From 77947c51c08179b8bc12347a7fbcb2c8d7908302 Mon Sep 17 00:00:00 2001
From: Tom Lane <tgl@sss.pgh.pa.us>
Date: Sun, 27 May 2007 03:50:39 +0000
Subject: [PATCH] Fix up pgstats counting of live and dead tuples to recognize
 that committed and aborted transactions have different effects; also teach it
 not to assume that prepared transactions are always committed.

Along the way, simplify the pgstats API by tying counting directly to
Relations; I cannot detect any redeeming social value in having stats
pointers in HeapScanDesc and IndexScanDesc structures.  And fix a few
corner cases in which counts might be missed because the relation's
pgstat_info pointer hadn't been set.
---
 src/backend/access/gin/ginscan.c           |   4 +-
 src/backend/access/gist/gistget.c          |   4 +-
 src/backend/access/hash/hashsearch.c       |   4 +-
 src/backend/access/heap/heapam.c           |  40 +-
 src/backend/access/index/genam.c           |   4 +-
 src/backend/access/index/indexam.c         |  12 +-
 src/backend/access/nbtree/nbtsearch.c      |   4 +-
 src/backend/access/transam/twophase.c      |   5 +-
 src/backend/access/transam/twophase_rmgr.c |  12 +-
 src/backend/access/transam/xact.c          |  15 +-
 src/backend/executor/nodeBitmapHeapscan.c  |   8 +-
 src/backend/postmaster/bgwriter.c          |  14 +-
 src/backend/postmaster/pgstat.c            | 836 +++++++++++++++------
 src/backend/storage/buffer/bufmgr.c        |  12 +-
 src/backend/utils/cache/relcache.c         |   4 +-
 src/include/access/heapam.h                |   6 +-
 src/include/access/relscan.h               |   6 +-
 src/include/access/twophase_rmgr.h         |   5 +-
 src/include/pgstat.h                       | 206 +++--
 src/include/utils/rel.h                    |  15 +-
 20 files changed, 802 insertions(+), 414 deletions(-)

diff --git a/src/backend/access/gin/ginscan.c b/src/backend/access/gin/ginscan.c
index 22896bc5d77..2eb1ba95b4b 100644
--- a/src/backend/access/gin/ginscan.c
+++ b/src/backend/access/gin/ginscan.c
@@ -8,7 +8,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *			$PostgreSQL: pgsql/src/backend/access/gin/ginscan.c,v 1.9 2007/01/31 15:09:45 teodor Exp $
+ *			$PostgreSQL: pgsql/src/backend/access/gin/ginscan.c,v 1.10 2007/05/27 03:50:38 tgl Exp $
  *-------------------------------------------------------------------------
  */
 
@@ -189,7 +189,7 @@ newScanKey(IndexScanDesc scan)
 				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
 				 errmsg("GIN index does not support search with void query")));
 
-	pgstat_count_index_scan(&scan->xs_pgstat_info);
+	pgstat_count_index_scan(scan->indexRelation);
 }
 
 Datum
diff --git a/src/backend/access/gist/gistget.c b/src/backend/access/gist/gistget.c
index 226812322aa..ed839de4034 100644
--- a/src/backend/access/gist/gistget.c
+++ b/src/backend/access/gist/gistget.c
@@ -8,7 +8,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/gist/gistget.c,v 1.65 2007/04/06 22:33:41 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/gist/gistget.c,v 1.66 2007/05/27 03:50:38 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -165,7 +165,7 @@ gistnext(IndexScanDesc scan, ScanDirection dir, ItemPointer tids,
 		stk->next = NULL;
 		stk->block = GIST_ROOT_BLKNO;
 
-		pgstat_count_index_scan(&scan->xs_pgstat_info);
+		pgstat_count_index_scan(scan->indexRelation);
 	}
 	else if (so->curbuf == InvalidBuffer)
 	{
diff --git a/src/backend/access/hash/hashsearch.c b/src/backend/access/hash/hashsearch.c
index 5de0f402297..104a0c14de3 100644
--- a/src/backend/access/hash/hashsearch.c
+++ b/src/backend/access/hash/hashsearch.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/hash/hashsearch.c,v 1.49 2007/05/03 16:45:58 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/hash/hashsearch.c,v 1.50 2007/05/27 03:50:38 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -127,7 +127,7 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
 	ItemPointer current;
 	OffsetNumber offnum;
 
-	pgstat_count_index_scan(&scan->xs_pgstat_info);
+	pgstat_count_index_scan(rel);
 
 	current = &(so->hashso_curpos);
 	ItemPointerSetInvalid(current);
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index ee2be7cfdb1..9edeaff1306 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/heap/heapam.c,v 1.232 2007/04/08 01:26:27 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/heap/heapam.c,v 1.233 2007/05/27 03:50:38 tgl Exp $
  *
  *
  * INTERFACE ROUTINES
@@ -100,7 +100,7 @@ initscan(HeapScanDesc scan, ScanKey key)
 	if (key != NULL)
 		memcpy(scan->rs_key, key, scan->rs_nkeys * sizeof(ScanKeyData));
 
-	pgstat_count_heap_scan(&scan->rs_pgstat_info);
+	pgstat_count_heap_scan(scan->rs_rd);
 }
 
 /*
@@ -701,6 +701,8 @@ relation_open(Oid relationId, LOCKMODE lockmode)
 	if (!RelationIsValid(r))
 		elog(ERROR, "could not open relation with OID %u", relationId);
 
+	pgstat_initstats(r);
+
 	return r;
 }
 
@@ -743,6 +745,8 @@ try_relation_open(Oid relationId, LOCKMODE lockmode)
 	if (!RelationIsValid(r))
 		elog(ERROR, "could not open relation with OID %u", relationId);
 
+	pgstat_initstats(r);
+
 	return r;
 }
 
@@ -787,6 +791,8 @@ relation_open_nowait(Oid relationId, LOCKMODE lockmode)
 	if (!RelationIsValid(r))
 		elog(ERROR, "could not open relation with OID %u", relationId);
 
+	pgstat_initstats(r);
+
 	return r;
 }
 
@@ -873,8 +879,6 @@ heap_open(Oid relationId, LOCKMODE lockmode)
 				 errmsg("\"%s\" is a composite type",
 						RelationGetRelationName(r))));
 
-	pgstat_initstats(&r->pgstat_info, r);
-
 	return r;
 }
 
@@ -903,8 +907,6 @@ heap_openrv(const RangeVar *relation, LOCKMODE lockmode)
 				 errmsg("\"%s\" is a composite type",
 						RelationGetRelationName(r))));
 
-	pgstat_initstats(&r->pgstat_info, r);
-
 	return r;
 }
 
@@ -954,8 +956,6 @@ heap_beginscan(Relation relation, Snapshot snapshot,
 	else
 		scan->rs_key = NULL;
 
-	pgstat_initstats(&scan->rs_pgstat_info, relation);
-
 	initscan(scan, key);
 
 	return scan;
@@ -1059,7 +1059,7 @@ heap_getnext(HeapScanDesc scan, ScanDirection direction)
 	 */
 	HEAPDEBUG_3;				/* heap_getnext returning tuple */
 
-	pgstat_count_heap_getnext(&scan->rs_pgstat_info);
+	pgstat_count_heap_getnext(scan->rs_rd);
 
 	return &(scan->rs_ctup);
 }
@@ -1086,6 +1086,10 @@ heap_getnext(HeapScanDesc scan, ScanDirection direction)
  * and return it in *userbuf (so the caller must eventually unpin it); when
  * keep_buf = false, the pin is released and *userbuf is set to InvalidBuffer.
  *
+ * stats_relation is the relation to charge the heap_fetch operation against
+ * for statistical purposes.  (This could be the heap rel itself, an
+ * associated index, or NULL to not count the fetch at all.)
+ *
  * It is somewhat inconsistent that we ereport() on invalid block number but
  * return false on invalid item number.  There are a couple of reasons though.
  * One is that the caller can relatively easily check the block number for
@@ -1101,12 +1105,12 @@ heap_fetch(Relation relation,
 		   HeapTuple tuple,
 		   Buffer *userbuf,
 		   bool keep_buf,
-		   PgStat_Info *pgstat_info)
+		   Relation stats_relation)
 {
 	/* Assume *userbuf is undefined on entry */
 	*userbuf = InvalidBuffer;
 	return heap_release_fetch(relation, snapshot, tuple,
-							  userbuf, keep_buf, pgstat_info);
+							  userbuf, keep_buf, stats_relation);
 }
 
 /*
@@ -1125,7 +1129,7 @@ heap_release_fetch(Relation relation,
 				   HeapTuple tuple,
 				   Buffer *userbuf,
 				   bool keep_buf,
-				   PgStat_Info *pgstat_info)
+				   Relation stats_relation)
 {
 	ItemPointer tid = &(tuple->t_self);
 	ItemId		lp;
@@ -1210,9 +1214,9 @@ heap_release_fetch(Relation relation,
 		 */
 		*userbuf = buffer;
 
-		/* Count the successful fetch in *pgstat_info, if given. */
-		if (pgstat_info != NULL)
-			pgstat_count_heap_fetch(pgstat_info);
+		/* Count the successful fetch against appropriate rel, if any */
+		if (stats_relation != NULL)
+			pgstat_count_heap_fetch(stats_relation);
 
 		return true;
 	}
@@ -1517,7 +1521,7 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
 	 */
 	CacheInvalidateHeapTuple(relation, heaptup);
 
-	pgstat_count_heap_insert(&relation->pgstat_info);
+	pgstat_count_heap_insert(relation);
 
 	/*
 	 * If heaptup is a private copy, release it.  Don't forget to copy t_self
@@ -1807,7 +1811,7 @@ heap_delete(Relation relation, ItemPointer tid,
 	if (have_tuple_lock)
 		UnlockTuple(relation, &(tp.t_self), ExclusiveLock);
 
-	pgstat_count_heap_delete(&relation->pgstat_info);
+	pgstat_count_heap_delete(relation);
 
 	return HeapTupleMayBeUpdated;
 }
@@ -2269,7 +2273,7 @@ heap_update(Relation relation, ItemPointer otid, HeapTuple newtup,
 	if (have_tuple_lock)
 		UnlockTuple(relation, &(oldtup.t_self), ExclusiveLock);
 
-	pgstat_count_heap_update(&relation->pgstat_info);
+	pgstat_count_heap_update(relation);
 
