Age Owner Branch data TLA Line data Source code
1 : : /*-------------------------------------------------------------------------
2 : : *
3 : : * heapam.h
4 : : * POSTGRES heap access method definitions.
5 : : *
6 : : *
7 : : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
8 : : * Portions Copyright (c) 1994, Regents of the University of California
9 : : *
10 : : * src/include/access/heapam.h
11 : : *
12 : : *-------------------------------------------------------------------------
13 : : */
14 : : #ifndef HEAPAM_H
15 : : #define HEAPAM_H
16 : :
17 : : #include "access/heapam_xlog.h"
18 : : #include "access/relation.h" /* for backward compatibility */
19 : : #include "access/relscan.h"
20 : : #include "access/sdir.h"
21 : : #include "access/skey.h"
22 : : #include "access/table.h" /* for backward compatibility */
23 : : #include "access/tableam.h"
24 : : #include "commands/vacuum.h"
25 : : #include "nodes/lockoptions.h"
26 : : #include "nodes/primnodes.h"
27 : : #include "storage/bufpage.h"
28 : : #include "storage/dsm.h"
29 : : #include "storage/lockdefs.h"
30 : : #include "storage/read_stream.h"
31 : : #include "storage/shm_toc.h"
32 : : #include "utils/relcache.h"
33 : : #include "utils/snapshot.h"
34 : :
35 : :
36 : : /* "options" flag bits for heap_insert */
37 : : #define HEAP_INSERT_SKIP_FSM TABLE_INSERT_SKIP_FSM
38 : : #define HEAP_INSERT_FROZEN TABLE_INSERT_FROZEN
39 : : #define HEAP_INSERT_NO_LOGICAL TABLE_INSERT_NO_LOGICAL
40 : : #define HEAP_INSERT_SPECULATIVE 0x0010
41 : :
42 : : /* "options" flag bits for heap_page_prune_and_freeze */
43 : : #define HEAP_PAGE_PRUNE_MARK_UNUSED_NOW (1 << 0)
44 : : #define HEAP_PAGE_PRUNE_FREEZE (1 << 1)
45 : :
46 : : typedef struct BulkInsertStateData *BulkInsertState;
47 : : typedef struct GlobalVisState GlobalVisState;
48 : : typedef struct TupleTableSlot TupleTableSlot;
49 : : struct VacuumCutoffs;
50 : :
51 : : #define MaxLockTupleMode LockTupleExclusive
52 : :
53 : : /*
54 : : * Descriptor for heap table scans.
55 : : */
56 : : typedef struct HeapScanDescData
57 : : {
58 : : TableScanDescData rs_base; /* AM independent part of the descriptor */
59 : :
60 : : /* state set up at initscan time */
61 : : BlockNumber rs_nblocks; /* total number of blocks in rel */
62 : : BlockNumber rs_startblock; /* block # to start at */
63 : : BlockNumber rs_numblocks; /* max number of blocks to scan */
64 : : /* rs_numblocks is usually InvalidBlockNumber, meaning "scan whole rel" */
65 : :
66 : : /* scan current state */
67 : : bool rs_inited; /* false = scan not init'd yet */
68 : : OffsetNumber rs_coffset; /* current offset # in non-page-at-a-time mode */
69 : : BlockNumber rs_cblock; /* current block # in scan, if any */
70 : : Buffer rs_cbuf; /* current buffer in scan, if any */
71 : : /* NB: if rs_cbuf is not InvalidBuffer, we hold a pin on that buffer */
72 : :
73 : : BufferAccessStrategy rs_strategy; /* access strategy for reads */
74 : :
75 : : HeapTupleData rs_ctup; /* current tuple in scan, if any */
76 : :
77 : : /* For scans that stream reads */
78 : : ReadStream *rs_read_stream;
79 : :
80 : : /*
81 : : * For sequential scans and TID range scans to stream reads. The read
82 : : * stream is allocated at the beginning of the scan and reset on rescan or
83 : : * when the scan direction changes. The scan direction is saved each time
84 : : * a new page is requested. If the scan direction changes from one page to
85 : : * the next, the read stream releases all previously pinned buffers and
86 : : * resets the prefetch block.
