDoPartialCleanup = (predlock->commitSeqNo <= PredXact->CanPartialClearThrough);
3774 : 0 : LWLockRelease(SerializableXactHashLock);
3775 : :
3776 : : /*
3777 : : * If this lock originally belonged to an old enough transaction, we
3778 : : * can release it.
3779 : : */
3780 [ # # ]: 0 : if (canDoPartialCleanup)
3781 : : {
3782 : : PREDICATELOCKTAG tag;
3783 : : PREDICATELOCKTARGET *target;
3784 : : PREDICATELOCKTARGETTAG targettag;
3785 : : uint32 targettaghash;
3786 : : LWLock *partitionLock;
3787 : :
3788 : 0 : tag = predlock->tag;
3789 : 0 : target = tag.myTarget;
3790 : 0 : targettag = target->tag;
3791 : 0 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
3792 : 0 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
3793 : :
3794 : 0 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
3795 : :
1063 andres@anarazel.de 3796 : 0 : dlist_delete(&(predlock->targetLink));
3797 : 0 : dlist_delete(&(predlock->xactLink));
3798 : :
5427 heikki.linnakangas@i 3799 : 0 : hash_search_with_hash_value(PredicateLockHash, &tag,
3101 tgl@sss.pgh.pa.us 3800 : 0 : PredicateLockHashCodeFromTargetHashCode(&tag,
3801 : : targettaghash),
3802 : : HASH_REMOVE, NULL);
5427 heikki.linnakangas@i 3803 : 0 : RemoveTargetIfNoLongerUsed(target, targettaghash);
3804 : :
3805 : 0 : LWLockRelease(partitionLock);
3806 : : }
3807 : : }
3808 : :
2042 tgl@sss.pgh.pa.us 3809 :CBC 868 : LWLockRelease(SerializablePredicateListLock);
5427 heikki.linnakangas@i 3810 : 868 : LWLockRelease(SerializableFinishedListLock);
3811 : 868 : }
3812 : :
3813 : : /*
3814 : : * This is the normal way to delete anything from any of the predicate
3815 : : * locking hash tables. Given a transaction which we know can be deleted:
3816 : : * delete all predicate locks held by that transaction and any predicate
3817 : : * lock targets which are now unreferenced by a lock; delete all conflicts
3818 : : * for the transaction; delete all xid values for the transaction; then
3819 : : * delete the transaction.
3820 : : *
3821 : : * When the partial flag is set, we can release all predicate locks and
3822 : : * in-conflict information -- we've established that there are no longer
3823 : : * any overlapping read write transactions for which this transaction could
3824 : : * matter -- but keep the transaction entry itself and any outConflicts.
3825 : : *
3826 : : * When the summarize flag is set, we've run short of room for sxact data
3827 : : * and must summarize to the SLRU. Predicate locks are transferred to a
3828 : : * dummy "old" transaction, with duplicate locks on a single target
3829 : : * collapsing to a single lock with the "latest" commitSeqNo from among
3830 : : * the conflicting locks..
3831 : : */
3832 : : static void
3833 : 1567 : ReleaseOneSerializableXact(SERIALIZABLEXACT *sxact, bool partial,
3834 : : bool summarize)
3835 : : {
3836 : : SERIALIZABLEXIDTAG sxidtag;
3837 : : dlist_mutable_iter iter;
3838 : :
3839 [ - + ]: 1567 : Assert(sxact != NULL);
5293 3840 [ + + - + ]: 1567 : Assert(SxactIsRolledBack(sxact) || SxactIsCommitted(sxact));
5082 3841 [ + + - + ]: 1567 : Assert(partial || !SxactIsOnFinishedList(sxact));
5427 3842 [ - + ]: 1567 : Assert(LWLockHeldByMe(SerializableFinishedListLock));
3843 : :
3844 : : /*
3845 : : * First release all the predicate locks held by this xact (or transfer
3846 : : * them to OldCommittedSxact if summarize is true)
3847 : : */
2042 tgl@sss.pgh.pa.us 3848 : 1567 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2469 tmunro@postgresql.or 3849 [ + + ]: 1567 : if (IsInParallelMode())
2042 tgl@sss.pgh.pa.us 3850 : 3 : LWLockAcquire(&sxact->perXactPredicateListLock, LW_EXCLUSIVE);
1063 andres@anarazel.de 3851 [ + - + + ]: 4432 : dlist_foreach_modify(iter, &sxact->predicateLocks)
3852 : : {
3853 : 2865 : PREDICATELOCK *predlock =
943 tgl@sss.pgh.pa.us 3854 : 2865 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
3855 : : PREDICATELOCKTAG tag;
3856 : : PREDICATELOCKTARGET *target;
3857 : : PREDICATELOCKTARGETTAG targettag;
3858 : : uint32 targettaghash;
3859 : : LWLock *partitionLock;
3860 : :
5427 heikki.linnakangas@i 3861 : 2865 : tag = predlock->tag;
3862 : 2865 : target = tag.myTarget;
3863 : 2865 : targettag = target->tag;
3864 : 2865 : targettaghash = PredicateLockTargetTagHashCode(&targettag);
3865 : 2865 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
3866 : :
3867 : 2865 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
3868 : :
1063 andres@anarazel.de 3869 : 2865 : dlist_delete(&predlock->targetLink);
3870 : :
5427 heikki.linnakangas@i 3871 : 2865 : hash_search_with_hash_value(PredicateLockHash, &tag,
3101 tgl@sss.pgh.pa.us 3872 : 2865 : PredicateLockHashCodeFromTargetHashCode(&tag,
3873 : : targettaghash),
3874 : : HASH_REMOVE, NULL);
5427 heikki.linnakangas@i 3875 [ - + ]: 2865 : if (summarize)
3876 : : {
3877 : : bool found;
3878 : :
3879 : : /* Fold into dummy transaction list. */
5427 heikki.linnakangas@i 3880 :UBC 0 : tag.myXact = OldCommittedSxact;
3881 : 0 : predlock = hash_search_with_hash_value(PredicateLockHash, &tag,
3101 tgl@sss.pgh.pa.us 3882 : 0 : PredicateLockHashCodeFromTargetHashCode(&tag,
3883 : : targettaghash),
3884 : : HASH_ENTER_NULL, &found);
5427 heikki.linnakangas@i 3885 [ # # ]: 0 : if (!predlock)
3886 [ # # ]: 0 : ereport(ERROR,
3887 : : (errcode(ERRCODE_OUT_OF_MEMORY),
3888 : : errmsg("out of shared memory"),
3889 : : errhint("You might need to increase \"%s\".", "max_pred_locks_per_transaction")));
3890 [ # # ]: 0 : if (found)
3891 : : {
5364 3892 [ # # ]: 0 : Assert(predlock->commitSeqNo != 0);
3893 [ # # ]: 0 : Assert(predlock->commitSeqNo != InvalidSerCommitSeqNo);
5427 3894 [ # # ]: 0 : if (predlock->commitSeqNo < sxact->commitSeqNo)
3895 : 0 : predlock->commitSeqNo = sxact->commitSeqNo;
3896 : : }
3897 : : else
3898 : : {
1063 andres@anarazel.de 3899 : 0 : dlist_push_tail(&target->predicateLocks,
3900 : : &predlock->targetLink);
3901 : 0 : dlist_push_tail(&OldCommittedSxact->predicateLocks,
3902 : : &predlock->xactLink);
5427 heikki.linnakangas@i 3903 : 0 : predlock->commitSeqNo = sxact->commitSeqNo;
3904 : : }
3905 : : }
3906 : : else
5427 heikki.linnakangas@i 3907 :CBC 2865 : RemoveTargetIfNoLongerUsed(target, targettaghash);
3908 : :
3909 : 2865 : LWLockRelease(partitionLock);
3910 : : }
3911 : :
3912 : : /*
3913 : : * Rather than retail removal, just re-init the head after we've run
3914 : : * through the list.