 	/*
 	 * If heaptup is a private copy, release it.  Don't forget to copy t_self
diff --git a/src/backend/access/index/genam.c b/src/backend/access/index/genam.c
index 49ffff6e51d..0009739180c 100644
--- a/src/backend/access/index/genam.c
+++ b/src/backend/access/index/genam.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/index/genam.c,v 1.61 2007/01/20 18:43:35 neilc Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/index/genam.c,v 1.62 2007/05/27 03:50:38 tgl Exp $
  *
  * NOTES
  *	  many of the old access method routines have been turned into
@@ -96,8 +96,6 @@ RelationGetIndexScan(Relation indexRelation,
 	scan->xs_ctup.t_data = NULL;
 	scan->xs_cbuf = InvalidBuffer;
 
-	pgstat_initstats(&scan->xs_pgstat_info, indexRelation);
-
 	/*
 	 * Let the AM fill in the key and any opaque data it wants.
 	 */
diff --git a/src/backend/access/index/indexam.c b/src/backend/access/index/indexam.c
index 23522ba740e..d905013a5fc 100644
--- a/src/backend/access/index/indexam.c
+++ b/src/backend/access/index/indexam.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/index/indexam.c,v 1.97 2007/01/05 22:19:23 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/index/indexam.c,v 1.98 2007/05/27 03:50:38 tgl Exp $
  *
  * INTERFACE ROUTINES
  *		index_open		- open an index relation by relation OID
@@ -145,8 +145,6 @@ index_open(Oid relationId, LOCKMODE lockmode)
 				 errmsg("\"%s\" is not an index",
 						RelationGetRelationName(r))));
 
-	pgstat_initstats(&r->pgstat_info, r);
-
 	return r;
 }
 
@@ -433,14 +431,14 @@ index_getnext(IndexScanDesc scan, ScanDirection direction)
 			return NULL;		/* failure exit */
 		}
 
-		pgstat_count_index_tuples(&scan->xs_pgstat_info, 1);
+		pgstat_count_index_tuples(scan->indexRelation, 1);
 
 		/*
 		 * Fetch the heap tuple and see if it matches the snapshot.
 		 */
 		if (heap_release_fetch(scan->heapRelation, scan->xs_snapshot,
 							   heapTuple, &scan->xs_cbuf, true,
-							   &scan->xs_pgstat_info))
+							   scan->indexRelation))
 			break;
 
 		/* Skip if no undeleted tuple at this location */
@@ -502,7 +500,7 @@ index_getnext_indexitem(IndexScanDesc scan,
 									   Int32GetDatum(direction)));
 
 	if (found)
-		pgstat_count_index_tuples(&scan->xs_pgstat_info, 1);
+		pgstat_count_index_tuples(scan->indexRelation, 1);
 
 	return found;
 }
@@ -543,7 +541,7 @@ index_getmulti(IndexScanDesc scan,
 									   Int32GetDatum(max_tids),
 									   PointerGetDatum(returned_tids)));
 
-	pgstat_count_index_tuples(&scan->xs_pgstat_info, *returned_tids);
+	pgstat_count_index_tuples(scan->indexRelation, *returned_tids);
 
 	return found;
 }
diff --git a/src/backend/access/nbtree/nbtsearch.c b/src/backend/access/nbtree/nbtsearch.c
index 036a97a8d04..b947d770aa2 100644
--- a/src/backend/access/nbtree/nbtsearch.c
+++ b/src/backend/access/nbtree/nbtsearch.c
@@ -8,7 +8,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/nbtree/nbtsearch.c,v 1.112 2007/04/06 22:33:42 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/nbtree/nbtsearch.c,v 1.113 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -453,7 +453,7 @@ _bt_first(IndexScanDesc scan, ScanDirection dir)
 	int			i;
 	StrategyNumber strat_total;
 
-	pgstat_count_index_scan(&scan->xs_pgstat_info);
+	pgstat_count_index_scan(rel);
 
 	/*
 	 * Examine the scan keys and eliminate any redundant keys; also mark the
diff --git a/src/backend/access/transam/twophase.c b/src/backend/access/transam/twophase.c
index 6f495a84087..7fdf5a7eed3 100644
--- a/src/backend/access/transam/twophase.c
+++ b/src/backend/access/transam/twophase.c
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *		$PostgreSQL: pgsql/src/backend/access/transam/twophase.c,v 1.30 2007/04/30 21:01:52 tgl Exp $
+ *		$PostgreSQL: pgsql/src/backend/access/transam/twophase.c,v 1.31 2007/05/27 03:50:39 tgl Exp $
  *
  * NOTES
  *		Each global transaction is associated with a global transaction
@@ -1211,7 +1211,8 @@ FinishPreparedTransaction(const char *gid, bool isCommit)
 	else
 		ProcessRecords(bufptr, xid, twophase_postabort_callbacks);
 
-	pgstat_count_xact_commit();
+	/* Count the prepared xact as committed or aborted */
+	AtEOXact_PgStat(isCommit);
 
 	/*
 	 * And now we can clean up our mess.
diff --git a/src/backend/access/transam/twophase_rmgr.c b/src/backend/access/transam/twophase_rmgr.c
index e93bac7b2d8..9c2f14a1a38 100644
--- a/src/backend/access/transam/twophase_rmgr.c
+++ b/src/backend/access/transam/twophase_rmgr.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/transam/twophase_rmgr.c,v 1.4 2007/01/05 22:19:23 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/transam/twophase_rmgr.c,v 1.5 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -16,6 +16,7 @@
 
 #include "access/twophase_rmgr.h"
 #include "commands/async.h"
+#include "pgstat.h"
 #include "storage/lock.h"
 #include "utils/flatfiles.h"
 #include "utils/inval.h"
@@ -27,7 +28,8 @@ const TwoPhaseCallback twophase_recover_callbacks[TWOPHASE_RM_MAX_ID + 1] =
 	lock_twophase_recover,		/* Lock */
 	NULL,						/* Inval */
 	NULL,						/* flat file update */
-	NULL						/* notify/listen */
+	NULL,						/* notify/listen */
+	NULL						/* pgstat */
 };
 
 const TwoPhaseCallback twophase_postcommit_callbacks[TWOPHASE_RM_MAX_ID + 1] =
@@ -36,7 +38,8 @@ const TwoPhaseCallback twophase_postcommit_callbacks[TWOPHASE_RM_MAX_ID + 1] =
 	lock_twophase_postcommit,	/* Lock */
 	inval_twophase_postcommit,	/* Inval */
 	flatfile_twophase_postcommit,		/* flat file update */
-	notify_twophase_postcommit	/* notify/listen */
+	notify_twophase_postcommit,	/* notify/listen */
+	pgstat_twophase_postcommit	/* pgstat */
 };
 
 const TwoPhaseCallback twophase_postabort_callbacks[TWOPHASE_RM_MAX_ID + 1] =
@@ -45,5 +48,6 @@ const TwoPhaseCallback twophase_postabort_callbacks[TWOPHASE_RM_MAX_ID + 1] =
 	lock_twophase_postabort,	/* Lock */
 	NULL,						/* Inval */
 	NULL,						/* flat file update */
-	NULL						/* notify/listen */
+	NULL,						/* notify/listen */
+	pgstat_twophase_postabort	/* pgstat */
 };
diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index f2685ee0b34..c16b4fa6be9 100644
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -10,7 +10,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/transam/xact.c,v 1.242 2007/04/30 21:01:52 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/transam/xact.c,v 1.243 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1661,8 +1661,7 @@ CommitTransaction(void)
 	AtEOXact_Files();
 	AtEOXact_ComboCid();
 	AtEOXact_HashTables(true);
-	pgstat_clear_snapshot();
-	pgstat_count_xact_commit();
+	AtEOXact_PgStat(true);
 	pgstat_report_txn_timestamp(0);
 
 	CurrentResourceOwner = NULL;
@@ -1796,6 +1795,7 @@ PrepareTransaction(void)
 	AtPrepare_UpdateFlatFiles();
 	AtPrepare_Inval();
 	AtPrepare_Locks();
+	AtPrepare_PgStat();
 
 	/*
 	 * Here is where we really truly prepare.
@@ -1853,6 +1853,8 @@ PrepareTransaction(void)
 
 	/* notify and flatfiles don't need a postprepare call */
 
+	PostPrepare_PgStat();
+
 	PostPrepare_Inval();
 
 	PostPrepare_smgr();
@@ -1880,7 +1882,7 @@ PrepareTransaction(void)
 	AtEOXact_Files();
 	AtEOXact_ComboCid();
 	AtEOXact_HashTables(true);
-	pgstat_clear_snapshot();
+	/* don't call AtEOXact_PgStat here */
 
 	CurrentResourceOwner = NULL;
 	ResourceOwnerDelete(TopTransactionResourceOwner);
@@ -2035,8 +2037,7 @@ AbortTransaction(void)
 	AtEOXact_Files();
 	AtEOXact_ComboCid();
 	AtEOXact_HashTables(false);
-	pgstat_clear_snapshot();
-	pgstat_count_xact_rollback();
+	AtEOXact_PgStat(false);
 	pgstat_report_txn_timestamp(0);
 
 	/*
@@ -3749,6 +3750,7 @@ CommitSubTransaction(void)
 	AtEOSubXact_Files(true, s->subTransactionId,
 					  s->parent->subTransactionId);
 	AtEOSubXact_HashTables(true, s->nestingLevel);
+	AtEOSubXact_PgStat(true, s->nestingLevel);
 
 	/*
 	 * We need to restore the upper transaction's read-only state, in case the
@@ -3861,6 +3863,7 @@ AbortSubTransaction(void)
 		AtEOSubXact_Files(false, s->subTransactionId,
 						  s->parent->subTransactionId);
 		AtEOSubXact_HashTables(false, s->nestingLevel);
+		AtEOSubXact_PgStat(false, s->nestingLevel);
 	}
 
 	/*
diff --git a/src/backend/executor/nodeBitmapHeapscan.c b/src/backend/executor/nodeBitmapHeapscan.c
index 3e9a91de2f5..07729da2be6 100644
--- a/src/backend/executor/nodeBitmapHeapscan.c
+++ b/src/backend/executor/nodeBitmapHeapscan.c
@@ -21,7 +21,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/executor/nodeBitmapHeapscan.c,v 1.16 2007/01/05 22:19:28 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/executor/nodeBitmapHeapscan.c,v 1.17 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -189,7 +189,7 @@ BitmapHeapNext(BitmapHeapScanState *node)
 		scan->rs_ctup.t_len = ItemIdGetLength(lp);
 		ItemPointerSet(&scan->rs_ctup.t_self, tbmres->blockno, targoffset);
 