87 : : */
88 : : ScanDirection rs_dir;
89 : : BlockNumber rs_prefetch_block;
90 : :
91 : : /*
92 : : * For parallel scans to store page allocation data. NULL when not
93 : : * performing a parallel scan.
94 : : */
95 : : ParallelBlockTableScanWorkerData *rs_parallelworkerdata;
96 : :
97 : : /* these fields only used in page-at-a-time mode and for bitmap scans */
98 : : uint32 rs_cindex; /* current tuple's index in vistuples */
99 : : uint32 rs_ntuples; /* number of visible tuples on page */
100 : : OffsetNumber rs_vistuples[MaxHeapTuplesPerPage]; /* their offsets */
101 : : } HeapScanDescData;
102 : : typedef struct HeapScanDescData *HeapScanDesc;
103 : :
104 : : typedef struct BitmapHeapScanDescData
105 : : {
106 : : HeapScanDescData rs_heap_base;
107 : :
108 : : /* Holds no data */
109 : : } BitmapHeapScanDescData;
110 : : typedef struct BitmapHeapScanDescData *BitmapHeapScanDesc;
111 : :
112 : : /*
113 : : * Descriptor for fetches from heap via an index.
114 : : */
115 : : typedef struct IndexFetchHeapData
116 : : {
117 : : IndexFetchTableData xs_base; /* AM independent part of the descriptor */
118 : :
119 : : Buffer xs_cbuf; /* current heap buffer in scan, if any */
120 : : /* NB: if xs_cbuf is not InvalidBuffer, we hold a pin on that buffer */
121 : : } IndexFetchHeapData;
122 : :
123 : : /* Result codes for HeapTupleSatisfiesVacuum */
124 : : typedef enum
125 : : {
126 : : HEAPTUPLE_DEAD, /* tuple is dead and deletable */
127 : : HEAPTUPLE_LIVE, /* tuple is live (committed, no deleter) */
128 : : HEAPTUPLE_RECENTLY_DEAD, /* tuple is dead, but not deletable yet */
129 : : HEAPTUPLE_INSERT_IN_PROGRESS, /* inserting xact is still in progress */
130 : : HEAPTUPLE_DELETE_IN_PROGRESS, /* deleting xact is still in progress */
131 : : } HTSV_Result;
132 : :
133 : : /*
134 : : * heap_prepare_freeze_tuple may request that heap_freeze_execute_prepared
135 : : * check any tuple's to-be-frozen xmin and/or xmax status using pg_xact
136 : : */
137 : : #define HEAP_FREEZE_CHECK_XMIN_COMMITTED 0x01
138 : : #define HEAP_FREEZE_CHECK_XMAX_ABORTED 0x02
139 : :
140 : : /* heap_prepare_freeze_tuple state describing how to freeze a tuple */
141 : : typedef struct HeapTupleFreeze
142 : : {
143 : : /* Fields describing how to process tuple */
144 : : TransactionId xmax;
145 : : uint16 t_infomask2;
146 : : uint16 t_infomask;
147 : : uint8 frzflags;
148 : :
149 : : /* xmin/xmax check flags */
150 : : uint8 checkflags;
151 : : /* Page offset number for tuple */
152 : : OffsetNumber offset;
153 : : } HeapTupleFreeze;
154 : :
155 : : /*
156 : : * State used by VACUUM to track the details of freezing all eligible tuples
157 : : * on a given heap page.
158 : : *
159 : : * VACUUM prepares freeze plans for each page via heap_prepare_freeze_tuple
160 : : * calls (every tuple with storage gets its own call). This page-level freeze
161 : : * state is updated across each call, which ultimately determines whether or
162 : : * not freezing the page is required.