3915 : : */
1063 andres@anarazel.de 3916 : 1567 : dlist_init(&sxact->predicateLocks);
3917 : :
2469 tmunro@postgresql.or 3918 [ + + ]: 1567 : if (IsInParallelMode())
2042 tgl@sss.pgh.pa.us 3919 : 3 : LWLockRelease(&sxact->perXactPredicateListLock);
3920 : 1567 : LWLockRelease(SerializablePredicateListLock);
3921 : :
5427 heikki.linnakangas@i 3922 : 1567 : sxidtag.xid = sxact->topXid;
3923 : 1567 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
3924 : :
3925 : : /* Release all outConflicts (unless 'partial' is true) */
3926 [ + + ]: 1567 : if (!partial)
3927 : : {
1063 andres@anarazel.de 3928 [ + - - + ]: 1558 : dlist_foreach_modify(iter, &sxact->outConflicts)
3929 : : {
1063 andres@anarazel.de 3930 :UBC 0 : RWConflict conflict =
943 tgl@sss.pgh.pa.us 3931 :EUB : dlist_container(RWConflictData, outLink, iter.cur);
3932 : :
5427 heikki.linnakangas@i 3933 [ # # ]:UBC 0 : if (summarize)
3934 : 0 : conflict->sxactIn->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
3935 : 0 : ReleaseRWConflict(conflict);
3936 : : }
3937 : : }
3938 : :
3939 : : /* Release all inConflicts. */
1063 andres@anarazel.de 3940 [ + - - + ]:CBC 1567 : dlist_foreach_modify(iter, &sxact->inConflicts)
3941 : : {
1063 andres@anarazel.de 3942 :UBC 0 : RWConflict conflict =
943 tgl@sss.pgh.pa.us 3943 : 0 : dlist_container(RWConflictData, inLink, iter.cur);
3944 : :
5427 heikki.linnakangas@i 3945 [ # # ]: 0 : if (summarize)
3946 : 0 : conflict->sxactOut->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
3947 : 0 : ReleaseRWConflict(conflict);
3948 : : }
3949 : :
3950 : : /* Finally, get rid of the xid and the record of the transaction itself. */
5427 heikki.linnakangas@i 3951 [ + + ]:CBC 1567 : if (!partial)
3952 : : {
3953 [ + + ]: 1558 : if (sxidtag.xid != InvalidTransactionId)
3954 : 1291 : hash_search(SerializableXidHash, &sxidtag, HASH_REMOVE, NULL);
3955 : 1558 : ReleasePredXact(sxact);
3956 : : }
3957 : :
3958 : 1567 : LWLockRelease(SerializableXactHashLock);
3959 : 1567 : }
3960 : :
3961 : : /*
3962 : : * Tests whether the given top level transaction is concurrent with
3963 : : * (overlaps) our current transaction.
3964 : : *
3965 : : * We need to identify the top level transaction for SSI, anyway, so pass
3966 : : * that to this function to save the overhead of checking the snapshot's
3967 : : * subxip array.
3968 : : */
3969 : : static bool
3970 : 536 : XidIsConcurrent(TransactionId xid)
3971 : : {
3972 : : Snapshot snap;
3973 : :
3974 [ - + ]: 536 : Assert(TransactionIdIsValid(xid));
3975 [ - + ]: 536 : Assert(!TransactionIdEquals(xid, GetTopTransactionIdIfAny()));
3976 : :
3977 : 536 : snap = GetTransactionSnapshot();
3978 : :
3979 [ - + ]: 536 : if (TransactionIdPrecedes(xid, snap->xmin))
5427 heikki.linnakangas@i 3980 :UBC 0 : return false;
3981 : :
5427 heikki.linnakangas@i 3982 [ + + ]:CBC 536 : if (TransactionIdFollowsOrEquals(xid, snap->xmax))
3983 : 526 : return true;
3984 : :
1182 michael@paquier.xyz 3985 : 10 : return pg_lfind32(xid, snap->xip, snap->xcnt);
3986 : : }
3987 : :
3988 : : bool
2150 tmunro@postgresql.or 3989 : 33220743 : CheckForSerializableConflictOutNeeded(Relation relation, Snapshot snapshot)
3990 : : {
3991 [ + + ]: 33220743 : if (!SerializationNeededForRead(relation, snapshot))
3992 : 33194737 : return false;
3993 : :
3994 : : /* Check if someone else has already decided that we need to die */
3995 [ - + ]: 26006 : if (SxactIsDoomed(MySerializableXact))
3996 : : {
2150 tmunro@postgresql.or 3997 [ # # ]:UBC 0 : ereport(ERROR,
3998 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
3999 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4000 : : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4001 : : errhint("The transaction might succeed if retried.")));
4002 : : }
4003 : :
2150 tmunro@postgresql.or 4004 :CBC 26006 : return true;
4005 : : }
4006 : :
4007 : : /*
4008 : : * CheckForSerializableConflictOut
4009 : : * A table AM is reading a tuple that has been modified. If it determines
4010 : : * that the tuple version it is reading is not visible to us, it should
4011 : : * pass in the top level xid of the transaction that created it.
4012 : : * Otherwise, if it determines that it is visible to us but it has been
4013 : : * deleted or there is a newer version available due to an update, it
4014 : : * should pass in the top level xid of the modifying transaction.
4015 : : *
4016 : : * This function will check for overlap with our own transaction. If the given
4017 : : * xid is also serializable and the transactions overlap (i.e., they cannot see
4018 : : * each other's writes), then we have a conflict out.
4019 : : */
4020 : : void
4021 : 571 : CheckForSerializableConflictOut(Relation relation, TransactionId xid, Snapshot snapshot)
4022 : : {
4023 : : SERIALIZABLEXIDTAG sxidtag;
4024 : : SERIALIZABLEXID *sxid;
4025 : : SERIALIZABLEXACT *sxact;
4026 : :
5299 heikki.linnakangas@i 4027 [ - + ]: 571 : if (!SerializationNeededForRead(relation, snapshot))
5427 4028 : 204 : return;
4029 : :
4030 : : /* Check if someone else has already decided that we need to die */
5299 4031 [ - + ]: 571 : if (SxactIsDoomed(MySerializableXact))
4032 : : {
5427 heikki.linnakangas@i 4033 [ # # ]:UBC 0 : ereport(ERROR,
4034 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4035 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4036 : : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict out checking."),
4037 : : errhint("The transaction might succeed if retried.")));
4038 : : }
5427 heikki.linnakangas@i 4039 [ - + ]:CBC 571 : Assert(TransactionIdIsValid(xid));
4040 : :
4041 [ - + ]: 571 : if (TransactionIdEquals(xid, GetTopTransactionIdIfAny()))
5427 heikki.linnakangas@i 4042 :UBC 0 : return;
4043 : :
4044 : : /*
4045 : : * Find sxact or summarized info for the top level xid.
4046 : : */
5427 heikki.linnakangas@i 4047 :CBC 571 : sxidtag.xid = xid;
4048 : 571 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4049 : : sxid = (SERIALIZABLEXID *)
4050 : 571 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
4051 [ + + ]: 571 : if (!sxid)
4052 : : {
4053 : : /*
4054 : : * Transaction not found in "normal" SSI structures. Check whether it
4055 : : * got pushed out to SLRU storage for "old committed" transactions.
4056 : : */
4057 : : SerCommitSeqNo conflictCommitSeqNo;
4058 : :
2042 tgl@sss.pgh.pa.us 4059 : 25 : conflictCommitSeqNo = SerialGetMinConflictCommitSeqNo(xid);
5427 heikki.linnakangas@i 4060 [ - + ]: 25 : if (conflictCommitSeqNo != 0)
4061 : : {
5427 heikki.linnakangas@i 4062 [ # # ]:UBC 0 : if (conflictCommitSeqNo != InvalidSerCommitSeqNo
4063 [ # # ]: 0 : && (!SxactIsReadOnly(MySerializableXact)
4064 : 0 : || conflictCommitSeqNo
4065 [ # # ]: 0 : <= MySerializableXact->SeqNo.lastCommitBeforeSnapshot))
4066 [ # # ]: 0 : ereport(ERROR,
4067 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4068 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4069 : : errdetail_internal("Reason code: Canceled on conflict out to old pivot %u.", xid),
4070 : : errhint("The transaction might succeed if retried.")));
4071 : :
4072 [ # # ]: 0 : if (SxactHasSummaryConflictIn(MySerializableXact)
1063 andres@anarazel.de 4073 [ # # ]: 0 : || !dlist_is_empty(&MySerializableXact->inConflicts))
5427 heikki.linnakangas@i 4074 [ # # ]: 0 : ereport(ERROR,
4075 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4076 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4077 : : errdetail_internal("Reason code: Canceled on identification as a pivot, with conflict out to old committed transaction %u.", xid),
4078 : : errhint("The transaction might succeed if retried.")));
4079 : :
4080 : 0 : MySerializableXact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4081 : : }
4082 : :
4083 : : /* It's not serializable or otherwise not important. */
5427 heikki.linnakangas@i 4084 :CBC 25 : LWLockRelease(SerializableXactHashLock);
4085 : 25 : return;
4086 : : }
4087 : 546 : sxact = sxid->myXact;
4088 [ - + ]: 546 : Assert(TransactionIdEquals(sxact->topXid, xid));
5299 4089 [ + - + + ]: 546 : if (sxact == MySerializableXact || SxactIsDoomed(sxact))
4090 : : {
4091 : : /* Can't conflict with ourself or a transaction that will roll back. */
5427 4092 : 4 : LWLockRelease(SerializableXactHashLock);
4093 : 4 : return;
4094 : : }
4095 : :
4096 : : /*
4097 : : * We have a conflict out to a transaction which has a conflict out to a
4098 : : * summarized transaction. That summarized transaction must have
4099 : : * committed first, and we can't tell when it committed in relation to our
4100 : : * snapshot acquisition, so something needs to be canceled.