-		pgstat_count_heap_fetch(&scan->rs_pgstat_info);
+		pgstat_count_heap_fetch(scan->rs_rd);
 
 		/*
 		 * Set up the result slot to point to this tuple. Note that the slot
@@ -389,7 +389,7 @@ ExecBitmapHeapReScan(BitmapHeapScanState *node, ExprContext *exprCtxt)
 	heap_rescan(node->ss.ss_currentScanDesc, NULL);
 
 	/* undo bogus "seq scan" count (see notes in ExecInitBitmapHeapScan) */
-	pgstat_discount_heap_scan(&node->ss.ss_currentScanDesc->rs_pgstat_info);
+	pgstat_discount_heap_scan(node->ss.ss_currentScanDesc->rs_rd);
 
 	if (node->tbm)
 		tbm_free(node->tbm);
@@ -535,7 +535,7 @@ ExecInitBitmapHeapScan(BitmapHeapScan *node, EState *estate, int eflags)
 	 * when we actually aren't doing any such thing.  Reverse out the added
 	 * scan count.	(Eventually we may want to count bitmap scans separately.)
 	 */
-	pgstat_discount_heap_scan(&scanstate->ss.ss_currentScanDesc->rs_pgstat_info);
+	pgstat_discount_heap_scan(scanstate->ss.ss_currentScanDesc->rs_rd);
 
 	/*
 	 * get the scan type from the relation descriptor.
diff --git a/src/backend/postmaster/bgwriter.c b/src/backend/postmaster/bgwriter.c
index 273588424eb..10f57f00b8f 100644
--- a/src/backend/postmaster/bgwriter.c
+++ b/src/backend/postmaster/bgwriter.c
@@ -37,7 +37,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/postmaster/bgwriter.c,v 1.37 2007/03/30 18:34:55 mha Exp $
+ *	  $PostgreSQL: pgsql/src/backend/postmaster/bgwriter.c,v 1.38 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -125,13 +125,6 @@ typedef struct
 
 static BgWriterShmemStruct *BgWriterShmem;
 
-/*
- * BgWriter statistics counters.
- * Stored directly in a stats message structure so it can be sent
- * without needing to copy things around.
- */
-PgStat_MsgBgWriter BgWriterStats;
-
 /*
  * GUC parameters
  */
@@ -250,11 +243,6 @@ BackgroundWriterMain(void)
 											 ALLOCSET_DEFAULT_MAXSIZE);
 	MemoryContextSwitchTo(bgwriter_context);
 
-	/*
-	 * Initialize statistics counters to zero
-	 */
-	memset(&BgWriterStats, 0, sizeof(BgWriterStats));
-
 	/*
 	 * If an exception is encountered, processing resumes here.
 	 *
diff --git a/src/backend/postmaster/pgstat.c b/src/backend/postmaster/pgstat.c
index 1fac5af284b..b41a16de44c 100644
--- a/src/backend/postmaster/pgstat.c
+++ b/src/backend/postmaster/pgstat.c
@@ -13,7 +13,7 @@
  *
  *	Copyright (c) 2001-2007, PostgreSQL Global Development Group
  *
- *	$PostgreSQL: pgsql/src/backend/postmaster/pgstat.c,v 1.155 2007/04/30 16:37:08 tgl Exp $
+ *	$PostgreSQL: pgsql/src/backend/postmaster/pgstat.c,v 1.156 2007/05/27 03:50:39 tgl Exp $
  * ----------
  */
 #include "postgres.h"
@@ -39,6 +39,7 @@
 
 #include "access/heapam.h"
 #include "access/transam.h"
+#include "access/twophase_rmgr.h"
 #include "access/xact.h"
 #include "catalog/pg_database.h"
 #include "libpq/ip.h"
@@ -98,6 +99,13 @@ bool		pgstat_collect_tuplelevel = false;
 bool		pgstat_collect_blocklevel = false;
 bool		pgstat_collect_querystring = false;
 
+/*
+ * BgWriter global statistics counters (unused in other processes).
+ * Stored directly in a stats message structure so it can be sent
+ * without needing to copy things around.  We assume this inits to zeroes.
+ */
+PgStat_MsgBgWriter BgWriterStats;
+
 /* ----------
  * Local data
  * ----------
@@ -111,43 +119,63 @@ static time_t last_pgstat_start_time;
 static bool pgStatRunningInCollector = false;
 
 /*
- * Place where backends store per-table info to be sent to the collector.
- * We store shared relations separately from non-shared ones, to be able to
- * send them in separate messages.
+ * Structures in which backends store per-table info that's waiting to be
+ * sent to the collector.
  *
- * NOTE: once allocated, a PgStat_MsgTabstat struct belonging to a
- * TabStatArray is never moved or deleted for the life of the backend.
- * Also, we zero out the t_id fields of the contained PgStat_TableEntry
- * structs whenever they are not actively in use.  This allows PgStat_Info
- * pointers to be treated as long-lived data, avoiding repeated searches in
- * pgstat_initstats() when a relation is repeatedly heap_open'd or
- * index_open'd during a transaction.
+ * NOTE: once allocated, TabStatusArray structures are never moved or deleted
+ * for the life of the backend.  Also, we zero out the t_id fields of the
+ * contained PgStat_TableStatus structs whenever they are not actively in use.
+ * This allows relcache pgstat_info pointers to be treated as long-lived data,
+ * avoiding repeated searches in pgstat_initstats() when a relation is
+ * repeatedly opened during a transaction.
  */
-typedef struct TabStatArray
+#define TABSTAT_QUANTUM		100			/* we alloc this many at a time */
+
+typedef struct TabStatusArray
 {
-	int			tsa_alloc;		/* num allocated */
-	int			tsa_used;		/* num actually used */
-	PgStat_MsgTabstat **tsa_messages;	/* the array itself */
-} TabStatArray;
+	struct TabStatusArray *tsa_next;	/* link to next array, if any */
+	int			tsa_used;				/* # entries currently used */
+	PgStat_TableStatus tsa_entries[TABSTAT_QUANTUM];	/* per-table data */
+} TabStatusArray;
 
-#define TABSTAT_QUANTUM		4	/* we alloc this many at a time */
+static TabStatusArray *pgStatTabList = NULL;
+
+/*
+ * Tuple insertion/deletion counts for an open transaction can't be propagated
+ * into PgStat_TableStatus counters until we know if it is going to commit
+ * or abort.  Hence, we keep these counts in per-subxact structs that live
+ * in TopTransactionContext.  This data structure is designed on the assumption
+ * that subxacts won't usually modify very many tables.
+ */
+typedef struct PgStat_SubXactStatus
+{
+	int			nest_level;				/* subtransaction nest level */
+	struct PgStat_SubXactStatus *prev;	/* higher-level subxact if any */
+	PgStat_TableXactStatus *first;		/* head of list for this subxact */
+} PgStat_SubXactStatus;
 
-static TabStatArray RegularTabStat = {0, 0, NULL};
-static TabStatArray SharedTabStat = {0, 0, NULL};
+static PgStat_SubXactStatus *pgStatXactStack = NULL;
 
 static int	pgStatXactCommit = 0;
 static int	pgStatXactRollback = 0;
 
+/* Record that's written to 2PC state file when pgstat state is persisted */
+typedef struct TwoPhasePgStatRecord
+{
+	PgStat_Counter tuples_inserted;	/* tuples inserted in xact */
+	PgStat_Counter tuples_deleted;	/* tuples deleted in xact */
+	Oid			t_id;				/* table's OID */
+	bool		t_shared;			/* is it a shared catalog? */
+} TwoPhasePgStatRecord;
+
+/*
+ * Info about current "snapshot" of stats file
+ */
 static MemoryContext pgStatLocalContext = NULL;
 static HTAB *pgStatDBHash = NULL;
 static PgBackendStatus *localBackendStatusTable = NULL;
 static int	localNumBackends = 0;
 
-/*
- * BgWriter global statistics counters, from bgwriter.c
- */
-extern PgStat_MsgBgWriter BgWriterStats;
-
 /*
  * Cluster wide statistics, kept in the stats collector.
  * Contains statistics that are not collected per database
@@ -177,9 +205,12 @@ static void pgstat_write_statsfile(void);
 static HTAB *pgstat_read_statsfile(Oid onlydb);
 static void backend_read_statsfile(void);
 static void pgstat_read_current_status(void);
-static void pgstat_report_one_tabstat(TabStatArray *tsarr, Oid dbid);
+
+static void pgstat_send_tabstat(PgStat_MsgTabstat *tsmsg);
 static HTAB *pgstat_collect_oids(Oid catalogid);
 
+static PgStat_TableStatus *get_tabstat_entry(Oid rel_id, bool isshared);
+
 static void pgstat_setup_memcxt(void);
 
 static void pgstat_setheader(PgStat_MsgHdr *hdr, StatMsgType mtype);
@@ -617,12 +648,19 @@ void allow_immediate_pgstat_restart(void)
 void
 pgstat_report_tabstat(bool force)
 {
+	/* we assume this inits to all zeroes: */
+	static const PgStat_TableCounts all_zeroes;
 	static TimestampTz last_report = 0;	
+
 	TimestampTz now;
+	PgStat_MsgTabstat regular_msg;
+	PgStat_MsgTabstat shared_msg;
+	TabStatusArray *tsa;
+	int			i;
 
 	/* Don't expend a clock check if nothing to do */
-	if (RegularTabStat.tsa_used == 0 &&
-		SharedTabStat.tsa_used == 0)
+	if (pgStatTabList == NULL ||
+		pgStatTabList->tsa_used == 0)
 		return;
 
 	/*
@@ -636,51 +674,101 @@ pgstat_report_tabstat(bool force)
 	last_report = now;
 