163 : : *
164 : : * Aside from the basic question of whether or not freezing will go ahead, the
165 : : * state also tracks the oldest extant XID/MXID in the table as a whole, for
166 : : * the purposes of advancing relfrozenxid/relminmxid values in pg_class later
167 : : * on. Each heap_prepare_freeze_tuple call pushes NewRelfrozenXid and/or
168 : : * NewRelminMxid back as required to avoid unsafe final pg_class values. Any
169 : : * and all unfrozen XIDs or MXIDs that remain after VACUUM finishes _must_
170 : : * have values >= the final relfrozenxid/relminmxid values in pg_class. This
171 : : * includes XIDs that remain as MultiXact members from any tuple's xmax.
172 : : *
173 : : * When 'freeze_required' flag isn't set after all tuples are examined, the
174 : : * final choice on freezing is made by vacuumlazy.c. It can decide to trigger
175 : : * freezing based on whatever criteria it deems appropriate. However, it is
176 : : * recommended that vacuumlazy.c avoid early freezing when freezing does not
177 : : * enable setting the target page all-frozen in the visibility map afterwards.
178 : : */
179 : : typedef struct HeapPageFreeze
180 : : {
181 : : /* Is heap_prepare_freeze_tuple caller required to freeze page? */
182 : : bool freeze_required;
183 : :
184 : : /*
185 : : * "Freeze" NewRelfrozenXid/NewRelminMxid trackers.
186 : : *
187 : : * Trackers used when heap_freeze_execute_prepared freezes, or when there
188 : : * are zero freeze plans for a page. It is always valid for vacuumlazy.c
189 : : * to freeze any page, by definition. This even includes pages that have
190 : : * no tuples with storage to consider in the first place. That way the
191 : : * 'totally_frozen' results from heap_prepare_freeze_tuple can always be
192 : : * used in the same way, even when no freeze plans need to be executed to
193 : : * "freeze the page". Only the "freeze" path needs to consider the need
194 : : * to set pages all-frozen in the visibility map under this scheme.
195 : : *
196 : : * When we freeze a page, we generally freeze all XIDs < OldestXmin, only
197 : : * leaving behind XIDs that are ineligible for freezing, if any. And so
198 : : * you might wonder why these trackers are necessary at all; why should
199 : : * _any_ page that VACUUM freezes _ever_ be left with XIDs/MXIDs that
200 : : * ratchet back the top-level NewRelfrozenXid/NewRelminMxid trackers?
201 : : *
202 : : * It is useful to use a definition of "freeze the page" that does not
203 : : * overspecify how MultiXacts are affected. heap_prepare_freeze_tuple
204 : : * generally prefers to remove Multis eagerly, but lazy processing is used
205 : : * in cases where laziness allows VACUUM to avoid allocating a new Multi.
206 : : * The "freeze the page" trackers enable this flexibility.
207 : : */
208 : : TransactionId FreezePageRelfrozenXid;
209 : : MultiXactId FreezePageRelminMxid;
210 : :
211 : : /*
212 : : * "No freeze" NewRelfrozenXid/NewRelminMxid trackers.
213 : : *
214 : : * These trackers are maintained in the same way as the trackers used when
215 : : * VACUUM scans a page that isn't cleanup locked. Both code paths are
216 : : * based on the same general idea (do less work for this page during the
217 : : * ongoing VACUUM, at the cost of having to accept older final values).
218 : : */
219 : : TransactionId NoFreezePageRelfrozenXid;
220 : : MultiXactId NoFreezePageRelminMxid;
221 : :
222 : : } HeapPageFreeze;
223 : :
224 : : /*
225 : : * Per-page state returned by heap_page_prune_and_freeze()
226 : : */
227 : : typedef struct PruneFreezeResult
228 : : {
229 : : int ndeleted; /* Number of tuples deleted from the page */
230 : : int nnewlpdead; /* Number of newly LP_DEAD items */
231 : : int nfrozen; /* Number of tuples we froze */
232 : :
233 : : /* Number of live and recently dead tuples on the page, after pruning */
234 : : int live_tuples;
235 : : int recently_dead_tuples;
236 : :
237 : : /*
238 : : * all_visible and all_frozen indicate if the all-visible and all-frozen
239 : : * bits in the visibility map can be set for this page, after pruning.
