Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * heapam.h
4 : * POSTGRES heap access method definitions.
5 : *
6 : *
7 : * Portions Copyright (c) 1996-2024, 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 "nodes/lockoptions.h"
25 : #include "nodes/primnodes.h"
26 : #include "storage/bufpage.h"
27 : #include "storage/dsm.h"
28 : #include "storage/lockdefs.h"
29 : #include "storage/read_stream.h"
30 : #include "storage/shm_toc.h"
31 : #include "utils/relcache.h"
32 : #include "utils/snapshot.h"
33 :
34 :
35 : /* "options" flag bits for heap_insert */
36 : #define HEAP_INSERT_SKIP_FSM TABLE_INSERT_SKIP_FSM
37 : #define HEAP_INSERT_FROZEN TABLE_INSERT_FROZEN
38 : #define HEAP_INSERT_NO_LOGICAL TABLE_INSERT_NO_LOGICAL
39 : #define HEAP_INSERT_SPECULATIVE 0x0010
40 :
41 : /* "options" flag bits for heap_page_prune_and_freeze */
42 : #define HEAP_PAGE_PRUNE_MARK_UNUSED_NOW (1 << 0)
43 : #define HEAP_PAGE_PRUNE_FREEZE (1 << 1)
44 :
45 : typedef struct BulkInsertStateData *BulkInsertState;
46 : struct TupleTableSlot;
47 : struct VacuumCutoffs;
48 :
49 : #define MaxLockTupleMode LockTupleExclusive
50 :
51 : /*
52 : * Descriptor for heap table scans.
53 : */
54 : typedef struct HeapScanDescData
55 : {
56 : TableScanDescData rs_base; /* AM independent part of the descriptor */
57 :
58 : /* state set up at initscan time */
59 : BlockNumber rs_nblocks; /* total number of blocks in rel */
60 : BlockNumber rs_startblock; /* block # to start at */
61 : BlockNumber rs_numblocks; /* max number of blocks to scan */
62 : /* rs_numblocks is usually InvalidBlockNumber, meaning "scan whole rel" */
63 :
64 : /* scan current state */
65 : bool rs_inited; /* false = scan not init'd yet */
66 : OffsetNumber rs_coffset; /* current offset # in non-page-at-a-time mode */
67 : BlockNumber rs_cblock; /* current block # in scan, if any */
68 : Buffer rs_cbuf; /* current buffer in scan, if any */
69 : /* NB: if rs_cbuf is not InvalidBuffer, we hold a pin on that buffer */
70 :
71 : BufferAccessStrategy rs_strategy; /* access strategy for reads */
72 :
73 : HeapTupleData rs_ctup; /* current tuple in scan, if any */
74 :
75 : /* For scans that stream reads */
76 : ReadStream *rs_read_stream;
77 :
78 : /*
79 : * For sequential scans and TID range scans to stream reads. The read
80 : * stream is allocated at the beginning of the scan and reset on rescan or
81 : * when the scan direction changes. The scan direction is saved each time
82 : * a new page is requested. If the scan direction changes from one page to
83 : * the next, the read stream releases all previously pinned buffers and
84 : * resets the prefetch block.
85 : */
86 : ScanDirection rs_dir;
87 : BlockNumber rs_prefetch_block;
88 :
89 : /*
90 : * For parallel scans to store page allocation data. NULL when not
91 : * performing a parallel scan.
92 : */
93 : ParallelBlockTableScanWorkerData *rs_parallelworkerdata;
94 :
95 : /*
96 : * These fields are only used for bitmap scans for the "skip fetch"
97 : * optimization. Bitmap scans needing no fields from the heap may skip
98 : * fetching an all visible block, instead using the number of tuples per
99 : * block reported by the bitmap to determine how many NULL-filled tuples
100 : * to return.
101 : */
102 : Buffer rs_vmbuffer;
103 : int rs_empty_tuples_pending;
104 :
105 : /* these fields only used in page-at-a-time mode and for bitmap scans */
106 : int rs_cindex; /* current tuple's index in vistuples */
107 : int rs_ntuples; /* number of visible tuples on page */
108 : OffsetNumber rs_vistuples[MaxHeapTuplesPerPage]; /* their offsets */
109 : } HeapScanDescData;
110 : typedef struct HeapScanDescData *HeapScanDesc;
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 : /*
285 : * HeapScanIsValid
286 : * True iff the heap scan is valid.