 	/*
-	 * For each message buffer used during the last queries, set the header
-	 * fields and send it out; then mark the entries unused.
+	 * Scan through the TabStatusArray struct(s) to find tables that actually
+	 * have counts, and build messages to send.  We have to separate shared
+	 * relations from regular ones because the databaseid field in the
+	 * message header has to depend on that.
 	 */
-	pgstat_report_one_tabstat(&RegularTabStat, MyDatabaseId);
-	pgstat_report_one_tabstat(&SharedTabStat, InvalidOid);
+	regular_msg.m_databaseid = MyDatabaseId;
+	shared_msg.m_databaseid = InvalidOid;
+	regular_msg.m_nentries = 0;
+	shared_msg.m_nentries = 0;
+
+	for (tsa = pgStatTabList; tsa != NULL; tsa = tsa->tsa_next)
+	{
+		for (i = 0; i < tsa->tsa_used; i++)
+		{
+			PgStat_TableStatus *entry = &tsa->tsa_entries[i];
+			PgStat_MsgTabstat *this_msg;
+			PgStat_TableEntry *this_ent;
+
+			/* Shouldn't have any pending transaction-dependent counts */
+			Assert(entry->trans == NULL);
+
+			/*
+			 * Ignore entries that didn't accumulate any actual counts,
+			 * such as indexes that were opened by the planner but not used.
+			 */
+			if (memcmp(&entry->t_counts, &all_zeroes,
+					   sizeof(PgStat_TableCounts)) == 0)
+				continue;
+			/*
+			 * OK, insert data into the appropriate message, and send if full.
+			 */
+			this_msg = entry->t_shared ? &shared_msg : &regular_msg;
+			this_ent = &this_msg->m_entry[this_msg->m_nentries];
+			this_ent->t_id = entry->t_id;
+			memcpy(&this_ent->t_counts, &entry->t_counts,
+				   sizeof(PgStat_TableCounts));
+			if (++this_msg->m_nentries >= PGSTAT_NUM_TABENTRIES)
+			{
+				pgstat_send_tabstat(this_msg);
+				this_msg->m_nentries = 0;
+			}
+		}
+		/* zero out TableStatus structs after use */
+		MemSet(tsa->tsa_entries, 0,
+			   tsa->tsa_used * sizeof(PgStat_TableStatus));
+		tsa->tsa_used = 0;
+	}
+
+	/*
+	 * Send partial messages.  If force is true, make sure that any pending
+	 * xact commit/abort gets counted, even if no table stats to send.
+	 */
+	if (regular_msg.m_nentries > 0 ||
+		(force && (pgStatXactCommit > 0 || pgStatXactRollback > 0)))
+		pgstat_send_tabstat(&regular_msg);
+	if (shared_msg.m_nentries > 0)
+		pgstat_send_tabstat(&shared_msg);
 }
 
+/*
+ * Subroutine for pgstat_report_tabstat: finish and send a tabstat message
+ */
 static void
-pgstat_report_one_tabstat(TabStatArray *tsarr, Oid dbid)
+pgstat_send_tabstat(PgStat_MsgTabstat *tsmsg)
 {
-	int			i;
-
-	for (i = 0; i < tsarr->tsa_used; i++)
-	{
-		PgStat_MsgTabstat *tsmsg = tsarr->tsa_messages[i];
-		int			n;
-		int			len;
+	int			n;
+	int			len;
 
-		n = tsmsg->m_nentries;
-		len = offsetof(PgStat_MsgTabstat, m_entry[0]) +
-			n * sizeof(PgStat_TableEntry);
+	/* It's unlikely we'd get here with no socket, but maybe not impossible */
+	if (pgStatSock < 0)
+		return;
 
+	/*
+	 * Report accumulated xact commit/rollback whenever we send a normal
+	 * tabstat message
+	 */
+	if (OidIsValid(tsmsg->m_databaseid))
+	{
 		tsmsg->m_xact_commit = pgStatXactCommit;
 		tsmsg->m_xact_rollback = pgStatXactRollback;
 		pgStatXactCommit = 0;
 		pgStatXactRollback = 0;
+	}
+	else
+	{
+		tsmsg->m_xact_commit = 0;
+		tsmsg->m_xact_rollback = 0;
+	}
 
-		/*
-		 * It's unlikely we'd get here with no socket, but maybe not
-		 * impossible
-		 */
-		if (pgStatSock >= 0)
-		{
-			pgstat_setheader(&tsmsg->m_hdr, PGSTAT_MTYPE_TABSTAT);
-			tsmsg->m_databaseid = dbid;
-			pgstat_send(tsmsg, len);
-		}
+	n = tsmsg->m_nentries;
+	len = offsetof(PgStat_MsgTabstat, m_entry[0]) +
+		n * sizeof(PgStat_TableEntry);
 
-		/*
-		 * Zero out the entries, to mark them unused and prepare them
-		 * for next use.
-		 */
-		MemSet(tsmsg, 0, len);
-	}
-	tsarr->tsa_used = 0;
+	pgstat_setheader(&tsmsg->m_hdr, PGSTAT_MTYPE_TABSTAT);
+	pgstat_send(tsmsg, len);
 }
 
 
@@ -1016,209 +1104,489 @@ pgstat_ping(void)
 	pgstat_send(&msg, sizeof(msg));
 }
 
-/*
- * Enlarge a TabStatArray
- */
-static void
-more_tabstat_space(TabStatArray *tsarr)
-{
-	PgStat_MsgTabstat *newMessages;
-	PgStat_MsgTabstat **msgArray;
-	int			newAlloc;
-	int			i;
-
-	AssertArg(PointerIsValid(tsarr));
-
-	newAlloc = tsarr->tsa_alloc + TABSTAT_QUANTUM;
-
-	/* Create (another) quantum of message buffers, and zero them */
-	newMessages = (PgStat_MsgTabstat *)
-		MemoryContextAllocZero(TopMemoryContext,
-							   sizeof(PgStat_MsgTabstat) * TABSTAT_QUANTUM);
-
-	/* Create or enlarge the pointer array */
-	if (tsarr->tsa_messages == NULL)
-		msgArray = (PgStat_MsgTabstat **)
-			MemoryContextAlloc(TopMemoryContext,
-							   sizeof(PgStat_MsgTabstat *) * newAlloc);
-	else
-		msgArray = (PgStat_MsgTabstat **)
-			repalloc(tsarr->tsa_messages,
-					 sizeof(PgStat_MsgTabstat *) * newAlloc);
-
-	for (i = 0; i < TABSTAT_QUANTUM; i++)
-		msgArray[tsarr->tsa_alloc + i] = newMessages++;
-	tsarr->tsa_messages = msgArray;
-	tsarr->tsa_alloc = newAlloc;
-
-	Assert(tsarr->tsa_used < tsarr->tsa_alloc);
-}
 
 /* ----------
  * pgstat_initstats() -
  *
- *	Called from various places usually dealing with initialization
- *	of Relation or Scan structures. The data placed into these
- *	structures from here tell where later to count for buffer reads,
- *	scans and tuples fetched.
- *
- *	NOTE: PgStat_Info pointers in scan structures are really redundant
- *	with those in relcache entries.  The passed stats pointer might point
- *	either to the Relation struct's own pgstat_info field, or to one in
- *	a scan structure; we'll set the Relation pg_statinfo and copy it to
- *	the scan struct.
+ *	Initialize a relcache entry to count access statistics.
+ *	Called whenever a relation is opened.
  *
  *	We assume that a relcache entry's pgstat_info field is zeroed by
  *	relcache.c when the relcache entry is made; thereafter it is long-lived
- *	data.  We can avoid repeated searches of the TabStat arrays when the
+ *	data.  We can avoid repeated searches of the TabStatus arrays when the
  *	same relation is touched repeatedly within a transaction.
  * ----------
  */
 void
-pgstat_initstats(PgStat_Info *stats, Relation rel)
+pgstat_initstats(Relation rel)
 {
 	Oid			rel_id = rel->rd_id;
-	PgStat_TableEntry *useent;
-	TabStatArray *tsarr;
-	PgStat_MsgTabstat *tsmsg;
-	int			mb;
-	int			i;
+	char		relkind = rel->rd_rel->relkind;
+
+	/* We only count stats for things that have storage */
+	if (!(relkind == RELKIND_RELATION ||
+		  relkind == RELKIND_INDEX ||
+		  relkind == RELKIND_TOASTVALUE))
+	{
+		rel->pgstat_info = NULL;
+		return;
+	}
 
 	if (pgStatSock < 0 ||
 		!(pgstat_collect_tuplelevel ||
 		  pgstat_collect_blocklevel))
 	{
-		/* We're not counting at all. */
-		stats->tabentry = NULL;
+		/* We're not counting at all */
+		rel->pgstat_info = NULL;
 		return;
 	}
 
 	/*
 	 * If we already set up this relation in the current transaction,
-	 * just copy the pointer.
+	 * nothing to do.
 	 */
-	if (rel->pgstat_info.tabentry != NULL &&
-		((PgStat_TableEntry *) rel->pgstat_info.tabentry)->t_id == rel_id)
-	{
-		stats->tabentry = rel->pgstat_info.tabentry;
+	if (rel->pgstat_info != NULL &&
+		rel->pgstat_info->t_id == rel_id)
 		return;
-	}
+
+	/* Else find or make the PgStat_TableStatus entry, and update link */
+	rel->pgstat_info = get_tabstat_entry(rel_id, rel->rd_rel->relisshared);
+}
+
+/*
+ * get_tabstat_entry - find or create a PgStat_TableStatus entry for rel
+ */
+static PgStat_TableStatus *
+get_tabstat_entry(Oid rel_id, bool isshared)
+{
+	PgStat_TableStatus *entry;
+	TabStatusArray *tsa;
+	TabStatusArray *prev_tsa;
+	int			i;
 
 	/*
-	 * Search the already-used message slots for this relation.
+	 * Search the already-used tabstat slots for this relation.
 	 */
-	tsarr = rel->rd_rel->relisshared ? &SharedTabStat : &RegularTabStat;
-
-	for (mb = 0; mb < tsarr->tsa_used; mb++)
+	prev_tsa = NULL;
+	for (tsa = pgStatTabList; tsa != NULL; prev_tsa = tsa, tsa = tsa->tsa_next)
 	{
-		tsmsg = tsarr->tsa_messages[mb];
-
-		for (i = tsmsg->m_nentries; --i >= 0;)
+		for (i = 0; i < tsa->tsa_used; i++)
 		{
-			if (tsmsg->m_entry[i].t_id == rel_id)
-			{
-				rel->pgstat_info.tabentry = (void *) &(tsmsg->m_entry[i]);
-				stats->tabentry = rel->pgstat_info.tabentry;
-				return;
-			}
+			entry = &tsa->tsa_entries[i];
+			if (entry->t_id == rel_id)
+				return entry;
 		}
 
-		if (tsmsg->m_nentries >= PGSTAT_NUM_TABENTRIES)
-			continue;
-
-		/*
-		 * Not found, but found a message buffer with an empty slot instead.
-		 * Fine, let's use this one.  We assume the entry was already zeroed,
-		 * either at creation or after last use.
-		 */
-		i = tsmsg->m_nentries++;
-		useent = &tsmsg->m_entry[i];
-		useent->t_id = rel_id;
-		rel->pgstat_info.tabentry = (void *) useent;
-		stats->tabentry = rel->pgstat_info.tabentry;
-		return;
+		if (tsa->tsa_used < TABSTAT_QUANTUM)
+		{
+			/*
+			 * It must not be present, but we found a free slot instead.
+			 * Fine, let's use this one.  We assume the entry was already
+			 * zeroed, either at creation or after last use.
+			 */
+			entry = &tsa->tsa_entries[tsa->tsa_used++];
+			entry->t_id = rel_id;
+			entry->t_shared = isshared;
+			return entry;
+		}
 	}
 