240 : : *
241 : : * vm_conflict_horizon is the newest xmin of live tuples on the page. The
242 : : * caller can use it as the conflict horizon when setting the VM bits. It
243 : : * is only valid if we froze some tuples (nfrozen > 0), and all_frozen is
244 : : * true.
245 : : *
246 : : * These are only set if the HEAP_PRUNE_FREEZE option is set.
247 : : */
248 : : bool all_visible;
249 : : bool all_frozen;
250 : : TransactionId vm_conflict_horizon;
251 : :
252 : : /*
253 : : * Whether or not the page makes rel truncation unsafe. This is set to
254 : : * 'true', even if the page contains LP_DEAD items. VACUUM will remove
255 : : * them before attempting to truncate.
256 : : */
257 : : bool hastup;
258 : :
259 : : /*
260 : : * LP_DEAD items on the page after pruning. Includes existing LP_DEAD
261 : : * items.
262 : : */
263 : : int lpdead_items;
264 : : OffsetNumber deadoffsets[MaxHeapTuplesPerPage];
265 : : } PruneFreezeResult;
266 : :
267 : : /* 'reason' codes for heap_page_prune_and_freeze() */
268 : : typedef enum
269 : : {
270 : : PRUNE_ON_ACCESS, /* on-access pruning */
271 : : PRUNE_VACUUM_SCAN, /* VACUUM 1st heap pass */
272 : : PRUNE_VACUUM_CLEANUP, /* VACUUM 2nd heap pass */
273 : : } PruneReason;
274 : :
275 : : /* ----------------
276 : : * function prototypes for heap access method
277 : : *
278 : : * heap_create, heap_create_with_catalog, and heap_drop_with_catalog
279 : : * are declared in catalog/heap.h
280 : : * ----------------
281 : : */
282 : :
283 : :
284 : : extern TableScanDesc heap_beginscan(Relation relation, Snapshot snapshot,
285 : : int nkeys, ScanKey key,
286 : : ParallelTableScanDesc parallel_scan,
287 : : uint32 flags);
288 : : extern void heap_setscanlimits(TableScanDesc sscan, BlockNumber startBlk,
289 : : BlockNumber numBlks);
290 : : extern void heap_prepare_pagescan(TableScanDesc sscan);
291 : : extern void heap_rescan(TableScanDesc sscan, ScanKey key, bool set_params,
292 : : bool allow_strat, bool allow_sync, bool allow_pagemode);
293 : : extern void heap_endscan(TableScanDesc sscan);
294 : : extern HeapTuple heap_getnext(TableScanDesc sscan, ScanDirection direction);
295 : : extern bool heap_getnextslot(TableScanDesc sscan,
296 : : ScanDirection direction, TupleTableSlot *slot);
297 : : extern void heap_set_tidrange(TableScanDesc sscan, ItemPointer mintid,
298 : : ItemPointer maxtid);
299 : : extern bool heap_getnextslot_tidrange(TableScanDesc sscan,
300 : : ScanDirection direction,
301 : : TupleTableSlot *slot);
302 : : extern bool heap_fetch(Relation relation, Snapshot snapshot,
303 : : HeapTuple tuple, Buffer *userbuf, bool keep_buf);
304 : : extern bool heap_hot_search_buffer(ItemPointer tid, Relation relation,
305 : : Buffer buffer, Snapshot snapshot, HeapTuple heapTuple,
306 : : bool *all_dead, bool first_call);
307 : :
308 : : extern void heap_get_latest_tid(TableScanDesc sscan, ItemPointer tid);
309 : :
310 : : extern BulkInsertState GetBulkInsertState(void);
311 : : extern void FreeBulkInsertState(BulkInsertState);
312 : : extern void ReleaseBulkInsertStatePin(BulkInsertState bistate);
313 : :
314 : : extern void heap_insert(Relation relation, HeapTuple tup, CommandId cid,
315 : : int options, BulkInsertState bistate);
316 : : extern void heap_multi_insert(Relation relation, TupleTableSlot **slots,
