Refactor XLogOpenRelation() and XLogReadBuffer() in preparation for relation

forks. XLogOpenRelation() and the associated light-weight relation cache in xlogutils.c is gone, and XLogReadBuffer() now takes a RelFileNode as argument, instead of Relation. For functions that still need a Relation struct during WAL replay, there's a new function called CreateFakeRelcacheEntry() that returns a fake entry like XLogOpenRelation() used to.
2025-11-10 17:42:29 +03:00 · 2008-06-12 09:12:31 +00:00
parent c4f2a0458d
commit a213f1ee6c
20 changed files with 295 additions and 427 deletions
--- a/src/backend/access/gin/ginxlog.c
+++ b/src/backend/access/gin/ginxlog.c
@@ -8,7 +8,7 @@
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
- *			 $PostgreSQL: pgsql/src/backend/access/gin/ginxlog.c,v 1.13 2008/05/12 00:00:44 alvherre Exp $
+ *			 $PostgreSQL: pgsql/src/backend/access/gin/ginxlog.c,v 1.14 2008/06/12 09:12:29 heikki Exp $
 *-------------------------------------------------------------------------
 */
 #include "postgres.h"
@@ -71,12 +71,10 @@ static void
 ginRedoCreateIndex(XLogRecPtr lsn, XLogRecord *record)
 {
 	RelFileNode *node = (RelFileNode *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

-	reln = XLogOpenRelation(*node);
-	buffer = XLogReadBuffer(reln, GIN_ROOT_BLKNO, true);
+	buffer = XLogReadBuffer(*node, GIN_ROOT_BLKNO, true);
 	Assert(BufferIsValid(buffer));
 	page = (Page) BufferGetPage(buffer);

@@ -94,12 +92,10 @@ ginRedoCreatePTree(XLogRecPtr lsn, XLogRecord *record)
 {
 	ginxlogCreatePostingTree *data = (ginxlogCreatePostingTree *) XLogRecGetData(record);
 	ItemPointerData *items = (ItemPointerData *) (XLogRecGetData(record) + sizeof(ginxlogCreatePostingTree));
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

-	reln = XLogOpenRelation(data->node);
-	buffer = XLogReadBuffer(reln, data->blkno, true);
+	buffer = XLogReadBuffer(data->node, data->blkno, true);
 	Assert(BufferIsValid(buffer));
 	page = (Page) BufferGetPage(buffer);

@@ -118,7 +114,6 @@ static void
 ginRedoInsert(XLogRecPtr lsn, XLogRecord *record)
 {
 	ginxlogInsert *data = (ginxlogInsert *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

@@ -126,8 +121,7 @@ ginRedoInsert(XLogRecPtr lsn, XLogRecord *record)
 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	reln = XLogOpenRelation(data->node);
-	buffer = XLogReadBuffer(reln, data->blkno, false);
+	buffer = XLogReadBuffer(data->node, data->blkno, false);
 	Assert(BufferIsValid(buffer));
 	page = (Page) BufferGetPage(buffer);

@@ -228,26 +222,23 @@ static void
 ginRedoSplit(XLogRecPtr lsn, XLogRecord *record)
 {
 	ginxlogSplit *data = (ginxlogSplit *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		lbuffer,
 				rbuffer;
 	Page		lpage,
 				rpage;
 	uint32		flags = 0;

-	reln = XLogOpenRelation(data->node);
-
 	if (data->isLeaf)
 		flags |= GIN_LEAF;
 	if (data->isData)
 		flags |= GIN_DATA;

-	lbuffer = XLogReadBuffer(reln, data->lblkno, data->isRootSplit);
+	lbuffer = XLogReadBuffer(data->node, data->lblkno, data->isRootSplit);
 	Assert(BufferIsValid(lbuffer));
 	lpage = (Page) BufferGetPage(lbuffer);
 	GinInitBuffer(lbuffer, flags);

-	rbuffer = XLogReadBuffer(reln, data->rblkno, true);
+	rbuffer = XLogReadBuffer(data->node, data->rblkno, true);
 	Assert(BufferIsValid(rbuffer));
 	rpage = (Page) BufferGetPage(rbuffer);
 	GinInitBuffer(rbuffer, flags);
@@ -319,7 +310,7 @@ ginRedoSplit(XLogRecPtr lsn, XLogRecord *record)

 	if (data->isRootSplit)
 	{
-		Buffer		rootBuf = XLogReadBuffer(reln, data->rootBlkno, false);
+		Buffer		rootBuf = XLogReadBuffer(data->node, data->rootBlkno, false);
 		Page		rootPage = BufferGetPage(rootBuf);

 		GinInitBuffer(rootBuf, flags & ~GIN_LEAF);
@@ -352,7 +343,6 @@ static void
 ginRedoVacuumPage(XLogRecPtr lsn, XLogRecord *record)
 {
 	ginxlogVacuumPage *data = (ginxlogVacuumPage *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

