PostgreSQL Source Code  git master
heapam.h File Reference
#include "access/relation.h"
#include "access/relscan.h"
#include "access/sdir.h"
#include "access/skey.h"
#include "access/table.h"
#include "access/tableam.h"
#include "nodes/lockoptions.h"
#include "nodes/primnodes.h"
#include "storage/bufpage.h"
#include "storage/dsm.h"
#include "storage/lockdefs.h"
#include "storage/shm_toc.h"
#include "utils/relcache.h"
#include "utils/snapshot.h"
Include dependency graph for heapam.h:
This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Data Structures

struct  HeapScanDescData
 
struct  IndexFetchHeapData
 

Macros

#define HEAP_INSERT_SKIP_FSM   TABLE_INSERT_SKIP_FSM
 
#define HEAP_INSERT_FROZEN   TABLE_INSERT_FROZEN
 
#define HEAP_INSERT_NO_LOGICAL   TABLE_INSERT_NO_LOGICAL
 
#define HEAP_INSERT_SPECULATIVE   0x0010
 
#define MaxLockTupleMode   LockTupleExclusive
 
#define HeapScanIsValid(scan)   PointerIsValid(scan)
 

Typedefs

typedef struct BulkInsertStateDataBulkInsertState
 
typedef struct HeapScanDescData HeapScanDescData
 
typedef struct HeapScanDescDataHeapScanDesc
 
typedef struct IndexFetchHeapData IndexFetchHeapData
 

Enumerations

enum  HTSV_Result {
  HEAPTUPLE_DEAD, HEAPTUPLE_LIVE, HEAPTUPLE_RECENTLY_DEAD, HEAPTUPLE_INSERT_IN_PROGRESS,
  HEAPTUPLE_DELETE_IN_PROGRESS
}
 

Functions

TableScanDesc heap_beginscan (Relation relation, Snapshot snapshot, int nkeys, ScanKey key, ParallelTableScanDesc parallel_scan, uint32 flags)
 
void heap_setscanlimits (TableScanDesc scan, BlockNumber startBlk, BlockNumber numBlks)
 
void heapgetpage (TableScanDesc scan, BlockNumber page)
 
void heap_rescan (TableScanDesc scan, ScanKey key, bool set_params, bool allow_strat, bool allow_sync, bool allow_pagemode)
 
void heap_endscan (TableScanDesc scan)
 
HeapTuple heap_getnext (TableScanDesc scan, ScanDirection direction)
 
bool heap_getnextslot (TableScanDesc sscan, ScanDirection direction, struct TupleTableSlot *slot)
 
void heap_set_tidrange (TableScanDesc sscan, ItemPointer mintid, ItemPointer maxtid)
 
bool heap_getnextslot_tidrange (TableScanDesc sscan, ScanDirection direction, TupleTableSlot *slot)
 
bool heap_fetch (Relation relation, Snapshot snapshot, HeapTuple tuple, Buffer *userbuf)
 
bool heap_hot_search_buffer (ItemPointer tid, Relation relation, Buffer buffer, Snapshot snapshot, HeapTuple heapTuple, bool *all_dead, bool first_call)
 
void heap_get_latest_tid (TableScanDesc scan, ItemPointer tid)
 
BulkInsertState GetBulkInsertState (void)
 
void FreeBulkInsertState (BulkInsertState)
 
void ReleaseBulkInsertStatePin (BulkInsertState bistate)
 
void heap_insert (Relation relation, HeapTuple tup, CommandId cid, int options, BulkInsertState bistate)
 
void heap_multi_insert (Relation relation, struct TupleTableSlot **slots, int ntuples, CommandId cid, int options, BulkInsertState bistate)
 
TM_Result heap_delete (Relation relation, ItemPointer tid, CommandId cid, Snapshot crosscheck, bool wait, struct TM_FailureData *tmfd, bool changingPart)
 
void heap_finish_speculative (Relation relation, ItemPointer tid)
 
void heap_abort_speculative (Relation relation, ItemPointer tid)
 
TM_Result heap_update (Relation relation, ItemPointer otid, HeapTuple newtup, CommandId cid, Snapshot crosscheck, bool wait, struct TM_FailureData *tmfd, LockTupleMode *lockmode)
 
TM_Result heap_lock_tuple (Relation relation, HeapTuple tuple, CommandId cid, LockTupleMode mode, LockWaitPolicy wait_policy, bool follow_update, Buffer *buffer, struct TM_FailureData *tmfd)
 
void heap_inplace_update (Relation relation, HeapTuple tuple)
 
bool heap_freeze_tuple (HeapTupleHeader tuple, TransactionId relfrozenxid, TransactionId relminmxid, TransactionId cutoff_xid, TransactionId cutoff_multi)
 
bool heap_tuple_needs_freeze (HeapTupleHeader tuple, TransactionId cutoff_xid, MultiXactId cutoff_multi, Buffer buf)
 
bool heap_tuple_needs_eventual_freeze (HeapTupleHeader tuple)
 
void simple_heap_insert (Relation relation, HeapTuple tup)
 
void simple_heap_delete (Relation relation, ItemPointer tid)
 
void simple_heap_update (Relation relation, ItemPointer otid, HeapTuple tup)
 
TransactionId heap_index_delete_tuples (Relation rel, TM_IndexDeleteOp *delstate)
 
void heap_page_prune_opt (Relation relation, Buffer buffer)
 
int heap_page_prune (Relation relation, Buffer buffer, struct GlobalVisState *vistest, TransactionId old_snap_xmin, TimestampTz old_snap_ts_ts, bool report_stats, OffsetNumber *off_loc)
 
void heap_page_prune_execute (Buffer buffer, OffsetNumber *redirected, int nredirected, OffsetNumber *nowdead, int ndead, OffsetNumber *nowunused, int nunused)
 
void heap_get_root_tuples (Page page, OffsetNumber *root_offsets)
 
void heap_vacuum_rel (Relation rel, struct VacuumParams *params, BufferAccessStrategy bstrategy)
 
void parallel_vacuum_main (dsm_segment *seg, shm_toc *toc)
 
bool HeapTupleSatisfiesVisibility (HeapTuple stup, Snapshot snapshot, Buffer buffer)
 
TM_Result HeapTupleSatisfiesUpdate (HeapTuple stup, CommandId curcid, Buffer buffer)
 
HTSV_Result HeapTupleSatisfiesVacuum (HeapTuple stup, TransactionId OldestXmin, Buffer buffer)
 
HTSV_Result HeapTupleSatisfiesVacuumHorizon (HeapTuple stup, Buffer buffer, TransactionId *dead_after)
 
void HeapTupleSetHintBits (HeapTupleHeader tuple, Buffer buffer, uint16 infomask, TransactionId xid)
 
bool HeapTupleHeaderIsOnlyLocked (HeapTupleHeader tuple)
 
bool XidInMVCCSnapshot (TransactionId xid, Snapshot snapshot)
 
bool HeapTupleIsSurelyDead (HeapTuple htup, struct GlobalVisState *vistest)
 
bool ResolveCminCmaxDuringDecoding (struct HTAB *tuplecid_data, Snapshot snapshot, HeapTuple htup, Buffer buffer, CommandId *cmin, CommandId *cmax)
 
void HeapCheckForSerializableConflictOut (bool valid, Relation relation, HeapTuple tuple, Buffer buffer, Snapshot snapshot)
 

Macro Definition Documentation

◆ HEAP_INSERT_FROZEN

#define HEAP_INSERT_FROZEN   TABLE_INSERT_FROZEN

Definition at line 35 of file heapam.h.

Referenced by heap_multi_insert(), heap_prepare_insert(), and RelationGetBufferForTuple().

◆ HEAP_INSERT_NO_LOGICAL

#define HEAP_INSERT_NO_LOGICAL   TABLE_INSERT_NO_LOGICAL

Definition at line 36 of file heapam.h.

Referenced by heap_insert(), heap_multi_insert(), and raw_heap_insert().

◆ HEAP_INSERT_SKIP_FSM

#define HEAP_INSERT_SKIP_FSM   TABLE_INSERT_SKIP_FSM

Definition at line 34 of file heapam.h.

Referenced by raw_heap_insert(), and RelationGetBufferForTuple().

◆ HEAP_INSERT_SPECULATIVE

#define HEAP_INSERT_SPECULATIVE   0x0010

◆ HeapScanIsValid

#define HeapScanIsValid (   scan)    PointerIsValid(scan)

Definition at line 115 of file heapam.h.

◆ MaxLockTupleMode

#define MaxLockTupleMode   LockTupleExclusive

Definition at line 42 of file heapam.h.

Typedef Documentation

◆ BulkInsertState

Definition at line 39 of file heapam.h.

◆ HeapScanDesc

typedef struct HeapScanDescData* HeapScanDesc

Definition at line 79 of file heapam.h.

◆ HeapScanDescData

◆ IndexFetchHeapData

Enumeration Type Documentation

◆ HTSV_Result

Enumerator
HEAPTUPLE_DEAD 
HEAPTUPLE_LIVE 
HEAPTUPLE_RECENTLY_DEAD 
HEAPTUPLE_INSERT_IN_PROGRESS 
HEAPTUPLE_DELETE_IN_PROGRESS 

Definition at line 93 of file heapam.h.

94 {
95  HEAPTUPLE_DEAD, /* tuple is dead and deletable */
96  HEAPTUPLE_LIVE, /* tuple is live (committed, no deleter) */
97  HEAPTUPLE_RECENTLY_DEAD, /* tuple is dead, but not deletable yet */
98  HEAPTUPLE_INSERT_IN_PROGRESS, /* inserting xact is still in progress */
99  HEAPTUPLE_DELETE_IN_PROGRESS /* deleting xact is still in progress */
100 } HTSV_Result;
HTSV_Result
Definition: heapam.h:93

Function Documentation

◆ FreeBulkInsertState()

void FreeBulkInsertState ( BulkInsertState  )

Definition at line 2021 of file heapam.c.

References BulkInsertStateData::current_buf, FreeAccessStrategy(), InvalidBuffer, pfree(), ReleaseBuffer(), and BulkInsertStateData::strategy.

Referenced by ATRewriteTable(), CopyFrom(), CopyMultiInsertBufferCleanup(), intorel_shutdown(), and transientrel_shutdown().

2022 {
2023  if (bistate->current_buf != InvalidBuffer)
2024  ReleaseBuffer(bistate->current_buf);
2025  FreeAccessStrategy(bistate->strategy);
2026  pfree(bistate);
2027 }
#define InvalidBuffer
Definition: buf.h:25
void ReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3768
void pfree(void *pointer)
Definition: mcxt.c:1169
void FreeAccessStrategy(BufferAccessStrategy strategy)
Definition: freelist.c:597

◆ GetBulkInsertState()

BulkInsertState GetBulkInsertState ( void  )

Definition at line 2007 of file heapam.c.

References BAS_BULKWRITE, BulkInsertStateData::current_buf, GetAccessStrategy(), InvalidBuffer, palloc(), and BulkInsertStateData::strategy.

Referenced by ATRewriteTable(), CopyFrom(), CopyMultiInsertBufferInit(), intorel_startup(), and transientrel_startup().

2008 {
2009  BulkInsertState bistate;
2010 
2011  bistate = (BulkInsertState) palloc(sizeof(BulkInsertStateData));
2013  bistate->current_buf = InvalidBuffer;
2014  return bistate;
2015 }
BufferAccessStrategy GetAccessStrategy(BufferAccessStrategyType btype)
Definition: freelist.c:542
#define InvalidBuffer
Definition: buf.h:25
struct BulkInsertStateData * BulkInsertState
Definition: heapam.h:39
BufferAccessStrategy strategy
Definition: hio.h:31
void * palloc(Size size)
Definition: mcxt.c:1062
Buffer current_buf
Definition: hio.h:32

◆ heap_abort_speculative()

void heap_abort_speculative ( Relation  relation,
ItemPointer  tid 
)

Definition at line 5838 of file heapam.c.

References Assert, BUFFER_LOCK_EXCLUSIVE, BUFFER_LOCK_UNLOCK, BufferGetPage, compute_infobits(), elog, END_CRIT_SECTION, ERROR, xl_heap_delete::flags, GetCurrentTransactionId(), HEAP_KEYS_UPDATED, HEAP_MOVED, heap_toast_delete(), HEAP_XMAX_BITS, HeapTupleHasExternal, HeapTupleHeaderIsHeapOnly, HeapTupleHeaderIsSpeculative, HeapTupleHeaderSetXmin, xl_heap_delete::infobits_set, InvalidTransactionId, IsToastRelation(), ItemIdGetLength, ItemIdIsNormal, ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, ItemPointerIsValid, LockBuffer(), MarkBufferDirty(), xl_heap_delete::offnum, PageGetItem, PageGetItemId, PageIsAllVisible, PageSetLSN, PageSetPrunable, pgstat_count_heap_delete(), RelationData::rd_rel, ReadBuffer(), REGBUF_STANDARD, RelationGetRelid, RelationNeedsWAL, ReleaseBuffer(), SizeOfHeapDelete, START_CRIT_SECTION, HeapTupleHeaderData::t_choice, HeapTupleHeaderData::t_ctid, HeapTupleData::t_data, HeapTupleHeaderData::t_heap, HeapTupleHeaderData::t_infomask, HeapTupleHeaderData::t_infomask2, HeapTupleData::t_len, HeapTupleData::t_self, HeapTupleData::t_tableOid, HeapTupleFields::t_xmin, TransactionIdIsValid, TransactionIdPrecedes(), TransactionXmin, XLH_DELETE_IS_SUPER, XLOG_HEAP_DELETE, XLogBeginInsert(), XLogInsert(), XLogRegisterBuffer(), XLogRegisterData(), and xl_heap_delete::xmax.

Referenced by heapam_tuple_complete_speculative(), and toast_delete_datum().

5839 {
5841  ItemId lp;
5842  HeapTupleData tp;
5843  Page page;
5844  BlockNumber block;
5845  Buffer buffer;
5846  TransactionId prune_xid;
5847 
5848  Assert(ItemPointerIsValid(tid));
5849 
5850  block = ItemPointerGetBlockNumber(tid);
5851  buffer = ReadBuffer(relation, block);
5852  page = BufferGetPage(buffer);
5853 
5855 
5856  /*
5857  * Page can't be all visible, we just inserted into it, and are still
5858  * running.
5859  */
5860  Assert(!PageIsAllVisible(page));
5861 
5862  lp = PageGetItemId(page, ItemPointerGetOffsetNumber(tid));
5863  Assert(ItemIdIsNormal(lp));
5864 
5865  tp.t_tableOid = RelationGetRelid(relation);
5866  tp.t_data = (HeapTupleHeader) PageGetItem(page, lp);
5867  tp.t_len = ItemIdGetLength(lp);
5868  tp.t_self = *tid;
5869 
5870  /*
5871  * Sanity check that the tuple really is a speculatively inserted tuple,
5872  * inserted by us.
5873  */
5874  if (tp.t_data->t_choice.t_heap.t_xmin != xid)
5875  elog(ERROR, "attempted to kill a tuple inserted by another transaction");
5876  if (!(IsToastRelation(relation) || HeapTupleHeaderIsSpeculative(tp.t_data)))
5877  elog(ERROR, "attempted to kill a non-speculative tuple");
5879 
5880  /*
5881  * No need to check for serializable conflicts here. There is never a
5882  * need for a combo CID, either. No need to extract replica identity, or
5883  * do anything special with infomask bits.
5884  */
5885 
5887 
5888  /*
5889  * The tuple will become DEAD immediately. Flag that this page is a
5890  * candidate for pruning by setting xmin to TransactionXmin. While not
5891  * immediately prunable, it is the oldest xid we can cheaply determine
5892  * that's safe against wraparound / being older than the table's
5893  * relfrozenxid. To defend against the unlikely case of a new relation
5894  * having a newer relfrozenxid than our TransactionXmin, use relfrozenxid
5895  * if so (vacuum can't subsequently move relfrozenxid to beyond
5896  * TransactionXmin, so there's no race here).
5897  */
5899  if (TransactionIdPrecedes(TransactionXmin, relation->rd_rel->relfrozenxid))
5900  prune_xid = relation->rd_rel->relfrozenxid;
5901  else
5902  prune_xid = TransactionXmin;
5903  PageSetPrunable(page, prune_xid);
5904 
5905  /* store transaction information of xact deleting the tuple */
5908 
5909  /*
5910  * Set the tuple header xmin to InvalidTransactionId. This makes the
5911  * tuple immediately invisible everyone. (In particular, to any
5912  * transactions waiting on the speculative token, woken up later.)
5913  */
5915 
5916  /* Clear the speculative insertion token too */
5917  tp.t_data->t_ctid = tp.t_self;
5918 
5919  MarkBufferDirty(buffer);
5920 
5921  /*
5922  * XLOG stuff
5923  *
5924  * The WAL records generated here match heap_delete(). The same recovery
5925  * routines are used.
5926  */
5927  if (RelationNeedsWAL(relation))
5928  {
5929  xl_heap_delete xlrec;
5930  XLogRecPtr recptr;
5931 
5932  xlrec.flags = XLH_DELETE_IS_SUPER;
5934  tp.t_data->t_infomask2);
5936  xlrec.xmax = xid;
5937 
5938  XLogBeginInsert();
5939  XLogRegisterData((char *) &xlrec, SizeOfHeapDelete);
5940  XLogRegisterBuffer(0, buffer, REGBUF_STANDARD);
5941 
5942  /* No replica identity & replication origin logged */
5943 
5944  recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_DELETE);
5945 
5946  PageSetLSN(page, recptr);
5947  }
5948 
5949  END_CRIT_SECTION();
5950 
5951  LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
5952 
5953  if (HeapTupleHasExternal(&tp))
5954  {
5955  Assert(!IsToastRelation(relation));
5956  heap_toast_delete(relation, &tp, true);
5957  }
5958 
5959  /*
5960  * Never need to mark tuple for invalidation, since catalogs don't support
5961  * speculative insertion
5962  */
5963 
5964  /* Now we can release the buffer */
5965  ReleaseBuffer(buffer);
5966 
5967  /* count deletion, as we counted the insertion too */
5968  pgstat_count_heap_delete(relation);
5969 }
#define ItemPointerIsValid(pointer)
Definition: itemptr.h:82
#define BUFFER_LOCK_UNLOCK
Definition: bufmgr.h:96
bool IsToastRelation(Relation relation)
Definition: catalog.c:146
#define HEAP_XMAX_BITS
Definition: htup_details.h:270
#define XLH_DELETE_IS_SUPER
Definition: heapam_xlog.h:99
static uint8 compute_infobits(uint16 infomask, uint16 infomask2)
Definition: heapam.c:2655
HeapTupleFields t_heap
Definition: htup_details.h:156
#define PageIsAllVisible(page)
Definition: bufpage.h:385
uint32 TransactionId
Definition: c.h:587
void MarkBufferDirty(Buffer buffer)
Definition: bufmgr.c:1565
void XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
Definition: xloginsert.c:232
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
#define END_CRIT_SECTION()
Definition: miscadmin.h:149
#define HeapTupleHeaderIsSpeculative(tup)
Definition: htup_details.h:429
#define PageSetPrunable(page, xid)
Definition: bufpage.h:392
#define START_CRIT_SECTION()
Definition: miscadmin.h:147
uint32 BlockNumber
Definition: block.h:31
void ReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3768
#define BUFFER_LOCK_EXCLUSIVE
Definition: bufmgr.h:98
Form_pg_class rd_rel
Definition: rel.h:109
union HeapTupleHeaderData::@43 t_choice
OffsetNumber offnum
Definition: heapam_xlog.h:110
TransactionId TransactionXmin
Definition: snapmgr.c:112
HeapTupleHeader t_data
Definition: htup.h:68
#define HeapTupleHeaderIsHeapOnly(tup)
Definition: htup_details.h:500
#define ItemIdGetLength(itemId)
Definition: itemid.h:59
#define ERROR
Definition: elog.h:46
ItemPointerData t_ctid
Definition: htup_details.h:160
ItemPointerData t_self
Definition: htup.h:65
TransactionId xmax
Definition: heapam_xlog.h:109
void heap_toast_delete(Relation rel, HeapTuple oldtup, bool is_speculative)
Definition: heaptoast.c:43
TransactionId GetCurrentTransactionId(void)
Definition: xact.c:438
uint32 t_len
Definition: htup.h:64
#define SizeOfHeapDelete
Definition: heapam_xlog.h:115
#define REGBUF_STANDARD
Definition: xloginsert.h:35
#define InvalidTransactionId
Definition: transam.h:31
Oid t_tableOid
Definition: htup.h:66
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
bool TransactionIdPrecedes(TransactionId id1, TransactionId id2)
Definition: transam.c:300
TransactionId t_xmin
Definition: htup_details.h:123
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
void XLogRegisterData(char *data, int len)
Definition: xloginsert.c:340
XLogRecPtr XLogInsert(RmgrId rmid, uint8 info)
Definition: xloginsert.c:432
void LockBuffer(Buffer buffer, int mode)
Definition: bufmgr.c:4007
#define HEAP_KEYS_UPDATED
Definition: htup_details.h:278
#define HEAP_MOVED
Definition: htup_details.h:216
uint64 XLogRecPtr
Definition: xlogdefs.h:21
#define Assert(condition)
Definition: c.h:804
uint8 infobits_set
Definition: heapam_xlog.h:111
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
Buffer ReadBuffer(Relation reln, BlockNumber blockNum)
Definition: bufmgr.c:694
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
#define RelationNeedsWAL(relation)
Definition: rel.h:601
void pgstat_count_heap_delete(Relation rel)
Definition: pgstat.c:2285
#define HeapTupleHasExternal(tuple)
Definition: htup_details.h:672
#define elog(elevel,...)
Definition: elog.h:232
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
#define TransactionIdIsValid(xid)
Definition: transam.h:41
void XLogBeginInsert(void)
Definition: xloginsert.c:135
#define PageSetLSN(page, lsn)
Definition: bufpage.h:368
int Buffer
Definition: buf.h:23
#define XLOG_HEAP_DELETE
Definition: heapam_xlog.h:33
#define RelationGetRelid(relation)
Definition: rel.h:477
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
Pointer Page
Definition: bufpage.h:78
#define HeapTupleHeaderSetXmin(tup, xid)
Definition: htup_details.h:319

◆ heap_beginscan()

TableScanDesc heap_beginscan ( Relation  relation,
Snapshot  snapshot,
int  nkeys,
ScanKey  key,
ParallelTableScanDesc  parallel_scan,
uint32  flags 
)

Definition at line 1185 of file heapam.c.

References Assert, initscan(), IsMVCCSnapshot, palloc(), PredicateLockRelation(), RelationGetRelid, RelationIncrementReferenceCount(), HeapScanDescData::rs_base, HeapScanDescData::rs_ctup, TableScanDescData::rs_flags, TableScanDescData::rs_key, TableScanDescData::rs_nkeys, TableScanDescData::rs_parallel, HeapScanDescData::rs_parallelworkerdata, TableScanDescData::rs_rd, TableScanDescData::rs_snapshot, HeapScanDescData::rs_strategy, SO_ALLOW_PAGEMODE, SO_TYPE_SAMPLESCAN, SO_TYPE_SEQSCAN, and HeapTupleData::t_tableOid.

Referenced by SampleHeapTupleVisible().

1189 {
1190  HeapScanDesc scan;
1191 
1192  /*
1193  * increment relation ref count while scanning relation
1194  *
1195  * This is just to make really sure the relcache entry won't go away while
1196  * the scan has a pointer to it. Caller should be holding the rel open
1197  * anyway, so this is redundant in all normal scenarios...
1198  */
1200 
1201  /*
1202  * allocate and initialize scan descriptor
1203  */
1204  scan = (HeapScanDesc) palloc(sizeof(HeapScanDescData));
1205 
1206  scan->rs_base.rs_rd = relation;
1207  scan->rs_base.rs_snapshot = snapshot;
1208  scan->rs_base.rs_nkeys = nkeys;
1209  scan->rs_base.rs_flags = flags;
1210  scan->rs_base.rs_parallel = parallel_scan;
1211  scan->rs_strategy = NULL; /* set in initscan */
1212 
1213  /*
1214  * Disable page-at-a-time mode if it's not a MVCC-safe snapshot.
1215  */
1216  if (!(snapshot && IsMVCCSnapshot(snapshot)))
1218 
1219  /*
1220  * For seqscan and sample scans in a serializable transaction, acquire a
1221  * predicate lock on the entire relation. This is required not only to
1222  * lock all the matching tuples, but also to conflict with new insertions
1223  * into the table. In an indexscan, we take page locks on the index pages
1224  * covering the range specified in the scan qual, but in a heap scan there
1225  * is nothing more fine-grained to lock. A bitmap scan is a different
1226  * story, there we have already scanned the index and locked the index
1227  * pages covering the predicate. But in that case we still have to lock
1228  * any matching heap tuples. For sample scan we could optimize the locking
1229  * to be at least page-level granularity, but we'd need to add per-tuple
1230  * locking for that.
1231  */
1233  {
1234  /*
1235  * Ensure a missing snapshot is noticed reliably, even if the
1236  * isolation mode means predicate locking isn't performed (and
1237  * therefore the snapshot isn't used here).
1238  */
1239  Assert(snapshot);
1240  PredicateLockRelation(relation, snapshot);
1241  }
1242 
1243  /* we only need to set this up once */
1244  scan->rs_ctup.t_tableOid = RelationGetRelid(relation);
1245 
1246  /*
1247  * Allocate memory to keep track of page allocation for parallel workers
1248  * when doing a parallel scan.
1249  */
1250  if (parallel_scan != NULL)
1252  else
1253  scan->rs_parallelworkerdata = NULL;
1254 
1255  /*
1256  * we do this here instead of in initscan() because heap_rescan also calls
1257  * initscan() and we don't want to allocate memory again
1258  */
1259  if (nkeys > 0)
1260  scan->rs_base.rs_key = (ScanKey) palloc(sizeof(ScanKeyData) * nkeys);
1261  else
1262  scan->rs_base.rs_key = NULL;
1263 
1264  initscan(scan, key, false);
1265 
1266  return (TableScanDesc) scan;
1267 }
TableScanDescData rs_base
Definition: heapam.h:49
void PredicateLockRelation(Relation relation, Snapshot snapshot)
Definition: predicate.c:2569
uint32 rs_flags
Definition: relscan.h:47
struct HeapScanDescData * HeapScanDesc
Definition: heapam.h:79
HeapTupleData rs_ctup
Definition: heapam.h:66
ScanKeyData * ScanKey
Definition: skey.h:75
ParallelBlockTableScanWorkerData * rs_parallelworkerdata
Definition: heapam.h:72
Oid t_tableOid
Definition: htup.h:66
struct ScanKeyData * rs_key
Definition: relscan.h:37
void RelationIncrementReferenceCount(Relation rel)
Definition: relcache.c:2068
BufferAccessStrategy rs_strategy
Definition: heapam.h:64
#define IsMVCCSnapshot(snapshot)
Definition: snapmgr.h:96
#define Assert(condition)
Definition: c.h:804
Relation rs_rd
Definition: relscan.h:34
struct SnapshotData * rs_snapshot
Definition: relscan.h:35
void * palloc(Size size)
Definition: mcxt.c:1062
struct ParallelTableScanDescData * rs_parallel
Definition: relscan.h:49
static void initscan(HeapScanDesc scan, ScanKey key, bool keep_startblock)
Definition: heapam.c:227
#define RelationGetRelid(relation)
Definition: rel.h:477

◆ heap_delete()

TM_Result heap_delete ( Relation  relation,
ItemPointer  tid,
CommandId  cid,
Snapshot  crosscheck,
bool  wait,
struct TM_FailureData tmfd,
bool  changingPart 
)

Definition at line 2700 of file heapam.c.