4101 : : */
4102 [ - + ]: 542 : if (SxactHasSummaryConflictOut(sxact))
4103 : : {
5427 heikki.linnakangas@i 4104 [ # # ]:UBC 0 : if (!SxactIsPrepared(sxact))
4105 : : {
5299 4106 : 0 : sxact->flags |= SXACT_FLAG_DOOMED;
5427 4107 : 0 : LWLockRelease(SerializableXactHashLock);
4108 : 0 : return;
4109 : : }
4110 : : else
4111 : : {
4112 : 0 : LWLockRelease(SerializableXactHashLock);
4113 [ # # ]: 0 : ereport(ERROR,
4114 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4115 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4116 : : errdetail_internal("Reason code: Canceled on conflict out to old pivot."),
4117 : : errhint("The transaction might succeed if retried.")));
4118 : : }
4119 : : }
4120 : :
4121 : : /*
4122 : : * If this is a read-only transaction and the writing transaction has
4123 : : * committed, and it doesn't have a rw-conflict to a transaction which
4124 : : * committed before it, no conflict.
4125 : : */
5427 heikki.linnakangas@i 4126 [ + + ]:CBC 542 : if (SxactIsReadOnly(MySerializableXact)
4127 [ + + ]: 119 : && SxactIsCommitted(sxact)
4128 [ + - ]: 8 : && !SxactHasSummaryConflictOut(sxact)
4129 [ + + ]: 8 : && (!SxactHasConflictOut(sxact)
4130 [ - + ]: 2 : || MySerializableXact->SeqNo.lastCommitBeforeSnapshot < sxact->SeqNo.earliestOutConflictCommit))
4131 : : {
4132 : : /* Read-only transaction will appear to run first. No conflict. */
4133 : 6 : LWLockRelease(SerializableXactHashLock);
4134 : 6 : return;
4135 : : }
4136 : :
4137 [ - + ]: 536 : if (!XidIsConcurrent(xid))
4138 : : {
4139 : : /* This write was already in our snapshot; no conflict. */
5427 heikki.linnakangas@i 4140 :UBC 0 : LWLockRelease(SerializableXactHashLock);
4141 : 0 : return;
4142 : : }
4143 : :
5304 heikki.linnakangas@i 4144 [ + + ]:CBC 536 : if (RWConflictExists(MySerializableXact, sxact))
4145 : : {
4146 : : /* We don't want duplicate conflict records in the list. */
5427 4147 : 169 : LWLockRelease(SerializableXactHashLock);
4148 : 169 : return;
4149 : : }
4150 : :
4151 : : /*
4152 : : * Flag the conflict. But first, if this conflict creates a dangerous
4153 : : * structure, ereport an error.
4154 : : */
5304 4155 : 367 : FlagRWConflict(MySerializableXact, sxact);
5427 4156 : 354 : LWLockRelease(SerializableXactHashLock);
4157 : : }
4158 : :
4159 : : /*
4160 : : * Check a particular target for rw-dependency conflict in. A subroutine of
4161 : : * CheckForSerializableConflictIn().
4162 : : */
4163 : : static void
4164 : 7540 : CheckTargetForConflictsIn(PREDICATELOCKTARGETTAG *targettag)
4165 : : {
4166 : : uint32 targettaghash;
4167 : : LWLock *partitionLock;
4168 : : PREDICATELOCKTARGET *target;
5326 rhaas@postgresql.org 4169 : 7540 : PREDICATELOCK *mypredlock = NULL;
4170 : : PREDICATELOCKTAG mypredlocktag;
4171 : : dlist_mutable_iter iter;
4172 : :
5427 heikki.linnakangas@i 4173 [ - + ]: 7540 : Assert(MySerializableXact != InvalidSerializableXact);
4174 : :
4175 : : /*
4176 : : * The same hash and LW lock apply to the lock target and the lock itself.
4177 : : */
4178 : 7540 : targettaghash = PredicateLockTargetTagHashCode(targettag);
4179 : 7540 : partitionLock = PredicateLockHashPartitionLock(targettaghash);
4180 : 7540 : LWLockAcquire(partitionLock, LW_SHARED);
4181 : : target = (PREDICATELOCKTARGET *)
4182 : 7540 : hash_search_with_hash_value(PredicateLockTargetHash,
4183 : : targettag, targettaghash,
4184 : : HASH_FIND, NULL);
4185 [ + + ]: 7540 : if (!target)
4186 : : {
4187 : : /* Nothing has this target locked; we're done here. */
4188 : 5653 : LWLockRelease(partitionLock);
5405 4189 : 5653 : return;
4190 : : }
4191 : :
4192 : : /*
4193 : : * Each lock for an overlapping transaction represents a conflict: a
4194 : : * rw-dependency in to this transaction.
4195 : : */
5427 4196 : 1887 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4197 : :
1063 andres@anarazel.de 4198 [ + - + + ]: 4254 : dlist_foreach_modify(iter, &target->predicateLocks)
4199 : : {
4200 : 2434 : PREDICATELOCK *predlock =
943 tgl@sss.pgh.pa.us 4201 : 2434 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
1063 andres@anarazel.de 4202 : 2434 : SERIALIZABLEXACT *sxact = predlock->tag.myXact;
4203 : :
5427 heikki.linnakangas@i 4204 [ + + ]: 2434 : if (sxact == MySerializableXact)
4205 : : {
4206 : : /*
4207 : : * If we're getting a write lock on a tuple, we don't need a
4208 : : * predicate (SIREAD) lock on the same tuple. We can safely remove
4209 : : * our SIREAD lock, but we'll defer doing so until after the loop
4210 : : * because that requires upgrading to an exclusive partition lock.
4211 : : *
4212 : : * We can't use this optimization within a subtransaction because
4213 : : * the subtransaction could roll back, and we would be left
4214 : : * without any lock at the top level.
4215 : : */
5367 rhaas@postgresql.org 4216 [ + - ]: 1576 : if (!IsSubTransaction()
4217 [ + + ]: 1576 : && GET_PREDICATELOCKTARGETTAG_OFFSET(*targettag))
4218 : : {
5326 4219 : 396 : mypredlock = predlock;
4220 : 396 : mypredlocktag = predlock->tag;
4221 : : }
4222 : : }
5299 heikki.linnakangas@i 4223 [ + - ]: 858 : else if (!SxactIsDoomed(sxact)
5427 4224 [ + + ]: 858 : && (!SxactIsCommitted(sxact)
4225 [ + + ]: 88 : || TransactionIdPrecedes(GetTransactionSnapshot()->xmin,
4226 : : sxact->finishedBefore))
5304 4227 [ + + ]: 849 : && !RWConflictExists(sxact, MySerializableXact))
4228 : : {
5427 4229 : 505 : LWLockRelease(SerializableXactHashLock);
4230 : 505 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4231 : :
4232 : : /*
4233 : : * Re-check after getting exclusive lock because the other
4234 : : * transaction may have flagged a conflict.
4235 : : */
5299 4236 [ + - ]: 505 : if (!SxactIsDoomed(sxact)
5370 rhaas@postgresql.org 4237 [ + + ]: 505 : && (!SxactIsCommitted(sxact)
4238 [ + - ]: 77 : || TransactionIdPrecedes(GetTransactionSnapshot()->xmin,
4239 : : sxact->finishedBefore))
5304 heikki.linnakangas@i 4240 [ + - ]: 505 : && !RWConflictExists(sxact, MySerializableXact))
4241 : : {
4242 : 505 : FlagRWConflict(sxact, MySerializableXact);
4243 : : }
4244 : :
5427 4245 : 438 : LWLockRelease(SerializableXactHashLock);
4246 : 438 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4247 : : }
4248 : : }
4249 : 1820 : LWLockRelease(SerializableXactHashLock);
4250 : 1820 : LWLockRelease(partitionLock);
4251 : :
4252 : : /*
4253 : : * If we found one of our own SIREAD locks to remove, remove it now.
4254 : : *
4255 : : * At this point our transaction already has a RowExclusiveLock on the
4256 : : * relation, so we are OK to drop the predicate lock on the tuple, if
4257 : : * found, without fearing that another write against the tuple will occur
4258 : : * before the MVCC information makes it to the buffer.