 	/*
-	 * If we ran out of message buffers, we just allocate more.
+	 * We ran out of tabstat slots, so allocate more.  Be sure they're zeroed.
 	 */
-	if (tsarr->tsa_used >= tsarr->tsa_alloc)
-		more_tabstat_space(tsarr);
+	tsa = (TabStatusArray *) MemoryContextAllocZero(TopMemoryContext,
+													sizeof(TabStatusArray));
+	if (prev_tsa)
+		prev_tsa->tsa_next = tsa;
+	else
+		pgStatTabList = tsa;
+
+	/*
+	 * Use the first entry of the new TabStatusArray.
+	 */
+	entry = &tsa->tsa_entries[tsa->tsa_used++];
+	entry->t_id = rel_id;
+	entry->t_shared = isshared;
+	return entry;
+}
+
+/*
+ * get_tabstat_stack_level - add a new (sub)transaction stack entry if needed
+ */
+static PgStat_SubXactStatus *
+get_tabstat_stack_level(int nest_level)
+{
+	PgStat_SubXactStatus *xact_state;
+
+	xact_state = pgStatXactStack;
+	if (xact_state == NULL || xact_state->nest_level != nest_level)
+	{
+		xact_state = (PgStat_SubXactStatus *)
+			MemoryContextAlloc(TopTransactionContext,
+							   sizeof(PgStat_SubXactStatus));
+		xact_state->nest_level = nest_level;
+		xact_state->prev = pgStatXactStack;
+		xact_state->first = NULL;
+		pgStatXactStack = xact_state;
+	}
+	return xact_state;
+}
+
+/*
+ * add_tabstat_xact_level - add a new (sub)transaction state record
+ */
+static void
+add_tabstat_xact_level(PgStat_TableStatus *pgstat_info, int nest_level)
+{
+	PgStat_SubXactStatus *xact_state;
+	PgStat_TableXactStatus *trans;
 
 	/*
-	 * Use the first entry of the next message buffer.
+	 * If this is the first rel to be modified at the current nest level,
+	 * we first have to push a transaction stack entry.
 	 */
-	mb = tsarr->tsa_used++;
-	tsmsg = tsarr->tsa_messages[mb];
-	tsmsg->m_nentries = 1;
-	useent = &tsmsg->m_entry[0];
-	useent->t_id = rel_id;
-	rel->pgstat_info.tabentry = (void *) useent;
-	stats->tabentry = rel->pgstat_info.tabentry;
+	xact_state = get_tabstat_stack_level(nest_level);
+
+	/* Now make a per-table stack entry */
+	trans = (PgStat_TableXactStatus *)
+		MemoryContextAllocZero(TopTransactionContext,
+							   sizeof(PgStat_TableXactStatus));
+	trans->nest_level = nest_level;
+	trans->upper = pgstat_info->trans;
+	trans->parent = pgstat_info;
+	trans->next = xact_state->first;
+	xact_state->first = trans;
+	pgstat_info->trans = trans;
+}
+
+/*
+ * pgstat_count_heap_insert - count a tuple insertion
+ */
+void
+pgstat_count_heap_insert(Relation rel)
+{
+	PgStat_TableStatus *pgstat_info = rel->pgstat_info;
+
+	if (pgstat_collect_tuplelevel && pgstat_info != NULL)
+	{
+		int		nest_level = GetCurrentTransactionNestLevel();
+
+		/* t_tuples_inserted is nontransactional, so just advance it */
+		pgstat_info->t_counts.t_tuples_inserted++;
+
+		/* We have to log the transactional effect at the proper level */
+		if (pgstat_info->trans == NULL ||
+			pgstat_info->trans->nest_level != nest_level)
+			add_tabstat_xact_level(pgstat_info, nest_level);
+
+		pgstat_info->trans->tuples_inserted++;
+	}
+}
+
+/*
+ * pgstat_count_heap_update - count a tuple update
+ */
+void
+pgstat_count_heap_update(Relation rel)
+{
+	PgStat_TableStatus *pgstat_info = rel->pgstat_info;
+
+	if (pgstat_collect_tuplelevel && pgstat_info != NULL)
+	{
+		int		nest_level = GetCurrentTransactionNestLevel();
+
+		/* t_tuples_updated is nontransactional, so just advance it */
+		pgstat_info->t_counts.t_tuples_updated++;
+
+		/* We have to log the transactional effect at the proper level */
+		if (pgstat_info->trans == NULL ||
+			pgstat_info->trans->nest_level != nest_level)
+			add_tabstat_xact_level(pgstat_info, nest_level);
+
+		/* An UPDATE both inserts a new tuple and deletes the old */
+		pgstat_info->trans->tuples_inserted++;
+		pgstat_info->trans->tuples_deleted++;
+	}
+}
+
+/*
+ * pgstat_count_heap_delete - count a tuple deletion
+ */
+void
+pgstat_count_heap_delete(Relation rel)
+{
+	PgStat_TableStatus *pgstat_info = rel->pgstat_info;
+
+	if (pgstat_collect_tuplelevel && pgstat_info != NULL)
+	{
+		int		nest_level = GetCurrentTransactionNestLevel();
+
+		/* t_tuples_deleted is nontransactional, so just advance it */
+		pgstat_info->t_counts.t_tuples_deleted++;
+
+		/* We have to log the transactional effect at the proper level */
+		if (pgstat_info->trans == NULL ||
+			pgstat_info->trans->nest_level != nest_level)
+			add_tabstat_xact_level(pgstat_info, nest_level);
+
+		pgstat_info->trans->tuples_deleted++;
+	}
 }
 
 
 /* ----------
- * pgstat_count_xact_commit() -
+ * AtEOXact_PgStat
  *
- *	Called from access/transam/xact.c to count transaction commits.
+ *	Called from access/transam/xact.c at top-level transaction commit/abort.
  * ----------
  */
 void
-pgstat_count_xact_commit(void)
+AtEOXact_PgStat(bool isCommit)
 {
-	if (!pgstat_collect_tuplelevel &&
-		!pgstat_collect_blocklevel)
-		return;
-
-	pgStatXactCommit++;
+	PgStat_SubXactStatus *xact_state;
 
 	/*
-	 * If there was no relation activity yet, just make one existing message
-	 * buffer used without slots, causing the next report to tell new
-	 * xact-counters.
+	 * Count transaction commit or abort.  (We use counters, not just bools,
+	 * in case the reporting message isn't sent right away.)
 	 */
-	if (RegularTabStat.tsa_alloc == 0)
-		more_tabstat_space(&RegularTabStat);
+	if (isCommit)
+		pgStatXactCommit++;
+	else
+		pgStatXactRollback++;
 
-	if (RegularTabStat.tsa_used == 0)
+	/*
+	 * Transfer transactional insert/update counts into the base tabstat
+	 * entries.  We don't bother to free any of the transactional state,
+	 * since it's all in TopTransactionContext and will go away anyway.
+	 */
+	xact_state = pgStatXactStack;
+	if (xact_state != NULL)
 	{
-		RegularTabStat.tsa_used++;
-		RegularTabStat.tsa_messages[0]->m_nentries = 0;
+		PgStat_TableXactStatus *trans;
+
+		Assert(xact_state->nest_level == 1);
+		Assert(xact_state->prev == NULL);
+		for (trans = xact_state->first; trans != NULL; trans = trans->next)
+		{
+			PgStat_TableStatus *tabstat;
+
+			Assert(trans->nest_level == 1);
+			Assert(trans->upper == NULL);
+			tabstat = trans->parent;
+			Assert(tabstat->trans == trans);
+			if (isCommit)
+			{
+				tabstat->t_counts.t_new_live_tuples += trans->tuples_inserted;
+				tabstat->t_counts.t_new_dead_tuples += trans->tuples_deleted;
+			}
+			else
+			{
+				/* inserted tuples are dead, deleted tuples are unaffected */
+				tabstat->t_counts.t_new_dead_tuples += trans->tuples_inserted;
+			}
+			tabstat->trans = NULL;
+		}
 	}
-}
+	pgStatXactStack = NULL;
 
+	/* Make sure any stats snapshot is thrown away */
+	pgstat_clear_snapshot();
+}
 
 /* ----------
- * pgstat_count_xact_rollback() -
+ * AtEOSubXact_PgStat
  *
- *	Called from access/transam/xact.c to count transaction rollbacks.
+ *	Called from access/transam/xact.c at subtransaction commit/abort.
  * ----------
  */
 void
-pgstat_count_xact_rollback(void)
+AtEOSubXact_PgStat(bool isCommit, int nestDepth)
 {
-	if (!pgstat_collect_tuplelevel &&
-		!pgstat_collect_blocklevel)
-		return;
-
-	pgStatXactRollback++;
+	PgStat_SubXactStatus *xact_state;
 