287 : */
288 : #define HeapScanIsValid(scan) PointerIsValid(scan)
289 :
290 : extern TableScanDesc heap_beginscan(Relation relation, Snapshot snapshot,
291 : int nkeys, ScanKey key,
292 : ParallelTableScanDesc parallel_scan,
293 : uint32 flags);
294 : extern void heap_setscanlimits(TableScanDesc sscan, BlockNumber startBlk,
295 : BlockNumber numBlks);
296 : extern void heap_prepare_pagescan(TableScanDesc sscan);
297 : extern void heap_rescan(TableScanDesc sscan, ScanKey key, bool set_params,
298 : bool allow_strat, bool allow_sync, bool allow_pagemode);
299 : extern void heap_endscan(TableScanDesc sscan);
300 : extern HeapTuple heap_getnext(TableScanDesc sscan, ScanDirection direction);
301 : extern bool heap_getnextslot(TableScanDesc sscan,
302 : ScanDirection direction, struct TupleTableSlot *slot);
303 : extern void heap_set_tidrange(TableScanDesc sscan, ItemPointer mintid,
304 : ItemPointer maxtid);
305 : extern bool heap_getnextslot_tidrange(TableScanDesc sscan,
306 : ScanDirection direction,
307 : TupleTableSlot *slot);
308 : extern bool heap_fetch(Relation relation, Snapshot snapshot,
309 : HeapTuple tuple, Buffer *userbuf, bool keep_buf);
310 : extern bool heap_hot_search_buffer(ItemPointer tid, Relation relation,
311 : Buffer buffer, Snapshot snapshot, HeapTuple heapTuple,
312 : bool *all_dead, bool first_call);
313 :
314 : extern void heap_get_latest_tid(TableScanDesc sscan, ItemPointer tid);
315 :
316 : extern BulkInsertState GetBulkInsertState(void);
317 : extern void FreeBulkInsertState(BulkInsertState);
318 : extern void ReleaseBulkInsertStatePin(BulkInsertState bistate);
319 :
320 : extern void heap_insert(Relation relation, HeapTuple tup, CommandId cid,
321 : int options, BulkInsertState bistate);
322 : extern void heap_multi_insert(Relation relation, struct TupleTableSlot **slots,
323 : int ntuples, CommandId cid, int options,
324 : BulkInsertState bistate);
325 : extern TM_Result heap_delete(Relation relation, ItemPointer tid,
326 : CommandId cid, Snapshot crosscheck, bool wait,
327 : struct TM_FailureData *tmfd, bool changingPart);
328 : extern void heap_finish_speculative(Relation relation, ItemPointer tid);
329 : extern void heap_abort_speculative(Relation relation, ItemPointer tid);
330 : extern TM_Result heap_update(Relation relation, ItemPointer otid,
331 : HeapTuple newtup,
332 : CommandId cid, Snapshot crosscheck, bool wait,
333 : struct TM_FailureData *tmfd, LockTupleMode *lockmode,
334 : TU_UpdateIndexes *update_indexes);
335 : extern TM_Result heap_lock_tuple(Relation relation, HeapTuple tuple,
336 : CommandId cid, LockTupleMode mode, LockWaitPolicy wait_policy,
337 : bool follow_updates,
338 : Buffer *buffer, struct TM_FailureData *tmfd);
339 :
340 : extern bool heap_inplace_lock(Relation relation,
341 : HeapTuple oldtup_ptr, Buffer buffer,
342 : void (*release_callback) (void *), void *arg);
343 : extern void heap_inplace_update_and_unlock(Relation relation,
344 : HeapTuple oldtup, HeapTuple tuple,
345 : Buffer buffer);
346 : extern void heap_inplace_unlock(Relation relation,
347 : HeapTuple oldtup, Buffer buffer);
348 : extern bool heap_prepare_freeze_tuple(HeapTupleHeader tuple,
349 : const struct VacuumCutoffs *cutoffs,
350 : HeapPageFreeze *pagefrz,
351 : HeapTupleFreeze *frz, bool *totally_frozen);
352 :
353 : extern void heap_pre_freeze_checks(Buffer buffer,
354 : HeapTupleFreeze *tuples, int ntuples);
355 : extern void heap_freeze_prepared_tuples(Buffer buffer,
356 : HeapTupleFreeze *tuples, int ntuples);
357 : extern bool heap_freeze_tuple(HeapTupleHeader tuple,
358 : TransactionId relfrozenxid, TransactionId relminmxid,
359 : TransactionId FreezeLimit, TransactionId MultiXactCutoff);
360 : extern bool heap_tuple_should_freeze(HeapTupleHeader tuple,
361 : const struct VacuumCutoffs *cutoffs,
362 : TransactionId *NoFreezePageRelfrozenXid,
363 : MultiXactId *NoFreezePageRelminMxid);
364 : extern bool heap_tuple_needs_eventual_freeze(HeapTupleHeader tuple);
365 :