4259 : : */
5326 rhaas@postgresql.org 4260 [ + + ]: 1820 : if (mypredlock != NULL)
4261 : : {
4262 : : uint32 predlockhashcode;
4263 : : PREDICATELOCK *rmpredlock;
4264 : :
2042 tgl@sss.pgh.pa.us 4265 : 389 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
2469 tmunro@postgresql.or 4266 [ - + ]: 389 : if (IsInParallelMode())
2042 tgl@sss.pgh.pa.us 4267 :UBC 0 : LWLockAcquire(&MySerializableXact->perXactPredicateListLock, LW_EXCLUSIVE);
5326 rhaas@postgresql.org 4268 :CBC 389 : LWLockAcquire(partitionLock, LW_EXCLUSIVE);
4269 : 389 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4270 : :
4271 : : /*
4272 : : * Remove the predicate lock from shared memory, if it wasn't removed
4273 : : * while the locks were released. One way that could happen is from
4274 : : * autovacuum cleaning up an index.
4275 : : */
4276 : 389 : predlockhashcode = PredicateLockHashCodeFromTargetHashCode
4277 : : (&mypredlocktag, targettaghash);
4278 : : rmpredlock = (PREDICATELOCK *)
4279 : 389 : hash_search_with_hash_value(PredicateLockHash,
4280 : : &mypredlocktag,
4281 : : predlockhashcode,
4282 : : HASH_FIND, NULL);
4283 [ + - ]: 389 : if (rmpredlock != NULL)
4284 : : {
4285 [ - + ]: 389 : Assert(rmpredlock == mypredlock);
4286 : :
1063 andres@anarazel.de 4287 : 389 : dlist_delete(&(mypredlock->targetLink));
4288 : 389 : dlist_delete(&(mypredlock->xactLink));
4289 : :
4290 : : rmpredlock = (PREDICATELOCK *)
5326 rhaas@postgresql.org 4291 : 389 : hash_search_with_hash_value(PredicateLockHash,
4292 : : &mypredlocktag,
4293 : : predlockhashcode,
4294 : : HASH_REMOVE, NULL);
4295 [ - + ]: 389 : Assert(rmpredlock == mypredlock);
4296 : :
4297 : 389 : RemoveTargetIfNoLongerUsed(target, targettaghash);
4298 : : }
4299 : :
4300 : 389 : LWLockRelease(SerializableXactHashLock);
4301 : 389 : LWLockRelease(partitionLock);
2469 tmunro@postgresql.or 4302 [ - + ]: 389 : if (IsInParallelMode())
2042 tgl@sss.pgh.pa.us 4303 :UBC 0 : LWLockRelease(&MySerializableXact->perXactPredicateListLock);
2042 tgl@sss.pgh.pa.us 4304 :CBC 389 : LWLockRelease(SerializablePredicateListLock);
4305 : :
5326 rhaas@postgresql.org 4306 [ + - ]: 389 : if (rmpredlock != NULL)
4307 : : {
4308 : : /*
4309 : : * Remove entry in local lock table if it exists. It's OK if it
4310 : : * doesn't exist; that means the lock was transferred to a new
4311 : : * target by a different backend.
4312 : : */
4313 : 389 : hash_search_with_hash_value(LocalPredicateLockHash,
4314 : : targettag, targettaghash,
4315 : : HASH_REMOVE, NULL);
4316 : :
4317 : 389 : DecrementParentLocks(targettag);
4318 : : }
4319 : : }
4320 : : }
4321 : :
4322 : : /*
4323 : : * CheckForSerializableConflictIn
4324 : : * We are writing the given tuple. If that indicates a rw-conflict
4325 : : * in from another serializable transaction, take appropriate action.
4326 : : *
4327 : : * Skip checking for any granularity for which a parameter is missing.
4328 : : *
4329 : : * A tuple update or delete is in conflict if we have a predicate lock
4330 : : * against the relation or page in which the tuple exists, or against the
4331 : : * tuple itself.
4332 : : */
4333 : : void
48 peter@eisentraut.org 4334 :GNC 16650512 : CheckForSerializableConflictIn(Relation relation, const ItemPointerData *tid, BlockNumber blkno)
4335 : : {
4336 : : PREDICATELOCKTARGETTAG targettag;
4337 : :
5299 heikki.linnakangas@i 4338 [ + + ]:CBC 16650512 : if (!SerializationNeededForWrite(relation))
5427 4339 : 16646047 : return;
4340 : :
4341 : : /* Check if someone else has already decided that we need to die */
5299 4342 [ + + ]: 4465 : if (SxactIsDoomed(MySerializableXact))
5427 4343 [ + - ]: 1 : ereport(ERROR,
4344 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4345 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4346 : : errdetail_internal("Reason code: Canceled on identification as a pivot, during conflict in checking."),
4347 : : errhint("The transaction might succeed if retried.")));
4348 : :
4349 : : /*
4350 : : * We're doing a write which might cause rw-conflicts now or later.
4351 : : * Memorize that fact.
4352 : : */
5304 4353 : 4464 : MyXactDidWrite = true;
4354 : :
4355 : : /*
4356 : : * It is important that we check for locks from the finest granularity to
4357 : : * the coarsest granularity, so that granularity promotion doesn't cause
4358 : : * us to miss a lock. The new (coarser) lock will be acquired before the
4359 : : * old (finer) locks are released.
4360 : : *
4361 : : * It is not possible to take and hold a lock across the checks for all
4362 : : * granularities because each target could be in a separate partition.
4363 : : */
2150 tmunro@postgresql.or 4364 [ + + ]: 4464 : if (tid != NULL)
4365 : : {
5427 heikki.linnakangas@i 4366 : 651 : SET_PREDICATELOCKTARGETTAG_TUPLE(targettag,
4367 : : relation->rd_locator.dbOid,
4368 : : relation->rd_id,
4369 : : ItemPointerGetBlockNumber(tid),
4370 : : ItemPointerGetOffsetNumber(tid));
4371 : 651 : CheckTargetForConflictsIn(&targettag);
4372 : : }
4373 : :
2150 tmunro@postgresql.or 4374 [ + + ]: 4441 : if (blkno != InvalidBlockNumber)
4375 : : {
5427 heikki.linnakangas@i 4376 : 2478 : SET_PREDICATELOCKTARGETTAG_PAGE(targettag,
4377 : : relation->rd_locator.dbOid,
4378 : : relation->rd_id,
4379 : : blkno);
4380 : 2478 : CheckTargetForConflictsIn(&targettag);
4381 : : }
4382 : :
4383 : 4411 : SET_PREDICATELOCKTARGETTAG_RELATION(targettag,
4384 : : relation->rd_locator.dbOid,
4385 : : relation->rd_id);
4386 : 4411 : CheckTargetForConflictsIn(&targettag);
4387 : : }
4388 : :
4389 : : /*
4390 : : * CheckTableForSerializableConflictIn
4391 : : * The entire table is going through a DDL-style logical mass delete
4392 : : * like TRUNCATE or DROP TABLE. If that causes a rw-conflict in from
4393 : : * another serializable transaction, take appropriate action.
4394 : : *
4395 : : * While these operations do not operate entirely within the bounds of
4396 : : * snapshot isolation, they can occur inside a serializable transaction, and
4397 : : * will logically occur after any reads which saw rows which were destroyed
4398 : : * by these operations, so we do what we can to serialize properly under
4399 : : * SSI.
4400 : : *
4401 : : * The relation passed in must be a heap relation. Any predicate lock of any
4402 : : * granularity on the heap will cause a rw-conflict in to this transaction.
4403 : : * Predicate locks on indexes do not matter because they only exist to guard
4404 : : * against conflicting inserts into the index, and this is a mass *delete*.
4405 : : * When a table is truncated or dropped, the index will also be truncated
4406 : : * or dropped, and we'll deal with locks on the index when that happens.
4407 : : *
4408 : : * Dropping or truncating a table also needs to drop any existing predicate
4409 : : * locks on heap tuples or pages, because they're about to go away. This
4410 : : * should be done before altering the predicate locks because the transaction
4411 : : * could be rolled back because of a conflict, in which case the lock changes
4412 : : * are not needed. (At the moment, we don't actually bother to drop the
4413 : : * existing locks on a dropped or truncated table at the moment. That might
4414 : : * lead to some false positives, but it doesn't seem worth the trouble.)
4415 : : */
4416 : : void
5292 4417 : 25352 : CheckTableForSerializableConflictIn(Relation relation)
4418 : : {
4419 : : HASH_SEQ_STATUS seqstat;
4420 : : PREDICATELOCKTARGET *target;
4421 : : Oid dbId;
4422 : : Oid heapId;
4423 : : int i;
4424 : :
4425 : : /*
4426 : : * Bail out quickly if there are no serializable transactions running.