References Assert, BUFFER_LOCK_EXCLUSIVE, BUFFER_LOCK_UNLOCK, BufferGetBlockNumber(), BufferGetPage, CacheInvalidateHeapTuple(), CheckForSerializableConflictIn(), TM_FailureData::cmax, compute_infobits(), compute_new_xmax_infomask(), TM_FailureData::ctid, DoesMultiXactIdConflict(), END_CRIT_SECTION, ereport, errcode(), errmsg(), ERROR, ExtractReplicaIdentity(), xl_heap_delete::flags, GetCurrentTransactionId(), heap_acquire_tuplock(), heap_freetuple(), HEAP_KEYS_UPDATED, HEAP_MOVED, heap_toast_delete(), HEAP_XMAX_BITS, HEAP_XMAX_INVALID, HEAP_XMAX_IS_LOCKED_ONLY, HEAP_XMAX_IS_MULTI, HeapTupleHasExternal, HeapTupleHeaderAdjustCmax(), HeapTupleHeaderClearHotUpdated, HeapTupleHeaderGetCmax(), HeapTupleHeaderGetRawXmax, HeapTupleHeaderGetUpdateXid, HeapTupleHeaderIsOnlyLocked(), HeapTupleHeaderSetCmax, HeapTupleHeaderSetMovedPartitions, HeapTupleHeaderSetXmax, HeapTupleSatisfiesUpdate(), HeapTupleSatisfiesVisibility(), xl_heap_delete::infobits_set, InvalidBuffer, InvalidCommandId, InvalidSnapshot, IsInParallelMode(), ItemIdGetLength, ItemIdIsNormal, ItemPointerEquals(), ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, ItemPointerIsValid, LockBuffer(), LockTupleExclusive, LockWaitBlock, log_heap_new_cid(), MarkBufferDirty(), MultiXactIdSetOldestMember(), MultiXactIdWait(), MultiXactStatusUpdate, xl_heap_delete::offnum, PageClearAllVisible, PageGetItem, PageGetItemId, PageIsAllVisible, PageSetLSN, PageSetPrunable, pgstat_count_heap_delete(), RelationData::rd_rel, ReadBuffer(), REGBUF_STANDARD, RelationGetRelid, RelationIsAccessibleInLogicalDecoding, RelationNeedsWAL, ReleaseBuffer(), SizeOfHeapDelete, SizeOfHeapHeader, SizeofHeapTupleHeader, START_CRIT_SECTION, HeapTupleHeaderData::t_ctid, HeapTupleData::t_data, xl_heap_header::t_hoff, HeapTupleHeaderData::t_hoff, xl_heap_header::t_infomask, HeapTupleHeaderData::t_infomask, xl_heap_header::t_infomask2, HeapTupleHeaderData::t_infomask2, HeapTupleData::t_len, HeapTupleData::t_self, HeapTupleData::t_tableOid, TM_BeingModified, TM_Deleted, TM_Invisible, TM_Ok, TM_SelfModified, TM_Updated, TransactionIdEquals, TransactionIdIsCurrentTransactionId(), UnlockReleaseBuffer(), UnlockTupleTuplock, UpdateXmaxHintBits(), visibilitymap_clear(), visibilitymap_pin(), VISIBILITYMAP_VALID_BITS, XactLockTableWait(), XLH_DELETE_ALL_VISIBLE_CLEARED, XLH_DELETE_CONTAINS_OLD_KEY, XLH_DELETE_CONTAINS_OLD_TUPLE, XLH_DELETE_IS_PARTITION_MOVE, XLOG_HEAP_DELETE, XLOG_INCLUDE_ORIGIN, XLogBeginInsert(), XLogInsert(), XLogRegisterBuffer(), XLogRegisterData(), XLogSetRecordFlags(), XLTW_Delete, xl_heap_delete::xmax, TM_FailureData::xmax, and xmax_infomask_changed().

Referenced by heapam_tuple_delete(), and simple_heap_delete().

2703 {
2704  TM_Result result;
2706  ItemId lp;
2707  HeapTupleData tp;
2708  Page page;
2709  BlockNumber block;
2710  Buffer buffer;
2711  Buffer vmbuffer = InvalidBuffer;
2712  TransactionId new_xmax;
2713  uint16 new_infomask,
2714  new_infomask2;
2715  bool have_tuple_lock = false;
2716  bool iscombo;
2717  bool all_visible_cleared = false;
2718  HeapTuple old_key_tuple = NULL; /* replica identity of the tuple */
2719  bool old_key_copied = false;
2720 
2721  Assert(ItemPointerIsValid(tid));
2722 
2723  /*
2724  * Forbid this during a parallel operation, lest it allocate a combo CID.
2725  * Other workers might need that combo CID for visibility checks, and we
2726  * have no provision for broadcasting it to them.
2727  */
2728  if (IsInParallelMode())
2729  ereport(ERROR,
2730  (errcode(ERRCODE_INVALID_TRANSACTION_STATE),
2731  errmsg("cannot delete tuples during a parallel operation")));
2732 
2733  block = ItemPointerGetBlockNumber(tid);
2734  buffer = ReadBuffer(relation, block);
2735  page = BufferGetPage(buffer);
2736 
2737  /*
2738  * Before locking the buffer, pin the visibility map page if it appears to
2739  * be necessary. Since we haven't got the lock yet, someone else might be
2740  * in the middle of changing this, so we'll need to recheck after we have
2741  * the lock.
2742  */
2743  if (PageIsAllVisible(page))
2744  visibilitymap_pin(relation, block, &vmbuffer);
2745 
2747 
2748  /*
2749  * If we didn't pin the visibility map page and the page has become all
2750  * visible while we were busy locking the buffer, we'll have to unlock and
2751  * re-lock, to avoid holding the buffer lock across an I/O. That's a bit
2752  * unfortunate, but hopefully shouldn't happen often.
2753  */
2754  if (vmbuffer == InvalidBuffer && PageIsAllVisible(page))
2755  {
2756  LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
2757  visibilitymap_pin(relation, block, &vmbuffer);
2759  }
2760 
2761  lp = PageGetItemId(page, ItemPointerGetOffsetNumber(tid));
2762  Assert(ItemIdIsNormal(lp));
2763 
2764  tp.t_tableOid = RelationGetRelid(relation);
2765  tp.t_data = (HeapTupleHeader) PageGetItem(page, lp);
2766  tp.t_len = ItemIdGetLength(lp);
2767  tp.t_self = *tid;
2768 
2769 l1:
2770  result = HeapTupleSatisfiesUpdate(&tp, cid, buffer);
2771 
2772  if (result == TM_Invisible)
2773  {
2774  UnlockReleaseBuffer(buffer);
2775  ereport(ERROR,
2776  (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
2777  errmsg("attempted to delete invisible tuple")));
2778  }
2779  else if (result == TM_BeingModified && wait)
2780  {
2781  TransactionId xwait;
2782  uint16 infomask;
2783 
2784  /* must copy state data before unlocking buffer */
2785  xwait = HeapTupleHeaderGetRawXmax(tp.t_data);
2786  infomask = tp.t_data->t_infomask;
2787 
2788  /*
2789  * Sleep until concurrent transaction ends -- except when there's a
2790  * single locker and it's our own transaction. Note we don't care
2791  * which lock mode the locker has, because we need the strongest one.
2792  *
2793  * Before sleeping, we need to acquire tuple lock to establish our
2794  * priority for the tuple (see heap_lock_tuple). LockTuple will
2795  * release us when we are next-in-line for the tuple.
2796  *
2797  * If we are forced to "start over" below, we keep the tuple lock;
2798  * this arranges that we stay at the head of the line while rechecking
2799  * tuple state.
2800  */
2801  if (infomask & HEAP_XMAX_IS_MULTI)
2802  {
2803  bool current_is_member = false;
2804 
2805  if (DoesMultiXactIdConflict((MultiXactId) xwait, infomask,
2806  LockTupleExclusive, &current_is_member))
2807  {
2808  LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
2809 
2810  /*
2811  * Acquire the lock, if necessary (but skip it when we're
2812  * requesting a lock and already have one; avoids deadlock).
2813  */
2814  if (!current_is_member)
2816  LockWaitBlock, &have_tuple_lock);
2817 
2818  /* wait for multixact */
2820  relation, &(tp.t_self), XLTW_Delete,
2821  NULL);
2823 
2824  /*
2825  * If xwait had just locked the tuple then some other xact
2826  * could update this tuple before we get to this point. Check
2827  * for xmax change, and start over if so.
2828  */
2829  if (xmax_infomask_changed(tp.t_data->t_infomask, infomask) ||
2831  xwait))
2832  goto l1;
2833  }
2834 
2835  /*
2836  * You might think the multixact is necessarily done here, but not
2837  * so: it could have surviving members, namely our own xact or
2838  * other subxacts of this backend. It is legal for us to delete
2839  * the tuple in either case, however (the latter case is
2840  * essentially a situation of upgrading our former shared lock to
2841  * exclusive). We don't bother changing the on-disk hint bits
2842  * since we are about to overwrite the xmax altogether.
2843  */
2844  }
2845  else if (!TransactionIdIsCurrentTransactionId(xwait))
2846  {
2847  /*
2848  * Wait for regular transaction to end; but first, acquire tuple
2849  * lock.
2850  */
2851  LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
2853  LockWaitBlock, &have_tuple_lock);
2854  XactLockTableWait(xwait, relation, &(tp.t_self), XLTW_Delete);
2856 
2857  /*
2858  * xwait is done, but if xwait had just locked the tuple then some
2859  * other xact could update this tuple before we get to this point.
2860  * Check for xmax change, and start over if so.
2861  */
2862  if (xmax_infomask_changed(tp.t_data->t_infomask, infomask) ||
2864  xwait))
2865  goto l1;
2866 
2867  /* Otherwise check if it committed or aborted */
2868  UpdateXmaxHintBits(tp.t_data, buffer, xwait);
2869  }
2870 
2871  /*
2872  * We may overwrite if previous xmax aborted, or if it committed but
2873  * only locked the tuple without updating it.
2874  */
2875  if ((tp.t_data->t_infomask & HEAP_XMAX_INVALID) ||
2878  result = TM_Ok;
2879  else if (!ItemPointerEquals(&tp.t_self, &tp.t_data->t_ctid))
2880  result = TM_Updated;
2881  else
2882  result = TM_Deleted;
2883  }
2884 
2885  if (crosscheck != InvalidSnapshot && result == TM_Ok)
2886  {
2887  /* Perform additional check for transaction-snapshot mode RI updates */
2888  if (!HeapTupleSatisfiesVisibility(&tp, crosscheck, buffer))
2889  result = TM_Updated;
2890  }
2891 
2892  if (result != TM_Ok)
2893  {
2894  Assert(result == TM_SelfModified ||
2895  result == TM_Updated ||
2896  result == TM_Deleted ||
2897  result == TM_BeingModified);
2899  Assert(result != TM_Updated ||
2900  !ItemPointerEquals(&tp.t_self, &tp.t_data->t_ctid));
2901  tmfd->ctid = tp.t_data->t_ctid;
2903  if (result == TM_SelfModified)
2904  tmfd->cmax = HeapTupleHeaderGetCmax(tp.t_data);
2905  else
2906  tmfd->cmax = InvalidCommandId;
2907  UnlockReleaseBuffer(buffer);
2908  if (have_tuple_lock)
2909  UnlockTupleTuplock(relation, &(tp.t_self), LockTupleExclusive);
2910  if (vmbuffer != InvalidBuffer)
2911  ReleaseBuffer(vmbuffer);
2912  return result;
2913  }
2914 
2915  /*
2916  * We're about to do the actual delete -- check for conflict first, to
2917  * avoid possibly having to roll back work we've just done.
2918  *
2919  * This is safe without a recheck as long as there is no possibility of
2920  * another process scanning the page between this check and the delete
2921  * being visible to the scan (i.e., an exclusive buffer content lock is
2922  * continuously held from this point until the tuple delete is visible).
2923  */
2924  CheckForSerializableConflictIn(relation, tid, BufferGetBlockNumber(buffer));
2925 
2926  /* replace cid with a combo CID if necessary */
2927  HeapTupleHeaderAdjustCmax(tp.t_data, &cid, &iscombo);
2928 
2929  /*
2930  * Compute replica identity tuple before entering the critical section so
2931  * we don't PANIC upon a memory allocation failure.
2932  */
2933  old_key_tuple = ExtractReplicaIdentity(relation, &tp, true, &old_key_copied);
2934 
2935  /*
2936  * If this is the first possibly-multixact-able operation in the current
2937  * transaction, set my per-backend OldestMemberMXactId setting. We can be
2938  * certain that the transaction will never become a member of any older
2939  * MultiXactIds than that. (We have to do this even if we end up just
2940  * using our own TransactionId below, since some other backend could
2941  * incorporate our XID into a MultiXact immediately afterwards.)
2942  */
2944 
2947  xid, LockTupleExclusive, true,
2948  &new_xmax, &new_infomask, &new_infomask2);
2949 
2951 
2952  /*
2953  * If this transaction commits, the tuple will become DEAD sooner or
2954  * later. Set flag that this page is a candidate for pruning once our xid
2955  * falls below the OldestXmin horizon. If the transaction finally aborts,
2956  * the subsequent page pruning will be a no-op and the hint will be
2957  * cleared.
2958  */
2959  PageSetPrunable(page, xid);
2960 
2961  if (PageIsAllVisible(page))
2962  {
2963  all_visible_cleared = true;
2964  PageClearAllVisible(page);
2965  visibilitymap_clear(relation, BufferGetBlockNumber(buffer),
2966  vmbuffer, VISIBILITYMAP_VALID_BITS);
2967  }
2968 
2969  /* store transaction information of xact deleting the tuple */
2972  tp.t_data->t_infomask |= new_infomask;
2973  tp.t_data->t_infomask2 |= new_infomask2;
2975  HeapTupleHeaderSetXmax(tp.t_data, new_xmax);
2976  HeapTupleHeaderSetCmax(tp.t_data, cid, iscombo);
2977  /* Make sure there is no forward chain link in t_ctid */
2978  tp.t_data->t_ctid = tp.t_self;
2979 
2980  /* Signal that this is actually a move into another partition */
2981  if (changingPart)
2983 
2984  MarkBufferDirty(buffer);
2985 
2986  /*
2987  * XLOG stuff
2988  *
2989  * NB: heap_abort_speculative() uses the same xlog record and replay
2990  * routines.
2991  */
2992  if (RelationNeedsWAL(relation))
2993  {
2994  xl_heap_delete xlrec;
2995  xl_heap_header xlhdr;
2996  XLogRecPtr recptr;
2997 
2998  /*
2999  * For logical decode we need combo CIDs to properly decode the
3000  * catalog
3001  */
3003  log_heap_new_cid(relation, &tp);
3004 
3005  xlrec.flags = 0;
3006  if (all_visible_cleared)
3008  if (changingPart)
3011  tp.t_data->t_infomask2);
3013  xlrec.xmax = new_xmax;
3014 
3015  if (old_key_tuple != NULL)
3016  {
3017  if (relation->rd_rel->relreplident == REPLICA_IDENTITY_FULL)
3019  else
3021  }
3022 
3023  XLogBeginInsert();
3024  XLogRegisterData((char *) &xlrec, SizeOfHeapDelete);
3025 
3026  XLogRegisterBuffer(0, buffer, REGBUF_STANDARD);
3027 
3028  /*
3029  * Log replica identity of the deleted tuple if there is one
3030  */
3031  if (old_key_tuple != NULL)
3032  {
3033  xlhdr.t_infomask2 = old_key_tuple->t_data->t_infomask2;
3034  xlhdr.t_infomask = old_key_tuple->t_data->t_infomask;
3035  xlhdr.t_hoff = old_key_tuple->t_data->t_hoff;
3036 
3037  XLogRegisterData((char *) &xlhdr, SizeOfHeapHeader);
3038  XLogRegisterData((char *) old_key_tuple->t_data
3040  old_key_tuple->t_len
3042  }
3043 
3044  /* filtering by origin on a row level is much more efficient */
3046 
3047  recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_DELETE);
3048 
3049  PageSetLSN(page, recptr);
3050  }
3051 
3052  END_CRIT_SECTION();
3053 
3054  LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
3055 
3056  if (vmbuffer != InvalidBuffer)
3057  ReleaseBuffer(vmbuffer);
3058 
3059  /*
3060  * If the tuple has toasted out-of-line attributes, we need to delete
3061  * those items too. We have to do this before releasing the buffer
3062  * because we need to look at the contents of the tuple, but it's OK to
3063  * release the content lock on the buffer first.
3064  */
3065  if (relation->rd_rel->relkind != RELKIND_RELATION &&
3066  relation->rd_rel->relkind != RELKIND_MATVIEW)
3067  {
3068  /* toast table entries should never be recursively toasted */
3070  }
3071  else if (HeapTupleHasExternal(&tp))
3072  heap_toast_delete(relation, &tp, false);
3073 
3074  /*
3075  * Mark tuple for invalidation from system caches at next command
3076  * boundary. We have to do this before releasing the buffer because we
3077  * need to look at the contents of the tuple.
3078  */
3079  CacheInvalidateHeapTuple(relation, &tp, NULL);
3080 
3081  /* Now we can release the buffer */
3082  ReleaseBuffer(buffer);
3083 
3084  /*
3085  * Release the lmgr tuple lock, if we had it.
3086  */
3087  if (have_tuple_lock)
3088  UnlockTupleTuplock(relation, &(tp.t_self), LockTupleExclusive);
3089 
3090  pgstat_count_heap_delete(relation);
3091 
3092  if (old_key_tuple != NULL && old_key_copied)
3093  heap_freetuple(old_key_tuple);
3094 
3095  return TM_Ok;
3096 }
#define HeapTupleHeaderGetUpdateXid(tup)
Definition: htup_details.h:365
#define ItemPointerIsValid(pointer)
Definition: itemptr.h:82
ItemPointerData ctid
Definition: tableam.h:126
#define BUFFER_LOCK_UNLOCK
Definition: bufmgr.h:96
#define SizeofHeapTupleHeader
Definition: htup_details.h:184
static XLogRecPtr log_heap_new_cid(Relation relation, HeapTuple tup)
Definition: heapam.c:8228
#define HEAP_XMAX_BITS
Definition: htup_details.h:270
static uint8 compute_infobits(uint16 infomask, uint16 infomask2)
Definition: heapam.c:2655
void CacheInvalidateHeapTuple(Relation relation, HeapTuple tuple, HeapTuple newtuple)
Definition: inval.c:1186
#define TransactionIdEquals(id1, id2)
Definition: transam.h:43
#define PageIsAllVisible(page)
Definition: bufpage.h:385
uint32 TransactionId
Definition: c.h:587
bool TransactionIdIsCurrentTransactionId(TransactionId xid)
Definition: xact.c:869
void visibilitymap_pin(Relation rel, BlockNumber heapBlk, Buffer *buf)
void MarkBufferDirty(Buffer buffer)
Definition: bufmgr.c:1565
void XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
Definition: xloginsert.c:232
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
static HeapTuple ExtractReplicaIdentity(Relation rel, HeapTuple tup, bool key_changed, bool *copy)
Definition: heapam.c:8310
static bool xmax_infomask_changed(uint16 new_infomask, uint16 old_infomask)
Definition: heapam.c:2677
#define HeapTupleHeaderClearHotUpdated(tup)
Definition: htup_details.h:495
#define END_CRIT_SECTION()
Definition: miscadmin.h:149
CommandId cmax
Definition: tableam.h:128
bool HeapTupleHeaderIsOnlyLocked(HeapTupleHeader tuple)
#define InvalidBuffer
Definition: buf.h:25
uint16 t_infomask2
Definition: heapam_xlog.h:146
#define PageSetPrunable(page, xid)
Definition: bufpage.h:392
#define START_CRIT_SECTION()
Definition: miscadmin.h:147
int errcode(int sqlerrcode)
Definition: elog.c:698
#define XLOG_INCLUDE_ORIGIN
Definition: xlog.h:213
uint32 BlockNumber
Definition: block.h:31
void ReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3768
#define BUFFER_LOCK_EXCLUSIVE
Definition: bufmgr.h:98
Form_pg_class rd_rel
Definition: rel.h:109
void heap_freetuple(HeapTuple htup)
Definition: heaptuple.c:1338
TM_Result HeapTupleSatisfiesUpdate(HeapTuple htup, CommandId curcid, Buffer buffer)
#define UnlockTupleTuplock(rel, tup, mode)
Definition: heapam.c:165
OffsetNumber offnum
Definition: heapam_xlog.h:110
void MultiXactIdSetOldestMember(void)
Definition: multixact.c:625
#define VISIBILITYMAP_VALID_BITS
Definition: visibilitymap.h:28
HeapTupleHeader t_data
Definition: htup.h:68
#define HeapTupleHeaderGetRawXmax(tup)
Definition: htup_details.h:375
unsigned short uint16
Definition: c.h:440
#define ItemIdGetLength(itemId)
Definition: itemid.h:59
bool IsInParallelMode(void)
Definition: xact.c:1012
bool visibilitymap_clear(Relation rel, BlockNumber heapBlk, Buffer buf, uint8 flags)
void UnlockReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3791
TransactionId xmax
Definition: tableam.h:127
#define ERROR
Definition: elog.h:46
#define HEAP_XMAX_INVALID
Definition: htup_details.h:207
ItemPointerData t_ctid
Definition: htup_details.h:160
#define HeapTupleHeaderSetMovedPartitions(tup)
Definition: htup_details.h:448
ItemPointerData t_self
Definition: htup.h:65
TransactionId xmax
Definition: heapam_xlog.h:109
static void MultiXactIdWait(MultiXactId multi, MultiXactStatus status, uint16 infomask, Relation rel, ItemPointer ctid, XLTW_Oper oper, int *remaining)
Definition: heapam.c:7009
void heap_toast_delete(Relation rel, HeapTuple oldtup, bool is_speculative)
Definition: heaptoast.c:43
TransactionId GetCurrentTransactionId(void)
Definition: xact.c:438
uint32 t_len
Definition: htup.h:64
#define SizeOfHeapDelete
Definition: heapam_xlog.h:115
#define REGBUF_STANDARD
Definition: xloginsert.h:35
#define XLH_DELETE_CONTAINS_OLD_KEY
Definition: heapam_xlog.h:98
#define HeapTupleHeaderSetXmax(tup, xid)
Definition: htup_details.h:380
Oid t_tableOid
Definition: htup.h:66
void XLogSetRecordFlags(uint8 flags)
Definition: xloginsert.c:414
#define HeapTupleHeaderSetCmax(tup, cid, iscombo)
Definition: htup_details.h:405
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
static void compute_new_xmax_infomask(TransactionId xmax, uint16 old_infomask, uint16 old_infomask2, TransactionId add_to_xmax, LockTupleMode mode, bool is_update, TransactionId *result_xmax, uint16 *result_infomask, uint16 *result_infomask2)
Definition: heapam.c:4985
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
#define InvalidSnapshot
Definition: snapshot.h:123
void XLogRegisterData(char *data, int len)
Definition: xloginsert.c:340
XLogRecPtr XLogInsert(RmgrId rmid, uint8 info)
Definition: xloginsert.c:432
TM_Result
Definition: tableam.h:71
#define RelationIsAccessibleInLogicalDecoding(relation)
Definition: rel.h:657
#define InvalidCommandId
Definition: c.h:604
#define HEAP_XMAX_IS_LOCKED_ONLY(infomask)
Definition: htup_details.h:230
void LockBuffer(Buffer buffer, int mode)
Definition: bufmgr.c:4007
#define HEAP_KEYS_UPDATED
Definition: htup_details.h:278
#define HEAP_XMAX_IS_MULTI
Definition: htup_details.h:208
void CheckForSerializableConflictIn(Relation relation, ItemPointer tid, BlockNumber blkno)
Definition: predicate.c:4446
static void UpdateXmaxHintBits(HeapTupleHeader tuple, Buffer buffer, TransactionId xid)
Definition: heapam.c:1985
#define HEAP_MOVED
Definition: htup_details.h:216
static bool heap_acquire_tuplock(Relation relation, ItemPointer tid, LockTupleMode mode, LockWaitPolicy wait_policy, bool *have_tuple_lock)
Definition: heapam.c:4936
#define ereport(elevel,...)
Definition: elog.h:157
TransactionId MultiXactId
Definition: c.h:597
#define PageClearAllVisible(page)
Definition: bufpage.h:389
void XactLockTableWait(TransactionId xid, Relation rel, ItemPointer ctid, XLTW_Oper oper)
Definition: lmgr.c:640
uint64 XLogRecPtr
Definition: xlogdefs.h:21
#define Assert(condition)
Definition: c.h:804
uint8 infobits_set
Definition: heapam_xlog.h:111
CommandId HeapTupleHeaderGetCmax(HeapTupleHeader tup)
Definition: combocid.c:118
Definition: tableam.h:77
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
Buffer ReadBuffer(Relation reln, BlockNumber blockNum)
Definition: bufmgr.c:694
uint16 t_infomask
Definition: heapam_xlog.h:147
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
static bool DoesMultiXactIdConflict(MultiXactId multi, uint16 infomask, LockTupleMode lockmode, bool *current_is_member)
Definition: heapam.c:6832
#define RelationNeedsWAL(relation)
Definition: rel.h:601
bool ItemPointerEquals(ItemPointer pointer1, ItemPointer pointer2)
Definition: itemptr.c:29
void pgstat_count_heap_delete(Relation rel)
Definition: pgstat.c:2285
void HeapTupleHeaderAdjustCmax(HeapTupleHeader tup, CommandId *cmax, bool *iscombo)
Definition: combocid.c:153
BlockNumber BufferGetBlockNumber(Buffer buffer)
Definition: bufmgr.c:2748
#define HeapTupleHasExternal(tuple)
Definition: htup_details.h:672
int errmsg(const char *fmt,...)
Definition: elog.c:909
#define XLH_DELETE_ALL_VISIBLE_CLEARED
Definition: heapam_xlog.h:96
#define XLH_DELETE_IS_PARTITION_MOVE
Definition: heapam_xlog.h:100
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
void XLogBeginInsert(void)
Definition: xloginsert.c:135
#define PageSetLSN(page, lsn)
Definition: bufpage.h:368
int Buffer
Definition: buf.h:23
#define XLOG_HEAP_DELETE
Definition: heapam_xlog.h:33
#define RelationGetRelid(relation)
Definition: rel.h:477
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
#define SizeOfHeapHeader
Definition: heapam_xlog.h:151
Pointer Page
Definition: bufpage.h:78
bool HeapTupleSatisfiesVisibility(HeapTuple tup, Snapshot snapshot, Buffer buffer)
#define XLH_DELETE_CONTAINS_OLD_TUPLE
Definition: heapam_xlog.h:97

◆ heap_endscan()

void heap_endscan ( TableScanDesc  scan)

Definition at line 1307 of file heapam.c.