 	/*
-	 * If there was no relation activity yet, just make one existing message
-	 * buffer used without slots, causing the next report to tell new
-	 * xact-counters.
+	 * Transfer transactional insert/update counts into the next higher
+	 * subtransaction state.
 	 */
-	if (RegularTabStat.tsa_alloc == 0)
-		more_tabstat_space(&RegularTabStat);
+	xact_state = pgStatXactStack;
+	if (xact_state != NULL &&
+		xact_state->nest_level >= nestDepth)
+	{
+		PgStat_TableXactStatus *trans;
+		PgStat_TableXactStatus *next_trans;
+
+		/* delink xact_state from stack immediately to simplify reuse case */
+		pgStatXactStack = xact_state->prev;
+
+		for (trans = xact_state->first; trans != NULL; trans = next_trans)
+		{
+			PgStat_TableStatus *tabstat;
+
+			next_trans = trans->next;
+			Assert(trans->nest_level == nestDepth);
+			tabstat = trans->parent;
+			Assert(tabstat->trans == trans);
+			if (isCommit)
+			{
+				if (trans->upper && trans->upper->nest_level == nestDepth - 1)
+				{
+					trans->upper->tuples_inserted += trans->tuples_inserted;
+					trans->upper->tuples_deleted += trans->tuples_deleted;
+					tabstat->trans = trans->upper;
+					pfree(trans);
+				}
+				else
+				{
+					/*
+					 * When there isn't an immediate parent state, we can
+					 * just reuse the record instead of going through a
+					 * palloc/pfree pushup (this works since it's all in
+					 * TopTransactionContext anyway).  We have to re-link
+					 * it into the parent level, though, and that might mean
+					 * pushing a new entry into the pgStatXactStack.
+					 */
+					PgStat_SubXactStatus *upper_xact_state;
+
+					upper_xact_state = get_tabstat_stack_level(nestDepth - 1);
+					trans->next = upper_xact_state->first;
+					upper_xact_state->first = trans;
+					trans->nest_level = nestDepth - 1;
+				}
+			}
+			else
+			{
+				/*
+				 * On abort, inserted tuples are dead (and can be bounced out
+				 * to the top-level tabstat), deleted tuples are unaffected
+				 */
+				tabstat->t_counts.t_new_dead_tuples += trans->tuples_inserted;
+				tabstat->trans = trans->upper;
+				pfree(trans);
+			}
+		}
+		pfree(xact_state);
+	}
+}
+
+
+/*
+ * AtPrepare_PgStat
+ *		Save the transactional stats state at 2PC transaction prepare.
+ *
+ * In this phase we just generate 2PC records for all the pending
+ * transaction-dependent stats work.
+ */
+void
+AtPrepare_PgStat(void)
+{
+	PgStat_SubXactStatus *xact_state;
 
-	if (RegularTabStat.tsa_used == 0)
+	xact_state = pgStatXactStack;
+	if (xact_state != NULL)
 	{
-		RegularTabStat.tsa_used++;
-		RegularTabStat.tsa_messages[0]->m_nentries = 0;
+		PgStat_TableXactStatus *trans;
+
+		Assert(xact_state->nest_level == 1);
+		Assert(xact_state->prev == NULL);
+		for (trans = xact_state->first; trans != NULL; trans = trans->next)
+		{
+			PgStat_TableStatus *tabstat;
+			TwoPhasePgStatRecord record;
+
+			Assert(trans->nest_level == 1);
+			Assert(trans->upper == NULL);
+			tabstat = trans->parent;
+			Assert(tabstat->trans == trans);
+
+			record.tuples_inserted = trans->tuples_inserted;
+			record.tuples_deleted = trans->tuples_deleted;
+			record.t_id = tabstat->t_id;
+			record.t_shared = tabstat->t_shared;
+
+			RegisterTwoPhaseRecord(TWOPHASE_RM_PGSTAT_ID, 0,
+								   &record, sizeof(TwoPhasePgStatRecord));
+		}
 	}
 }
 
+/*
+ * PostPrepare_PgStat
+ *		Clean up after successful PREPARE.
+ *
+ * All we need do here is unlink the transaction stats state from the
+ * nontransactional state.  The nontransactional action counts will be
+ * reported to the stats collector immediately, while the effects on live
+ * and dead tuple counts are preserved in the 2PC state file.
+ *
+ * Note: AtEOXact_PgStat is not called during PREPARE.
+ */
+void
+PostPrepare_PgStat(void)
+{
+	PgStat_SubXactStatus *xact_state;
+
+	/*
+	 * We don't bother to free any of the transactional state,
+	 * since it's all in TopTransactionContext and will go away anyway.
+	 */
+	xact_state = pgStatXactStack;
+	if (xact_state != NULL)
+	{
+		PgStat_TableXactStatus *trans;
+
+		for (trans = xact_state->first; trans != NULL; trans = trans->next)
+		{
+			PgStat_TableStatus *tabstat;
+
+			tabstat = trans->parent;
+			tabstat->trans = NULL;
+		}
+	}
+	pgStatXactStack = NULL;
+
+	/* Make sure any stats snapshot is thrown away */
+	pgstat_clear_snapshot();
+}
+
+/*
+ * 2PC processing routine for COMMIT PREPARED case.
+ *
+ * Load the saved counts into our local pgstats state.
+ */
+void
+pgstat_twophase_postcommit(TransactionId xid, uint16 info,
+						   void *recdata, uint32 len)
+{
+	TwoPhasePgStatRecord *rec = (TwoPhasePgStatRecord *) recdata;
+	PgStat_TableStatus *pgstat_info;
+
+	/* Find or create a tabstat entry for the rel */
+	pgstat_info = get_tabstat_entry(rec->t_id, rec->t_shared);
+
+	pgstat_info->t_counts.t_new_live_tuples += rec->tuples_inserted;
+	pgstat_info->t_counts.t_new_dead_tuples += rec->tuples_deleted;
+}
+
+/*
+ * 2PC processing routine for ROLLBACK PREPARED case.
+ *
+ * Load the saved counts into our local pgstats state, but treat them
+ * as aborted.
+ */
+void
+pgstat_twophase_postabort(TransactionId xid, uint16 info,
+						  void *recdata, uint32 len)
+{
+	TwoPhasePgStatRecord *rec = (TwoPhasePgStatRecord *) recdata;
+	PgStat_TableStatus *pgstat_info;
+
+	/* Find or create a tabstat entry for the rel */
+	pgstat_info = get_tabstat_entry(rec->t_id, rec->t_shared);
+
+	/* inserted tuples are dead, deleted tuples are no-ops */
+	pgstat_info->t_counts.t_new_dead_tuples += rec->tuples_inserted;
+}
+
 
 /* ----------
  * pgstat_fetch_stat_dbentry() -
@@ -1725,18 +2093,15 @@ pgstat_send(void *msg, int len)
 void
 pgstat_send_bgwriter(void)
 {
+	/* We assume this initializes to zeroes */
+	static const PgStat_MsgBgWriter all_zeroes;
+
 	/*
 	 * This function can be called even if nothing at all has happened.
 	 * In this case, avoid sending a completely empty message to
 	 * the stats collector.
 	 */
-	if (BgWriterStats.m_timed_checkpoints == 0 &&
-		BgWriterStats.m_requested_checkpoints == 0 &&
-		BgWriterStats.m_buf_written_checkpoints == 0 &&
-		BgWriterStats.m_buf_written_lru == 0 &&
-		BgWriterStats.m_buf_written_all == 0 &&
-		BgWriterStats.m_maxwritten_lru == 0 &&
-		BgWriterStats.m_maxwritten_all == 0)
+	if (memcmp(&BgWriterStats, &all_zeroes, sizeof(PgStat_MsgBgWriter)) == 0)
 		return;
 
 	/*
@@ -1746,10 +2111,9 @@ pgstat_send_bgwriter(void)
 	pgstat_send(&BgWriterStats, sizeof(BgWriterStats));
 
 	/*
-	 * Clear out the bgwriter statistics buffer, so it can be
-	 * re-used.
+	 * Clear out the statistics buffer, so it can be re-used.
 	 */
-	memset(&BgWriterStats, 0, sizeof(BgWriterStats));
+	MemSet(&BgWriterStats, 0, sizeof(BgWriterStats));
 }
 
 
@@ -2509,60 +2873,50 @@ pgstat_recv_tabstat(PgStat_MsgTabstat *msg, int len)
 			 * If it's a new table entry, initialize counters to the values we
 			 * just got.
 			 */
-			tabentry->numscans = tabmsg[i].t_numscans;
-			tabentry->tuples_returned = tabmsg[i].t_tuples_returned;
-			tabentry->tuples_fetched = tabmsg[i].t_tuples_fetched;
-			tabentry->tuples_inserted = tabmsg[i].t_tuples_inserted;
-			tabentry->tuples_updated = tabmsg[i].t_tuples_updated;
-			tabentry->tuples_deleted = tabmsg[i].t_tuples_deleted;
-
-			tabentry->n_live_tuples = tabmsg[i].t_tuples_inserted;
-			tabentry->n_dead_tuples = tabmsg[i].t_tuples_updated +
-				tabmsg[i].t_tuples_deleted;
+			tabentry->numscans = tabmsg[i].t_counts.t_numscans;
+			tabentry->tuples_returned = tabmsg[i].t_counts.t_tuples_returned;
+			tabentry->tuples_fetched = tabmsg[i].t_counts.t_tuples_fetched;
+			tabentry->tuples_inserted = tabmsg[i].t_counts.t_tuples_inserted;
+			tabentry->tuples_updated = tabmsg[i].t_counts.t_tuples_updated;
+			tabentry->tuples_deleted = tabmsg[i].t_counts.t_tuples_deleted;
+			tabentry->n_live_tuples = tabmsg[i].t_counts.t_new_live_tuples;
+			tabentry->n_dead_tuples = tabmsg[i].t_counts.t_new_dead_tuples;
+			tabentry->blocks_fetched = tabmsg[i].t_counts.t_blocks_fetched;
+			tabentry->blocks_hit = tabmsg[i].t_counts.t_blocks_hit;
+
 			tabentry->last_anl_tuples = 0;
 			tabentry->vacuum_timestamp = 0;
 			tabentry->autovac_vacuum_timestamp = 0;
 			tabentry->analyze_timestamp = 0;
 			tabentry->autovac_analyze_timestamp = 0;
-
-			tabentry->blocks_fetched = tabmsg[i].t_blocks_fetched;
-			tabentry->blocks_hit = tabmsg[i].t_blocks_hit;
 		}
 		else
 		{
 			/*
 			 * Otherwise add the values to the existing entry.
 			 */
-			tabentry->numscans += tabmsg[i].t_numscans;
-			tabentry->tuples_returned += tabmsg[i].t_tuples_returned;
-			tabentry->tuples_fetched += tabmsg[i].t_tuples_fetched;
-			tabentry->tuples_inserted += tabmsg[i].t_tuples_inserted;
-			tabentry->tuples_updated += tabmsg[i].t_tuples_updated;
-			tabentry->tuples_deleted += tabmsg[i].t_tuples_deleted;
-
-			tabentry->n_live_tuples += tabmsg[i].t_tuples_inserted -
-				tabmsg[i].t_tuples_deleted;
-			tabentry->n_dead_tuples += tabmsg[i].t_tuples_updated +
-				tabmsg[i].t_tuples_deleted;
-
-			tabentry->blocks_fetched += tabmsg[i].t_blocks_fetched;
-			tabentry->blocks_hit += tabmsg[i].t_blocks_hit;
+			tabentry->numscans += tabmsg[i].t_counts.t_numscans;
+			tabentry->tuples_returned += tabmsg[i].t_counts.t_tuples_returned;
+			tabentry->tuples_fetched += tabmsg[i].t_counts.t_tuples_fetched;
+			tabentry->tuples_inserted += tabmsg[i].t_counts.t_tuples_inserted;
+			tabentry->tuples_updated += tabmsg[i].t_counts.t_tuples_updated;
+			tabentry->tuples_deleted += tabmsg[i].t_counts.t_tuples_deleted;
+			tabentry->n_live_tuples += tabmsg[i].t_counts.t_new_live_tuples;
+			tabentry->n_dead_tuples += tabmsg[i].t_counts.t_new_dead_tuples;
+			tabentry->blocks_fetched += tabmsg[i].t_counts.t_blocks_fetched;
+			tabentry->blocks_hit += tabmsg[i].t_counts.t_blocks_hit;
 		}
 