4427 : : * It's safe to check this without taking locks because the caller is
4428 : : * holding an ACCESS EXCLUSIVE lock on the relation. No new locks which
4429 : : * would matter here can be acquired while that is held.
4430 : : */
5306 4431 [ + + ]: 25352 : if (!TransactionIdIsValid(PredXact->SxactGlobalXmin))
4432 : 25331 : return;
4433 : :
5299 4434 [ + + ]: 155 : if (!SerializationNeededForWrite(relation))
5306 4435 : 134 : return;
4436 : :
4437 : : /*
4438 : : * We're doing a write which might cause rw-conflicts now or later.
4439 : : * Memorize that fact.
4440 : : */
5304 4441 : 21 : MyXactDidWrite = true;
4442 : :
5306 4443 [ - + ]: 21 : Assert(relation->rd_index == NULL); /* not an index relation */
4444 : :
1260 rhaas@postgresql.org 4445 : 21 : dbId = relation->rd_locator.dbOid;
5306 heikki.linnakangas@i 4446 : 21 : heapId = relation->rd_id;
4447 : :
2042 tgl@sss.pgh.pa.us 4448 : 21 : LWLockAcquire(SerializablePredicateListLock, LW_EXCLUSIVE);
5306 heikki.linnakangas@i 4449 [ + + ]: 357 : for (i = 0; i < NUM_PREDICATELOCK_PARTITIONS; i++)
4342 rhaas@postgresql.org 4450 : 336 : LWLockAcquire(PredicateLockHashPartitionLockByIndex(i), LW_SHARED);
3700 kgrittn@postgresql.o 4451 : 21 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4452 : :
4453 : : /* Scan through target list */
5306 heikki.linnakangas@i 4454 : 21 : hash_seq_init(&seqstat, PredicateLockTargetHash);
4455 : :
4456 [ + + ]: 72 : while ((target = (PREDICATELOCKTARGET *) hash_seq_search(&seqstat)))
4457 : : {
4458 : : dlist_mutable_iter iter;
4459 : :
4460 : : /*
4461 : : * Check whether this is a target which needs attention.
4462 : : */
4463 [ + + ]: 51 : if (GET_PREDICATELOCKTARGETTAG_RELATION(target->tag) != heapId)
4464 : 42 : continue; /* wrong relation id */
5306 heikki.linnakangas@i 4465 [ - + ]:GBC 9 : if (GET_PREDICATELOCKTARGETTAG_DB(target->tag) != dbId)
5306 heikki.linnakangas@i 4466 :UBC 0 : continue; /* wrong database id */
4467 : :
4468 : : /*
4469 : : * Loop through locks for this target and flag conflicts.
4470 : : */
1063 andres@anarazel.de 4471 [ + - + + ]:GBC 18 : dlist_foreach_modify(iter, &target->predicateLocks)
4472 : : {
4473 : 9 : PREDICATELOCK *predlock =
943 tgl@sss.pgh.pa.us 4474 : 9 : dlist_container(PREDICATELOCK, targetLink, iter.cur);
4475 : :
5306 heikki.linnakangas@i 4476 [ - + ]: 9 : if (predlock->tag.myXact != MySerializableXact
3101 tgl@sss.pgh.pa.us 4477 [ # # ]:UBC 0 : && !RWConflictExists(predlock->tag.myXact, MySerializableXact))
4478 : : {
5304 heikki.linnakangas@i 4479 : 0 : FlagRWConflict(predlock->tag.myXact, MySerializableXact);
4480 : : }
4481 : : }
4482 : : }
4483 : :
4484 : : /* Release locks in reverse order */
5306 heikki.linnakangas@i 4485 :CBC 21 : LWLockRelease(SerializableXactHashLock);
4486 [ + + ]: 357 : for (i = NUM_PREDICATELOCK_PARTITIONS - 1; i >= 0; i--)
4342 rhaas@postgresql.org 4487 : 336 : LWLockRelease(PredicateLockHashPartitionLockByIndex(i));
2042 tgl@sss.pgh.pa.us 4488 : 21 : LWLockRelease(SerializablePredicateListLock);
4489 : : }
4490 : :
4491 : :
4492 : : /*
4493 : : * Flag a rw-dependency between two serializable transactions.
4494 : : *
4495 : : * The caller is responsible for ensuring that we have a LW lock on
4496 : : * the transaction hash table.
4497 : : */
4498 : : static void
5427 heikki.linnakangas@i 4499 : 872 : FlagRWConflict(SERIALIZABLEXACT *reader, SERIALIZABLEXACT *writer)
4500 : : {
4501 [ - + ]: 872 : Assert(reader != writer);
4502 : :
4503 : : /* First, see if this conflict causes failure. */
4504 : 872 : OnConflict_CheckForSerializationFailure(reader, writer);
4505 : :
4506 : : /* Actually do the conflict flagging. */
4507 [ - + ]: 792 : if (reader == OldCommittedSxact)
5427 heikki.linnakangas@i 4508 :UBC 0 : writer->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
5427 heikki.linnakangas@i 4509 [ - + ]:CBC 792 : else if (writer == OldCommittedSxact)
5427 heikki.linnakangas@i 4510 :UBC 0 : reader->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4511 : : else
5427 heikki.linnakangas@i 4512 :CBC 792 : SetRWConflict(reader, writer);
4513 : 792 : }
4514 : :
4515 : : /*----------------------------------------------------------------------------
4516 : : * We are about to add a RW-edge to the dependency graph - check that we don't
4517 : : * introduce a dangerous structure by doing so, and abort one of the
4518 : : * transactions if so.
4519 : : *
4520 : : * A serialization failure can only occur if there is a dangerous structure
4521 : : * in the dependency graph:
4522 : : *
4523 : : * Tin ------> Tpivot ------> Tout
4524 : : * rw rw
4525 : : *
4526 : : * Furthermore, Tout must commit first.
4527 : : *
4528 : : * One more optimization is that if Tin is declared READ ONLY (or commits
4529 : : * without writing), we can only have a problem if Tout committed before Tin
4530 : : * acquired its snapshot.
4531 : : *----------------------------------------------------------------------------
4532 : : */
4533 : : static void
5292 tgl@sss.pgh.pa.us 4534 : 872 : OnConflict_CheckForSerializationFailure(const SERIALIZABLEXACT *reader,
4535 : : SERIALIZABLEXACT *writer)
4536 : : {
4537 : : bool failure;
4538 : :
5427 heikki.linnakangas@i 4539 [ - + ]: 872 : Assert(LWLockHeldByMe(SerializableXactHashLock));
4540 : :
4541 : 872 : failure = false;
4542 : :
4543 : : /*------------------------------------------------------------------------
4544 : : * Check for already-committed writer with rw-conflict out flagged
4545 : : * (conflict-flag on W means that T2 committed before W):
4546 : : *
4547 : : * R ------> W ------> T2
4548 : : * rw rw
4549 : : *
4550 : : * That is a dangerous structure, so we must abort. (Since the writer
4551 : : * has already committed, we must be the reader)
4552 : : *------------------------------------------------------------------------
4553 : : */
4554 [ + + ]: 872 : if (SxactIsCommitted(writer)
3101 tgl@sss.pgh.pa.us 4555 [ + + - + ]: 18 : && (SxactHasConflictOut(writer) || SxactHasSummaryConflictOut(writer)))
5427 heikki.linnakangas@i 4556 : 2 : failure = true;
4557 : :
4558 : : /*------------------------------------------------------------------------
4559 : : * Check whether the writer has become a pivot with an out-conflict
4560 : : * committed transaction (T2), and T2 committed first:
4561 : : *
4562 : : * R ------> W ------> T2
4563 : : * rw rw
4564 : : *
4565 : : * Because T2 must've committed first, there is no anomaly if:
4566 : : * - the reader committed before T2
4567 : : * - the writer committed before T2
4568 : : * - the reader is a READ ONLY transaction and the reader was concurrent
4569 : : * with T2 (= reader acquired its snapshot before T2 committed)
4570 : : *
4571 : : * We also handle the case that T2 is prepared but not yet committed
4572 : : * here. In that case T2 has already checked for conflicts, so if it
4573 : : * commits first, making the above conflict real, it's too late for it
4574 : : * to abort.