References BufferIsValid, FreeAccessStrategy(), pfree(), RelationDecrementReferenceCount(), ReleaseBuffer(), HeapScanDescData::rs_base, HeapScanDescData::rs_cbuf, TableScanDescData::rs_flags, TableScanDescData::rs_key, HeapScanDescData::rs_parallelworkerdata, TableScanDescData::rs_rd, TableScanDescData::rs_snapshot, HeapScanDescData::rs_strategy, SO_TEMP_SNAPSHOT, and UnregisterSnapshot().

Referenced by SampleHeapTupleVisible().

1308 {
1309  HeapScanDesc scan = (HeapScanDesc) sscan;
1310 
1311  /* Note: no locking manipulations needed */
1312 
1313  /*
1314  * unpin scan buffers
1315  */
1316  if (BufferIsValid(scan->rs_cbuf))
1317  ReleaseBuffer(scan->rs_cbuf);
1318 
1319  /*
1320  * decrement relation reference count and free scan descriptor storage
1321  */
1323 
1324  if (scan->rs_base.rs_key)
1325  pfree(scan->rs_base.rs_key);
1326 
1327  if (scan->rs_strategy != NULL)
1329 
1330  if (scan->rs_parallelworkerdata != NULL)
1332 
1333  if (scan->rs_base.rs_flags & SO_TEMP_SNAPSHOT)
1335 
1336  pfree(scan);
1337 }
TableScanDescData rs_base
Definition: heapam.h:49
void ReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3768
uint32 rs_flags
Definition: relscan.h:47
struct HeapScanDescData * HeapScanDesc
Definition: heapam.h:79
void pfree(void *pointer)
Definition: mcxt.c:1169
void RelationDecrementReferenceCount(Relation rel)
Definition: relcache.c:2081
ParallelBlockTableScanWorkerData * rs_parallelworkerdata
Definition: heapam.h:72
struct ScanKeyData * rs_key
Definition: relscan.h:37
void UnregisterSnapshot(Snapshot snapshot)
Definition: snapmgr.c:852
BufferAccessStrategy rs_strategy
Definition: heapam.h:64
Buffer rs_cbuf
Definition: heapam.h:60
void FreeAccessStrategy(BufferAccessStrategy strategy)
Definition: freelist.c:597
#define BufferIsValid(bufnum)
Definition: bufmgr.h:123
Relation rs_rd
Definition: relscan.h:34
struct SnapshotData * rs_snapshot
Definition: relscan.h:35

◆ heap_fetch()

bool heap_fetch ( Relation  relation,
Snapshot  snapshot,
HeapTuple  tuple,
Buffer userbuf 
)

Definition at line 1595 of file heapam.c.

References BUFFER_LOCK_SHARE, BUFFER_LOCK_UNLOCK, BufferGetPage, HeapCheckForSerializableConflictOut(), HeapTupleHeaderGetXmin, HeapTupleSatisfiesVisibility(), InvalidBuffer, ItemIdGetLength, ItemIdIsNormal, ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, LockBuffer(), PageGetItem, PageGetItemId, PageGetMaxOffsetNumber, PredicateLockTID(), ReadBuffer(), RelationGetRelid, ReleaseBuffer(), HeapTupleData::t_data, HeapTupleData::t_len, HeapTupleData::t_self, HeapTupleData::t_tableOid, and TestForOldSnapshot().

Referenced by heap_lock_updated_tuple_rec(), heapam_fetch_row_version(), and heapam_tuple_lock().

1599 {
1600  ItemPointer tid = &(tuple->t_self);
1601  ItemId lp;
1602  Buffer buffer;
1603  Page page;
1604  OffsetNumber offnum;
1605  bool valid;
1606 
1607  /*
1608  * Fetch and pin the appropriate page of the relation.
1609  */
1610  buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid));
1611 
1612  /*
1613  * Need share lock on buffer to examine tuple commit status.
1614  */
1615  LockBuffer(buffer, BUFFER_LOCK_SHARE);
1616  page = BufferGetPage(buffer);
1617  TestForOldSnapshot(snapshot, relation, page);
1618 
1619  /*
1620  * We'd better check for out-of-range offnum in case of VACUUM since the
1621  * TID was obtained.
1622  */
1623  offnum = ItemPointerGetOffsetNumber(tid);
1624  if (offnum < FirstOffsetNumber || offnum > PageGetMaxOffsetNumber(page))
1625  {
1626  LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
1627  ReleaseBuffer(buffer);
1628  *userbuf = InvalidBuffer;
1629  tuple->t_data = NULL;
1630  return false;
1631  }
1632 
1633  /*
1634  * get the item line pointer corresponding to the requested tid
1635  */
1636  lp = PageGetItemId(page, offnum);
1637 
1638  /*
1639  * Must check for deleted tuple.
1640  */
1641  if (!ItemIdIsNormal(lp))
1642  {
1643  LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
1644  ReleaseBuffer(buffer);
1645  *userbuf = InvalidBuffer;
1646  tuple->t_data = NULL;
1647  return false;
1648  }
1649 
1650  /*
1651  * fill in *tuple fields
1652  */
1653  tuple->t_data = (HeapTupleHeader) PageGetItem(page, lp);
1654  tuple->t_len = ItemIdGetLength(lp);
1655  tuple->t_tableOid = RelationGetRelid(relation);
1656 
1657  /*
1658  * check tuple visibility, then release lock
1659  */
1660  valid = HeapTupleSatisfiesVisibility(tuple, snapshot, buffer);
1661 
1662  if (valid)
1663  PredicateLockTID(relation, &(tuple->t_self), snapshot,
1664  HeapTupleHeaderGetXmin(tuple->t_data));
1665 
1666  HeapCheckForSerializableConflictOut(valid, relation, tuple, buffer, snapshot);
1667 
1668  LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
1669 
1670  if (valid)
1671  {
1672  /*
1673  * All checks passed, so return the tuple as valid. Caller is now
1674  * responsible for releasing the buffer.
1675  */
1676  *userbuf = buffer;
1677 
1678  return true;
1679  }
1680 
1681  /* Tuple failed time qual */
1682  ReleaseBuffer(buffer);
1683  *userbuf = InvalidBuffer;
1684 
1685  return false;
1686 }
#define BUFFER_LOCK_UNLOCK
Definition: bufmgr.h:96
static void TestForOldSnapshot(Snapshot snapshot, Relation relation, Page page)
Definition: bufmgr.h:278
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
#define InvalidBuffer
Definition: buf.h:25
void ReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3768
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
uint16 OffsetNumber
Definition: off.h:24
HeapTupleHeader t_data
Definition: htup.h:68
#define ItemIdGetLength(itemId)
Definition: itemid.h:59
ItemPointerData t_self
Definition: htup.h:65
uint32 t_len
Definition: htup.h:64
Oid t_tableOid
Definition: htup.h:66
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
void LockBuffer(Buffer buffer, int mode)
Definition: bufmgr.c:4007
void PredicateLockTID(Relation relation, ItemPointer tid, Snapshot snapshot, TransactionId tuple_xid)
Definition: predicate.c:2614
void HeapCheckForSerializableConflictOut(bool visible, Relation relation, HeapTuple tuple, Buffer buffer, Snapshot snapshot)
Definition: heapam.c:9789
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
#define HeapTupleHeaderGetXmin(tup)
Definition: htup_details.h:313
Buffer ReadBuffer(Relation reln, BlockNumber blockNum)
Definition: bufmgr.c:694
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
#define BUFFER_LOCK_SHARE
Definition: bufmgr.h:97
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
int Buffer
Definition: buf.h:23
#define RelationGetRelid(relation)
Definition: rel.h:477
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
Pointer Page
Definition: bufpage.h:78
bool HeapTupleSatisfiesVisibility(HeapTuple tup, Snapshot snapshot, Buffer buffer)

◆ heap_finish_speculative()

void heap_finish_speculative ( Relation  relation,
ItemPointer  tid 
)

Definition at line 5747 of file heapam.c.

References Assert, BUFFER_LOCK_EXCLUSIVE, BufferGetPage, elog, END_CRIT_SECTION, ERROR, HeapTupleHeaderIsSpeculative, ItemIdIsNormal, ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, LockBuffer(), MarkBufferDirty(), MaxOffsetNumber, xl_heap_confirm::offnum, PageGetItem, PageGetItemId, PageGetMaxOffsetNumber, PageSetLSN, ReadBuffer(), REGBUF_STANDARD, RelationNeedsWAL, SizeOfHeapConfirm, SpecTokenOffsetNumber, START_CRIT_SECTION, StaticAssertStmt, HeapTupleHeaderData::t_ctid, UnlockReleaseBuffer(), XLOG_HEAP_CONFIRM, XLOG_INCLUDE_ORIGIN, XLogBeginInsert(), XLogInsert(), XLogRegisterBuffer(), XLogRegisterData(), and XLogSetRecordFlags().

Referenced by heapam_tuple_complete_speculative().

5748 {
5749  Buffer buffer;
5750  Page page;
5751  OffsetNumber offnum;
5752  ItemId lp = NULL;
5753  HeapTupleHeader htup;
5754 
5755  buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid));
5757  page = (Page) BufferGetPage(buffer);
5758 
5759  offnum = ItemPointerGetOffsetNumber(tid);
5760  if (PageGetMaxOffsetNumber(page) >= offnum)
5761  lp = PageGetItemId(page, offnum);
5762 
5763  if (PageGetMaxOffsetNumber(page) < offnum || !ItemIdIsNormal(lp))
5764  elog(ERROR, "invalid lp");
5765 
5766  htup = (HeapTupleHeader) PageGetItem(page, lp);
5767 
5768  /* SpecTokenOffsetNumber should be distinguishable from any real offset */
5770  "invalid speculative token constant");
5771 
5772  /* NO EREPORT(ERROR) from here till changes are logged */
5774 
5776 
5777  MarkBufferDirty(buffer);
5778 
5779  /*
5780  * Replace the speculative insertion token with a real t_ctid, pointing to
5781  * itself like it does on regular tuples.
5782  */
5783  htup->t_ctid = *tid;
5784 
5785  /* XLOG stuff */
5786  if (RelationNeedsWAL(relation))
5787  {
5788  xl_heap_confirm xlrec;
5789  XLogRecPtr recptr;
5790 
5791  xlrec.offnum = ItemPointerGetOffsetNumber(tid);
5792 
5793  XLogBeginInsert();
5794 
5795  /* We want the same filtering on this as on a plain insert */
5797 
5798  XLogRegisterData((char *) &xlrec, SizeOfHeapConfirm);
5799  XLogRegisterBuffer(0, buffer, REGBUF_STANDARD);
5800 
5801  recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_CONFIRM);
5802 
5803  PageSetLSN(page, recptr);
5804  }
5805 
5806  END_CRIT_SECTION();
5807 
5808  UnlockReleaseBuffer(buffer);
5809 }
OffsetNumber offnum
Definition: heapam_xlog.h:302
void MarkBufferDirty(Buffer buffer)
Definition: bufmgr.c:1565
void XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
Definition: xloginsert.c:232
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
#define MaxOffsetNumber
Definition: off.h:28
#define END_CRIT_SECTION()
Definition: miscadmin.h:149
#define HeapTupleHeaderIsSpeculative(tup)
Definition: htup_details.h:429
#define START_CRIT_SECTION()
Definition: miscadmin.h:147
#define XLOG_INCLUDE_ORIGIN
Definition: xlog.h:213
#define BUFFER_LOCK_EXCLUSIVE
Definition: bufmgr.h:98
#define SpecTokenOffsetNumber
Definition: itemptr.h:63
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
uint16 OffsetNumber
Definition: off.h:24
#define StaticAssertStmt(condition, errmessage)
Definition: c.h:918
void UnlockReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3791
#define ERROR
Definition: elog.h:46
ItemPointerData t_ctid
Definition: htup_details.h:160
#define REGBUF_STANDARD
Definition: xloginsert.h:35
void XLogSetRecordFlags(uint8 flags)
Definition: xloginsert.c:414
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
#define SizeOfHeapConfirm
Definition: heapam_xlog.h:305
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
void XLogRegisterData(char *data, int len)
Definition: xloginsert.c:340
XLogRecPtr XLogInsert(RmgrId rmid, uint8 info)
Definition: xloginsert.c:432
void LockBuffer(Buffer buffer, int mode)
Definition: bufmgr.c:4007
uint64 XLogRecPtr
Definition: xlogdefs.h:21
#define Assert(condition)
Definition: c.h:804
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
Buffer ReadBuffer(Relation reln, BlockNumber blockNum)
Definition: bufmgr.c:694
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
#define RelationNeedsWAL(relation)
Definition: rel.h:601
#define elog(elevel,...)
Definition: elog.h:232
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
void XLogBeginInsert(void)
Definition: xloginsert.c:135
#define PageSetLSN(page, lsn)
Definition: bufpage.h:368
int Buffer
Definition: buf.h:23
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
Pointer Page
Definition: bufpage.h:78
#define XLOG_HEAP_CONFIRM
Definition: heapam_xlog.h:37

◆ heap_freeze_tuple()

bool heap_freeze_tuple ( HeapTupleHeader  tuple,
TransactionId  relfrozenxid,
TransactionId  relminmxid,
TransactionId  cutoff_xid,
TransactionId  cutoff_multi 
)

Definition at line 6652 of file heapam.c.

References heap_execute_freeze_tuple(), and heap_prepare_freeze_tuple().

Referenced by rewrite_heap_tuple().

6655 {
6657  bool do_freeze;
6658  bool tuple_totally_frozen;
6659 
6660  do_freeze = heap_prepare_freeze_tuple(tuple,
6661  relfrozenxid, relminmxid,
6662  cutoff_xid, cutoff_multi,
6663  &frz, &tuple_totally_frozen);
6664 
6665  /*
6666  * Note that because this is not a WAL-logged operation, we don't need to
6667  * fill in the offset in the freeze record.
6668  */
6669 
6670  if (do_freeze)
6671  heap_execute_freeze_tuple(tuple, &frz);
6672  return do_freeze;
6673 }
bool heap_prepare_freeze_tuple(HeapTupleHeader tuple, TransactionId relfrozenxid, TransactionId relminmxid, TransactionId cutoff_xid, TransactionId cutoff_multi, xl_heap_freeze_tuple *frz, bool *totally_frozen_p)
Definition: heapam.c:6402
void heap_execute_freeze_tuple(HeapTupleHeader tuple, xl_heap_freeze_tuple *frz)
Definition: heapam.c:6631

◆ heap_get_latest_tid()

void heap_get_latest_tid ( TableScanDesc  scan,
ItemPointer  tid 
)

Definition at line 1862 of file heapam.c.

References Assert, BUFFER_LOCK_SHARE, BufferGetPage, HEAP_XMAX_INVALID, HeapCheckForSerializableConflictOut(), HeapTupleHeaderGetUpdateXid, HeapTupleHeaderGetXmin, HeapTupleHeaderIndicatesMovedPartitions, HeapTupleHeaderIsOnlyLocked(), HeapTupleSatisfiesVisibility(), InvalidTransactionId, ItemIdGetLength, ItemIdIsNormal, ItemPointerEquals(), ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, ItemPointerIsValid, LockBuffer(), PageGetItem, PageGetItemId, PageGetMaxOffsetNumber, ReadBuffer(), RelationGetRelid, TableScanDescData::rs_rd, TableScanDescData::rs_snapshot, HeapTupleHeaderData::t_ctid, HeapTupleData::t_data, HeapTupleHeaderData::t_infomask, HeapTupleData::t_len, HeapTupleData::t_self, HeapTupleData::t_tableOid, TestForOldSnapshot(), TransactionIdEquals, TransactionIdIsValid, and UnlockReleaseBuffer().

Referenced by SampleHeapTupleVisible().

1864 {
1865  Relation relation = sscan->rs_rd;
1866  Snapshot snapshot = sscan->rs_snapshot;
1867  ItemPointerData ctid;
1868  TransactionId priorXmax;
1869 
1870  /*
1871  * table_tuple_get_latest_tid() verified that the passed in tid is valid.
1872  * Assume that t_ctid links are valid however - there shouldn't be invalid
1873  * ones in the table.
1874  */
1875  Assert(ItemPointerIsValid(tid));
1876 
1877  /*
1878  * Loop to chase down t_ctid links. At top of loop, ctid is the tuple we
1879  * need to examine, and *tid is the TID we will return if ctid turns out
1880  * to be bogus.
1881  *
1882  * Note that we will loop until we reach the end of the t_ctid chain.
1883  * Depending on the snapshot passed, there might be at most one visible
1884  * version of the row, but we don't try to optimize for that.
1885  */
1886  ctid = *tid;
1887  priorXmax = InvalidTransactionId; /* cannot check first XMIN */
1888  for (;;)
1889  {
1890  Buffer buffer;
1891  Page page;
1892  OffsetNumber offnum;
1893  ItemId lp;
1894  HeapTupleData tp;
1895  bool valid;
1896 
1897  /*
1898  * Read, pin, and lock the page.
1899  */
1900  buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(&ctid));
1901  LockBuffer(buffer, BUFFER_LOCK_SHARE);
1902  page = BufferGetPage(buffer);
1903  TestForOldSnapshot(snapshot, relation, page);
1904 
1905  /*
1906  * Check for bogus item number. This is not treated as an error
1907  * condition because it can happen while following a t_ctid link. We
1908  * just assume that the prior tid is OK and return it unchanged.
1909  */
1910  offnum = ItemPointerGetOffsetNumber(&ctid);
1911  if (offnum < FirstOffsetNumber || offnum > PageGetMaxOffsetNumber(page))
1912  {
1913  UnlockReleaseBuffer(buffer);
1914  break;
1915  }
1916  lp = PageGetItemId(page, offnum);
1917  if (!ItemIdIsNormal(lp))
1918  {
1919  UnlockReleaseBuffer(buffer);
1920  break;
1921  }
1922 
1923  /* OK to access the tuple */
1924  tp.t_self = ctid;
1925  tp.t_data = (HeapTupleHeader) PageGetItem(page, lp);
1926  tp.t_len = ItemIdGetLength(lp);
1927  tp.t_tableOid = RelationGetRelid(relation);
1928 
1929  /*
1930  * After following a t_ctid link, we might arrive at an unrelated
1931  * tuple. Check for XMIN match.
1932  */
1933  if (TransactionIdIsValid(priorXmax) &&
1935  {
1936  UnlockReleaseBuffer(buffer);
1937  break;
1938  }
1939 
1940  /*
1941  * Check tuple visibility; if visible, set it as the new result
1942  * candidate.
1943  */
1944  valid = HeapTupleSatisfiesVisibility(&tp, snapshot, buffer);
1945  HeapCheckForSerializableConflictOut(valid, relation, &tp, buffer, snapshot);
1946  if (valid)
1947  *tid = ctid;
1948 
1949  /*
1950  * If there's a valid t_ctid link, follow it, else we're done.
1951  */
1952  if ((tp.t_data->t_infomask & HEAP_XMAX_INVALID) ||
1956  {
1957  UnlockReleaseBuffer(buffer);
1958  break;
1959  }
1960 
1961  ctid = tp.t_data->t_ctid;
1962  priorXmax = HeapTupleHeaderGetUpdateXid(tp.t_data);
1963  UnlockReleaseBuffer(buffer);
1964  } /* end of loop */
1965 }
#define HeapTupleHeaderGetUpdateXid(tup)
Definition: htup_details.h:365
#define ItemPointerIsValid(pointer)
Definition: itemptr.h:82
static void TestForOldSnapshot(Snapshot snapshot, Relation relation, Page page)
Definition: bufmgr.h:278
#define TransactionIdEquals(id1, id2)
Definition: transam.h:43
uint32 TransactionId
Definition: c.h:587
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
bool HeapTupleHeaderIsOnlyLocked(HeapTupleHeader tuple)
#define HeapTupleHeaderIndicatesMovedPartitions(tup)
Definition: htup_details.h:445
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
uint16 OffsetNumber
Definition: off.h:24
HeapTupleHeader t_data
Definition: htup.h:68
#define ItemIdGetLength(itemId)
Definition: itemid.h:59
void UnlockReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3791
#define HEAP_XMAX_INVALID
Definition: htup_details.h:207
ItemPointerData t_ctid
Definition: htup_details.h:160
ItemPointerData t_self
Definition: htup.h:65
uint32 t_len
Definition: htup.h:64
#define InvalidTransactionId
Definition: transam.h:31
Oid t_tableOid
Definition: htup.h:66
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
void LockBuffer(Buffer buffer, int mode)
Definition: bufmgr.c:4007
void HeapCheckForSerializableConflictOut(bool visible, Relation relation, HeapTuple tuple, Buffer buffer, Snapshot snapshot)
Definition: heapam.c:9789
#define Assert(condition)
Definition: c.h:804
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
#define HeapTupleHeaderGetXmin(tup)
Definition: htup_details.h:313
Buffer ReadBuffer(Relation reln, BlockNumber blockNum)
Definition: bufmgr.c:694
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
bool ItemPointerEquals(ItemPointer pointer1, ItemPointer pointer2)
Definition: itemptr.c:29
#define BUFFER_LOCK_SHARE
Definition: bufmgr.h:97
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
#define TransactionIdIsValid(xid)
Definition: transam.h:41
int Buffer
Definition: buf.h:23
#define RelationGetRelid(relation)
Definition: rel.h:477
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
Pointer Page
Definition: bufpage.h:78
bool HeapTupleSatisfiesVisibility(HeapTuple tup, Snapshot snapshot, Buffer buffer)

◆ heap_get_root_tuples()

void heap_get_root_tuples ( Page  page,
OffsetNumber root_offsets 
)

Definition at line 900 of file pruneheap.c.

References Assert, FirstOffsetNumber, HeapTupleHeaderGetUpdateXid, HeapTupleHeaderGetXmin, HeapTupleHeaderIndicatesMovedPartitions, HeapTupleHeaderIsHeapOnly, HeapTupleHeaderIsHotUpdated, InvalidOffsetNumber, InvalidTransactionId, ItemIdGetRedirect, ItemIdIsDead, ItemIdIsNormal, ItemIdIsRedirected, ItemIdIsUsed, ItemPointerGetOffsetNumber, MaxHeapTuplesPerPage, MemSet, OffsetNumberNext, PageGetItem, PageGetItemId, PageGetMaxOffsetNumber, HeapTupleHeaderData::t_ctid, TransactionIdEquals, and TransactionIdIsValid.

Referenced by heapam_index_build_range_scan(), and heapam_index_validate_scan().