317 : : int ntuples, CommandId cid, int options,
318 : : BulkInsertState bistate);
319 : : extern TM_Result heap_delete(Relation relation, ItemPointer tid,
320 : : CommandId cid, Snapshot crosscheck, bool wait,
321 : : TM_FailureData *tmfd, bool changingPart);
322 : : extern void heap_finish_speculative(Relation relation, ItemPointer tid);
323 : : extern void heap_abort_speculative(Relation relation, ItemPointer tid);
324 : : extern TM_Result heap_update(Relation relation, ItemPointer otid,
325 : : HeapTuple newtup,
326 : : CommandId cid, Snapshot crosscheck, bool wait,
327 : : TM_FailureData *tmfd, LockTupleMode *lockmode,
328 : : TU_UpdateIndexes *update_indexes);
329 : : extern TM_Result heap_lock_tuple(Relation relation, HeapTuple tuple,
330 : : CommandId cid, LockTupleMode mode, LockWaitPolicy wait_policy,
331 : : bool follow_updates,
332 : : Buffer *buffer, TM_FailureData *tmfd);
333 : :
334 : : extern bool heap_inplace_lock(Relation relation,
335 : : HeapTuple oldtup_ptr, Buffer buffer,
336 : : void (*release_callback) (void *), void *arg);
337 : : extern void heap_inplace_update_and_unlock(Relation relation,
338 : : HeapTuple oldtup, HeapTuple tuple,
339 : : Buffer buffer);
340 : : extern void heap_inplace_unlock(Relation relation,
341 : : HeapTuple oldtup, Buffer buffer);
342 : : extern bool heap_prepare_freeze_tuple(HeapTupleHeader tuple,
343 : : const struct VacuumCutoffs *cutoffs,
344 : : HeapPageFreeze *pagefrz,
345 : : HeapTupleFreeze *frz, bool *totally_frozen);
346 : :
347 : : extern void heap_pre_freeze_checks(Buffer buffer,
348 : : HeapTupleFreeze *tuples, int ntuples);
349 : : extern void heap_freeze_prepared_tuples(Buffer buffer,
350 : : HeapTupleFreeze *tuples, int ntuples);
351 : : extern bool heap_freeze_tuple(HeapTupleHeader tuple,
352 : : TransactionId relfrozenxid, TransactionId relminmxid,
353 : : TransactionId FreezeLimit, TransactionId MultiXactCutoff);
354 : : extern bool heap_tuple_should_freeze(HeapTupleHeader tuple,
355 : : const struct VacuumCutoffs *cutoffs,
356 : : TransactionId *NoFreezePageRelfrozenXid,
357 : : MultiXactId *NoFreezePageRelminMxid);
358 : : extern bool heap_tuple_needs_eventual_freeze(HeapTupleHeader tuple);
359 : :
360 : : extern void simple_heap_insert(Relation relation, HeapTuple tup);
361 : : extern void simple_heap_delete(Relation relation, ItemPointer tid);
362 : : extern void simple_heap_update(Relation relation, ItemPointer otid,
363 : : HeapTuple tup, TU_UpdateIndexes *update_indexes);
364 : :
365 : : extern TransactionId heap_index_delete_tuples(Relation rel,
366 : : TM_IndexDeleteOp *delstate);
367 : :
368 : : /* in heap/pruneheap.c */
369 : : extern void heap_page_prune_opt(Relation relation, Buffer buffer);
370 : : extern void heap_page_prune_and_freeze(Relation relation, Buffer buffer,
371 : : GlobalVisState *vistest,
372 : : int options,
373 : : struct VacuumCutoffs *cutoffs,
374 : : PruneFreezeResult *presult,
375 : : PruneReason reason,
376 : : OffsetNumber *off_loc,
377 : : TransactionId *new_relfrozen_xid,
378 : : MultiXactId *new_relmin_mxid);
379 : : extern void heap_page_prune_execute(Buffer buffer, bool lp_truncate_only,
380 : : OffsetNumber *redirected, int nredirected,
381 : : OffsetNumber *nowdead, int ndead,