4575 : : *------------------------------------------------------------------------
4576 : : */
1063 andres@anarazel.de 4577 [ + + - + ]: 872 : if (!failure && SxactHasSummaryConflictOut(writer))
1063 andres@anarazel.de 4578 :UBC 0 : failure = true;
1063 andres@anarazel.de 4579 [ + + ]:CBC 872 : else if (!failure)
4580 : : {
4581 : : dlist_iter iter;
4582 : :
4583 [ + - + + ]: 1087 : dlist_foreach(iter, &writer->outConflicts)
4584 : : {
4585 : 292 : RWConflict conflict =
943 tgl@sss.pgh.pa.us 4586 :ECB (288) : dlist_container(RWConflictData, outLink, iter.cur);
5315 heikki.linnakangas@i 4587 :CBC 292 : SERIALIZABLEXACT *t2 = conflict->sxactIn;
4588 : :
5277 4589 [ + + ]: 292 : if (SxactIsPrepared(t2)
5315 4590 [ + + ]: 81 : && (!SxactIsCommitted(reader)
5277 4591 [ + - ]: 64 : || t2->prepareSeqNo <= reader->commitSeqNo)
5315 4592 [ - + ]: 81 : && (!SxactIsCommitted(writer)
5277 heikki.linnakangas@i 4593 [ # # ]:UBC 0 : || t2->prepareSeqNo <= writer->commitSeqNo)
5315 heikki.linnakangas@i 4594 [ + + ]:CBC 81 : && (!SxactIsReadOnly(reader)
3101 tgl@sss.pgh.pa.us 4595 [ + + ]: 12 : || t2->prepareSeqNo <= reader->SeqNo.lastCommitBeforeSnapshot))
4596 : : {
5427 heikki.linnakangas@i 4597 : 75 : failure = true;
4598 : 75 : break;
4599 : : }
4600 : : }
4601 : : }
4602 : :
4603 : : /*------------------------------------------------------------------------
4604 : : * Check whether the reader has become a pivot with a writer
4605 : : * that's committed (or prepared):
4606 : : *
4607 : : * T0 ------> R ------> W
4608 : : * rw rw
4609 : : *
4610 : : * Because W must've committed first for an anomaly to occur, there is no
4611 : : * anomaly if:
4612 : : * - T0 committed before the writer
4613 : : * - T0 is READ ONLY, and overlaps the writer
4614 : : *------------------------------------------------------------------------
4615 : : */
5277 4616 [ + + + + : 872 : if (!failure && SxactIsPrepared(writer) && !SxactIsReadOnly(reader))
+ - ]
4617 : : {
5315 4618 [ - + ]: 18 : if (SxactHasSummaryConflictIn(reader))
4619 : : {
5427 heikki.linnakangas@i 4620 :UBC 0 : failure = true;
4621 : : }
4622 : : else
4623 : : {
4624 : : dlist_iter iter;
4625 : :
4626 : : /*
4627 : : * The unconstify is needed as we have no const version of
4628 : : * dlist_foreach().
4629 : : */
1063 andres@anarazel.de 4630 [ + - + + ]:CBC 18 : dlist_foreach(iter, &unconstify(SERIALIZABLEXACT *, reader)->inConflicts)
4631 : : {
4632 : 11 : const RWConflict conflict =
943 tgl@sss.pgh.pa.us 4633 : 11 : dlist_container(RWConflictData, inLink, iter.cur);
1063 andres@anarazel.de 4634 : 11 : const SERIALIZABLEXACT *t0 = conflict->sxactOut;
4635 : :
4636 [ + - ]: 11 : if (!SxactIsDoomed(t0)
4637 [ + - ]: 11 : && (!SxactIsCommitted(t0)
4638 [ + - ]: 11 : || t0->commitSeqNo >= writer->prepareSeqNo)
4639 [ - + ]: 11 : && (!SxactIsReadOnly(t0)
1063 andres@anarazel.de 4640 [ # # ]:UBC 0 : || t0->SeqNo.lastCommitBeforeSnapshot >= writer->prepareSeqNo))
4641 : : {
1063 andres@anarazel.de 4642 :CBC 11 : failure = true;
4643 : 11 : break;
4644 : : }
4645 : : }
4646 : : }
4647 : : }
4648 : :
5427 heikki.linnakangas@i 4649 [ + + ]: 872 : if (failure)
4650 : : {
4651 : : /*
4652 : : * We have to kill a transaction to avoid a possible anomaly from
4653 : : * occurring. If the writer is us, we can just ereport() to cause a
4654 : : * transaction abort. Otherwise we flag the writer for termination,
4655 : : * causing it to abort when it tries to commit. However, if the writer
4656 : : * is a prepared transaction, already prepared, we can't abort it
4657 : : * anymore, so we have to kill the reader instead.
4658 : : */
4659 [ + + ]: 88 : if (MySerializableXact == writer)
4660 : : {
4661 : 67 : LWLockRelease(SerializableXactHashLock);
4662 [ + - ]: 67 : ereport(ERROR,
4663 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4664 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4665 : : errdetail_internal("Reason code: Canceled on identification as a pivot, during write."),
4666 : : errhint("The transaction might succeed if retried.")));
4667 : : }
4668 [ + + ]: 21 : else if (SxactIsPrepared(writer))
4669 : : {
4670 : 13 : LWLockRelease(SerializableXactHashLock);
4671 : :
4672 : : /* if we're not the writer, we have to be the reader */
5315 4673 [ - + ]: 13 : Assert(MySerializableXact == reader);
5427 4674 [ + - ]: 13 : ereport(ERROR,
4675 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4676 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4677 : : errdetail_internal("Reason code: Canceled on conflict out to pivot %u, during read.", writer->topXid),
4678 : : errhint("The transaction might succeed if retried.")));
4679 : : }
5299 4680 : 8 : writer->flags |= SXACT_FLAG_DOOMED;
4681 : : }
5427 4682 : 792 : }
4683 : :
4684 : : /*
4685 : : * PreCommit_CheckForSerializationFailure
4686 : : * Check for dangerous structures in a serializable transaction
4687 : : * at commit.
4688 : : *
4689 : : * We're checking for a dangerous structure as each conflict is recorded.
4690 : : * The only way we could have a problem at commit is if this is the "out"
4691 : : * side of a pivot, and neither the "in" side nor the pivot has yet
4692 : : * committed.
4693 : : *
4694 : : * If a dangerous structure is found, the pivot (the near conflict) is
4695 : : * marked for death, because rolling back another transaction might mean
4696 : : * that we fail without ever making progress. This transaction is
4697 : : * committing writes, so letting it commit ensures progress. If we
4698 : : * canceled the far conflict, it might immediately fail again on retry.
4699 : : */
4700 : : void
4701 : 304558 : PreCommit_CheckForSerializationFailure(void)
4702 : : {
4703 : : dlist_iter near_iter;
4704 : :
4705 [ + + ]: 304558 : if (MySerializableXact == InvalidSerializableXact)
4706 : 303155 : return;
4707 : :
4708 [ - + ]: 1403 : Assert(IsolationIsSerializable());
4709 : :
4710 : 1403 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4711 : :
4712 : : /*
4713 : : * Check if someone else has already decided that we need to die. Since
4714 : : * we set our own DOOMED flag when partially releasing, ignore in that
4715 : : * case.
4716 : : */
1017 tmunro@postgresql.or 4717 [ + + ]: 1403 : if (SxactIsDoomed(MySerializableXact) &&
4718 [ + + ]: 156 : !SxactIsPartiallyReleased(MySerializableXact))
4719 : : {
5427 heikki.linnakangas@i 4720 : 155 : LWLockRelease(SerializableXactHashLock);
4721 [ + - ]: 155 : ereport(ERROR,
4722 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4723 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4724 : : errdetail_internal("Reason code: Canceled on identification as a pivot, during commit attempt."),
4725 : : errhint("The transaction might succeed if retried.")));
4726 : : }
4727 : :
1063 andres@anarazel.de 4728 [ + - + + ]: 1854 : dlist_foreach(near_iter, &MySerializableXact->inConflicts)
4729 : : {
4730 : 606 : RWConflict nearConflict =
943 tgl@sss.pgh.pa.us 4731 : 606 : dlist_container(RWConflictData, inLink, near_iter.cur);
4732 : :
5427 heikki.linnakangas@i 4733 [ + + ]: 606 : if (!SxactIsCommitted(nearConflict->sxactOut)
5299 4734 [ + - ]: 421 : && !SxactIsDoomed(nearConflict->sxactOut))
4735 : : {
4736 : : dlist_iter far_iter;
4737 : :
1063 andres@anarazel.de 4738 [ + - + + ]: 451 : dlist_foreach(far_iter, &nearConflict->sxactOut->inConflicts)
4739 : : {
4740 : 182 : RWConflict farConflict =
943 tgl@sss.pgh.pa.us 4741 : 182 : dlist_container(RWConflictData, inLink, far_iter.cur);
4742 : :
5427 heikki.linnakangas@i 4743 [ + + ]: 182 : if (farConflict->sxactOut == MySerializableXact
4744 [ + + ]: 42 : || (!SxactIsCommitted(farConflict->sxactOut)
4745 [ + + ]: 24 : && !SxactIsReadOnly(farConflict->sxactOut)
5299 4746 [ + - ]: 12 : && !SxactIsDoomed(farConflict->sxactOut)))
4747 : : {
4748 : : /*
4749 : : * Normally, we kill the pivot transaction to make sure we
4750 : : * make progress if the failing transaction is retried.