901 {
902  OffsetNumber offnum,
903  maxoff;
904 
905  MemSet(root_offsets, InvalidOffsetNumber,
907 
908  maxoff = PageGetMaxOffsetNumber(page);
909  for (offnum = FirstOffsetNumber; offnum <= maxoff; offnum = OffsetNumberNext(offnum))
910  {
911  ItemId lp = PageGetItemId(page, offnum);
912  HeapTupleHeader htup;
913  OffsetNumber nextoffnum;
914  TransactionId priorXmax;
915 
916  /* skip unused and dead items */
917  if (!ItemIdIsUsed(lp) || ItemIdIsDead(lp))
918  continue;
919 
920  if (ItemIdIsNormal(lp))
921  {
922  htup = (HeapTupleHeader) PageGetItem(page, lp);
923 
924  /*
925  * Check if this tuple is part of a HOT-chain rooted at some other
926  * tuple. If so, skip it for now; we'll process it when we find
927  * its root.
928  */
929  if (HeapTupleHeaderIsHeapOnly(htup))
930  continue;
931 
932  /*
933  * This is either a plain tuple or the root of a HOT-chain.
934  * Remember it in the mapping.
935  */
936  root_offsets[offnum - 1] = offnum;
937 
938  /* If it's not the start of a HOT-chain, we're done with it */
939  if (!HeapTupleHeaderIsHotUpdated(htup))
940  continue;
941 
942  /* Set up to scan the HOT-chain */
943  nextoffnum = ItemPointerGetOffsetNumber(&htup->t_ctid);
944  priorXmax = HeapTupleHeaderGetUpdateXid(htup);
945  }
946  else
947  {
948  /* Must be a redirect item. We do not set its root_offsets entry */
950  /* Set up to scan the HOT-chain */
951  nextoffnum = ItemIdGetRedirect(lp);
952  priorXmax = InvalidTransactionId;
953  }
954 
955  /*
956  * Now follow the HOT-chain and collect other tuples in the chain.
957  *
958  * Note: Even though this is a nested loop, the complexity of the
959  * function is O(N) because a tuple in the page should be visited not
960  * more than twice, once in the outer loop and once in HOT-chain
961  * chases.
962  */
963  for (;;)
964  {
965  /* Sanity check */
966  if (nextoffnum < FirstOffsetNumber || nextoffnum > maxoff)
967  break;
968 
969  lp = PageGetItemId(page, nextoffnum);
970 
971  /* Check for broken chains */
972  if (!ItemIdIsNormal(lp))
973  break;
974 
975  htup = (HeapTupleHeader) PageGetItem(page, lp);
976 
977  if (TransactionIdIsValid(priorXmax) &&
978  !TransactionIdEquals(priorXmax, HeapTupleHeaderGetXmin(htup)))
979  break;
980 
981  /* Remember the root line pointer for this item */
982  root_offsets[nextoffnum - 1] = offnum;
983 
984  /* Advance to next chain member, if any */
985  if (!HeapTupleHeaderIsHotUpdated(htup))
986  break;
987 
988  /* HOT implies it can't have moved to different partition */
990 
991  nextoffnum = ItemPointerGetOffsetNumber(&htup->t_ctid);
992  priorXmax = HeapTupleHeaderGetUpdateXid(htup);
993  }
994  }
995 }
#define HeapTupleHeaderGetUpdateXid(tup)
Definition: htup_details.h:365
#define ItemIdIsRedirected(itemId)
Definition: itemid.h:106
#define TransactionIdEquals(id1, id2)
Definition: transam.h:43
uint32 TransactionId
Definition: c.h:587
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
#define ItemIdGetRedirect(itemId)
Definition: itemid.h:78
#define ItemIdIsUsed(itemId)
Definition: itemid.h:92
#define MaxHeapTuplesPerPage
Definition: htup_details.h:573
#define MemSet(start, val, len)
Definition: c.h:1008
#define HeapTupleHeaderIndicatesMovedPartitions(tup)
Definition: htup_details.h:445
#define ItemIdIsDead(itemId)
Definition: itemid.h:113
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
uint16 OffsetNumber
Definition: off.h:24
#define HeapTupleHeaderIsHeapOnly(tup)
Definition: htup_details.h:500
ItemPointerData t_ctid
Definition: htup_details.h:160
#define FirstOffsetNumber
Definition: off.h:27
#define InvalidTransactionId
Definition: transam.h:31
#define HeapTupleHeaderIsHotUpdated(tup)
Definition: htup_details.h:483
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
#define InvalidOffsetNumber
Definition: off.h:26
#define Assert(condition)
Definition: c.h:804
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
#define HeapTupleHeaderGetXmin(tup)
Definition: htup_details.h:313
#define OffsetNumberNext(offsetNumber)
Definition: off.h:52
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
#define TransactionIdIsValid(xid)
Definition: transam.h:41
#define PageGetItem(page, itemId)
Definition: bufpage.h:340

◆ heap_getnext()

HeapTuple heap_getnext ( TableScanDesc  scan,
ScanDirection  direction 
)

Definition at line 1340 of file heapam.c.

References bsysscan, CheckXidAlive, elog, ereport, errcode(), errmsg_internal(), ERROR, GetHeapamTableAmRoutine(), heapgettup(), heapgettup_pagemode(), pgstat_count_heap_getnext, RelationData::rd_tableam, HeapScanDescData::rs_base, HeapScanDescData::rs_ctup, TableScanDescData::rs_flags, TableScanDescData::rs_key, TableScanDescData::rs_nkeys, TableScanDescData::rs_rd, SO_ALLOW_PAGEMODE, HeapTupleData::t_data, TransactionIdIsValid, and unlikely.

Referenced by AlterTableMoveAll(), AlterTableSpaceOptions(), check_db_file_conflict(), createdb(), do_autovacuum(), DropSetting(), DropTableSpace(), find_typed_table_dependencies(), get_all_vacuum_rels(), get_database_list(), get_subscription_list(), get_tables_to_cluster(), get_tablespace_name(), get_tablespace_oid(), GetAllTablesPublicationRelations(), getRelationsInNamespace(), heapam_index_build_range_scan(), heapam_index_validate_scan(), index_update_stats(), objectsInSchemaToOids(), pgrowlocks(), pgstat_collect_oids(), pgstat_heap(), populate_typ_list(), ReindexMultipleTables(), remove_dbtablespaces(), RemoveSubscriptionRel(), RenameTableSpace(), ThereIsAtLeastOneRole(), and vac_truncate_clog().

1341 {
1342  HeapScanDesc scan = (HeapScanDesc) sscan;
1343 
1344  /*
1345  * This is still widely used directly, without going through table AM, so
1346  * add a safety check. It's possible we should, at a later point,
1347  * downgrade this to an assert. The reason for checking the AM routine,
1348  * rather than the AM oid, is that this allows to write regression tests
1349  * that create another AM reusing the heap handler.
1350  */
1351  if (unlikely(sscan->rs_rd->rd_tableam != GetHeapamTableAmRoutine()))
1352  ereport(ERROR,
1353  (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
1354  errmsg_internal("only heap AM is supported")));
1355 
1356  /*
1357  * We don't expect direct calls to heap_getnext with valid CheckXidAlive
1358  * for catalog or regular tables. See detailed comments in xact.c where
1359  * these variables are declared. Normally we have such a check at tableam
1360  * level API but this is called from many places so we need to ensure it
1361  * here.
1362  */
1364  elog(ERROR, "unexpected heap_getnext call during logical decoding");
1365 
1366  /* Note: no locking manipulations needed */
1367 
1368  if (scan->rs_base.rs_flags & SO_ALLOW_PAGEMODE)
1369  heapgettup_pagemode(scan, direction,
1370  scan->rs_base.rs_nkeys, scan->rs_base.rs_key);
1371  else
1372  heapgettup(scan, direction,
1373  scan->rs_base.rs_nkeys, scan->rs_base.rs_key);
1374 
1375  if (scan->rs_ctup.t_data == NULL)
1376  return NULL;
1377 
1378  /*
1379  * if we get here it means we have a new current scan tuple, so point to
1380  * the proper return buffer and return the tuple.
1381  */
1382 
1384 
1385  return &scan->rs_ctup;
1386 }
TableScanDescData rs_base
Definition: heapam.h:49
int errcode(int sqlerrcode)
Definition: elog.c:698
uint32 rs_flags
Definition: relscan.h:47
struct HeapScanDescData * HeapScanDesc
Definition: heapam.h:79
HeapTupleData rs_ctup
Definition: heapam.h:66
HeapTupleHeader t_data
Definition: htup.h:68
#define ERROR
Definition: elog.h:46
bool bsysscan
Definition: xact.c:96
struct ScanKeyData * rs_key
Definition: relscan.h:37
TransactionId CheckXidAlive
Definition: xact.c:95
static void heapgettup(HeapScanDesc scan, ScanDirection dir, int nkeys, ScanKey key)
Definition: heapam.c:506
#define ereport(elevel,...)
Definition: elog.h:157
int errmsg_internal(const char *fmt,...)
Definition: elog.c:996
Relation rs_rd
Definition: relscan.h:34
#define elog(elevel,...)
Definition: elog.h:232
#define unlikely(x)
Definition: c.h:273
#define pgstat_count_heap_getnext(rel)
Definition: pgstat.h:1055
#define TransactionIdIsValid(xid)
Definition: transam.h:41
static void heapgettup_pagemode(HeapScanDesc scan, ScanDirection dir, int nkeys, ScanKey key)
Definition: heapam.c:844
const TableAmRoutine * GetHeapamTableAmRoutine(void)

◆ heap_getnextslot()

bool heap_getnextslot ( TableScanDesc  sscan,
ScanDirection  direction,
struct TupleTableSlot slot 
)

Definition at line 1389 of file heapam.c.

References ExecClearTuple(), ExecStoreBufferHeapTuple(), heapgettup(), heapgettup_pagemode(), pgstat_count_heap_getnext, HeapScanDescData::rs_base, HeapScanDescData::rs_cbuf, HeapScanDescData::rs_ctup, TableScanDescData::rs_flags, TableScanDescData::rs_key, TableScanDescData::rs_nkeys, TableScanDescData::rs_rd, SO_ALLOW_PAGEMODE, and HeapTupleData::t_data.

Referenced by SampleHeapTupleVisible().

1390 {
1391  HeapScanDesc scan = (HeapScanDesc) sscan;
1392 
1393  /* Note: no locking manipulations needed */
1394 
1395  if (sscan->rs_flags & SO_ALLOW_PAGEMODE)
1396  heapgettup_pagemode(scan, direction, sscan->rs_nkeys, sscan->rs_key);
1397  else
1398  heapgettup(scan, direction, sscan->rs_nkeys, sscan->rs_key);
1399 
1400  if (scan->rs_ctup.t_data == NULL)
1401  {
1402  ExecClearTuple(slot);
1403  return false;
1404  }
1405 
1406  /*
1407  * if we get here it means we have a new current scan tuple, so point to
1408  * the proper return buffer and return the tuple.
1409  */
1410 
1412 
1413  ExecStoreBufferHeapTuple(&scan->rs_ctup, slot,
1414  scan->rs_cbuf);
1415  return true;
1416 }
static TupleTableSlot * ExecClearTuple(TupleTableSlot *slot)
Definition: tuptable.h:425
TableScanDescData rs_base
Definition: heapam.h:49
uint32 rs_flags
Definition: relscan.h:47
struct HeapScanDescData * HeapScanDesc
Definition: heapam.h:79
HeapTupleData rs_ctup
Definition: heapam.h:66
HeapTupleHeader t_data
Definition: htup.h:68
struct ScanKeyData * rs_key
Definition: relscan.h:37
static void heapgettup(HeapScanDesc scan, ScanDirection dir, int nkeys, ScanKey key)
Definition: heapam.c:506
TupleTableSlot * ExecStoreBufferHeapTuple(HeapTuple tuple, TupleTableSlot *slot, Buffer buffer)
Definition: execTuples.c:1392
Buffer rs_cbuf
Definition: heapam.h:60
Relation rs_rd
Definition: relscan.h:34
#define pgstat_count_heap_getnext(rel)
Definition: pgstat.h:1055
static void heapgettup_pagemode(HeapScanDesc scan, ScanDirection dir, int nkeys, ScanKey key)
Definition: heapam.c:844

◆ heap_getnextslot_tidrange()

bool heap_getnextslot_tidrange ( TableScanDesc  sscan,
ScanDirection  direction,
TupleTableSlot slot 
)

Definition at line 1492 of file heapam.c.

References ExecClearTuple(), ExecStoreBufferHeapTuple(), heapgettup(), heapgettup_pagemode(), ItemPointerCompare(), pgstat_count_heap_getnext, HeapScanDescData::rs_base, HeapScanDescData::rs_cbuf, HeapScanDescData::rs_ctup, TableScanDescData::rs_flags, TableScanDescData::rs_key, TableScanDescData::rs_maxtid, TableScanDescData::rs_mintid, TableScanDescData::rs_nkeys, TableScanDescData::rs_rd, ScanDirectionIsBackward, ScanDirectionIsForward, SO_ALLOW_PAGEMODE, HeapTupleData::t_data, and HeapTupleData::t_self.

Referenced by SampleHeapTupleVisible().

1494 {
1495  HeapScanDesc scan = (HeapScanDesc) sscan;
1496  ItemPointer mintid = &sscan->rs_mintid;
1497  ItemPointer maxtid = &sscan->rs_maxtid;
1498 
1499  /* Note: no locking manipulations needed */
1500  for (;;)
1501  {
1502  if (sscan->rs_flags & SO_ALLOW_PAGEMODE)
1503  heapgettup_pagemode(scan, direction, sscan->rs_nkeys, sscan->rs_key);
1504  else
1505  heapgettup(scan, direction, sscan->rs_nkeys, sscan->rs_key);
1506 
1507  if (scan->rs_ctup.t_data == NULL)
1508  {
1509  ExecClearTuple(slot);
1510  return false;
1511  }
1512 
1513  /*
1514  * heap_set_tidrange will have used heap_setscanlimits to limit the
1515  * range of pages we scan to only ones that can contain the TID range
1516  * we're scanning for. Here we must filter out any tuples from these
1517  * pages that are outwith that range.
1518  */
1519  if (ItemPointerCompare(&scan->rs_ctup.t_self, mintid) < 0)
1520  {
1521  ExecClearTuple(slot);
1522 
1523  /*
1524  * When scanning backwards, the TIDs will be in descending order.
1525  * Future tuples in this direction will be lower still, so we can
1526  * just return false to indicate there will be no more tuples.
1527  */
1528  if (ScanDirectionIsBackward(direction))
1529  return false;
1530 
1531  continue;
1532  }
1533 
1534  /*
1535  * Likewise for the final page, we must filter out TIDs greater than
1536  * maxtid.
1537  */
1538  if (ItemPointerCompare(&scan->rs_ctup.t_self, maxtid) > 0)
1539  {
1540  ExecClearTuple(slot);
1541 
1542  /*
1543  * When scanning forward, the TIDs will be in ascending order.
1544  * Future tuples in this direction will be higher still, so we can
1545  * just return false to indicate there will be no more tuples.
1546  */
1547  if (ScanDirectionIsForward(direction))
1548  return false;
1549  continue;
1550  }
1551 
1552  break;
1553  }
1554 
1555  /*
1556  * if we get here it means we have a new current scan tuple, so point to
1557  * the proper return buffer and return the tuple.
1558  */
1560 
1561  ExecStoreBufferHeapTuple(&scan->rs_ctup, slot, scan->rs_cbuf);
1562  return true;
1563 }
int32 ItemPointerCompare(ItemPointer arg1, ItemPointer arg2)
Definition: itemptr.c:52
ItemPointerData rs_mintid
Definition: relscan.h:40
static TupleTableSlot * ExecClearTuple(TupleTableSlot *slot)
Definition: tuptable.h:425
#define ScanDirectionIsForward(direction)
Definition: sdir.h:55
TableScanDescData rs_base
Definition: heapam.h:49
uint32 rs_flags
Definition: relscan.h:47
struct HeapScanDescData * HeapScanDesc
Definition: heapam.h:79
HeapTupleData rs_ctup
Definition: heapam.h:66
HeapTupleHeader t_data
Definition: htup.h:68
#define ScanDirectionIsBackward(direction)
Definition: sdir.h:41
ItemPointerData t_self
Definition: htup.h:65
struct ScanKeyData * rs_key
Definition: relscan.h:37
ItemPointerData rs_maxtid
Definition: relscan.h:41
static void heapgettup(HeapScanDesc scan, ScanDirection dir, int nkeys, ScanKey key)
Definition: heapam.c:506
TupleTableSlot * ExecStoreBufferHeapTuple(HeapTuple tuple, TupleTableSlot *slot, Buffer buffer)
Definition: execTuples.c:1392
Buffer rs_cbuf
Definition: heapam.h:60
Relation rs_rd
Definition: relscan.h:34
#define pgstat_count_heap_getnext(rel)
Definition: pgstat.h:1055
static void heapgettup_pagemode(HeapScanDesc scan, ScanDirection dir, int nkeys, ScanKey key)
Definition: heapam.c:844

◆ heap_hot_search_buffer()

bool heap_hot_search_buffer ( ItemPointer  tid,
Relation  relation,
Buffer  buffer,
Snapshot  snapshot,
HeapTuple  heapTuple,
bool all_dead,
bool  first_call 
)

Definition at line 1710 of file heapam.c.

References Assert, BufferGetBlockNumber(), BufferGetPage, GlobalVisTestFor(), HeapCheckForSerializableConflictOut(), HeapTupleHeaderGetUpdateXid, HeapTupleHeaderGetXmin, HeapTupleIsHeapOnly, HeapTupleIsHotUpdated, HeapTupleIsSurelyDead(), HeapTupleSatisfiesVisibility(), InvalidTransactionId, ItemIdGetLength, ItemIdGetRedirect, ItemIdIsNormal, ItemIdIsRedirected, ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, ItemPointerSet, ItemPointerSetOffsetNumber, PageGetItem, PageGetItemId, PageGetMaxOffsetNumber, PredicateLockTID(), RecentXmin, RelationGetRelid, skip, HeapTupleHeaderData::t_ctid, HeapTupleData::t_data, HeapTupleData::t_len, HeapTupleData::t_self, HeapTupleData::t_tableOid, TransactionIdEquals, and TransactionIdIsValid.

Referenced by heap_index_delete_tuples(), heapam_index_fetch_tuple(), and heapam_scan_bitmap_next_block().

1713 {
1714  Page dp = (Page) BufferGetPage(buffer);
1715  TransactionId prev_xmax = InvalidTransactionId;
1716  BlockNumber blkno;
1717  OffsetNumber offnum;
1718  bool at_chain_start;
1719  bool valid;
1720  bool skip;
1721  GlobalVisState *vistest = NULL;
1722 
1723  /* If this is not the first call, previous call returned a (live!) tuple */
1724  if (all_dead)
1725  *all_dead = first_call;
1726 
1727  blkno = ItemPointerGetBlockNumber(tid);
1728  offnum = ItemPointerGetOffsetNumber(tid);
1729  at_chain_start = first_call;
1730  skip = !first_call;
1731 
1732  /* XXX: we should assert that a snapshot is pushed or registered */
1734  Assert(BufferGetBlockNumber(buffer) == blkno);
1735 
1736  /* Scan through possible multiple members of HOT-chain */
1737  for (;;)
1738  {
1739  ItemId lp;
1740 
1741  /* check for bogus TID */
1742  if (offnum < FirstOffsetNumber || offnum > PageGetMaxOffsetNumber(dp))
1743  break;
1744 
1745  lp = PageGetItemId(dp, offnum);
1746 
1747  /* check for unused, dead, or redirected items */
1748  if (!ItemIdIsNormal(lp))
1749  {
1750  /* We should only see a redirect at start of chain */
1751  if (ItemIdIsRedirected(lp) && at_chain_start)
1752  {
1753  /* Follow the redirect */
1754  offnum = ItemIdGetRedirect(lp);
1755  at_chain_start = false;
1756  continue;
1757  }
1758  /* else must be end of chain */
1759  break;
1760  }
1761 
1762  /*
1763  * Update heapTuple to point to the element of the HOT chain we're
1764  * currently investigating. Having t_self set correctly is important
1765  * because the SSI checks and the *Satisfies routine for historical
1766  * MVCC snapshots need the correct tid to decide about the visibility.
1767  */
1768  heapTuple->t_data = (HeapTupleHeader) PageGetItem(dp, lp);
1769  heapTuple->t_len = ItemIdGetLength(lp);
1770  heapTuple->t_tableOid = RelationGetRelid(relation);
1771  ItemPointerSet(&heapTuple->t_self, blkno, offnum);
1772 
1773  /*
1774  * Shouldn't see a HEAP_ONLY tuple at chain start.
1775  */
1776  if (at_chain_start && HeapTupleIsHeapOnly(heapTuple))
1777  break;
1778 
1779  /*
1780  * The xmin should match the previous xmax value, else chain is
1781  * broken.
1782  */
1783  if (TransactionIdIsValid(prev_xmax) &&
1784  !TransactionIdEquals(prev_xmax,
1785  HeapTupleHeaderGetXmin(heapTuple->t_data)))
1786  break;
1787 
1788  /*
1789  * When first_call is true (and thus, skip is initially false) we'll
1790  * return the first tuple we find. But on later passes, heapTuple
1791  * will initially be pointing to the tuple we returned last time.
1792  * Returning it again would be incorrect (and would loop forever), so
1793  * we skip it and return the next match we find.
1794  */
1795  if (!skip)
1796  {
1797  /* If it's visible per the snapshot, we must return it */
1798  valid = HeapTupleSatisfiesVisibility(heapTuple, snapshot, buffer);
1799  HeapCheckForSerializableConflictOut(valid, relation, heapTuple,
1800  buffer, snapshot);
1801 
1802  if (valid)
1803  {
1804  ItemPointerSetOffsetNumber(tid, offnum);
1805  PredicateLockTID(relation, &heapTuple->t_self, snapshot,
1806  HeapTupleHeaderGetXmin(heapTuple->t_data));
1807  if (all_dead)
1808  *all_dead = false;
1809  return true;
1810  }
1811  }
1812  skip = false;
1813 
1814  /*
1815  * If we can't see it, maybe no one else can either. At caller
1816  * request, check whether all chain members are dead to all
1817  * transactions.
1818  *
1819  * Note: if you change the criterion here for what is "dead", fix the
1820  * planner's get_actual_variable_range() function to match.
1821  */
1822  if (all_dead && *all_dead)
1823  {
1824  if (!vistest)
1825  vistest = GlobalVisTestFor(relation);
1826 
1827  if (!HeapTupleIsSurelyDead(heapTuple, vistest))
1828  *all_dead = false;
1829  }
1830 
1831  /*
1832  * Check to see if HOT chain continues past this tuple; if so fetch
1833  * the next offnum and loop around.
1834  */
1835  if (HeapTupleIsHotUpdated(heapTuple))
1836  {
1838  blkno);
1839  offnum = ItemPointerGetOffsetNumber(&heapTuple->t_data->t_ctid);
1840  at_chain_start = false;
1841  prev_xmax = HeapTupleHeaderGetUpdateXid(heapTuple->t_data);
1842  }
1843  else
1844  break; /* end of chain */
1845  }
1846 
1847  return false;
1848 }
#define HeapTupleHeaderGetUpdateXid(tup)
Definition: htup_details.h:365
#define ItemIdIsRedirected(itemId)
Definition: itemid.h:106
#define TransactionIdEquals(id1, id2)
Definition: transam.h:43
uint32 TransactionId
Definition: c.h:587
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
#define ItemIdGetRedirect(itemId)
Definition: itemid.h:78
static const struct exclude_list_item skip[]
Definition: pg_checksums.c:116
TransactionId RecentXmin
Definition: snapmgr.c:113
uint32 BlockNumber
Definition: block.h:31
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
uint16 OffsetNumber
Definition: off.h:24
HeapTupleHeader t_data
Definition: htup.h:68
#define HeapTupleIsHotUpdated(tuple)
Definition: htup_details.h:675
GlobalVisState * GlobalVisTestFor(Relation rel)
Definition: procarray.c:4029
#define ItemIdGetLength(itemId)
Definition: itemid.h:59
ItemPointerData t_ctid
Definition: htup_details.h:160
ItemPointerData t_self
Definition: htup.h:65
uint32 t_len
Definition: htup.h:64
#define InvalidTransactionId
Definition: transam.h:31
Oid t_tableOid
Definition: htup.h:66
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
void PredicateLockTID(Relation relation, ItemPointer tid, Snapshot snapshot, TransactionId tuple_xid)
Definition: predicate.c:2614
void HeapCheckForSerializableConflictOut(bool visible, Relation relation, HeapTuple tuple, Buffer buffer, Snapshot snapshot)
Definition: heapam.c:9789
#define HeapTupleIsHeapOnly(tuple)
Definition: htup_details.h:684
#define Assert(condition)
Definition: c.h:804
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
#define HeapTupleHeaderGetXmin(tup)
Definition: htup_details.h:313
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
#define ItemPointerSetOffsetNumber(pointer, offsetNumber)
Definition: itemptr.h:148
BlockNumber BufferGetBlockNumber(Buffer buffer)
Definition: bufmgr.c:2748
bool HeapTupleIsSurelyDead(HeapTuple htup, GlobalVisState *vistest)
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
#define TransactionIdIsValid(xid)
Definition: transam.h:41
#define RelationGetRelid(relation)
Definition: rel.h:477
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
Pointer Page
Definition: bufpage.h:78
#define ItemPointerSet(pointer, blockNumber, offNum)
Definition: itemptr.h:127
bool HeapTupleSatisfiesVisibility(HeapTuple tup, Snapshot snapshot, Buffer buffer)

◆ heap_index_delete_tuples()

TransactionId heap_index_delete_tuples ( Relation  rel,
TM_IndexDeleteOp delstate 
)

Definition at line 7269 of file heapam.c.