 		/*
-		 * Add table stats to the database entry.
-		 */
-		dbentry->n_tuples_returned += tabmsg[i].t_tuples_returned;
-		dbentry->n_tuples_fetched += tabmsg[i].t_tuples_fetched;
-		dbentry->n_tuples_inserted += tabmsg[i].t_tuples_inserted;
-		dbentry->n_tuples_updated += tabmsg[i].t_tuples_updated;
-		dbentry->n_tuples_deleted += tabmsg[i].t_tuples_deleted;
-
-		/*
-		 * And add the block IO to the database entry.
+		 * Add per-table stats to the per-database entry, too.
 		 */
-		dbentry->n_blocks_fetched += tabmsg[i].t_blocks_fetched;
-		dbentry->n_blocks_hit += tabmsg[i].t_blocks_hit;
+		dbentry->n_tuples_returned += tabmsg[i].t_counts.t_tuples_returned;
+		dbentry->n_tuples_fetched += tabmsg[i].t_counts.t_tuples_fetched;
+		dbentry->n_tuples_inserted += tabmsg[i].t_counts.t_tuples_inserted;
+		dbentry->n_tuples_updated += tabmsg[i].t_counts.t_tuples_updated;
+		dbentry->n_tuples_deleted += tabmsg[i].t_counts.t_tuples_deleted;
+		dbentry->n_blocks_fetched += tabmsg[i].t_counts.t_blocks_fetched;
+		dbentry->n_blocks_hit += tabmsg[i].t_counts.t_blocks_hit;
 	}
 }
 
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 9f4876a6050..e2cfc870e2e 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/storage/buffer/bufmgr.c,v 1.218 2007/05/02 23:34:48 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/storage/buffer/bufmgr.c,v 1.219 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -88,12 +88,6 @@ static bool IsForInput;
 /* local state for LockBufferForCleanup */
 static volatile BufferDesc *PinCountWaitBuf = NULL;
 
-/*
- * Global statistics for the bgwriter. The contents of this variable
- * only makes sense in the bgwriter process.
- */
-extern PgStat_MsgBgWriter BgWriterStats;
-
 
 static Buffer ReadBuffer_common(Relation reln, BlockNumber blockNum,
 								bool zeroPage);
@@ -174,7 +168,7 @@ ReadBuffer_common(Relation reln, BlockNumber blockNum, bool zeroPage)
 	if (isExtend)
 		blockNum = smgrnblocks(reln->rd_smgr);
 
-	pgstat_count_buffer_read(&reln->pgstat_info, reln);
+	pgstat_count_buffer_read(reln);
 
 	if (isLocalBuf)
 	{
@@ -204,7 +198,7 @@ ReadBuffer_common(Relation reln, BlockNumber blockNum, bool zeroPage)
 		if (!isExtend)
 		{
 			/* Just need to update stats before we exit */
-			pgstat_count_buffer_hit(&reln->pgstat_info, reln);
+			pgstat_count_buffer_hit(reln);
 
 			if (VacuumCostActive)
 				VacuumCostBalance += VacuumCostPageHit;
diff --git a/src/backend/utils/cache/relcache.c b/src/backend/utils/cache/relcache.c
index 7d554c2ada2..45cb103adee 100644
--- a/src/backend/utils/cache/relcache.c
+++ b/src/backend/utils/cache/relcache.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/cache/relcache.c,v 1.260 2007/05/02 21:08:46 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/cache/relcache.c,v 1.261 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1802,6 +1802,7 @@ RelationClearRelation(Relation relation, bool rebuild)
 		int			old_refcnt = relation->rd_refcnt;
 		SubTransactionId old_createSubid = relation->rd_createSubid;
 		SubTransactionId old_newRelfilenodeSubid = relation->rd_newRelfilenodeSubid;
+		struct PgStat_TableStatus *old_pgstat_info = relation->pgstat_info;
 		TupleDesc	old_att = relation->rd_att;
 		RuleLock   *old_rules = relation->rd_rules;
 		MemoryContext old_rulescxt = relation->rd_rulescxt;
@@ -1821,6 +1822,7 @@ RelationClearRelation(Relation relation, bool rebuild)
 		relation->rd_refcnt = old_refcnt;
 		relation->rd_createSubid = old_createSubid;
 		relation->rd_newRelfilenodeSubid = old_newRelfilenodeSubid;
+		relation->pgstat_info = old_pgstat_info;
 
 		if (equalTupleDescs(old_att, relation->rd_att))
 		{
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 5ea66e74672..ebb2e984c24 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2007, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/access/heapam.h,v 1.123 2007/04/08 01:26:33 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/access/heapam.h,v 1.124 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -147,10 +147,10 @@ extern HeapTuple heap_getnext(HeapScanDesc scan, ScanDirection direction);
 
 extern bool heap_fetch(Relation relation, Snapshot snapshot,
 		   HeapTuple tuple, Buffer *userbuf, bool keep_buf,
-		   PgStat_Info *pgstat_info);
+		   Relation stats_relation);
 extern bool heap_release_fetch(Relation relation, Snapshot snapshot,
 				   HeapTuple tuple, Buffer *userbuf, bool keep_buf,
-				   PgStat_Info *pgstat_info);
+				   Relation stats_relation);
 
 extern void heap_get_latest_tid(Relation relation, Snapshot snapshot,
 					ItemPointer tid);
diff --git a/src/include/access/relscan.h b/src/include/access/relscan.h
index 77bca6be482..7a1ea39352a 100644
--- a/src/include/access/relscan.h
+++ b/src/include/access/relscan.h
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2007, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/access/relscan.h,v 1.52 2007/01/20 18:43:35 neilc Exp $
+ * $PostgreSQL: pgsql/src/include/access/relscan.h,v 1.53 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -37,8 +37,6 @@ typedef struct HeapScanDescData
 	/* NB: if rs_cbuf is not InvalidBuffer, we hold a pin on that buffer */
 	ItemPointerData rs_mctid;	/* marked scan position, if any */
 
-	PgStat_Info rs_pgstat_info; /* statistics collector hook */
-
 	/* these fields only used in page-at-a-time mode */
 	int			rs_cindex;		/* current tuple's index in vistuples */
 	int			rs_mindex;		/* marked tuple's saved index */
@@ -78,8 +76,6 @@ typedef struct IndexScanDescData
 	HeapTupleData xs_ctup;		/* current heap tuple, if any */
 	Buffer		xs_cbuf;		/* current heap buffer in scan, if any */
 	/* NB: if xs_cbuf is not InvalidBuffer, we hold a pin on that buffer */
-
-	PgStat_Info xs_pgstat_info; /* statistics collector hook */
 } IndexScanDescData;
 
 typedef IndexScanDescData *IndexScanDesc;
diff --git a/src/include/access/twophase_rmgr.h b/src/include/access/twophase_rmgr.h
index 0dbcd226fbd..e98ad7cb375 100644
--- a/src/include/access/twophase_rmgr.h
+++ b/src/include/access/twophase_rmgr.h
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2007, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/access/twophase_rmgr.h,v 1.4 2007/01/05 22:19:51 momjian Exp $
+ * $PostgreSQL: pgsql/src/include/access/twophase_rmgr.h,v 1.5 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -26,7 +26,8 @@ typedef uint8 TwoPhaseRmgrId;
 #define TWOPHASE_RM_INVAL_ID		2
 #define TWOPHASE_RM_FLATFILES_ID	3
 #define TWOPHASE_RM_NOTIFY_ID		4
-#define TWOPHASE_RM_MAX_ID			TWOPHASE_RM_NOTIFY_ID
+#define TWOPHASE_RM_PGSTAT_ID		5
+#define TWOPHASE_RM_MAX_ID			TWOPHASE_RM_PGSTAT_ID
 
 extern const TwoPhaseCallback twophase_recover_callbacks[];
 extern const TwoPhaseCallback twophase_postcommit_callbacks[];
diff --git a/src/include/pgstat.h b/src/include/pgstat.h
index 694ee44db19..476fd47dc7b 100644
--- a/src/include/pgstat.h
+++ b/src/include/pgstat.h
@@ -5,7 +5,7 @@
  *
  *	Copyright (c) 2001-2007, PostgreSQL Global Development Group
  *
- *	$PostgreSQL: pgsql/src/include/pgstat.h,v 1.58 2007/04/30 16:37:08 tgl Exp $
+ *	$PostgreSQL: pgsql/src/include/pgstat.h,v 1.59 2007/05/27 03:50:39 tgl Exp $
  * ----------
  */
 #ifndef PGSTAT_H
@@ -40,6 +40,90 @@ typedef enum StatMsgType
  */
 typedef int64 PgStat_Counter;
 