4751 : : * However, we can't kill it if it's already prepared, so
4752 : : * in that case we commit suicide instead.
4753 : : */
5293 4754 [ - + ]: 152 : if (SxactIsPrepared(nearConflict->sxactOut))
4755 : : {
5293 heikki.linnakangas@i 4756 :UBC 0 : LWLockRelease(SerializableXactHashLock);
4757 [ # # ]: 0 : ereport(ERROR,
4758 : : (errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
4759 : : errmsg("could not serialize access due to read/write dependencies among transactions"),
4760 : : errdetail_internal("Reason code: Canceled on commit attempt with conflict in from prepared pivot."),
4761 : : errhint("The transaction might succeed if retried.")));
4762 : : }
5299 heikki.linnakangas@i 4763 :CBC 152 : nearConflict->sxactOut->flags |= SXACT_FLAG_DOOMED;
5427 4764 : 152 : break;
4765 : : }
4766 : : }
4767 : : }
4768 : : }
4769 : :
5277 4770 : 1248 : MySerializableXact->prepareSeqNo = ++(PredXact->LastSxactCommitSeqNo);
5427 4771 : 1248 : MySerializableXact->flags |= SXACT_FLAG_PREPARED;
4772 : :
4773 : 1248 : LWLockRelease(SerializableXactHashLock);
4774 : : }
4775 : :
4776 : : /*------------------------------------------------------------------------*/
4777 : :
4778 : : /*
4779 : : * Two-phase commit support
4780 : : */
4781 : :
4782 : : /*
4783 : : * AtPrepare_Locks
4784 : : * Do the preparatory work for a PREPARE: make 2PC state file
4785 : : * records for all predicate locks currently held.
4786 : : */
4787 : : void
4788 : 298 : AtPrepare_PredicateLocks(void)
4789 : : {
4790 : : SERIALIZABLEXACT *sxact;
4791 : : TwoPhasePredicateRecord record;
4792 : : TwoPhasePredicateXactRecord *xactRecord;
4793 : : TwoPhasePredicateLockRecord *lockRecord;
4794 : : dlist_iter iter;
4795 : :
5304 4796 : 298 : sxact = MySerializableXact;
5427 4797 : 298 : xactRecord = &(record.data.xactRecord);
4798 : 298 : lockRecord = &(record.data.lockRecord);
4799 : :
4800 [ + + ]: 298 : if (MySerializableXact == InvalidSerializableXact)
4801 : 286 : return;
4802 : :
4803 : : /* Generate an xact record for our SERIALIZABLEXACT */
4804 : 12 : record.type = TWOPHASEPREDICATERECORD_XACT;
4805 : 12 : xactRecord->xmin = MySerializableXact->xmin;
4806 : 12 : xactRecord->flags = MySerializableXact->flags;
4807 : :
4808 : : /*
4809 : : * Note that we don't include the list of conflicts in our out in the
4810 : : * statefile, because new conflicts can be added even after the
4811 : : * transaction prepares. We'll just make a conservative assumption during
4812 : : * recovery instead.
4813 : : */
4814 : :
4815 : 12 : RegisterTwoPhaseRecord(TWOPHASE_RM_PREDICATELOCK_ID, 0,
4816 : : &record, sizeof(record));
4817 : :
4818 : : /*
4819 : : * Generate a lock record for each lock.
4820 : : *
4821 : : * To do this, we need to walk the predicate lock list in our sxact rather
4822 : : * than using the local predicate lock table because the latter is not
4823 : : * guaranteed to be accurate.
4824 : : */
2042 tgl@sss.pgh.pa.us 4825 : 12 : LWLockAcquire(SerializablePredicateListLock, LW_SHARED);
4826 : :
4827 : : /*
4828 : : * No need to take sxact->perXactPredicateListLock in parallel mode
4829 : : * because there cannot be any parallel workers running while we are
4830 : : * preparing a transaction.
4831 : : */
2469 tmunro@postgresql.or 4832 [ + - - + ]: 12 : Assert(!IsParallelWorker() && !ParallelContextActive());
4833 : :
1063 andres@anarazel.de 4834 [ + - + + ]: 22 : dlist_foreach(iter, &sxact->predicateLocks)
4835 : : {
4836 : 10 : PREDICATELOCK *predlock =
943 tgl@sss.pgh.pa.us 4837 : 10 : dlist_container(PREDICATELOCK, xactLink, iter.cur);
4838 : :
5427 heikki.linnakangas@i 4839 : 10 : record.type = TWOPHASEPREDICATERECORD_LOCK;
4840 : 10 : lockRecord->target = predlock->tag.myTarget->tag;
4841 : :
4842 : 10 : RegisterTwoPhaseRecord(TWOPHASE_RM_PREDICATELOCK_ID, 0,
4843 : : &record, sizeof(record));
4844 : : }
4845 : :
2042 tgl@sss.pgh.pa.us 4846 : 12 : LWLockRelease(SerializablePredicateListLock);
4847 : : }
4848 : :
4849 : : /*
4850 : : * PostPrepare_Locks
4851 : : * Clean up after successful PREPARE. Unlike the non-predicate
4852 : : * lock manager, we do not need to transfer locks to a dummy
4853 : : * PGPROC because our SERIALIZABLEXACT will stay around
4854 : : * anyway. We only need to clean up our local state.
4855 : : */
4856 : : void
163 michael@paquier.xyz 4857 :GNC 298 : PostPrepare_PredicateLocks(FullTransactionId fxid)
4858 : : {
5427 heikki.linnakangas@i 4859 [ + + ]:CBC 298 : if (MySerializableXact == InvalidSerializableXact)
4860 : 286 : return;
4861 : :
4862 [ - + ]: 12 : Assert(SxactIsPrepared(MySerializableXact));
4863 : :
4864 : 12 : MySerializableXact->pid = 0;
654 4865 : 12 : MySerializableXact->pgprocno = INVALID_PROC_NUMBER;
4866 : :
5427 4867 : 12 : hash_destroy(LocalPredicateLockHash);
4868 : 12 : LocalPredicateLockHash = NULL;
4869 : :
4870 : 12 : MySerializableXact = InvalidSerializableXact;
5304 4871 : 12 : MyXactDidWrite = false;
4872 : : }
4873 : :
4874 : : /*
4875 : : * PredicateLockTwoPhaseFinish
4876 : : * Release a prepared transaction's predicate locks once it
4877 : : * commits or aborts.
4878 : : */
4879 : : void
163 michael@paquier.xyz 4880 :GNC 306 : PredicateLockTwoPhaseFinish(FullTransactionId fxid, bool isCommit)
4881 : : {
4882 : : SERIALIZABLEXID *sxid;
4883 : : SERIALIZABLEXIDTAG sxidtag;
4884 : :
4885 : 306 : sxidtag.xid = XidFromFullTransactionId(fxid);
4886 : :
5427 heikki.linnakangas@i 4887 :CBC 306 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
4888 : : sxid = (SERIALIZABLEXID *)
4889 : 306 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
4890 : 306 : LWLockRelease(SerializableXactHashLock);
4891 : :
4892 : : /* xid will not be found if it wasn't a serializable transaction */
4893 [ + + ]: 306 : if (sxid == NULL)
4894 : 294 : return;
4895 : :
4896 : : /* Release its locks */
4897 : 12 : MySerializableXact = sxid->myXact;
5304 4898 : 12 : MyXactDidWrite = true; /* conservatively assume that we wrote
4899 : : * something */
2469 tmunro@postgresql.or 4900 : 12 : ReleasePredicateLocks(isCommit, false);
4901 : : }
4902 : :
4903 : : /*
4904 : : * Re-acquire a predicate lock belonging to a transaction that was prepared.