References Assert, TM_IndexDeleteOp::bottomup, BOTTOMUP_MAX_NBLOCKS, bottomup_sort_and_shrink(), TM_IndexDeleteOp::bottomupfreespace, buf, BUFFER_LOCK_SHARE, BufferGetPage, BufferIsValid, TM_IndexDeleteOp::deltids, TM_IndexStatus::freespace, get_tablespace_maintenance_io_concurrency(), GlobalVisTestFor(), heap_hot_search_buffer(), HeapTupleHeaderAdvanceLatestRemovedXid(), HeapTupleHeaderGetUpdateXid, HeapTupleHeaderGetXmin, HeapTupleHeaderIsHotUpdated, i, TM_IndexDelete::id, index_delete_sort(), InitNonVacuumableSnapshot, InvalidBlockNumber, InvalidBuffer, InvalidOffsetNumber, InvalidTransactionId, IsCatalogRelation(), ItemIdGetRedirect, ItemIdIsNormal, ItemIdIsRedirected, ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, TM_IndexStatus::knowndeletable, LockBuffer(), maintenance_io_concurrency, Min, TM_IndexDeleteOp::ndeltids, PageGetItem, PageGetItemId, PageGetMaxOffsetNumber, TM_IndexStatus::promising, RelationData::rd_rel, ReadBuffer(), TM_IndexDeleteOp::status, HeapTupleHeaderData::t_ctid, TM_IndexDelete::tid, TransactionIdEquals, TransactionIdIsValid, and UnlockReleaseBuffer().

Referenced by SampleHeapTupleVisible().

7270 {
7271  /* Initial assumption is that earlier pruning took care of conflict */
7272  TransactionId latestRemovedXid = InvalidTransactionId;
7275  Page page = NULL;
7277  TransactionId priorXmax;
7278 #ifdef USE_PREFETCH
7279  IndexDeletePrefetchState prefetch_state;
7280  int prefetch_distance;
7281 #endif
7282  SnapshotData SnapshotNonVacuumable;
7283  int finalndeltids = 0,
7284  nblocksaccessed = 0;
7285 
7286  /* State that's only used in bottom-up index deletion case */
7287  int nblocksfavorable = 0;
7288  int curtargetfreespace = delstate->bottomupfreespace,
7289  lastfreespace = 0,
7290  actualfreespace = 0;
7291  bool bottomup_final_block = false;
7292 
7293  InitNonVacuumableSnapshot(SnapshotNonVacuumable, GlobalVisTestFor(rel));
7294 
7295  /* Sort caller's deltids array by TID for further processing */
7296  index_delete_sort(delstate);
7297 
7298  /*
7299  * Bottom-up case: resort deltids array in an order attuned to where the
7300  * greatest number of promising TIDs are to be found, and determine how
7301  * many blocks from the start of sorted array should be considered
7302  * favorable. This will also shrink the deltids array in order to
7303  * eliminate completely unfavorable blocks up front.
7304  */
7305  if (delstate->bottomup)
7306  nblocksfavorable = bottomup_sort_and_shrink(delstate);
7307 
7308 #ifdef USE_PREFETCH
7309  /* Initialize prefetch state. */
7310  prefetch_state.cur_hblkno = InvalidBlockNumber;
7311  prefetch_state.next_item = 0;
7312  prefetch_state.ndeltids = delstate->ndeltids;
7313  prefetch_state.deltids = delstate->deltids;
7314 
7315  /*
7316  * Determine the prefetch distance that we will attempt to maintain.
7317  *
7318  * Since the caller holds a buffer lock somewhere in rel, we'd better make
7319  * sure that isn't a catalog relation before we call code that does
7320  * syscache lookups, to avoid risk of deadlock.
7321  */
7322  if (IsCatalogRelation(rel))
7323  prefetch_distance = maintenance_io_concurrency;
7324  else
7325  prefetch_distance =
7327 
7328  /* Cap initial prefetch distance for bottom-up deletion caller */
7329  if (delstate->bottomup)
7330  {
7331  Assert(nblocksfavorable >= 1);
7332  Assert(nblocksfavorable <= BOTTOMUP_MAX_NBLOCKS);
7333  prefetch_distance = Min(prefetch_distance, nblocksfavorable);
7334  }
7335 
7336  /* Start prefetching. */
7337  index_delete_prefetch_buffer(rel, &prefetch_state, prefetch_distance);
7338 #endif
7339 
7340  /* Iterate over deltids, determine which to delete, check their horizon */
7341  Assert(delstate->ndeltids > 0);
7342  for (int i = 0; i < delstate->ndeltids; i++)
7343  {
7344  TM_IndexDelete *ideltid = &delstate->deltids[i];
7345  TM_IndexStatus *istatus = delstate->status + ideltid->id;
7346  ItemPointer htid = &ideltid->tid;
7347  OffsetNumber offnum;
7348 
7349  /*
7350  * Read buffer, and perform required extra steps each time a new block
7351  * is encountered. Avoid refetching if it's the same block as the one
7352  * from the last htid.
7353  */
7354  if (blkno == InvalidBlockNumber ||
7355  ItemPointerGetBlockNumber(htid) != blkno)
7356  {
7357  /*
7358  * Consider giving up early for bottom-up index deletion caller
7359  * first. (Only prefetch next-next block afterwards, when it
7360  * becomes clear that we're at least going to access the next
7361  * block in line.)
7362  *
7363  * Sometimes the first block frees so much space for bottom-up
7364  * caller that the deletion process can end without accessing any
7365  * more blocks. It is usually necessary to access 2 or 3 blocks
7366  * per bottom-up deletion operation, though.
7367  */
7368  if (delstate->bottomup)
7369  {
7370  /*
7371  * We often allow caller to delete a few additional items
7372  * whose entries we reached after the point that space target
7373  * from caller was satisfied. The cost of accessing the page
7374  * was already paid at that point, so it made sense to finish
7375  * it off. When that happened, we finalize everything here
7376  * (by finishing off the whole bottom-up deletion operation
7377  * without needlessly paying the cost of accessing any more
7378  * blocks).
7379  */
7380  if (bottomup_final_block)
7381  break;
7382 
7383  /*
7384  * Give up when we didn't enable our caller to free any
7385  * additional space as a result of processing the page that we
7386  * just finished up with. This rule is the main way in which
7387  * we keep the cost of bottom-up deletion under control.
7388  */
7389  if (nblocksaccessed >= 1 && actualfreespace == lastfreespace)
7390  break;
7391  lastfreespace = actualfreespace; /* for next time */
7392 
7393  /*
7394  * Deletion operation (which is bottom-up) will definitely
7395  * access the next block in line. Prepare for that now.
7396  *
7397  * Decay target free space so that we don't hang on for too
7398  * long with a marginal case. (Space target is only truly
7399  * helpful when it allows us to recognize that we don't need
7400  * to access more than 1 or 2 blocks to satisfy caller due to
7401  * agreeable workload characteristics.)
7402  *
7403  * We are a bit more patient when we encounter contiguous
7404  * blocks, though: these are treated as favorable blocks. The
7405  * decay process is only applied when the next block in line
7406  * is not a favorable/contiguous block. This is not an
7407  * exception to the general rule; we still insist on finding
7408  * at least one deletable item per block accessed. See
7409  * bottomup_nblocksfavorable() for full details of the theory
7410  * behind favorable blocks and heap block locality in general.
7411  *
7412  * Note: The first block in line is always treated as a
7413  * favorable block, so the earliest possible point that the
7414  * decay can be applied is just before we access the second
7415  * block in line. The Assert() verifies this for us.
7416  */
7417  Assert(nblocksaccessed > 0 || nblocksfavorable > 0);
7418  if (nblocksfavorable > 0)
7419  nblocksfavorable--;
7420  else
7421  curtargetfreespace /= 2;
7422  }
7423 
7424  /* release old buffer */
7425  if (BufferIsValid(buf))
7426  UnlockReleaseBuffer(buf);
7427 
7428  blkno = ItemPointerGetBlockNumber(htid);
7429  buf = ReadBuffer(rel, blkno);
7430  nblocksaccessed++;
7431  Assert(!delstate->bottomup ||
7432  nblocksaccessed <= BOTTOMUP_MAX_NBLOCKS);
7433 
7434 #ifdef USE_PREFETCH
7435 
7436  /*
7437  * To maintain the prefetch distance, prefetch one more page for
7438  * each page we read.
7439  */
7440  index_delete_prefetch_buffer(rel, &prefetch_state, 1);
7441 #endif
7442 
7444 
7445  page = BufferGetPage(buf);
7446  maxoff = PageGetMaxOffsetNumber(page);
7447  }
7448 
7449  if (istatus->knowndeletable)
7450  Assert(!delstate->bottomup && !istatus->promising);
7451  else
7452  {
7453  ItemPointerData tmp = *htid;
7454  HeapTupleData heapTuple;
7455 
7456  /* Are any tuples from this HOT chain non-vacuumable? */
7457  if (heap_hot_search_buffer(&tmp, rel, buf, &SnapshotNonVacuumable,
7458  &heapTuple, NULL, true))
7459  continue; /* can't delete entry */
7460 
7461  /* Caller will delete, since whole HOT chain is vacuumable */
7462  istatus->knowndeletable = true;
7463 
7464  /* Maintain index free space info for bottom-up deletion case */
7465  if (delstate->bottomup)
7466  {
7467  Assert(istatus->freespace > 0);
7468  actualfreespace += istatus->freespace;
7469  if (actualfreespace >= curtargetfreespace)
7470  bottomup_final_block = true;
7471  }
7472  }
7473 
7474  /*
7475  * Maintain latestRemovedXid value for deletion operation as a whole
7476  * by advancing current value using heap tuple headers. This is
7477  * loosely based on the logic for pruning a HOT chain.
7478  */
7479  offnum = ItemPointerGetOffsetNumber(htid);
7480  priorXmax = InvalidTransactionId; /* cannot check first XMIN */
7481  for (;;)
7482  {
7483  ItemId lp;
7484  HeapTupleHeader htup;
7485 
7486  /* Some sanity checks */
7487  if (offnum < FirstOffsetNumber || offnum > maxoff)
7488  {
7489  Assert(false);
7490  break;
7491  }
7492 
7493  lp = PageGetItemId(page, offnum);
7494  if (ItemIdIsRedirected(lp))
7495  {
7496  offnum = ItemIdGetRedirect(lp);
7497  continue;
7498  }
7499 
7500  /*
7501  * We'll often encounter LP_DEAD line pointers (especially with an
7502  * entry marked knowndeletable by our caller up front). No heap
7503  * tuple headers get examined for an htid that leads us to an
7504  * LP_DEAD item. This is okay because the earlier pruning
7505  * operation that made the line pointer LP_DEAD in the first place
7506  * must have considered the original tuple header as part of
7507  * generating its own latestRemovedXid value.
7508  *
7509  * Relying on XLOG_HEAP2_PRUNE records like this is the same
7510  * strategy that index vacuuming uses in all cases. Index VACUUM
7511  * WAL records don't even have a latestRemovedXid field of their
7512  * own for this reason.
7513  */
7514  if (!ItemIdIsNormal(lp))
7515  break;
7516 
7517  htup = (HeapTupleHeader) PageGetItem(page, lp);
7518 
7519  /*
7520  * Check the tuple XMIN against prior XMAX, if any
7521  */
7522  if (TransactionIdIsValid(priorXmax) &&
7523  !TransactionIdEquals(HeapTupleHeaderGetXmin(htup), priorXmax))
7524  break;
7525 
7526  HeapTupleHeaderAdvanceLatestRemovedXid(htup, &latestRemovedXid);
7527 
7528  /*
7529  * If the tuple is not HOT-updated, then we are at the end of this
7530  * HOT-chain. No need to visit later tuples from the same update
7531  * chain (they get their own index entries) -- just move on to
7532  * next htid from index AM caller.
7533  */
7534  if (!HeapTupleHeaderIsHotUpdated(htup))
7535  break;
7536 
7537  /* Advance to next HOT chain member */
7538  Assert(ItemPointerGetBlockNumber(&htup->t_ctid) == blkno);
7539  offnum = ItemPointerGetOffsetNumber(&htup->t_ctid);
7540  priorXmax = HeapTupleHeaderGetUpdateXid(htup);
7541  }
7542 
7543  /* Enable further/final shrinking of deltids for caller */
7544  finalndeltids = i + 1;
7545  }
7546 
7547  UnlockReleaseBuffer(buf);
7548 
7549  /*
7550  * Shrink deltids array to exclude non-deletable entries at the end. This
7551  * is not just a minor optimization. Final deltids array size might be
7552  * zero for a bottom-up caller. Index AM is explicitly allowed to rely on
7553  * ndeltids being zero in all cases with zero total deletable entries.
7554  */
7555  Assert(finalndeltids > 0 || delstate->bottomup);
7556  delstate->ndeltids = finalndeltids;
7557 
7558  return latestRemovedXid;
7559 }
#define HeapTupleHeaderGetUpdateXid(tup)
Definition: htup_details.h:365
void HeapTupleHeaderAdvanceLatestRemovedXid(HeapTupleHeader tuple, TransactionId *latestRemovedXid)
Definition: heapam.c:7179
TM_IndexDelete * deltids
Definition: tableam.h:228
bool IsCatalogRelation(Relation relation)
Definition: catalog.c:104
int maintenance_io_concurrency
Definition: bufmgr.c:150
#define ItemIdIsRedirected(itemId)
Definition: itemid.h:106
#define TransactionIdEquals(id1, id2)
Definition: transam.h:43
uint32 TransactionId
Definition: c.h:587
static int bottomup_sort_and_shrink(TM_IndexDeleteOp *delstate)
Definition: heapam.c:7814
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
#define ItemIdGetRedirect(itemId)
Definition: itemid.h:78
#define Min(x, y)
Definition: c.h:986
#define InvalidBuffer
Definition: buf.h:25
bool knowndeletable
Definition: tableam.h:196
#define InitNonVacuumableSnapshot(snapshotdata, vistestp)
Definition: snapmgr.h:82
uint32 BlockNumber
Definition: block.h:31
Form_pg_class rd_rel
Definition: rel.h:109
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
uint16 OffsetNumber
Definition: off.h:24
bool heap_hot_search_buffer(ItemPointer tid, Relation relation, Buffer buffer, Snapshot snapshot, HeapTuple heapTuple, bool *all_dead, bool first_call)
Definition: heapam.c:1710
GlobalVisState * GlobalVisTestFor(Relation rel)
Definition: procarray.c:4029
int get_tablespace_maintenance_io_concurrency(Oid spcid)
Definition: spccache.c:228
void UnlockReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3791
ItemPointerData t_ctid
Definition: htup_details.h:160
static char * buf
Definition: pg_test_fsync.c:68
#define BOTTOMUP_MAX_NBLOCKS
Definition: heapam.c:185
#define InvalidTransactionId
Definition: transam.h:31
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
bool promising
Definition: tableam.h:199
#define HeapTupleHeaderIsHotUpdated(tup)
Definition: htup_details.h:483
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
TM_IndexStatus * status
Definition: tableam.h:229
void LockBuffer(Buffer buffer, int mode)
Definition: bufmgr.c:4007
ItemPointerData tid
Definition: tableam.h:189
#define InvalidOffsetNumber
Definition: off.h:26
static void index_delete_sort(TM_IndexDeleteOp *delstate)
Definition: heapam.c:7601
#define Assert(condition)
Definition: c.h:804
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
#define HeapTupleHeaderGetXmin(tup)
Definition: htup_details.h:313
Buffer ReadBuffer(Relation reln, BlockNumber blockNum)
Definition: bufmgr.c:694
#define InvalidBlockNumber
Definition: block.h:33
#define BufferIsValid(bufnum)
Definition: bufmgr.h:123
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
int16 freespace
Definition: tableam.h:200
int i
#define BUFFER_LOCK_SHARE
Definition: bufmgr.h:97
int bottomupfreespace
Definition: tableam.h:224
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
#define TransactionIdIsValid(xid)
Definition: transam.h:41
int Buffer
Definition: buf.h:23
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
Pointer Page
Definition: bufpage.h:78

◆ heap_inplace_update()

void heap_inplace_update ( Relation  relation,
HeapTuple  tuple 
)

Definition at line 5991 of file heapam.c.

References BUFFER_LOCK_EXCLUSIVE, BufferGetPage, CacheInvalidateHeapTuple(), elog, END_CRIT_SECTION, ereport, errcode(), errmsg(), ERROR, IsBootstrapProcessingMode, IsInParallelMode(), ItemIdGetLength, ItemIdIsNormal, ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, LockBuffer(), MarkBufferDirty(), xl_heap_inplace::offnum, PageGetItem, PageGetItemId, PageGetMaxOffsetNumber, PageSetLSN, ReadBuffer(), REGBUF_STANDARD, RelationNeedsWAL, SizeOfHeapInplace, START_CRIT_SECTION, HeapTupleData::t_data, HeapTupleHeaderData::t_hoff, HeapTupleData::t_len, HeapTupleData::t_self, UnlockReleaseBuffer(), XLOG_HEAP_INPLACE, XLogBeginInsert(), XLogInsert(), XLogRegisterBufData(), XLogRegisterBuffer(), and XLogRegisterData().

Referenced by create_toast_table(), index_update_stats(), vac_update_datfrozenxid(), and vac_update_relstats().

5992 {
5993  Buffer buffer;
5994  Page page;
5995  OffsetNumber offnum;
5996  ItemId lp = NULL;
5997  HeapTupleHeader htup;
5998  uint32 oldlen;
5999  uint32 newlen;
6000 
6001  /*
6002  * For now, we don't allow parallel updates. Unlike a regular update,
6003  * this should never create a combo CID, so it might be possible to relax
6004  * this restriction, but not without more thought and testing. It's not
6005  * clear that it would be useful, anyway.
6006  */
6007  if (IsInParallelMode())
6008  ereport(ERROR,
6009  (errcode(ERRCODE_INVALID_TRANSACTION_STATE),
6010  errmsg("cannot update tuples during a parallel operation")));
6011 
6012  buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(&(tuple->t_self)));
6014  page = (Page) BufferGetPage(buffer);
6015 
6016  offnum = ItemPointerGetOffsetNumber(&(tuple->t_self));
6017  if (PageGetMaxOffsetNumber(page) >= offnum)
6018  lp = PageGetItemId(page, offnum);
6019 
6020  if (PageGetMaxOffsetNumber(page) < offnum || !ItemIdIsNormal(lp))
6021  elog(ERROR, "invalid lp");
6022 
6023  htup = (HeapTupleHeader) PageGetItem(page, lp);
6024 
6025  oldlen = ItemIdGetLength(lp) - htup->t_hoff;
6026  newlen = tuple->t_len - tuple->t_data->t_hoff;
6027  if (oldlen != newlen || htup->t_hoff != tuple->t_data->t_hoff)
6028  elog(ERROR, "wrong tuple length");
6029 
6030  /* NO EREPORT(ERROR) from here till changes are logged */
6032 
6033  memcpy((char *) htup + htup->t_hoff,
6034  (char *) tuple->t_data + tuple->t_data->t_hoff,
6035  newlen);
6036 
6037  MarkBufferDirty(buffer);
6038 
6039  /* XLOG stuff */
6040  if (RelationNeedsWAL(relation))
6041  {
6042  xl_heap_inplace xlrec;
6043  XLogRecPtr recptr;
6044 
6045  xlrec.offnum = ItemPointerGetOffsetNumber(&tuple->t_self);
6046 
6047  XLogBeginInsert();
6048  XLogRegisterData((char *) &xlrec, SizeOfHeapInplace);
6049 
6050  XLogRegisterBuffer(0, buffer, REGBUF_STANDARD);
6051  XLogRegisterBufData(0, (char *) htup + htup->t_hoff, newlen);
6052 
6053  /* inplace updates aren't decoded atm, don't log the origin */
6054 
6055  recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_INPLACE);
6056 
6057  PageSetLSN(page, recptr);
6058  }
6059 
6060  END_CRIT_SECTION();
6061 
6062  UnlockReleaseBuffer(buffer);
6063 
6064  /*
6065  * Send out shared cache inval if necessary. Note that because we only
6066  * pass the new version of the tuple, this mustn't be used for any
6067  * operations that could change catcache lookup keys. But we aren't
6068  * bothering with index updates either, so that's true a fortiori.
6069  */
6071  CacheInvalidateHeapTuple(relation, tuple, NULL);
6072 }
void XLogRegisterBufData(uint8 block_id, char *data, int len)
Definition: xloginsert.c:378
void CacheInvalidateHeapTuple(Relation relation, HeapTuple tuple, HeapTuple newtuple)
Definition: inval.c:1186
void MarkBufferDirty(Buffer buffer)
Definition: bufmgr.c:1565
void XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
Definition: xloginsert.c:232
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
#define END_CRIT_SECTION()
Definition: miscadmin.h:149
#define SizeOfHeapInplace
Definition: heapam_xlog.h:314
#define START_CRIT_SECTION()
Definition: miscadmin.h:147
int errcode(int sqlerrcode)
Definition: elog.c:698
#define BUFFER_LOCK_EXCLUSIVE
Definition: bufmgr.h:98
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
uint16 OffsetNumber
Definition: off.h:24
HeapTupleHeader t_data
Definition: htup.h:68
#define ItemIdGetLength(itemId)
Definition: itemid.h:59
bool IsInParallelMode(void)
Definition: xact.c:1012
void UnlockReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3791
#define ERROR
Definition: elog.h:46
ItemPointerData t_self
Definition: htup.h:65
uint32 t_len
Definition: htup.h:64
#define REGBUF_STANDARD
Definition: xloginsert.h:35
unsigned int uint32
Definition: c.h:441
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
void XLogRegisterData(char *data, int len)
Definition: xloginsert.c:340
XLogRecPtr XLogInsert(RmgrId rmid, uint8 info)
Definition: xloginsert.c:432
OffsetNumber offnum
Definition: heapam_xlog.h:310
void LockBuffer(Buffer buffer, int mode)
Definition: bufmgr.c:4007
#define ereport(elevel,...)
Definition: elog.h:157
uint64 XLogRecPtr
Definition: xlogdefs.h:21
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
Buffer ReadBuffer(Relation reln, BlockNumber blockNum)
Definition: bufmgr.c:694
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
#define XLOG_HEAP_INPLACE
Definition: heapam_xlog.h:39
#define RelationNeedsWAL(relation)
Definition: rel.h:601
#define IsBootstrapProcessingMode()
Definition: miscadmin.h:406
int errmsg(const char *fmt,...)
Definition: elog.c:909
#define elog(elevel,...)
Definition: elog.h:232
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
void XLogBeginInsert(void)
Definition: xloginsert.c:135
#define PageSetLSN(page, lsn)
Definition: bufpage.h:368
int Buffer
Definition: buf.h:23
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
Pointer Page
Definition: bufpage.h:78

◆ heap_insert()

void heap_insert ( Relation  relation,
HeapTuple  tup,
CommandId  cid,
int  options,
BulkInsertState  bistate 
)

Definition at line 2060 of file heapam.c.

References Assert, BufferGetBlockNumber(), BufferGetPage, CacheInvalidateHeapTuple(), CheckForSerializableConflictIn(), END_CRIT_SECTION, FirstOffsetNumber, xl_heap_insert::flags, GetCurrentTransactionId(), heap_freetuple(), HEAP_INSERT_NO_LOGICAL, HEAP_INSERT_SPECULATIVE, heap_prepare_insert(), HeapTupleHeaderGetNatts, InvalidBlockNumber, InvalidBuffer, IsToastRelation(), ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, log_heap_new_cid(), MarkBufferDirty(), xl_heap_insert::offnum, PageClearAllVisible, PageGetMaxOffsetNumber, PageIsAllVisible, PageSetLSN, pgstat_count_heap_insert(), REGBUF_KEEP_DATA, REGBUF_STANDARD, REGBUF_WILL_INIT, RelationGetBufferForTuple(), RelationGetNumberOfAttributes, RelationIsAccessibleInLogicalDecoding, RelationIsLogicallyLogged, RelationNeedsWAL, RelationPutHeapTuple(), ReleaseBuffer(), SizeOfHeapHeader, SizeOfHeapInsert, SizeofHeapTupleHeader, START_CRIT_SECTION, HeapTupleData::t_data, xl_heap_header::t_hoff, HeapTupleHeaderData::t_hoff, xl_heap_header::t_infomask, HeapTupleHeaderData::t_infomask, xl_heap_header::t_infomask2, HeapTupleHeaderData::t_infomask2, HeapTupleData::t_len, HeapTupleData::t_self, UnlockReleaseBuffer(), visibilitymap_clear(), VISIBILITYMAP_VALID_BITS, XLH_INSERT_ALL_VISIBLE_CLEARED, XLH_INSERT_CONTAINS_NEW_TUPLE, XLH_INSERT_IS_SPECULATIVE, XLH_INSERT_ON_TOAST_RELATION, XLOG_HEAP_INIT_PAGE, XLOG_HEAP_INSERT, XLOG_INCLUDE_ORIGIN, XLogBeginInsert(), XLogInsert(), XLogRegisterBufData(), XLogRegisterBuffer(), XLogRegisterData(), and XLogSetRecordFlags().

Referenced by heapam_tuple_insert(), heapam_tuple_insert_speculative(), simple_heap_insert(), and toast_save_datum().