+/* ----------
+ * PgStat_TableCounts			The actual per-table counts kept by a backend
+ *
+ * This struct should contain only actual event counters, because we memcmp
+ * it against zeroes to detect whether there are any counts to transmit.
+ * It is a component of PgStat_TableStatus (within-backend state) and
+ * PgStat_TableEntry (the transmitted message format).
+ *
+ * Note: for a table, tuples_returned is the number of tuples successfully
+ * fetched by heap_getnext, while tuples_fetched is the number of tuples
+ * successfully fetched by heap_fetch under the control of bitmap indexscans.
+ * For an index, tuples_returned is the number of index entries returned by
+ * the index AM, while tuples_fetched is the number of tuples successfully
+ * fetched by heap_fetch under the control of simple indexscans for this index.
+ *
+ * tuples_inserted/tuples_updated/tuples_deleted count attempted actions,
+ * regardless of whether the transaction committed.  new_live_tuples and
+ * new_dead_tuples are properly adjusted depending on commit or abort.
+ * ----------
+ */
+typedef struct PgStat_TableCounts
+{
+	PgStat_Counter t_numscans;
+
+	PgStat_Counter t_tuples_returned;
+	PgStat_Counter t_tuples_fetched;
+
+	PgStat_Counter t_tuples_inserted;
+	PgStat_Counter t_tuples_updated;
+	PgStat_Counter t_tuples_deleted;
+
+	PgStat_Counter t_new_live_tuples;
+	PgStat_Counter t_new_dead_tuples;
+
+	PgStat_Counter t_blocks_fetched;
+	PgStat_Counter t_blocks_hit;
+} PgStat_TableCounts;
+
+
+/* ------------------------------------------------------------
+ * Structures kept in backend local memory while accumulating counts
+ * ------------------------------------------------------------
+ */
+
+
+/* ----------
+ * PgStat_TableStatus			Per-table status within a backend
+ *
+ * Most of the event counters are nontransactional, ie, we count events
+ * in committed and aborted transactions alike.  For these, we just count
+ * directly in the PgStat_TableStatus.  However, new_live_tuples and
+ * new_dead_tuples must be derived from tuple insertion and deletion counts
+ * with awareness of whether the transaction or subtransaction committed or
+ * aborted.  Hence, we also keep a stack of per-(sub)transaction status
+ * records for every table modified in the current transaction.  At commit
+ * or abort, we propagate tuples_inserted and tuples_deleted up to the
+ * parent subtransaction level, or out to the parent PgStat_TableStatus,
+ * as appropriate.
+ * ----------
+ */
+typedef struct PgStat_TableStatus
+{
+	Oid			t_id;				/* table's OID */
+	bool		t_shared;			/* is it a shared catalog? */
+	struct PgStat_TableXactStatus *trans;	/* lowest subxact's counts */
+	PgStat_TableCounts t_counts;	/* event counts to be sent */
+} PgStat_TableStatus;
+
+/* ----------
+ * PgStat_TableXactStatus		Per-table, per-subtransaction status
+ * ----------
+ */
+typedef struct PgStat_TableXactStatus
+{
+	PgStat_Counter tuples_inserted;	/* tuples inserted in (sub)xact */
+	PgStat_Counter tuples_deleted;	/* tuples deleted in (sub)xact */
+	int			nest_level;			/* subtransaction nest level */
+	/* links to other structs for same relation: */
+	struct PgStat_TableXactStatus *upper;	/* next higher subxact if any */
+	PgStat_TableStatus *parent;				/* per-table status */
+	/* structs of same subxact level are linked here: */
+	struct PgStat_TableXactStatus *next;	/* next of same subxact */
+} PgStat_TableXactStatus;
+
 
 /* ------------------------------------------------------------
  * Message formats follow
@@ -78,30 +162,12 @@ typedef struct PgStat_MsgDummy
 
 /* ----------
  * PgStat_TableEntry			Per-table info in a MsgTabstat
- *
- * Note: for a table, tuples_returned is the number of tuples successfully
- * fetched by heap_getnext, while tuples_fetched is the number of tuples
- * successfully fetched by heap_fetch under the control of bitmap indexscans.
- * For an index, tuples_returned is the number of index entries returned by
- * the index AM, while tuples_fetched is the number of tuples successfully
- * fetched by heap_fetch under the control of simple indexscans for this index.
  * ----------
  */
 typedef struct PgStat_TableEntry
 {
 	Oid			t_id;
-
-	PgStat_Counter t_numscans;
-
-	PgStat_Counter t_tuples_returned;
-	PgStat_Counter t_tuples_fetched;
-
-	PgStat_Counter t_tuples_inserted;
-	PgStat_Counter t_tuples_updated;
-	PgStat_Counter t_tuples_deleted;
-
-	PgStat_Counter t_blocks_fetched;
-	PgStat_Counter t_blocks_hit;
+	PgStat_TableCounts t_counts;
 } PgStat_TableEntry;
 
 /* ----------
@@ -393,6 +459,10 @@ extern bool pgstat_collect_tuplelevel;
 extern bool pgstat_collect_blocklevel;
 extern bool pgstat_collect_querystring;
 
+/*
+ * BgWriter statistics counters are updated directly by bgwriter and bufmgr
+ */
+extern PgStat_MsgBgWriter BgWriterStats;
 
 /* ----------
  * Functions called from postmaster
@@ -436,83 +506,67 @@ extern void pgstat_report_activity(const char *what);
 extern void pgstat_report_txn_timestamp(TimestampTz tstamp);
 extern void pgstat_report_waiting(bool waiting);
 
-extern void pgstat_initstats(PgStat_Info *stats, Relation rel);
+extern void pgstat_initstats(Relation rel);
 
+/* nontransactional event counts are simple enough to inline */
 
-#define pgstat_count_heap_scan(s)										\
+#define pgstat_count_heap_scan(rel)										\
 	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_numscans++;		\
+		if (pgstat_collect_tuplelevel && (rel)->pgstat_info != NULL)	\
+			(rel)->pgstat_info->t_counts.t_numscans++;					\
 	} while (0)
 /* kluge for bitmap scans: */
-#define pgstat_discount_heap_scan(s)									\
+#define pgstat_discount_heap_scan(rel)									\
 	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_numscans--;		\
+		if (pgstat_collect_tuplelevel && (rel)->pgstat_info != NULL)	\
+			(rel)->pgstat_info->t_counts.t_numscans--;					\
 	} while (0)
-#define pgstat_count_heap_getnext(s)									\
+#define pgstat_count_heap_getnext(rel)									\
 	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_tuples_returned++; \
+		if (pgstat_collect_tuplelevel && (rel)->pgstat_info != NULL)	\
+			(rel)->pgstat_info->t_counts.t_tuples_returned++;			\
 	} while (0)
-#define pgstat_count_heap_fetch(s)										\
+#define pgstat_count_heap_fetch(rel)									\
 	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_tuples_fetched++; \
+		if (pgstat_collect_tuplelevel && (rel)->pgstat_info != NULL)	\
+			(rel)->pgstat_info->t_counts.t_tuples_fetched++;			\
 	} while (0)
-#define pgstat_count_heap_insert(s)										\
+#define pgstat_count_index_scan(rel)									\
 	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_tuples_inserted++; \
+		if (pgstat_collect_tuplelevel && (rel)->pgstat_info != NULL)	\
+			(rel)->pgstat_info->t_counts.t_numscans++;					\
 	} while (0)
-#define pgstat_count_heap_update(s)										\
+#define pgstat_count_index_tuples(rel, n)								\
 	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_tuples_updated++; \
+		if (pgstat_collect_tuplelevel && (rel)->pgstat_info != NULL)	\
+			(rel)->pgstat_info->t_counts.t_tuples_returned += (n);		\
 	} while (0)
-#define pgstat_count_heap_delete(s)										\
+#define pgstat_count_buffer_read(rel)									\
 	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_tuples_deleted++; \
+		if (pgstat_collect_blocklevel && (rel)->pgstat_info != NULL)	\
+			(rel)->pgstat_info->t_counts.t_blocks_fetched++;			\
 	} while (0)
-#define pgstat_count_index_scan(s)										\
+#define pgstat_count_buffer_hit(rel)									\
 	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_numscans++;		\
-	} while (0)
-#define pgstat_count_index_tuples(s, n)									\
-	do {																\
-		if (pgstat_collect_tuplelevel && (s)->tabentry != NULL)			\
-			((PgStat_TableEntry *)((s)->tabentry))->t_tuples_returned += (n); \
-	} while (0)
-#define pgstat_count_buffer_read(s,r)									\
-	do {																\
-		if (pgstat_collect_blocklevel) {								\
-			if ((s)->tabentry != NULL)									\
-				((PgStat_TableEntry *)((s)->tabentry))->t_blocks_fetched++; \
-			else {														\
-				pgstat_initstats((s), (r));								\
-				if ((s)->tabentry != NULL)								\
-					((PgStat_TableEntry *)((s)->tabentry))->t_blocks_fetched++; \
-			}															\
-		}																\
-	} while (0)
-#define pgstat_count_buffer_hit(s,r)									\
-	do {																\
-		if (pgstat_collect_blocklevel) {								\
-			if ((s)->tabentry != NULL)									\
-				((PgStat_TableEntry *)((s)->tabentry))->t_blocks_hit++; \
-			else {														\
-				pgstat_initstats((s), (r));								\
-				if ((s)->tabentry != NULL)								\
-					((PgStat_TableEntry *)((s)->tabentry))->t_blocks_hit++; \
-			}															\
-		}																\
+		if (pgstat_collect_blocklevel && (rel)->pgstat_info != NULL)	\
+			(rel)->pgstat_info->t_counts.t_blocks_hit++;				\
 	} while (0)
 
+extern void pgstat_count_heap_insert(Relation rel);
+extern void pgstat_count_heap_update(Relation rel);
+extern void pgstat_count_heap_delete(Relation rel);
+
+extern void AtEOXact_PgStat(bool isCommit);
+extern void AtEOSubXact_PgStat(bool isCommit, int nestDepth);
+
+extern void AtPrepare_PgStat(void);
+extern void PostPrepare_PgStat(void);
+
+extern void pgstat_twophase_postcommit(TransactionId xid, uint16 info,
+									   void *recdata, uint32 len);
+extern void pgstat_twophase_postabort(TransactionId xid, uint16 info,
+									  void *recdata, uint32 len);
 
-extern void pgstat_count_xact_commit(void);
-extern void pgstat_count_xact_rollback(void);
 extern void pgstat_send_bgwriter(void);
 
 /* ----------
diff --git a/src/include/utils/rel.h b/src/include/utils/rel.h
index 33795de2bf8..bc6bf190b86 100644
--- a/src/include/utils/rel.h
+++ b/src/include/utils/rel.h
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2007, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/utils/rel.h,v 1.100 2007/03/29 00:15:39 tgl Exp $
+ * $PostgreSQL: pgsql/src/include/utils/rel.h,v 1.101 2007/05/27 03:50:39 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -89,15 +89,6 @@ typedef struct TriggerDesc
 } TriggerDesc;
 
 
-/*
- * Same for the statistics collector data in Relation and scan data.
- */
-typedef struct PgStat_Info
-{
-	void	   *tabentry;
-} PgStat_Info;
-
-
 /*
  * Cached lookup information for the index access method functions defined
  * by the pg_am row associated with an index relation.
@@ -200,8 +191,8 @@ typedef struct RelationData
 	List	   *rd_indpred;		/* index predicate tree, if any */
 	void	   *rd_amcache;		/* available for use by index AM */
 
-	/* statistics collection area */
-	PgStat_Info pgstat_info;
+	/* use "struct" here to avoid needing to include pgstat.h: */
+	struct PgStat_TableStatus *pgstat_info;	/* statistics collection area */
 } RelationData;
 
 typedef RelationData *Relation;