366 : extern void simple_heap_insert(Relation relation, HeapTuple tup);
367 : extern void simple_heap_delete(Relation relation, ItemPointer tid);
368 : extern void simple_heap_update(Relation relation, ItemPointer otid,
369 : HeapTuple tup, TU_UpdateIndexes *update_indexes);
370 :
371 : extern TransactionId heap_index_delete_tuples(Relation rel,
372 : TM_IndexDeleteOp *delstate);
373 :
374 : /* in heap/pruneheap.c */
375 : struct GlobalVisState;
376 : extern void heap_page_prune_opt(Relation relation, Buffer buffer);
377 : extern void heap_page_prune_and_freeze(Relation relation, Buffer buffer,
378 : struct GlobalVisState *vistest,
379 : int options,
380 : struct VacuumCutoffs *cutoffs,
381 : PruneFreezeResult *presult,
382 : PruneReason reason,
383 : OffsetNumber *off_loc,
384 : TransactionId *new_relfrozen_xid,
385 : MultiXactId *new_relmin_mxid);
386 : extern void heap_page_prune_execute(Buffer buffer, bool lp_truncate_only,
387 : OffsetNumber *redirected, int nredirected,
388 : OffsetNumber *nowdead, int ndead,
389 : OffsetNumber *nowunused, int nunused);
390 : extern void heap_get_root_tuples(Page page, OffsetNumber *root_offsets);
391 : extern void log_heap_prune_and_freeze(Relation relation, Buffer buffer,
392 : TransactionId conflict_xid,
393 : bool cleanup_lock,
394 : PruneReason reason,
395 : HeapTupleFreeze *frozen, int nfrozen,
396 : OffsetNumber *redirected, int nredirected,
397 : OffsetNumber *dead, int ndead,
398 : OffsetNumber *unused, int nunused);
399 :
400 : /* in heap/vacuumlazy.c */
401 : struct VacuumParams;
402 : extern void heap_vacuum_rel(Relation rel,
403 : struct VacuumParams *params, BufferAccessStrategy bstrategy);
404 :
405 : /* in heap/heapam_visibility.c */
406 : extern bool HeapTupleSatisfiesVisibility(HeapTuple htup, Snapshot snapshot,
407 : Buffer buffer);
408 : extern TM_Result HeapTupleSatisfiesUpdate(HeapTuple htup, CommandId curcid,
409 : Buffer buffer);
410 : extern HTSV_Result HeapTupleSatisfiesVacuum(HeapTuple htup, TransactionId OldestXmin,
411 : Buffer buffer);
412 : extern HTSV_Result HeapTupleSatisfiesVacuumHorizon(HeapTuple htup, Buffer buffer,
413 : TransactionId *dead_after);
414 : extern void HeapTupleSetHintBits(HeapTupleHeader tuple, Buffer buffer,
415 : uint16 infomask, TransactionId xid);
416 : extern bool HeapTupleHeaderIsOnlyLocked(HeapTupleHeader tuple);
417 : extern bool HeapTupleIsSurelyDead(HeapTuple htup,
418 : struct GlobalVisState *vistest);
419 :
420 : /*
421 : * To avoid leaking too much knowledge about reorderbuffer implementation
422 : * details this is implemented in reorderbuffer.c not heapam_visibility.c
423 : */
424 : struct HTAB;
425 : extern bool ResolveCminCmaxDuringDecoding(struct HTAB *tuplecid_data,
426 : Snapshot snapshot,
427 : HeapTuple htup,
428 : Buffer buffer,
429 : CommandId *cmin, CommandId *cmax);
430 : extern void HeapCheckForSerializableConflictOut(bool visible, Relation relation, HeapTuple tuple,
431 : Buffer buffer, Snapshot snapshot);
432 :
433 : /*
434 : * heap_execute_freeze_tuple
435 : * Execute the prepared freezing of a tuple with caller's freeze plan.
436 : *
437 : * Caller is responsible for ensuring that no other backend can access the
438 : * storage underlying this tuple, either by holding an exclusive lock on the
439 : * buffer containing it (which is what lazy VACUUM does), or by having it be
440 : * in private storage (which is what CLUSTER and friends do).
441 : */
442 : static inline void
443 1970270 : heap_execute_freeze_tuple(HeapTupleHeader tuple, HeapTupleFreeze *frz)
444 : {
445 1970270 : HeapTupleHeaderSetXmax(tuple, frz->xmax);
446 :
447 1970270 : if (frz->frzflags & XLH_FREEZE_XVAC)
448 0 : HeapTupleHeaderSetXvac(tuple, FrozenTransactionId);
449 :
450 1970270 : if (frz->frzflags & XLH_INVALID_XVAC)
451 0 : HeapTupleHeaderSetXvac(tuple, InvalidTransactionId);
452 :
453 1970270 : tuple->t_infomask = frz->t_infomask;
454 1970270 : tuple->t_infomask2 = frz->t_infomask2;
455 1970270 : }
456 :
457 : #endif /* HEAPAM_H */
|