2062 {
2064  HeapTuple heaptup;
2065  Buffer buffer;
2066  Buffer vmbuffer = InvalidBuffer;
2067  bool all_visible_cleared = false;
2068 
2069  /* Cheap, simplistic check that the tuple matches the rel's rowtype. */
2071  RelationGetNumberOfAttributes(relation));
2072 
2073  /*
2074  * Fill in tuple header fields and toast the tuple if necessary.
2075  *
2076  * Note: below this point, heaptup is the data we actually intend to store
2077  * into the relation; tup is the caller's original untoasted data.
2078  */
2079  heaptup = heap_prepare_insert(relation, tup, xid, cid, options);
2080 
2081  /*
2082  * Find buffer to insert this tuple into. If the page is all visible,
2083  * this will also pin the requisite visibility map page.
2084  */
2085  buffer = RelationGetBufferForTuple(relation, heaptup->t_len,
2086  InvalidBuffer, options, bistate,
2087  &vmbuffer, NULL);
2088 
2089  /*
2090  * We're about to do the actual insert -- but check for conflict first, to
2091  * avoid possibly having to roll back work we've just done.
2092  *
2093  * This is safe without a recheck as long as there is no possibility of
2094  * another process scanning the page between this check and the insert
2095  * being visible to the scan (i.e., an exclusive buffer content lock is
2096  * continuously held from this point until the tuple insert is visible).
2097  *
2098  * For a heap insert, we only need to check for table-level SSI locks. Our
2099  * new tuple can't possibly conflict with existing tuple locks, and heap
2100  * page locks are only consolidated versions of tuple locks; they do not
2101  * lock "gaps" as index page locks do. So we don't need to specify a
2102  * buffer when making the call, which makes for a faster check.
2103  */
2105 
2106  /* NO EREPORT(ERROR) from here till changes are logged */
2108 
2109  RelationPutHeapTuple(relation, buffer, heaptup,
2110  (options & HEAP_INSERT_SPECULATIVE) != 0);
2111 
2112  if (PageIsAllVisible(BufferGetPage(buffer)))
2113  {
2114  all_visible_cleared = true;
2116  visibilitymap_clear(relation,
2117  ItemPointerGetBlockNumber(&(heaptup->t_self)),
2118  vmbuffer, VISIBILITYMAP_VALID_BITS);
2119  }
2120 
2121  /*
2122  * XXX Should we set PageSetPrunable on this page ?
2123  *
2124  * The inserting transaction may eventually abort thus making this tuple
2125  * DEAD and hence available for pruning. Though we don't want to optimize
2126  * for aborts, if no other tuple in this page is UPDATEd/DELETEd, the
2127  * aborted tuple will never be pruned until next vacuum is triggered.
2128  *
2129  * If you do add PageSetPrunable here, add it in heap_xlog_insert too.
2130  */
2131 
2132  MarkBufferDirty(buffer);
2133 
2134  /* XLOG stuff */
2135  if (RelationNeedsWAL(relation))
2136  {
2137  xl_heap_insert xlrec;
2138  xl_heap_header xlhdr;
2139  XLogRecPtr recptr;
2140  Page page = BufferGetPage(buffer);
2141  uint8 info = XLOG_HEAP_INSERT;
2142  int bufflags = 0;
2143 
2144  /*
2145  * If this is a catalog, we need to transmit combo CIDs to properly
2146  * decode, so log that as well.
2147  */
2149  log_heap_new_cid(relation, heaptup);
2150 
2151  /*
2152  * If this is the single and first tuple on page, we can reinit the
2153  * page instead of restoring the whole thing. Set flag, and hide
2154  * buffer references from XLogInsert.
2155  */
2156  if (ItemPointerGetOffsetNumber(&(heaptup->t_self)) == FirstOffsetNumber &&
2158  {
2159  info |= XLOG_HEAP_INIT_PAGE;
2160  bufflags |= REGBUF_WILL_INIT;
2161  }
2162 
2163  xlrec.offnum = ItemPointerGetOffsetNumber(&heaptup->t_self);
2164  xlrec.flags = 0;
2165  if (all_visible_cleared)
2170 
2171  /*
2172  * For logical decoding, we need the tuple even if we're doing a full
2173  * page write, so make sure it's included even if we take a full-page
2174  * image. (XXX We could alternatively store a pointer into the FPW).
2175  */
2176  if (RelationIsLogicallyLogged(relation) &&
2178  {
2180  bufflags |= REGBUF_KEEP_DATA;
2181 
2182  if (IsToastRelation(relation))
2184  }
2185 
2186  XLogBeginInsert();
2187  XLogRegisterData((char *) &xlrec, SizeOfHeapInsert);
2188 
2189  xlhdr.t_infomask2 = heaptup->t_data->t_infomask2;
2190  xlhdr.t_infomask = heaptup->t_data->t_infomask;
2191  xlhdr.t_hoff = heaptup->t_data->t_hoff;
2192 
2193  /*
2194  * note we mark xlhdr as belonging to buffer; if XLogInsert decides to
2195  * write the whole page to the xlog, we don't need to store
2196  * xl_heap_header in the xlog.
2197  */
2198  XLogRegisterBuffer(0, buffer, REGBUF_STANDARD | bufflags);
2199  XLogRegisterBufData(0, (char *) &xlhdr, SizeOfHeapHeader);
2200  /* PG73FORMAT: write bitmap [+ padding] [+ oid] + data */
2202  (char *) heaptup->t_data + SizeofHeapTupleHeader,
2203  heaptup->t_len - SizeofHeapTupleHeader);
2204 
2205  /* filtering by origin on a row level is much more efficient */
2207 
2208  recptr = XLogInsert(RM_HEAP_ID, info);
2209 
2210  PageSetLSN(page, recptr);
2211  }
2212 
2213  END_CRIT_SECTION();
2214 
2215  UnlockReleaseBuffer(buffer);
2216  if (vmbuffer != InvalidBuffer)
2217  ReleaseBuffer(vmbuffer);
2218 
2219  /*
2220  * If tuple is cachable, mark it for invalidation from the caches in case
2221  * we abort. Note it is OK to do this after releasing the buffer, because
2222  * the heaptup data structure is all in local memory, not in the shared
2223  * buffer.
2224  */
2225  CacheInvalidateHeapTuple(relation, heaptup, NULL);
2226 
2227  /* Note: speculative insertions are counted too, even if aborted later */
2228  pgstat_count_heap_insert(relation, 1);
2229 
2230  /*
2231  * If heaptup is a private copy, release it. Don't forget to copy t_self
2232  * back to the caller's image, too.
2233  */
2234  if (heaptup != tup)
2235  {
2236  tup->t_self = heaptup->t_self;
2237  heap_freetuple(heaptup);
2238  }
2239 }
void XLogRegisterBufData(uint8 block_id, char *data, int len)
Definition: xloginsert.c:378
#define SizeofHeapTupleHeader
Definition: htup_details.h:184
bool IsToastRelation(Relation relation)
Definition: catalog.c:146
#define XLOG_HEAP_INSERT
Definition: heapam_xlog.h:32
static XLogRecPtr log_heap_new_cid(Relation relation, HeapTuple tup)
Definition: heapam.c:8228
void CacheInvalidateHeapTuple(Relation relation, HeapTuple tuple, HeapTuple newtuple)
Definition: inval.c:1186
static HeapTuple heap_prepare_insert(Relation relation, HeapTuple tup, TransactionId xid, CommandId cid, int options)
Definition: heapam.c:2248
#define PageIsAllVisible(page)
Definition: bufpage.h:385
uint32 TransactionId
Definition: c.h:587
#define RelationGetNumberOfAttributes(relation)
Definition: rel.h:483
void MarkBufferDirty(Buffer buffer)
Definition: bufmgr.c:1565
void XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
Definition: xloginsert.c:232
#define END_CRIT_SECTION()
Definition: miscadmin.h:149
unsigned char uint8
Definition: c.h:439
#define XLH_INSERT_IS_SPECULATIVE
Definition: heapam_xlog.h:68
#define InvalidBuffer
Definition: buf.h:25
#define REGBUF_WILL_INIT
Definition: xloginsert.h:33
uint16 t_infomask2
Definition: heapam_xlog.h:146
#define START_CRIT_SECTION()
Definition: miscadmin.h:147
#define XLH_INSERT_ON_TOAST_RELATION
Definition: heapam_xlog.h:70
#define XLOG_INCLUDE_ORIGIN
Definition: xlog.h:213
void ReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3768
#define RelationIsLogicallyLogged(relation)
Definition: rel.h:674
void heap_freetuple(HeapTuple htup)
Definition: heaptuple.c:1338
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
void RelationPutHeapTuple(Relation relation, Buffer buffer, HeapTuple tuple, bool token)
Definition: hio.c:36
#define XLOG_HEAP_INIT_PAGE
Definition: heapam_xlog.h:46
#define HEAP_INSERT_SPECULATIVE
Definition: heapam.h:37
#define VISIBILITYMAP_VALID_BITS
Definition: visibilitymap.h:28
HeapTupleHeader t_data
Definition: htup.h:68
bool visibilitymap_clear(Relation rel, BlockNumber heapBlk, Buffer buf, uint8 flags)
void UnlockReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3791
#define XLH_INSERT_CONTAINS_NEW_TUPLE
Definition: heapam_xlog.h:69
#define HeapTupleHeaderGetNatts(tup)
Definition: htup_details.h:530
ItemPointerData t_self
Definition: htup.h:65
TransactionId GetCurrentTransactionId(void)
Definition: xact.c:438
uint32 t_len
Definition: htup.h:64
#define FirstOffsetNumber
Definition: off.h:27
#define REGBUF_STANDARD
Definition: xloginsert.h:35
Buffer RelationGetBufferForTuple(Relation relation, Size len, Buffer otherBuffer, int options, BulkInsertState bistate, Buffer *vmbuffer, Buffer *vmbuffer_other)
Definition: hio.c:333
void XLogSetRecordFlags(uint8 flags)
Definition: xloginsert.c:414
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
void XLogRegisterData(char *data, int len)
Definition: xloginsert.c:340
XLogRecPtr XLogInsert(RmgrId rmid, uint8 info)
Definition: xloginsert.c:432
#define RelationIsAccessibleInLogicalDecoding(relation)
Definition: rel.h:657
#define REGBUF_KEEP_DATA
Definition: xloginsert.h:38
void CheckForSerializableConflictIn(Relation relation, ItemPointer tid, BlockNumber blkno)
Definition: predicate.c:4446
#define PageClearAllVisible(page)
Definition: bufpage.h:389
uint64 XLogRecPtr
Definition: xlogdefs.h:21
#define Assert(condition)
Definition: c.h:804
uint16 t_infomask
Definition: heapam_xlog.h:147
#define InvalidBlockNumber
Definition: block.h:33
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
#define RelationNeedsWAL(relation)
Definition: rel.h:601
#define SizeOfHeapInsert
Definition: heapam_xlog.h:162
#define XLH_INSERT_ALL_VISIBLE_CLEARED
Definition: heapam_xlog.h:66
BlockNumber BufferGetBlockNumber(Buffer buffer)
Definition: bufmgr.c:2748
void pgstat_count_heap_insert(Relation rel, PgStat_Counter n)
Definition: pgstat.c:2239
#define HEAP_INSERT_NO_LOGICAL
Definition: heapam.h:36
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
void XLogBeginInsert(void)
Definition: xloginsert.c:135
#define PageSetLSN(page, lsn)
Definition: bufpage.h:368
int Buffer
Definition: buf.h:23
OffsetNumber offnum
Definition: heapam_xlog.h:156
#define SizeOfHeapHeader
Definition: heapam_xlog.h:151
Pointer Page
Definition: bufpage.h:78

◆ heap_lock_tuple()

TM_Result heap_lock_tuple ( Relation  relation,
HeapTuple  tuple,
CommandId  cid,
LockTupleMode  mode,
LockWaitPolicy  wait_policy,
bool  follow_update,
Buffer buffer,
struct TM_FailureData tmfd 
)

Definition at line 4247 of file heapam.c.

References Assert, BUFFER_LOCK_EXCLUSIVE, BUFFER_LOCK_UNLOCK, BufferGetPage, BufferIsValid, TM_FailureData::cmax, compute_infobits(), compute_new_xmax_infomask(), ConditionalMultiXactIdWait(), ConditionalXactLockTableWait(), TM_FailureData::ctid, DoesMultiXactIdConflict(), elog, END_CRIT_SECTION, ereport, errcode(), errmsg(), ERROR, xl_heap_lock::flags, get_mxact_status_for_lock(), GetCurrentTransactionId(), GetMultiXactIdMembers(), heap_acquire_tuplock(), HEAP_KEYS_UPDATED, heap_lock_updated_tuple(), HEAP_XMAX_BITS, HEAP_XMAX_INVALID, HEAP_XMAX_IS_EXCL_LOCKED, HEAP_XMAX_IS_KEYSHR_LOCKED, HEAP_XMAX_IS_LOCKED_ONLY, HEAP_XMAX_IS_MULTI, HEAP_XMAX_IS_SHR_LOCKED, HeapTupleHeaderClearHotUpdated, HeapTupleHeaderGetCmax(), HeapTupleHeaderGetRawXmax, HeapTupleHeaderGetUpdateXid, HeapTupleHeaderIsOnlyLocked(), HeapTupleHeaderSetXmax, HeapTupleSatisfiesUpdate(), i, xl_heap_lock::infobits_set, InvalidBuffer, InvalidCommandId, ItemIdGetLength, ItemIdIsNormal, ItemPointerCopy, ItemPointerEquals(), ItemPointerGetBlockNumber, ItemPointerGetOffsetNumber, LockBuffer(), xl_heap_lock::locking_xid, LockTupleExclusive, LockTupleKeyShare, LockTupleNoKeyExclusive, LockTupleShare, LockWaitBlock, LockWaitError, LockWaitSkip, MarkBufferDirty(), MultiXactIdSetOldestMember(), MultiXactIdWait(), MultiXactStatusNoKeyUpdate, xl_heap_lock::offnum, PageGetItem, PageGetItemId, PageIsAllVisible, PageSetLSN, pfree(), ReadBuffer(), REGBUF_STANDARD, RelationGetRelationName, RelationGetRelid, RelationNeedsWAL, ReleaseBuffer(), SizeOfHeapLock, START_CRIT_SECTION, status(), HeapTupleHeaderData::t_ctid, HeapTupleData::t_data, HeapTupleHeaderData::t_infomask, HeapTupleHeaderData::t_infomask2, HeapTupleData::t_len, HeapTupleData::t_self, HeapTupleData::t_tableOid, TM_BeingModified, TM_Deleted, TM_Invisible, TM_Ok, TM_SelfModified, TM_Updated, TM_WouldBlock, TransactionIdEquals, TransactionIdIsCurrentTransactionId(), TUPLOCK_from_mxstatus, UnlockTupleTuplock, UpdateXmaxHintBits(), VISIBILITYMAP_ALL_FROZEN, visibilitymap_clear(), visibilitymap_pin(), XactLockTableWait(), XLH_LOCK_ALL_FROZEN_CLEARED, XLOG_HEAP_LOCK, XLogBeginInsert(), XLogInsert(), XLogRegisterBuffer(), XLogRegisterData(), XLTW_Lock, TM_FailureData::xmax, and xmax_infomask_changed().

Referenced by heapam_tuple_lock().