382 : : OffsetNumber *nowunused, int nunused);
383 : : extern void heap_get_root_tuples(Page page, OffsetNumber *root_offsets);
384 : : extern void log_heap_prune_and_freeze(Relation relation, Buffer buffer,
385 : : Buffer vmbuffer, uint8 vmflags,
386 : : TransactionId conflict_xid,
387 : : bool cleanup_lock,
388 : : PruneReason reason,
389 : : HeapTupleFreeze *frozen, int nfrozen,
390 : : OffsetNumber *redirected, int nredirected,
391 : : OffsetNumber *dead, int ndead,
392 : : OffsetNumber *unused, int nunused);
393 : :
394 : : /* in heap/vacuumlazy.c */
395 : : extern void heap_vacuum_rel(Relation rel,
396 : : const VacuumParams params, BufferAccessStrategy bstrategy);
397 : :
398 : : /* in heap/heapam_visibility.c */
399 : : extern bool HeapTupleSatisfiesVisibility(HeapTuple htup, Snapshot snapshot,
400 : : Buffer buffer);
401 : : extern TM_Result HeapTupleSatisfiesUpdate(HeapTuple htup, CommandId curcid,
402 : : Buffer buffer);
403 : : extern HTSV_Result HeapTupleSatisfiesVacuum(HeapTuple htup, TransactionId OldestXmin,
404 : : Buffer buffer);
405 : : extern HTSV_Result HeapTupleSatisfiesVacuumHorizon(HeapTuple htup, Buffer buffer,
406 : : TransactionId *dead_after);
407 : : extern void HeapTupleSetHintBits(HeapTupleHeader tuple, Buffer buffer,
408 : : uint16 infomask, TransactionId xid);
409 : : extern bool HeapTupleHeaderIsOnlyLocked(HeapTupleHeader tuple);
410 : : extern bool HeapTupleIsSurelyDead(HeapTuple htup,
411 : : GlobalVisState *vistest);
412 : :
413 : : /*
414 : : * To avoid leaking too much knowledge about reorderbuffer implementation
415 : : * details this is implemented in reorderbuffer.c not heapam_visibility.c
416 : : */
417 : : struct HTAB;
418 : : extern bool ResolveCminCmaxDuringDecoding(struct HTAB *tuplecid_data,
419 : : Snapshot snapshot,
420 : : HeapTuple htup,
421 : : Buffer buffer,
422 : : CommandId *cmin, CommandId *cmax);
423 : : extern void HeapCheckForSerializableConflictOut(bool visible, Relation relation, HeapTuple tuple,
424 : : Buffer buffer, Snapshot snapshot);
425 : :
426 : : /*
427 : : * heap_execute_freeze_tuple
428 : : * Execute the prepared freezing of a tuple with caller's freeze plan.
429 : : *
430 : : * Caller is responsible for ensuring that no other backend can access the
431 : : * storage underlying this tuple, either by holding an exclusive lock on the
432 : : * buffer containing it (which is what lazy VACUUM does), or by having it be
433 : : * in private storage (which is what CLUSTER and friends do).
434 : : */
435 : : static inline void
410 michael@paquier.xyz 436 :CBC 1240912 : heap_execute_freeze_tuple(HeapTupleHeader tuple, HeapTupleFreeze *frz)
437 : : {
438 : 1240912 : HeapTupleHeaderSetXmax(tuple, frz->xmax);
439 : :
440 [ - + ]: 1240912 : if (frz->frzflags & XLH_FREEZE_XVAC)
410 michael@paquier.xyz 441 :UBC 0 : HeapTupleHeaderSetXvac(tuple, FrozenTransactionId);
442 : :
410 michael@paquier.xyz 443 [ - + ]:CBC 1240912 : if (frz->frzflags & XLH_INVALID_XVAC)
410 michael@paquier.xyz 444 :UBC 0 : HeapTupleHeaderSetXvac(tuple, InvalidTransactionId);
445 : :
410 michael@paquier.xyz 446 :CBC 1240912 : tuple->t_infomask = frz->t_infomask;
447 : 1240912 : tuple->t_infomask2 = frz->t_infomask2;
448 : 1240912 : }
449 : :
450 : : #endif /* HEAPAM_H */
|