4905 : : */
4906 : : void
163 michael@paquier.xyz 4907 :UNC 0 : predicatelock_twophase_recover(FullTransactionId fxid, uint16 info,
4908 : : void *recdata, uint32 len)
4909 : : {
4910 : : TwoPhasePredicateRecord *record;
4911 : 0 : TransactionId xid = XidFromFullTransactionId(fxid);
4912 : :
5427 heikki.linnakangas@i 4913 [ # # ]:UBC 0 : Assert(len == sizeof(TwoPhasePredicateRecord));
4914 : :
4915 : 0 : record = (TwoPhasePredicateRecord *) recdata;
4916 : :
4917 [ # # # # ]: 0 : Assert((record->type == TWOPHASEPREDICATERECORD_XACT) ||
4918 : : (record->type == TWOPHASEPREDICATERECORD_LOCK));
4919 : :
4920 [ # # ]: 0 : if (record->type == TWOPHASEPREDICATERECORD_XACT)
4921 : : {
4922 : : /* Per-transaction record. Set up a SERIALIZABLEXACT. */
4923 : : TwoPhasePredicateXactRecord *xactRecord;
4924 : : SERIALIZABLEXACT *sxact;
4925 : : SERIALIZABLEXID *sxid;
4926 : : SERIALIZABLEXIDTAG sxidtag;
4927 : : bool found;
4928 : :
4929 : 0 : xactRecord = (TwoPhasePredicateXactRecord *) &record->data.xactRecord;
4930 : :
4931 : 0 : LWLockAcquire(SerializableXactHashLock, LW_EXCLUSIVE);
4932 : 0 : sxact = CreatePredXact();
4933 [ # # ]: 0 : if (!sxact)
4934 [ # # ]: 0 : ereport(ERROR,
4935 : : (errcode(ERRCODE_OUT_OF_MEMORY),
4936 : : errmsg("out of shared memory")));
4937 : :
4938 : : /* vxid for a prepared xact is INVALID_PROC_NUMBER/xid; no pid */
654 4939 : 0 : sxact->vxid.procNumber = INVALID_PROC_NUMBER;
5427 4940 : 0 : sxact->vxid.localTransactionId = (LocalTransactionId) xid;
4941 : 0 : sxact->pid = 0;
654 4942 : 0 : sxact->pgprocno = INVALID_PROC_NUMBER;
4943 : :
4944 : : /* a prepared xact hasn't committed yet */
5277 4945 : 0 : sxact->prepareSeqNo = RecoverySerCommitSeqNo;
5427 4946 : 0 : sxact->commitSeqNo = InvalidSerCommitSeqNo;
4947 : 0 : sxact->finishedBefore = InvalidTransactionId;
4948 : :
4949 : 0 : sxact->SeqNo.lastCommitBeforeSnapshot = RecoverySerCommitSeqNo;
4950 : :
4951 : : /*
4952 : : * Don't need to track this; no transactions running at the time the
4953 : : * recovered xact started are still active, except possibly other
4954 : : * prepared xacts and we don't care whether those are RO_SAFE or not.
4955 : : */
1063 andres@anarazel.de 4956 : 0 : dlist_init(&(sxact->possibleUnsafeConflicts));
4957 : :
4958 : 0 : dlist_init(&(sxact->predicateLocks));
4959 : 0 : dlist_node_init(&sxact->finishedLink);
4960 : :
5427 heikki.linnakangas@i 4961 : 0 : sxact->topXid = xid;
4962 : 0 : sxact->xmin = xactRecord->xmin;
4963 : 0 : sxact->flags = xactRecord->flags;
4964 [ # # ]: 0 : Assert(SxactIsPrepared(sxact));
4965 [ # # ]: 0 : if (!SxactIsReadOnly(sxact))
4966 : : {
4967 : 0 : ++(PredXact->WritableSxactCount);
4968 [ # # ]: 0 : Assert(PredXact->WritableSxactCount <=
4969 : : (MaxBackends + max_prepared_xacts));
4970 : : }
4971 : :
4972 : : /*
4973 : : * We don't know whether the transaction had any conflicts or not, so
4974 : : * we'll conservatively assume that it had both a conflict in and a
4975 : : * conflict out, and represent that with the summary conflict flags.
4976 : : */
1063 andres@anarazel.de 4977 : 0 : dlist_init(&(sxact->outConflicts));
4978 : 0 : dlist_init(&(sxact->inConflicts));
5040 heikki.linnakangas@i 4979 : 0 : sxact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_IN;
4980 : 0 : sxact->flags |= SXACT_FLAG_SUMMARY_CONFLICT_OUT;
4981 : :
4982 : : /* Register the transaction's xid */
5427 4983 : 0 : sxidtag.xid = xid;
4984 : 0 : sxid = (SERIALIZABLEXID *) hash_search(SerializableXidHash,
4985 : : &sxidtag,
4986 : : HASH_ENTER, &found);
5368 rhaas@postgresql.org 4987 [ # # ]: 0 : Assert(sxid != NULL);
5427 heikki.linnakangas@i 4988 [ # # ]: 0 : Assert(!found);
15 peter@eisentraut.org 4989 :UNC 0 : sxid->myXact = sxact;
4990 : :
4991 : : /*
4992 : : * Update global xmin. Note that this is a special case compared to
4993 : : * registering a normal transaction, because the global xmin might go
4994 : : * backwards. That's OK, because until recovery is over we're not
4995 : : * going to complete any transactions or create any non-prepared
4996 : : * transactions, so there's no danger of throwing away.
4997 : : */
5427 heikki.linnakangas@i 4998 [ # # # # ]:UBC 0 : if ((!TransactionIdIsValid(PredXact->SxactGlobalXmin)) ||
4999 : 0 : (TransactionIdFollows(PredXact->SxactGlobalXmin, sxact->xmin)))
5000 : : {
5001 : 0 : PredXact->SxactGlobalXmin = sxact->xmin;
5002 : 0 : PredXact->SxactGlobalXminCount = 1;
2042 tgl@sss.pgh.pa.us 5003 : 0 : SerialSetActiveSerXmin(sxact->xmin);
5004 : : }
5427 heikki.linnakangas@i 5005 [ # # ]: 0 : else if (TransactionIdEquals(sxact->xmin, PredXact->SxactGlobalXmin))
5006 : : {
5007 [ # # ]: 0 : Assert(PredXact->SxactGlobalXminCount > 0);
5008 : 0 : PredXact->SxactGlobalXminCount++;
5009 : : }
5010 : :
5011 : 0 : LWLockRelease(SerializableXactHashLock);
5012 : : }
5013 [ # # ]: 0 : else if (record->type == TWOPHASEPREDICATERECORD_LOCK)
5014 : : {
5015 : : /* Lock record. Recreate the PREDICATELOCK */
5016 : : TwoPhasePredicateLockRecord *lockRecord;
5017 : : SERIALIZABLEXID *sxid;
5018 : : SERIALIZABLEXACT *sxact;
5019 : : SERIALIZABLEXIDTAG sxidtag;
5020 : : uint32 targettaghash;
5021 : :
5022 : 0 : lockRecord = (TwoPhasePredicateLockRecord *) &record->data.lockRecord;
5023 : 0 : targettaghash = PredicateLockTargetTagHashCode(&lockRecord->target);
5024 : :
5025 : 0 : LWLockAcquire(SerializableXactHashLock, LW_SHARED);
5026 : 0 : sxidtag.xid = xid;
5027 : : sxid = (SERIALIZABLEXID *)
5028 : 0 : hash_search(SerializableXidHash, &sxidtag, HASH_FIND, NULL);
5029 : 0 : LWLockRelease(SerializableXactHashLock);
5030 : :
5031 [ # # ]: 0 : Assert(sxid != NULL);
5032 : 0 : sxact = sxid->myXact;
5033 [ # # ]: 0 : Assert(sxact != InvalidSerializableXact);
5034 : :
5035 : 0 : CreatePredicateLock(&lockRecord->target, targettaghash, sxact);
5036 : : }
5037 : 0 : }
5038 : :
5039 : : /*
5040 : : * Prepare to share the current SERIALIZABLEXACT with parallel workers.
5041 : : * Return a handle object that can be used by AttachSerializableXact() in a
5042 : : * parallel worker.
5043 : : */
5044 : : SerializableXactHandle
2469 tmunro@postgresql.or 5045 :CBC 477 : ShareSerializableXact(void)
5046 : : {
5047 : 477 : return MySerializableXact;
5048 : : }
5049 : :
5050 : : /*
5051 : : * Allow parallel workers to import the leader's SERIALIZABLEXACT.
5052 : : */
5053 : : void
5054 : 1440 : AttachSerializableXact(SerializableXactHandle handle)
5055 : : {
5056 : :
5057 [ - + ]: 1440 : Assert(MySerializableXact == InvalidSerializableXact);
5058 : :
5059 : 1440 : MySerializableXact = (SERIALIZABLEXACT *) handle;
5060 [ + + ]: 1440 : if (MySerializableXact != InvalidSerializableXact)
5061 : 13 : CreateLocalPredicateLockHash();
5062 : 1440 : }
|