4251 {
4252  TM_Result result;
4253  ItemPointer tid = &(tuple->t_self);
4254  ItemId lp;
4255  Page page;
4256  Buffer vmbuffer = InvalidBuffer;
4257  BlockNumber block;
4258  TransactionId xid,
4259  xmax;
4260  uint16 old_infomask,
4261  new_infomask,
4262  new_infomask2;
4263  bool first_time = true;
4264  bool skip_tuple_lock = false;
4265  bool have_tuple_lock = false;
4266  bool cleared_all_frozen = false;
4267 
4268  *buffer = ReadBuffer(relation, ItemPointerGetBlockNumber(tid));
4269  block = ItemPointerGetBlockNumber(tid);
4270 
4271  /*
4272  * Before locking the buffer, pin the visibility map page if it appears to
4273  * be necessary. Since we haven't got the lock yet, someone else might be
4274  * in the middle of changing this, so we'll need to recheck after we have
4275  * the lock.
4276  */
4277  if (PageIsAllVisible(BufferGetPage(*buffer)))
4278  visibilitymap_pin(relation, block, &vmbuffer);
4279 
4281 
4282  page = BufferGetPage(*buffer);
4283  lp = PageGetItemId(page, ItemPointerGetOffsetNumber(tid));
4284  Assert(ItemIdIsNormal(lp));
4285 
4286  tuple->t_data = (HeapTupleHeader) PageGetItem(page, lp);
4287  tuple->t_len = ItemIdGetLength(lp);
4288  tuple->t_tableOid = RelationGetRelid(relation);
4289 
4290 l3:
4291  result = HeapTupleSatisfiesUpdate(tuple, cid, *buffer);
4292 
4293  if (result == TM_Invisible)
4294  {
4295  /*
4296  * This is possible, but only when locking a tuple for ON CONFLICT
4297  * UPDATE. We return this value here rather than throwing an error in
4298  * order to give that case the opportunity to throw a more specific
4299  * error.
4300  */
4301  result = TM_Invisible;
4302  goto out_locked;
4303  }
4304  else if (result == TM_BeingModified ||
4305  result == TM_Updated ||
4306  result == TM_Deleted)
4307  {
4308  TransactionId xwait;
4309  uint16 infomask;
4310  uint16 infomask2;
4311  bool require_sleep;
4312  ItemPointerData t_ctid;
4313 
4314  /* must copy state data before unlocking buffer */
4315  xwait = HeapTupleHeaderGetRawXmax(tuple->t_data);
4316  infomask = tuple->t_data->t_infomask;
4317  infomask2 = tuple->t_data->t_infomask2;
4318  ItemPointerCopy(&tuple->t_data->t_ctid, &t_ctid);
4319 
4320  LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
4321 
4322  /*
4323  * If any subtransaction of the current top transaction already holds
4324  * a lock as strong as or stronger than what we're requesting, we
4325  * effectively hold the desired lock already. We *must* succeed
4326  * without trying to take the tuple lock, else we will deadlock
4327  * against anyone wanting to acquire a stronger lock.
4328  *
4329  * Note we only do this the first time we loop on the HTSU result;
4330  * there is no point in testing in subsequent passes, because
4331  * evidently our own transaction cannot have acquired a new lock after
4332  * the first time we checked.
4333  */
4334  if (first_time)
4335  {
4336  first_time = false;
4337 
4338  if (infomask & HEAP_XMAX_IS_MULTI)
4339  {
4340  int i;
4341  int nmembers;
4342  MultiXactMember *members;
4343 
4344  /*
4345  * We don't need to allow old multixacts here; if that had
4346  * been the case, HeapTupleSatisfiesUpdate would have returned
4347  * MayBeUpdated and we wouldn't be here.
4348  */
4349  nmembers =
4350  GetMultiXactIdMembers(xwait, &members, false,
4351  HEAP_XMAX_IS_LOCKED_ONLY(infomask));
4352 
4353  for (i = 0; i < nmembers; i++)
4354  {
4355  /* only consider members of our own transaction */
4356  if (!TransactionIdIsCurrentTransactionId(members[i].xid))
4357  continue;
4358 
4359  if (TUPLOCK_from_mxstatus(members[i].status) >= mode)
4360  {
4361  pfree(members);
4362  result = TM_Ok;
4363  goto out_unlocked;
4364  }
4365  else
4366  {
4367  /*
4368  * Disable acquisition of the heavyweight tuple lock.
4369  * Otherwise, when promoting a weaker lock, we might
4370  * deadlock with another locker that has acquired the
4371  * heavyweight tuple lock and is waiting for our
4372  * transaction to finish.
4373  *
4374  * Note that in this case we still need to wait for
4375  * the multixact if required, to avoid acquiring
4376  * conflicting locks.
4377  */
4378  skip_tuple_lock = true;
4379  }
4380  }
4381 
4382  if (members)
4383  pfree(members);
4384  }
4385  else if (TransactionIdIsCurrentTransactionId(xwait))
4386  {
4387  switch (mode)
4388  {
4389  case LockTupleKeyShare:
4390  Assert(HEAP_XMAX_IS_KEYSHR_LOCKED(infomask) ||
4391  HEAP_XMAX_IS_SHR_LOCKED(infomask) ||
4392  HEAP_XMAX_IS_EXCL_LOCKED(infomask));
4393  result = TM_Ok;
4394  goto out_unlocked;
4395  case LockTupleShare:
4396  if (HEAP_XMAX_IS_SHR_LOCKED(infomask) ||
4397  HEAP_XMAX_IS_EXCL_LOCKED(infomask))
4398  {
4399  result = TM_Ok;
4400  goto out_unlocked;
4401  }
4402  break;
4404  if (HEAP_XMAX_IS_EXCL_LOCKED(infomask))
4405  {
4406  result = TM_Ok;
4407  goto out_unlocked;
4408  }
4409  break;
4410  case LockTupleExclusive:
4411  if (HEAP_XMAX_IS_EXCL_LOCKED(infomask) &&
4412  infomask2 & HEAP_KEYS_UPDATED)
4413  {
4414  result = TM_Ok;
4415  goto out_unlocked;
4416  }
4417  break;
4418  }
4419  }
4420  }
4421 
4422  /*
4423  * Initially assume that we will have to wait for the locking
4424  * transaction(s) to finish. We check various cases below in which
4425  * this can be turned off.
4426  */
4427  require_sleep = true;
4428  if (mode == LockTupleKeyShare)
4429  {
4430  /*
4431  * If we're requesting KeyShare, and there's no update present, we
4432  * don't need to wait. Even if there is an update, we can still
4433  * continue if the key hasn't been modified.
4434  *
4435  * However, if there are updates, we need to walk the update chain
4436  * to mark future versions of the row as locked, too. That way,
4437  * if somebody deletes that future version, we're protected
4438  * against the key going away. This locking of future versions
4439  * could block momentarily, if a concurrent transaction is
4440  * deleting a key; or it could return a value to the effect that
4441  * the transaction deleting the key has already committed. So we
4442  * do this before re-locking the buffer; otherwise this would be
4443  * prone to deadlocks.
4444  *
4445  * Note that the TID we're locking was grabbed before we unlocked
4446  * the buffer. For it to change while we're not looking, the
4447  * other properties we're testing for below after re-locking the
4448  * buffer would also change, in which case we would restart this
4449  * loop above.
4450  */
4451  if (!(infomask2 & HEAP_KEYS_UPDATED))
4452  {
4453  bool updated;
4454 
4455  updated = !HEAP_XMAX_IS_LOCKED_ONLY(infomask);
4456 
4457  /*
4458  * If there are updates, follow the update chain; bail out if
4459  * that cannot be done.
4460  */
4461  if (follow_updates && updated)
4462  {
4463  TM_Result res;
4464 
4465  res = heap_lock_updated_tuple(relation, tuple, &t_ctid,
4467  mode);
4468  if (res != TM_Ok)
4469  {
4470  result = res;
4471  /* recovery code expects to have buffer lock held */
4473  goto failed;
4474  }
4475  }
4476 
4478 
4479  /*
4480  * Make sure it's still an appropriate lock, else start over.
4481  * Also, if it wasn't updated before we released the lock, but
4482  * is updated now, we start over too; the reason is that we
4483  * now need to follow the update chain to lock the new
4484  * versions.
4485  */
4486  if (!HeapTupleHeaderIsOnlyLocked(tuple->t_data) &&
4487  ((tuple->t_data->t_infomask2 & HEAP_KEYS_UPDATED) ||
4488  !updated))
4489  goto l3;
4490 
4491  /* Things look okay, so we can skip sleeping */
4492  require_sleep = false;
4493 
4494  /*
4495  * Note we allow Xmax to change here; other updaters/lockers
4496  * could have modified it before we grabbed the buffer lock.
4497  * However, this is not a problem, because with the recheck we
4498  * just did we ensure that they still don't conflict with the
4499  * lock we want.
4500  */
4501  }
4502  }
4503  else if (mode == LockTupleShare)
4504  {
4505  /*
4506  * If we're requesting Share, we can similarly avoid sleeping if
4507  * there's no update and no exclusive lock present.
4508  */
4509  if (HEAP_XMAX_IS_LOCKED_ONLY(infomask) &&
4510  !HEAP_XMAX_IS_EXCL_LOCKED(infomask))
4511  {
4513 
4514  /*
4515  * Make sure it's still an appropriate lock, else start over.
4516  * See above about allowing xmax to change.
4517  */
4518  if (!HEAP_XMAX_IS_LOCKED_ONLY(tuple->t_data->t_infomask) ||
4520  goto l3;
4521  require_sleep = false;
4522  }
4523  }
4524  else if (mode == LockTupleNoKeyExclusive)
4525  {
4526  /*
4527  * If we're requesting NoKeyExclusive, we might also be able to
4528  * avoid sleeping; just ensure that there no conflicting lock
4529  * already acquired.
4530  */
4531  if (infomask & HEAP_XMAX_IS_MULTI)
4532  {
4533  if (!DoesMultiXactIdConflict((MultiXactId) xwait, infomask,
4534  mode, NULL))
4535  {
4536  /*
4537  * No conflict, but if the xmax changed under us in the
4538  * meantime, start over.
4539  */
4541  if (xmax_infomask_changed(tuple->t_data->t_infomask, infomask) ||
4543  xwait))
4544  goto l3;
4545 
4546  /* otherwise, we're good */
4547  require_sleep = false;
4548  }
4549  }
4550  else if (HEAP_XMAX_IS_KEYSHR_LOCKED(infomask))
4551  {
4553 
4554  /* if the xmax changed in the meantime, start over */
4555  if (xmax_infomask_changed(tuple->t_data->t_infomask, infomask) ||
4557  xwait))
4558  goto l3;
4559  /* otherwise, we're good */
4560  require_sleep = false;
4561  }
4562  }
4563 
4564  /*
4565  * As a check independent from those above, we can also avoid sleeping
4566  * if the current transaction is the sole locker of the tuple. Note
4567  * that the strength of the lock already held is irrelevant; this is
4568  * not about recording the lock in Xmax (which will be done regardless
4569  * of this optimization, below). Also, note that the cases where we
4570  * hold a lock stronger than we are requesting are already handled
4571  * above by not doing anything.
4572  *
4573  * Note we only deal with the non-multixact case here; MultiXactIdWait
4574  * is well equipped to deal with this situation on its own.
4575  */
4576  if (require_sleep && !(infomask & HEAP_XMAX_IS_MULTI) &&
4578  {
4579  /* ... but if the xmax changed in the meantime, start over */
4581  if (xmax_infomask_changed(tuple->t_data->t_infomask, infomask) ||
4583  xwait))
4584  goto l3;
4586  require_sleep = false;
4587  }
4588 
4589  /*
4590  * Time to sleep on the other transaction/multixact, if necessary.
4591  *
4592  * If the other transaction is an update/delete that's already
4593  * committed, then sleeping cannot possibly do any good: if we're
4594  * required to sleep, get out to raise an error instead.
4595  *
4596  * By here, we either have already acquired the buffer exclusive lock,
4597  * or we must wait for the locking transaction or multixact; so below
4598  * we ensure that we grab buffer lock after the sleep.
4599  */
4600  if (require_sleep && (result == TM_Updated || result == TM_Deleted))
4601  {
4603  goto failed;
4604  }
4605  else if (require_sleep)
4606  {
4607  /*
4608  * Acquire tuple lock to establish our priority for the tuple, or
4609  * die trying. LockTuple will release us when we are next-in-line
4610  * for the tuple. We must do this even if we are share-locking,
4611  * but not if we already have a weaker lock on the tuple.
4612  *
4613  * If we are forced to "start over" below, we keep the tuple lock;
4614  * this arranges that we stay at the head of the line while
4615  * rechecking tuple state.
4616  */
4617  if (!skip_tuple_lock &&
4618  !heap_acquire_tuplock(relation, tid, mode, wait_policy,
4619  &have_tuple_lock))
4620  {
4621  /*
4622  * This can only happen if wait_policy is Skip and the lock
4623  * couldn't be obtained.
4624  */
4625  result = TM_WouldBlock;
4626  /* recovery code expects to have buffer lock held */
4628  goto failed;
4629  }
4630 
4631  if (infomask & HEAP_XMAX_IS_MULTI)
4632  {
4634 
4635  /* We only ever lock tuples, never update them */
4636  if (status >= MultiXactStatusNoKeyUpdate)
4637  elog(ERROR, "invalid lock mode in heap_lock_tuple");
4638 
4639  /* wait for multixact to end, or die trying */
4640  switch (wait_policy)
4641  {
4642  case LockWaitBlock:
4643  MultiXactIdWait((MultiXactId) xwait, status, infomask,
4644  relation, &tuple->t_self, XLTW_Lock, NULL);
4645  break;
4646  case LockWaitSkip:
4648  status, infomask, relation,
4649  NULL))
4650  {
4651  result = TM_WouldBlock;
4652  /* recovery code expects to have buffer lock held */
4654  goto failed;
4655  }
4656  break;
4657  case LockWaitError:
4659  status, infomask, relation,
4660  NULL))
4661  ereport(ERROR,
4662  (errcode(ERRCODE_LOCK_NOT_AVAILABLE),
4663  errmsg("could not obtain lock on row in relation \"%s\"",
4664  RelationGetRelationName(relation))));
4665 
4666  break;
4667  }
4668 
4669  /*
4670  * Of course, the multixact might not be done here: if we're
4671  * requesting a light lock mode, other transactions with light
4672  * locks could still be alive, as well as locks owned by our
4673  * own xact or other subxacts of this backend. We need to
4674  * preserve the surviving MultiXact members. Note that it
4675  * isn't absolutely necessary in the latter case, but doing so
4676  * is simpler.
4677  */
4678  }
4679  else
4680  {
4681  /* wait for regular transaction to end, or die trying */
4682  switch (wait_policy)
4683  {
4684  case LockWaitBlock:
4685  XactLockTableWait(xwait, relation, &tuple->t_self,
4686  XLTW_Lock);
4687  break;
4688  case LockWaitSkip:
4689  if (!ConditionalXactLockTableWait(xwait))
4690  {
4691  result = TM_WouldBlock;
4692  /* recovery code expects to have buffer lock held */
4694  goto failed;
4695  }
4696  break;
4697  case LockWaitError:
4698  if (!ConditionalXactLockTableWait(xwait))
4699  ereport(ERROR,
4700  (errcode(ERRCODE_LOCK_NOT_AVAILABLE),
4701  errmsg("could not obtain lock on row in relation \"%s\"",
4702  RelationGetRelationName(relation))));
4703  break;
4704  }
4705  }
4706 
4707  /* if there are updates, follow the update chain */
4708  if (follow_updates && !HEAP_XMAX_IS_LOCKED_ONLY(infomask))
4709  {
4710  TM_Result res;
4711 
4712  res = heap_lock_updated_tuple(relation, tuple, &t_ctid,
4714  mode);
4715  if (res != TM_Ok)
4716  {
4717  result = res;
4718  /* recovery code expects to have buffer lock held */
4720  goto failed;
4721  }
4722  }
4723 
4725 
4726  /*
4727  * xwait is done, but if xwait had just locked the tuple then some
4728  * other xact could update this tuple before we get to this point.
4729  * Check for xmax change, and start over if so.
4730  */
4731  if (xmax_infomask_changed(tuple->t_data->t_infomask, infomask) ||
4733  xwait))
4734  goto l3;
4735 
4736  if (!(infomask & HEAP_XMAX_IS_MULTI))
4737  {
4738  /*
4739  * Otherwise check if it committed or aborted. Note we cannot
4740  * be here if the tuple was only locked by somebody who didn't
4741  * conflict with us; that would have been handled above. So
4742  * that transaction must necessarily be gone by now. But
4743  * don't check for this in the multixact case, because some
4744  * locker transactions might still be running.
4745  */
4746  UpdateXmaxHintBits(tuple->t_data, *buffer, xwait);
4747  }
4748  }
4749 
4750  /* By here, we're certain that we hold buffer exclusive lock again */
4751 
4752  /*
4753  * We may lock if previous xmax aborted, or if it committed but only
4754  * locked the tuple without updating it; or if we didn't have to wait
4755  * at all for whatever reason.
4756  */
4757  if (!require_sleep ||
4758  (tuple->t_data->t_infomask & HEAP_XMAX_INVALID) ||
4761  result = TM_Ok;
4762  else if (!ItemPointerEquals(&tuple->t_self, &tuple->t_data->t_ctid))
4763  result = TM_Updated;
4764  else
4765  result = TM_Deleted;
4766  }
4767 
4768 failed:
4769  if (result != TM_Ok)
4770  {
4771  Assert(result == TM_SelfModified || result == TM_Updated ||
4772  result == TM_Deleted || result == TM_WouldBlock);
4773  Assert(!(tuple->t_data->t_infomask & HEAP_XMAX_INVALID));
4774  Assert(result != TM_Updated ||
4775  !ItemPointerEquals(&tuple->t_self, &tuple->t_data->t_ctid));
4776  tmfd->ctid = tuple->t_data->t_ctid;
4777  tmfd->xmax = HeapTupleHeaderGetUpdateXid(tuple->t_data);
4778  if (result == TM_SelfModified)
4779  tmfd->cmax = HeapTupleHeaderGetCmax(tuple->t_data);
4780  else
4781  tmfd->cmax = InvalidCommandId;
4782  goto out_locked;
4783  }
4784 
4785  /*
4786  * If we didn't pin the visibility map page and the page has become all
4787  * visible while we were busy locking the buffer, or during some
4788  * subsequent window during which we had it unlocked, we'll have to unlock
4789  * and re-lock, to avoid holding the buffer lock across I/O. That's a bit
4790  * unfortunate, especially since we'll now have to recheck whether the
4791  * tuple has been locked or updated under us, but hopefully it won't
4792  * happen very often.
4793  */
4794  if (vmbuffer == InvalidBuffer && PageIsAllVisible(page))
4795  {
4796  LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
4797  visibilitymap_pin(relation, block, &vmbuffer);
4799  goto l3;
4800  }
4801 
4802  xmax = HeapTupleHeaderGetRawXmax(tuple->t_data);
4803  old_infomask = tuple->t_data->t_infomask;
4804 
4805  /*
4806  * If this is the first possibly-multixact-able operation in the current
4807  * transaction, set my per-backend OldestMemberMXactId setting. We can be
4808  * certain that the transaction will never become a member of any older
4809  * MultiXactIds than that. (We have to do this even if we end up just
4810  * using our own TransactionId below, since some other backend could
4811  * incorporate our XID into a MultiXact immediately afterwards.)
4812  */
4814 
4815  /*
4816  * Compute the new xmax and infomask to store into the tuple. Note we do
4817  * not modify the tuple just yet, because that would leave it in the wrong
4818  * state if multixact.c elogs.
4819  */
4820  compute_new_xmax_infomask(xmax, old_infomask, tuple->t_data->t_infomask2,
4821  GetCurrentTransactionId(), mode, false,
4822  &xid, &new_infomask, &new_infomask2);
4823 
4825 
4826  /*
4827  * Store transaction information of xact locking the tuple.
4828  *
4829  * Note: Cmax is meaningless in this context, so don't set it; this avoids
4830  * possibly generating a useless combo CID. Moreover, if we're locking a
4831  * previously updated tuple, it's important to preserve the Cmax.
4832  *
4833  * Also reset the HOT UPDATE bit, but only if there's no update; otherwise
4834  * we would break the HOT chain.
4835  */
4836  tuple->t_data->t_infomask &= ~HEAP_XMAX_BITS;
4837  tuple->t_data->t_infomask2 &= ~HEAP_KEYS_UPDATED;
4838  tuple->t_data->t_infomask |= new_infomask;
4839  tuple->t_data->t_infomask2 |= new_infomask2;
4840  if (HEAP_XMAX_IS_LOCKED_ONLY(new_infomask))
4842  HeapTupleHeaderSetXmax(tuple->t_data, xid);
4843 
4844  /*
4845  * Make sure there is no forward chain link in t_ctid. Note that in the
4846  * cases where the tuple has been updated, we must not overwrite t_ctid,
4847  * because it was set by the updater. Moreover, if the tuple has been
4848  * updated, we need to follow the update chain to lock the new versions of
4849  * the tuple as well.
4850  */
4851  if (HEAP_XMAX_IS_LOCKED_ONLY(new_infomask))
4852  tuple->t_data->t_ctid = *tid;
4853 
4854  /* Clear only the all-frozen bit on visibility map if needed */
4855  if (PageIsAllVisible(page) &&
4856  visibilitymap_clear(relation, block, vmbuffer,
4858  cleared_all_frozen = true;
4859 
4860 
4861  MarkBufferDirty(*buffer);
4862 
4863  /*
4864  * XLOG stuff. You might think that we don't need an XLOG record because
4865  * there is no state change worth restoring after a crash. You would be
4866  * wrong however: we have just written either a TransactionId or a
4867  * MultiXactId that may never have been seen on disk before, and we need
4868  * to make sure that there are XLOG entries covering those ID numbers.
4869  * Else the same IDs might be re-used after a crash, which would be
4870  * disastrous if this page made it to disk before the crash. Essentially
4871  * we have to enforce the WAL log-before-data rule even in this case.
4872  * (Also, in a PITR log-shipping or 2PC environment, we have to have XLOG
4873  * entries for everything anyway.)
4874  */
4875  if (RelationNeedsWAL(relation))
4876  {
4877  xl_heap_lock xlrec;
4878  XLogRecPtr recptr;
4879 
4880  XLogBeginInsert();
4881  XLogRegisterBuffer(0, *buffer, REGBUF_STANDARD);
4882 
4883  xlrec.offnum = ItemPointerGetOffsetNumber(&tuple->t_self);
4884  xlrec.locking_xid = xid;
4885  xlrec.infobits_set = compute_infobits(new_infomask,
4886  tuple->t_data->t_infomask2);
4887  xlrec.flags = cleared_all_frozen ? XLH_LOCK_ALL_FROZEN_CLEARED : 0;
4888  XLogRegisterData((char *) &xlrec, SizeOfHeapLock);
4889 
4890  /* we don't decode row locks atm, so no need to log the origin */
4891 
4892  recptr = XLogInsert(RM_HEAP_ID, XLOG_HEAP_LOCK);
4893 
4894  PageSetLSN(page, recptr);
4895  }
4896 
4897  END_CRIT_SECTION();
4898 
4899  result = TM_Ok;
4900 
4901 out_locked:
4902  LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
4903 
4904 out_unlocked:
4905  if (BufferIsValid(vmbuffer))
4906  ReleaseBuffer(vmbuffer);
4907 
4908  /*
4909  * Don't update the visibility map here. Locking a tuple doesn't change
4910  * visibility info.
4911  */
4912 
4913  /*
4914  * Now that we have successfully marked the tuple as locked, we can
4915  * release the lmgr tuple lock, if we had it.
4916  */
4917  if (have_tuple_lock)
4918  UnlockTupleTuplock(relation, tid, mode);
4919 
4920  return result;
4921 }
#define HeapTupleHeaderGetUpdateXid(tup)
Definition: htup_details.h:365
ItemPointerData ctid
Definition: tableam.h:126
static PgChecksumMode mode
Definition: pg_checksums.c:65
MultiXactStatus
Definition: multixact.h:41
#define BUFFER_LOCK_UNLOCK
Definition: bufmgr.h:96
#define HEAP_XMAX_BITS
Definition: htup_details.h:270
OffsetNumber offnum
Definition: heapam_xlog.h:281
static uint8 compute_infobits(uint16 infomask, uint16 infomask2)
Definition: heapam.c:2655
#define TransactionIdEquals(id1, id2)
Definition: transam.h:43
#define PageIsAllVisible(page)
Definition: bufpage.h:385
#define XLH_LOCK_ALL_FROZEN_CLEARED
Definition: heapam_xlog.h:275
TransactionId locking_xid
Definition: heapam_xlog.h:280
uint32 TransactionId
Definition: c.h:587
bool TransactionIdIsCurrentTransactionId(TransactionId xid)
Definition: xact.c:869
void visibilitymap_pin(Relation rel, BlockNumber heapBlk, Buffer *buf)
void MarkBufferDirty(Buffer buffer)
Definition: bufmgr.c:1565
void XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
Definition: xloginsert.c:232
HeapTupleHeaderData * HeapTupleHeader
Definition: htup.h:23
#define VISIBILITYMAP_ALL_FROZEN
Definition: visibilitymap.h:27
static bool xmax_infomask_changed(uint16 new_infomask, uint16 old_infomask)
Definition: heapam.c:2677
#define HeapTupleHeaderClearHotUpdated(tup)
Definition: htup_details.h:495
#define END_CRIT_SECTION()
Definition: miscadmin.h:149
CommandId cmax
Definition: tableam.h:128
bool HeapTupleHeaderIsOnlyLocked(HeapTupleHeader tuple)
#define InvalidBuffer
Definition: buf.h:25
#define START_CRIT_SECTION()
Definition: miscadmin.h:147
int errcode(int sqlerrcode)
Definition: elog.c:698
uint32 BlockNumber
Definition: block.h:31
void ReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3768
#define BUFFER_LOCK_EXCLUSIVE
Definition: bufmgr.h:98
TM_Result HeapTupleSatisfiesUpdate(HeapTuple htup, CommandId curcid, Buffer buffer)
#define UnlockTupleTuplock(rel, tup, mode)
Definition: heapam.c:165
bool ConditionalXactLockTableWait(TransactionId xid)
Definition: lmgr.c:713
#define HEAP_XMAX_IS_SHR_LOCKED(infomask)
Definition: htup_details.h:262
void MultiXactIdSetOldestMember(void)
Definition: multixact.c:625
HeapTupleHeader t_data
Definition: htup.h:68
#define HeapTupleHeaderGetRawXmax(tup)
Definition: htup_details.h:375
unsigned short uint16
Definition: c.h:440
void pfree(void *pointer)
Definition: mcxt.c:1169
#define ItemIdGetLength(itemId)
Definition: itemid.h:59
bool visibilitymap_clear(Relation rel, BlockNumber heapBlk, Buffer buf, uint8 flags)
TransactionId xmax
Definition: tableam.h:127
#define ERROR
Definition: elog.h:46
#define HEAP_XMAX_INVALID
Definition: htup_details.h:207
int8 infobits_set
Definition: heapam_xlog.h:282
ItemPointerData t_ctid
Definition: htup_details.h:160
ItemPointerData t_self
Definition: htup.h:65
static void MultiXactIdWait(MultiXactId multi, MultiXactStatus status, uint16 infomask, Relation rel, ItemPointer ctid, XLTW_Oper oper, int *remaining)
Definition: heapam.c:7009
TransactionId GetCurrentTransactionId(void)
Definition: xact.c:438
uint32 t_len
Definition: htup.h:64
#define REGBUF_STANDARD
Definition: xloginsert.h:35
#define HeapTupleHeaderSetXmax(tup, xid)
Definition: htup_details.h:380
#define RelationGetRelationName(relation)
Definition: rel.h:511
Oid t_tableOid
Definition: htup.h:66
#define SizeOfHeapLock
Definition: heapam_xlog.h:286
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
static void compute_new_xmax_infomask(TransactionId xmax, uint16 old_infomask, uint16 old_infomask2, TransactionId add_to_xmax, LockTupleMode mode, bool is_update, TransactionId *result_xmax, uint16 *result_infomask, uint16 *result_infomask2)
Definition: heapam.c:4985
#define PageGetItemId(page, offsetNumber)
Definition: bufpage.h:235
void XLogRegisterData(char *data, int len)
Definition: xloginsert.c:340
XLogRecPtr XLogInsert(RmgrId rmid, uint8 info)
Definition: xloginsert.c:432
TM_Result
Definition: tableam.h:71
#define InvalidCommandId
Definition: c.h:604
#define HEAP_XMAX_IS_LOCKED_ONLY(infomask)
Definition: htup_details.h:230
void LockBuffer(Buffer buffer, int mode)
Definition: bufmgr.c:4007
#define HEAP_KEYS_UPDATED
Definition: htup_details.h:278
#define HEAP_XMAX_IS_MULTI
Definition: htup_details.h:208
static void UpdateXmaxHintBits(HeapTupleHeader tuple, Buffer buffer, TransactionId xid)
Definition: heapam.c:1985
static bool heap_acquire_tuplock(Relation relation, ItemPointer tid, LockTupleMode mode, LockWaitPolicy wait_policy, bool *have_tuple_lock)
Definition: heapam.c:4936
#define ereport(elevel,...)
Definition: elog.h:157
TransactionId MultiXactId
Definition: c.h:597
void XactLockTableWait(TransactionId xid, Relation rel, ItemPointer ctid, XLTW_Oper oper)
Definition: lmgr.c:640
uint64 XLogRecPtr
Definition: xlogdefs.h:21
#define Assert(condition)
Definition: c.h:804
static TM_Result heap_lock_updated_tuple(Relation rel, HeapTuple tuple, ItemPointer ctid, TransactionId xid, LockTupleMode mode)
Definition: heapam.c:5702
Definition: lmgr.h:29
CommandId HeapTupleHeaderGetCmax(HeapTupleHeader tup)
Definition: combocid.c:118
Definition: tableam.h:77
#define ItemIdIsNormal(itemId)
Definition: itemid.h:99
Buffer ReadBuffer(Relation reln, BlockNumber blockNum)
Definition: bufmgr.c:694
#define BufferIsValid(bufnum)
Definition: bufmgr.h:123
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
static bool DoesMultiXactIdConflict(MultiXactId multi, uint16 infomask, LockTupleMode lockmode, bool *current_is_member)
Definition: heapam.c:6832
#define RelationNeedsWAL(relation)
Definition: rel.h:601
bool ItemPointerEquals(ItemPointer pointer1, ItemPointer pointer2)
Definition: itemptr.c:29
#define TUPLOCK_from_mxstatus(status)
Definition: heapam.c:214
int errmsg(const char *fmt,...)
Definition: elog.c:909
#define XLOG_HEAP_LOCK
Definition: heapam_xlog.h:38
#define elog(elevel,...)
Definition: elog.h:232
int i
static MultiXactStatus get_mxact_status_for_lock(LockTupleMode mode, bool is_update)
Definition: heapam.c:4199
#define HEAP_XMAX_IS_EXCL_LOCKED(infomask)
Definition: htup_details.h:264
int GetMultiXactIdMembers(MultiXactId multi, MultiXactMember **members, bool from_pgupgrade, bool onlyLock)
Definition: multixact.c:1223
#define ItemPointerGetBlockNumber(pointer)
Definition: itemptr.h:98
#define HEAP_XMAX_IS_KEYSHR_LOCKED(infomask)
Definition: htup_details.h:266
static void static void status(const char *fmt,...) pg_attribute_printf(1
Definition: pg_regress.c:227
void XLogBeginInsert(void)
Definition: xloginsert.c:135
#define PageSetLSN(page, lsn)
Definition: bufpage.h:368
int Buffer
Definition: buf.h:23
#define RelationGetRelid(relation)
Definition: rel.h:477
static bool ConditionalMultiXactIdWait(MultiXactId multi, MultiXactStatus status, uint16 infomask, Relation rel, int *remaining)
Definition: heapam.c:7031
#define PageGetItem(page, itemId)
Definition: bufpage.h:340
Pointer Page
Definition: bufpage.h:78
#define ItemPointerCopy(fromPointer, toPointer)
Definition: itemptr.h:161

◆ heap_multi_insert()

void heap_multi_insert ( Relation  relation,
struct TupleTableSlot **  slots,
int  ntuples,
CommandId  cid,
int  options,
BulkInsertState  bistate 
)

Definition at line 2302 of file heapam.c.

References Assert, AssertArg, BufferGetBlockNumber(), BufferGetPage, CacheInvalidateHeapTuple(), CHECK_FOR_INTERRUPTS, CheckForSerializableConflictIn(), PGAlignedBlock::data, xl_multi_insert_tuple::datalen, END_CRIT_SECTION, ExecFetchSlotHeapTuple(), xl_heap_multi_insert::flags, GetCurrentTransactionId(), HEAP_DEFAULT_FILLFACTOR, HEAP_INSERT_FROZEN, HEAP_INSERT_NO_LOGICAL, heap_prepare_insert(), i, init, InvalidBlockNumber, InvalidBuffer, InvalidTransactionId, InvalidXLogRecPtr, IsCatalogRelation(), ItemPointerGetOffsetNumber, log_heap_new_cid(), MarkBufferDirty(), MAXALIGN, xl_heap_multi_insert::ntuples, xl_heap_multi_insert::offsets, PageClearAllVisible, PageGetHeapFreeSpace(), PageGetMaxOffsetNumber, PageIsAllVisible, PageSetAllVisible, PageSetLSN, palloc(), pgstat_count_heap_insert(), REGBUF_KEEP_DATA, REGBUF_STANDARD, REGBUF_WILL_INIT, RelationGetBufferForTuple(), RelationGetRelid, RelationGetTargetPageFreeSpace, RelationIsAccessibleInLogicalDecoding, RelationIsLogicallyLogged, RelationNeedsWAL, RelationPutHeapTuple(), ReleaseBuffer(), SHORTALIGN, SizeOfHeapMultiInsert, SizeofHeapTupleHeader, SizeOfMultiInsertTuple, START_CRIT_SECTION, HeapTupleData::t_data, HeapTupleHeaderData::t_hoff, xl_multi_insert_tuple::t_hoff, HeapTupleHeaderData::t_infomask, xl_multi_insert_tuple::t_infomask, HeapTupleHeaderData::t_infomask2, xl_multi_insert_tuple::t_infomask2, HeapTupleData::t_len, HeapTupleData::t_self, HeapTupleData::t_tableOid, TupleTableSlot::tts_tableOid, UnlockReleaseBuffer(), VISIBILITYMAP_ALL_FROZEN, VISIBILITYMAP_ALL_VISIBLE, visibilitymap_clear(), visibilitymap_pin_ok(), visibilitymap_set(), VISIBILITYMAP_VALID_BITS, XLH_INSERT_ALL_FROZEN_SET, XLH_INSERT_ALL_VISIBLE_CLEARED, XLH_INSERT_CONTAINS_NEW_TUPLE, XLH_INSERT_LAST_IN_MULTI, XLOG_HEAP2_MULTI_INSERT, XLOG_HEAP_INIT_PAGE, XLOG_INCLUDE_ORIGIN, XLogBeginInsert(), XLogInsert(), XLogRegisterBufData(), XLogRegisterBuffer(), XLogRegisterData(), and XLogSetRecordFlags().

Referenced by CatalogTuplesMultiInsertWithInfo(), and SampleHeapTupleVisible().

2304 {
2306  HeapTuple *heaptuples;
2307  int i;
2308  int ndone;
2309  PGAlignedBlock scratch;
2310  Page page;
2311  Buffer vmbuffer = InvalidBuffer;
2312  bool needwal;
2313  Size saveFreeSpace;
2314  bool need_tuple_data = RelationIsLogicallyLogged(relation);
2315  bool need_cids = RelationIsAccessibleInLogicalDecoding(relation);
2316 
2317  /* currently not needed (thus unsupported) for heap_multi_insert() */
2319 
2320  needwal = RelationNeedsWAL(relation);
2321  saveFreeSpace = RelationGetTargetPageFreeSpace(relation,
2323 
2324  /* Toast and set header data in all the slots */
2325  heaptuples = palloc(ntuples * sizeof(HeapTuple));
2326  for (i = 0; i < ntuples; i++)
2327  {
2328  HeapTuple tuple;
2329 
2330  tuple = ExecFetchSlotHeapTuple(slots[i], true, NULL);
2331  slots[i]->tts_tableOid = RelationGetRelid(relation);
2332  tuple->t_tableOid = slots[i]->tts_tableOid;
2333  heaptuples[i] = heap_prepare_insert(relation, tuple, xid, cid,
2334  options);
2335  }
2336 
2337  /*
2338  * We're about to do the actual inserts -- but check for conflict first,
2339  * to minimize the possibility of having to roll back work we've just
2340  * done.
2341  *
2342  * A check here does not definitively prevent a serialization anomaly;
2343  * that check MUST be done at least past the point of acquiring an
2344  * exclusive buffer content lock on every buffer that will be affected,
2345  * and MAY be done after all inserts are reflected in the buffers and
2346  * those locks are released; otherwise there is a race condition. Since
2347  * multiple buffers can be locked and unlocked in the loop below, and it
2348  * would not be feasible to identify and lock all of those buffers before
2349  * the loop, we must do a final check at the end.
2350  *
2351  * The check here could be omitted with no loss of correctness; it is
2352  * present strictly as an optimization.
2353  *
2354  * For heap inserts, we only need to check for table-level SSI locks. Our
2355  * new tuples can't possibly conflict with existing tuple locks, and heap
2356  * page locks are only consolidated versions of tuple locks; they do not
2357  * lock "gaps" as index page locks do. So we don't need to specify a
2358  * buffer when making the call, which makes for a faster check.
2359  */
2361 
2362  ndone = 0;
2363  while (ndone < ntuples)
2364  {
2365  Buffer buffer;
2366  bool starting_with_empty_page;
2367  bool all_visible_cleared = false;
2368  bool all_frozen_set = false;
2369  int nthispage;
2370 
2372 
2373  /*
2374  * Find buffer where at least the next tuple will fit. If the page is
2375  * all-visible, this will also pin the requisite visibility map page.
2376  *
2377  * Also pin visibility map page if COPY FREEZE inserts tuples into an
2378  * empty page. See all_frozen_set below.
2379  */
2380  buffer = RelationGetBufferForTuple(relation, heaptuples[ndone]->t_len,
2381  InvalidBuffer, options, bistate,
2382  &vmbuffer, NULL);
2383  page = BufferGetPage(buffer);
2384 
2385  starting_with_empty_page = PageGetMaxOffsetNumber(page) == 0;
2386 
2387  if (starting_with_empty_page && (options & HEAP_INSERT_FROZEN))
2388  all_frozen_set = true;
2389 
2390  /* NO EREPORT(ERROR) from here till changes are logged */
2392 
2393  /*
2394  * RelationGetBufferForTuple has ensured that the first tuple fits.
2395  * Put that on the page, and then as many other tuples as fit.
2396  */
2397  RelationPutHeapTuple(relation, buffer, heaptuples[ndone], false);
2398 
2399  /*
2400  * For logical decoding we need combo CIDs to properly decode the
2401  * catalog.
2402  */
2403  if (needwal && need_cids)
2404  log_heap_new_cid(relation, heaptuples[ndone]);
2405 
2406  for (nthispage = 1; ndone + nthispage < ntuples; nthispage++)
2407  {
2408  HeapTuple heaptup = heaptuples[ndone + nthispage];
2409 
2410  if (PageGetHeapFreeSpace(page) < MAXALIGN(heaptup->t_len) + saveFreeSpace)
2411  break;
2412 
2413  RelationPutHeapTuple(relation, buffer, heaptup, false);
2414 
2415  /*
2416  * For logical decoding we need combo CIDs to properly decode the
2417  * catalog.
2418  */
2419  if (needwal && need_cids)
2420  log_heap_new_cid(relation, heaptup);
2421  }
2422 
2423  /*
2424  * If the page is all visible, need to clear that, unless we're only
2425  * going to add further frozen rows to it.
2426  *
2427  * If we're only adding already frozen rows to a previously empty
2428  * page, mark it as all-visible.
2429  */
2430  if (PageIsAllVisible(page) && !(options & HEAP_INSERT_FROZEN))
2431  {
2432  all_visible_cleared = true;
2433  PageClearAllVisible(page);
2434  visibilitymap_clear(relation,
2435  BufferGetBlockNumber(buffer),
2436  vmbuffer, VISIBILITYMAP_VALID_BITS);
2437  }
2438  else if (all_frozen_set)
2439  PageSetAllVisible(page);
2440 
2441  /*
2442  * XXX Should we set PageSetPrunable on this page ? See heap_insert()
2443  */
2444 
2445  MarkBufferDirty(buffer);
2446 
2447  /* XLOG stuff */
2448  if (needwal)
2449  {
2450  XLogRecPtr recptr;
2451  xl_heap_multi_insert *xlrec;
2453  char *tupledata;
2454  int totaldatalen;
2455  char *scratchptr = scratch.data;
2456  bool init;
2457  int bufflags = 0;
2458 
2459  /*
2460  * If the page was previously empty, we can reinit the page
2461  * instead of restoring the whole thing.
2462  */
2463  init = starting_with_empty_page;
2464 
2465  /* allocate xl_heap_multi_insert struct from the scratch area */
2466  xlrec = (xl_heap_multi_insert *) scratchptr;
2467  scratchptr += SizeOfHeapMultiInsert;
2468 
2469  /*
2470  * Allocate offsets array. Unless we're reinitializing the page,
2471  * in that case the tuples are stored in order starting at
2472  * FirstOffsetNumber and we don't need to store the offsets
2473  * explicitly.
2474  */
2475  if (!init)
2476  scratchptr += nthispage * sizeof(OffsetNumber);
2477 
2478  /* the rest of the scratch space is used for tuple data */
2479  tupledata = scratchptr;
2480 
2481  /* check that the mutually exclusive flags are not both set */
2482  Assert(!(all_visible_cleared && all_frozen_set));
2483 
2484  xlrec->flags = 0;
2485  if (all_visible_cleared)
2487  if (all_frozen_set)
2489 
2490  xlrec->ntuples = nthispage;
2491 
2492  /*
2493  * Write out an xl_multi_insert_tuple and the tuple data itself
2494  * for each tuple.
2495  */
2496  for (i = 0; i < nthispage; i++)
2497  {
2498  HeapTuple heaptup = heaptuples[ndone + i];
2499  xl_multi_insert_tuple *tuphdr;
2500  int datalen;
2501 
2502  if (!init)
2503  xlrec->offsets[i] = ItemPointerGetOffsetNumber(&heaptup->t_self);
2504  /* xl_multi_insert_tuple needs two-byte alignment. */
2505  tuphdr = (xl_multi_insert_tuple *) SHORTALIGN(scratchptr);
2506  scratchptr = ((char *) tuphdr) + SizeOfMultiInsertTuple;
2507 
2508  tuphdr->t_infomask2 = heaptup->t_data->t_infomask2;
2509  tuphdr->t_infomask = heaptup->t_data->t_infomask;
2510  tuphdr->t_hoff = heaptup->t_data->t_hoff;
2511 
2512  /* write bitmap [+ padding] [+ oid] + data */
2513  datalen = heaptup->t_len - SizeofHeapTupleHeader;
2514  memcpy(scratchptr,
2515  (char *) heaptup->t_data + SizeofHeapTupleHeader,
2516  datalen);
2517  tuphdr->datalen = datalen;
2518  scratchptr += datalen;
2519  }
2520  totaldatalen = scratchptr - tupledata;
2521  Assert((scratchptr - scratch.data) < BLCKSZ);
2522 
2523  if (need_tuple_data)
2525 
2526  /*
2527  * Signal that this is the last xl_heap_multi_insert record
2528  * emitted by this call to heap_multi_insert(). Needed for logical
2529  * decoding so it knows when to cleanup temporary data.
2530  */
2531  if (ndone + nthispage == ntuples)
2532  xlrec->flags |= XLH_INSERT_LAST_IN_MULTI;
2533 
2534  if (init)
2535  {
2536  info |= XLOG_HEAP_INIT_PAGE;
2537  bufflags |= REGBUF_WILL_INIT;
2538  }
2539 
2540  /*
2541  * If we're doing logical decoding, include the new tuple data
2542  * even if we take a full-page image of the page.
2543  */
2544  if (need_tuple_data)
2545  bufflags |= REGBUF_KEEP_DATA;
2546 
2547  XLogBeginInsert();
2548  XLogRegisterData((char *) xlrec, tupledata - scratch.data);
2549  XLogRegisterBuffer(0, buffer, REGBUF_STANDARD | bufflags);
2550 
2551  XLogRegisterBufData(0, tupledata, totaldatalen);
2552 
2553  /* filtering by origin on a row level is much more efficient */
2555 
2556  recptr = XLogInsert(RM_HEAP2_ID, info);
2557 
2558  PageSetLSN(page, recptr);
2559  }
2560 
2561  END_CRIT_SECTION();
2562 
2563  /*
2564  * If we've frozen everything on the page, update the visibilitymap.
2565  * We're already holding pin on the vmbuffer.
2566  */
2567  if (all_frozen_set)
2568  {
2569  Assert(PageIsAllVisible(page));
2570  Assert(visibilitymap_pin_ok(BufferGetBlockNumber(buffer), vmbuffer));
2571 
2572  /*
2573  * It's fine to use InvalidTransactionId here - this is only used
2574  * when HEAP_INSERT_FROZEN is specified, which intentionally
2575  * violates visibility rules.
2576  */
2577  visibilitymap_set(relation, BufferGetBlockNumber(buffer), buffer,
2578  InvalidXLogRecPtr, vmbuffer,
2581  }
2582 
2583  UnlockReleaseBuffer(buffer);
2584  ndone += nthispage;
2585 
2586  /*
2587  * NB: Only release vmbuffer after inserting all tuples - it's fairly
2588  * likely that we'll insert into subsequent heap pages that are likely
2589  * to use the same vm page.
2590  */
2591  }
2592 
2593  /* We're done with inserting all tuples, so release the last vmbuffer. */
2594  if (vmbuffer != InvalidBuffer)
2595  ReleaseBuffer(vmbuffer);
2596 
2597  /*
2598  * We're done with the actual inserts. Check for conflicts again, to
2599  * ensure that all rw-conflicts in to these inserts are detected. Without
2600  * this final check, a sequential scan of the heap may have locked the
2601  * table after the "before" check, missing one opportunity to detect the
2602  * conflict, and then scanned the table before the new tuples were there,
2603  * missing the other chance to detect the conflict.
2604  *
2605  * For heap inserts, we only need to check for table-level SSI locks. Our
2606  * new tuples can't possibly conflict with existing tuple locks, and heap
2607  * page locks are only consolidated versions of tuple locks; they do not
2608  * lock "gaps" as index page locks do. So we don't need to specify a
2609  * buffer when making the call.
2610  */
2612 
2613  /*
2614  * If tuples are cachable, mark them for invalidation from the caches in
2615  * case we abort. Note it is OK to do this after releasing the buffer,
2616  * because the heaptuples data structure is all in local memory, not in
2617  * the shared buffer.
2618  */
2619  if (IsCatalogRelation(relation))
2620  {
2621  for (i = 0; i < ntuples; i++)
2622  CacheInvalidateHeapTuple(relation, heaptuples[i], NULL);
2623  }
2624 
2625  /* copy t_self fields back to the caller's slots */
2626  for (i = 0; i < ntuples; i++)
2627  slots[i]->tts_tid = heaptuples[i]->t_self;
2628 
2629  pgstat_count_heap_insert(relation, ntuples);
2630 }
void XLogRegisterBufData(uint8 block_id, char *data, int len)
Definition: xloginsert.c:378
Oid tts_tableOid
Definition: tuptable.h:131
#define SizeofHeapTupleHeader
Definition: htup_details.h:184
bool IsCatalogRelation(Relation relation)
Definition: catalog.c:104
static XLogRecPtr log_heap_new_cid(Relation relation, HeapTuple tup)
Definition: heapam.c:8228
#define InvalidXLogRecPtr
Definition: xlogdefs.h:28
void CacheInvalidateHeapTuple(Relation relation, HeapTuple tuple, HeapTuple newtuple)
Definition: inval.c:1186
static HeapTuple heap_prepare_insert(Relation relation, HeapTuple tup, TransactionId xid, CommandId cid, int options)
Definition: heapam.c:2248
#define XLH_INSERT_ALL_FROZEN_SET
Definition: heapam_xlog.h:73
#define PageIsAllVisible(page)
Definition: bufpage.h:385
uint32 TransactionId
Definition: c.h:587
#define HEAP_INSERT_FROZEN
Definition: heapam.h:35
void visibilitymap_set(Relation rel, BlockNumber heapBlk, Buffer heapBuf, XLogRecPtr recptr, Buffer vmBuf, TransactionId cutoff_xid, uint8 flags)
OffsetNumber offsets[FLEXIBLE_ARRAY_MEMBER]
Definition: heapam_xlog.h:179
void MarkBufferDirty(Buffer buffer)
Definition: bufmgr.c:1565
void XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
Definition: xloginsert.c:232
#define VISIBILITYMAP_ALL_FROZEN
Definition: visibilitymap.h:27
#define END_CRIT_SECTION()
Definition: miscadmin.h:149
unsigned char uint8
Definition: c.h:439
#define InvalidBuffer
Definition: buf.h:25
#define SizeOfHeapMultiInsert
Definition: heapam_xlog.h:182
#define REGBUF_WILL_INIT
Definition: xloginsert.h:33
#define START_CRIT_SECTION()
Definition: miscadmin.h:147
#define XLOG_INCLUDE_ORIGIN
Definition: xlog.h:213
void ReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3768
#define RelationIsLogicallyLogged(relation)
Definition: rel.h:674
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
void RelationPutHeapTuple(Relation relation, Buffer buffer, HeapTuple tuple, bool token)
Definition: hio.c:36
#define XLOG_HEAP_INIT_PAGE
Definition: heapam_xlog.h:46
#define XLOG_HEAP2_MULTI_INSERT
Definition: heapam_xlog.h:58
uint16 OffsetNumber
Definition: off.h:24
#define VISIBILITYMAP_VALID_BITS
Definition: visibilitymap.h:28
HeapTupleHeader t_data
Definition: htup.h:68
char data[BLCKSZ]
Definition: c.h:1141
bool visibilitymap_clear(Relation rel, BlockNumber heapBlk, Buffer buf, uint8 flags)
void UnlockReleaseBuffer(Buffer buffer)
Definition: bufmgr.c:3791
Size PageGetHeapFreeSpace(Page page)
Definition: bufpage.c:984
#define XLH_INSERT_CONTAINS_NEW_TUPLE
Definition: heapam_xlog.h:69
ItemPointerData t_self
Definition: htup.h:65
TransactionId GetCurrentTransactionId(void)
Definition: xact.c:438
uint32 t_len
Definition: htup.h:64
#define PageSetAllVisible(page)
Definition: bufpage.h:387
#define REGBUF_STANDARD
Definition: xloginsert.h:35
Buffer RelationGetBufferForTuple(Relation relation, Size len, Buffer otherBuffer, int options, BulkInsertState bistate, Buffer *vmbuffer, Buffer *vmbuffer_other)
Definition: hio.c:333
#define InvalidTransactionId
Definition: transam.h:31
Oid t_tableOid
Definition: htup.h:66
void XLogSetRecordFlags(uint8 flags)
Definition: xloginsert.c:414
#define init()
#define BufferGetPage(buffer)
Definition: bufmgr.h:169
#define AssertArg(condition)
Definition: c.h:806
HeapTuple ExecFetchSlotHeapTuple(TupleTableSlot *slot, bool materialize, bool *shouldFree)
Definition: execTuples.c:1644
#define RelationGetTargetPageFreeSpace(relation, defaultff)
Definition: rel.h:361
#define XLH_INSERT_LAST_IN_MULTI
Definition: heapam_xlog.h:67
void XLogRegisterData(char *data, int len)
Definition: xloginsert.c:340
XLogRecPtr XLogInsert(RmgrId rmid, uint8 info)
Definition: xloginsert.c:432
#define RelationIsAccessibleInLogicalDecoding(relation)
Definition: rel.h:657
#define REGBUF_KEEP_DATA
Definition: xloginsert.h:38
void CheckForSerializableConflictIn(Relation relation, ItemPointer tid, BlockNumber blkno)
Definition: predicate.c:4446
#define PageClearAllVisible(page)
Definition: bufpage.h:389
uint64 XLogRecPtr
Definition: xlogdefs.h:21
#define Assert(condition)
Definition: c.h:804
size_t Size
Definition: c.h:540
#define InvalidBlockNumber
Definition: block.h:33
#define MAXALIGN(LEN)
Definition: c.h:757
#define ItemPointerGetOffsetNumber(pointer)
Definition: itemptr.h:117
#define RelationNeedsWAL(relation)
Definition: rel.h:601
#define VISIBILITYMAP_ALL_VISIBLE
Definition: visibilitymap.h:26
#define XLH_INSERT_ALL_VISIBLE_CLEARED
Definition: heapam_xlog.h:66
BlockNumber BufferGetBlockNumber(Buffer buffer)
Definition: bufmgr.c:2748
void pgstat_count_heap_insert(Relation rel, PgStat_Counter n)
Definition: pgstat.c:2239
void * palloc(Size size)
Definition: mcxt.c:1062
#define HEAP_INSERT_NO_LOGICAL
Definition: heapam.h:36
int i
#define CHECK_FOR_INTERRUPTS()
Definition: miscadmin.h:120
bool visibilitymap_pin_ok(BlockNumber heapBlk, Buffer buf)
#define HEAP_DEFAULT_FILLFACTOR
Definition: rel.h:332
#define SHORTALIGN(LEN)
Definition: c.h:753
void XLogBeginInsert(void)
Definition: xloginsert.c:135
#define PageSetLSN(page, lsn)
Definition: bufpage.h:368
int Buffer
Definition: buf.h:23
#define RelationGetRelid(relation)
Definition: rel.h:477
Pointer Page
Definition: bufpage.h:78
#define SizeOfMultiInsertTuple
Definition: heapam_xlog.h:193

◆ heap_page_prune()

int heap_page_prune ( Relation  relation,
Buffer  buffer,
struct GlobalVisState vistest,
TransactionId  old_snap_xmin,
TimestampTz  old_snap_ts_ts,
bool  report_stats,
OffsetNumber off_loc 
)

Definition at line 219 of file pruneheap.c.

References BufferGetPage, END_CRIT_SECTION, FirstOffsetNumber, heap_page_prune_execute(), heap_prune_chain(), InvalidOffsetNumber, InvalidTransactionId, ItemIdIsDead, ItemIdIsUsed, PruneState::latestRemovedXid, xl_heap_prune::latestRemovedXid, MarkBufferDirty(), MarkBufferDirtyHint(), PruneState::marked, PruneState::ndead, xl_heap_prune::ndead, PruneState::new_prune_xid, PruneState::nowdead, PruneState::nowunused, PruneState::nredirected, xl_heap_prune::nredirected, PruneState::nunused, OffsetNumberNext, PruneState::old_snap_ts, PruneState::old_snap_used, PruneState::old_snap_xmin, PageClearFull, PageGetItemId, PageGetMaxOffsetNumber, PageIsFull, PageSetLSN, pgstat_update_heap_dead_tuples(), PruneState::redirected, REGBUF_STANDARD, PruneState::rel, RelationNeedsWAL, SizeOfHeapPrune, START_CRIT_SECTION, PruneState::vistest, XLOG_HEAP2_PRUNE, XLogBeginInsert(), XLogInsert(), XLogRegisterBufData(), XLogRegisterBuffer(), and XLogRegisterData().

Referenced by heap_page_prune_opt(), and lazy_scan_prune().

225 {
226  int ndeleted = 0;
227  Page page = BufferGetPage(buffer);
228  OffsetNumber offnum,
229  maxoff;
230  PruneState prstate;
231 
232  /*
233  * Our strategy is to scan the page and make lists of items to change,
234  * then apply the changes within a critical section. This keeps as much
235  * logic as possible out of the critical section, and also ensures that
236  * WAL replay will work the same as the normal case.
237  *
238  * First, initialize the new pd_prune_xid value to zero (indicating no
239  * prunable tuples). If we find any tuples which may soon become
240  * prunable, we will save the lowest relevant XID in new_prune_xid. Also
241  * initialize the rest of our working state.
242  */
244  prstate.rel = relation;
245  prstate.vistest = vistest;
246  prstate.old_snap_xmin = old_snap_xmin;
247  prstate.old_snap_ts = old_snap_ts;
248  prstate.old_snap_used = false;
250  prstate.nredirected = prstate.ndead = prstate.nunused = 0;
251  memset(prstate.marked, 0, sizeof(prstate.marked));
252 
253  /* Scan the page */
254  maxoff = PageGetMaxOffsetNumber(page);
255  for (offnum = FirstOffsetNumber;
256  offnum <= maxoff;
257  offnum = OffsetNumberNext(offnum))
258  {
259  ItemId itemid;
260 
261  /* Ignore items already processed as part of an earlier chain */
262  if (prstate.marked[offnum])
263  continue;
264 
265  /*
266  * Set the offset number so that we can display it along with any
267  * error that occurred while processing this tuple.
268  */
269  if (off_loc)
270  *off_loc = offnum;
271 
272  /* Nothing to do if slot is empty or already dead */
273  itemid = PageGetItemId(page, offnum);
274  if (!ItemIdIsUsed(itemid) || ItemIdIsDead(itemid))
275  continue;
276 
277  /* Process this item or chain of items */
278  ndeleted += heap_prune_chain(buffer, offnum, &prstate);
279  }
280 
281  /* Clear the offset information once we have processed the given page. */
282  if (off_loc)
283  *off_loc = InvalidOffsetNumber;
284 
285  /* Any error while applying the changes is critical */
287 
288  /* Have we found any prunable items? */
289  if (prstate.nredirected > 0 || prstate.ndead > 0 || prstate.nunused > 0)
290  {
291  /*
292  * Apply the planned item changes, then repair page fragmentation, and
293  * update the page's hint bit about whether it has free line pointers.
294  */
296  prstate.redirected, prstate.nredirected,
297  prstate.nowdead, prstate.ndead,
298  prstate.nowunused, prstate.nunused);
299 
300  /*
301  * Update the page's pd_prune_xid field to either zero, or the lowest
302  * XID of any soon-prunable tuple.
303  */
304  ((PageHeader) page)->pd_prune_xid = prstate.new_prune_xid;
305 
306  /*
307  * Also clear the "page is full" flag, since there's no point in
308  * repeating the prune/defrag process until something else happens to
309  * the page.
310  */
311  PageClearFull(page);
312 
313  MarkBufferDirty(buffer);
314 
315  /*
316  * Emit a WAL XLOG_HEAP2_PRUNE record showing what we did
317  */
318  if (RelationNeedsWAL(relation))
319  {
320  xl_heap_prune xlrec;
321  XLogRecPtr recptr;
322 
323  xlrec.latestRemovedXid = prstate.latestRemovedXid;
324  xlrec.nredirected = prstate.nredirected;
325  xlrec.ndead = prstate.ndead;
326 
327  XLogBeginInsert();
328  XLogRegisterData((char *) &xlrec, SizeOfHeapPrune);
329 
331 
332  /*
333  * The OffsetNumber arrays are not actually in the buffer, but we
334  * pretend that they are. When XLogInsert stores the whole
335  * buffer, the offset arrays need not be stored too.
336  */
337  if (prstate.nredirected > 0)
338  XLogRegisterBufData(0, (char *) prstate.redirected,
339  prstate.nredirected *
340  sizeof(OffsetNumber) * 2);
341 
342  if (prstate.ndead > 0)
343  XLogRegisterBufData(0, (char *) prstate.nowdead,
344  prstate.ndead * sizeof(OffsetNumber));
345 
346  if (prstate.nunused > 0)
347  XLogRegisterBufData(0, (char *) prstate.nowunused,
348  prstate.nunused * sizeof(OffsetNumber));
349 
350  recptr = XLogInsert(RM_HEAP2_ID, XLOG_HEAP2_PRUNE);
351 
352  PageSetLSN(BufferGetPage(buffer), recptr);
353  }
354  }
355  else
356  {
357  /*
358  * If we didn't prune anything, but have found a new value for the
359  * pd_prune_xid field, update it and mark the buffer dirty. This is
360  * treated as a non-WAL-logged hint.
361  *
362  * Also clear the "page is full" flag if it is set, since there's no
363  * point in repeating the prune/defrag process until something else
364  * happens to the page.
365  */
366  if (((PageHeader) page)->pd_prune_xid != prstate.new_prune_xid ||
367  PageIsFull(page))
368  {
369  ((PageHeader) page)->pd_prune_xid = prstate.new_prune_xid;
370  PageClearFull(page);
371  MarkBufferDirtyHint(buffer, true);
372  }
373  }
374 
376 
377  /*
378  * If requested, report the number of tuples reclaimed to pgstats. This is
379  * ndeleted minus ndead, because we don't want to count a now-DEAD root
380  * item as a deletion for this purpose.
381  */
382  if (report_stats && ndeleted > prstate.ndead)
383  pgstat_update_heap_dead_tuples(relation, ndeleted - prstate.ndead);
384 
385  /*
386  * XXX Should we update the FSM information of this page ?
387  *
388  * There are two schools of thought here. We may not want to update FSM
389  * information so that the page is not used for unrelated UPDATEs/INSERTs
390  * and any free space in this page will remain available for further
391  * UPDATEs in *this* page, thus improving chances for doing HOT updates.
392  *
393  * But for a large table and where a page does not receive further UPDATEs
394  * for a long time, we might waste this space by not updating the FSM
395  * information. The relation may get extended and fragmented further.
396  *
397  * One possibility is to leave "fillfactor" worth of space in this page
398  * and update FSM with the remaining space.
399  */
400 
401  return ndeleted;
402 }
void XLogRegisterBufData(uint8 block_id, char *data, int len)
Definition: xloginsert.c:378
int nredirected
Definition: pruneheap.c:52
static int heap_prune_chain(Buffer buffer, OffsetNumber rootoffnum, PruneState *prstate)
Definition: pruneheap.c:510
void pgstat_update_heap_dead_tuples(Relation rel, int delta)
Definition: pgstat.c:2369
void MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
Definition: bufmgr.c:3838
void MarkBufferDirty(Buffer buffer)
Definition: bufmgr.c:1565
void XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
Definition: xloginsert.c:232
bool old_snap_used
Definition: pruneheap.c:48
Relation rel
Definition: pruneheap.c:33
#define END_CRIT_SECTION()
Definition: miscadmin.h:149
#define ItemIdIsUsed(itemId)
Definition: itemid.h:92
TimestampTz old_snap_ts
Definition: pruneheap.c:46
#define START_CRIT_SECTION()
Definition: miscadmin.h:147
#define XLOG_HEAP2_PRUNE
Definition: heapam_xlog.h:54
OffsetNumber nowdead[MaxHeapTuplesPerPage]
Definition: pruneheap.c:57
#define ItemIdIsDead(itemId)
Definition: itemid.h:113
bool marked[MaxHeapTuplesPerPage+1]
Definition: pruneheap.c:60
#define PageGetMaxOffsetNumber(page)
Definition: bufpage.h:357
TransactionId new_prune_xid
Definit