@@ -360,8 +350,7 @@ ginRedoVacuumPage(XLogRecPtr lsn, XLogRecord *record)
 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	reln = XLogOpenRelation(data->node);
-	buffer = XLogReadBuffer(reln, data->blkno, false);
+	buffer = XLogReadBuffer(data->node, data->blkno, false);
 	Assert(BufferIsValid(buffer));
 	page = (Page) BufferGetPage(buffer);

@@ -403,15 +392,12 @@ static void
 ginRedoDeletePage(XLogRecPtr lsn, XLogRecord *record)
 {
 	ginxlogDeletePage *data = (ginxlogDeletePage *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

-	reln = XLogOpenRelation(data->node);
-
 	if (!(record->xl_info & XLR_BKP_BLOCK_1))
 	{
-		buffer = XLogReadBuffer(reln, data->blkno, false);
+		buffer = XLogReadBuffer(data->node, data->blkno, false);
 		page = BufferGetPage(buffer);
 		Assert(GinPageIsData(page));
 		GinPageGetOpaque(page)->flags = GIN_DELETED;
@@ -423,7 +409,7 @@ ginRedoDeletePage(XLogRecPtr lsn, XLogRecord *record)

 	if (!(record->xl_info & XLR_BKP_BLOCK_2))
 	{
-		buffer = XLogReadBuffer(reln, data->parentBlkno, false);
+		buffer = XLogReadBuffer(data->node, data->parentBlkno, false);
 		page = BufferGetPage(buffer);
 		Assert(GinPageIsData(page));
 		Assert(!GinPageIsLeaf(page));
@@ -436,7 +422,7 @@ ginRedoDeletePage(XLogRecPtr lsn, XLogRecord *record)

 	if (!(record->xl_info & XLR_BKP_BLOCK_3) && data->leftBlkno != InvalidBlockNumber)
 	{
-		buffer = XLogReadBuffer(reln, data->leftBlkno, false);
+		buffer = XLogReadBuffer(data->node, data->leftBlkno, false);
 		page = BufferGetPage(buffer);
 		Assert(GinPageIsData(page));
 		GinPageGetOpaque(page)->rightlink = data->rightLink;
@@ -557,9 +543,9 @@ ginContinueSplit(ginIncompleteSplit *split)
 	 * elog(NOTICE,"ginContinueSplit root:%u l:%u r:%u",  split->rootBlkno,
 	 * split->leftBlkno, split->rightBlkno);
 	 */
-	reln = XLogOpenRelation(split->node);
+	buffer = XLogReadBuffer(split->node, split->leftBlkno, false);

-	buffer = XLogReadBuffer(reln, split->leftBlkno, false);
+	reln = CreateFakeRelcacheEntry(split->node);

 	if (split->rootBlkno == GIN_ROOT_BLKNO)
 	{
@@ -581,6 +567,8 @@ ginContinueSplit(ginIncompleteSplit *split)
 									   GinPageGetOpaque(page)->maxoff))->key;
 	}

+	FreeFakeRelcacheEntry(reln);
+
 	btree.rightblkno = split->rightBlkno;

 	stack.blkno = split->leftBlkno;
--- a/src/backend/access/gist/gist.c
+++ b/src/backend/access/gist/gist.c
@@ -8,7 +8,7 @@
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/gist/gist.c,v 1.150 2008/05/12 00:00:44 alvherre Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/gist/gist.c,v 1.151 2008/06/12 09:12:29 heikki Exp $
 *
 *-------------------------------------------------------------------------
 */
@@ -462,7 +462,7 @@ gistplacetopage(GISTInsertState *state, GISTSTATE *giststate)

 		if (!is_leaf)
 			PageIndexTupleDelete(state->stack->page, state->stack->childoffnum);
-		gistfillbuffer(state->r, state->stack->page, state->itup, state->ituplen, InvalidOffsetNumber);
+		gistfillbuffer(state->stack->page, state->itup, state->ituplen, InvalidOffsetNumber);

 		MarkBufferDirty(state->stack->buffer);

@@ -1008,7 +1008,7 @@ gistnewroot(Relation r, Buffer buffer, IndexTuple *itup, int len, ItemPointer ke
 	START_CRIT_SECTION();

 	GISTInitBuffer(buffer, 0);
-	gistfillbuffer(r, page, itup, len, FirstOffsetNumber);
+	gistfillbuffer(page, itup, len, FirstOffsetNumber);

 	MarkBufferDirty(buffer);

--- a/src/backend/access/gist/gistutil.c
+++ b/src/backend/access/gist/gistutil.c
@@ -8,7 +8,7 @@
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
- *			$PostgreSQL: pgsql/src/backend/access/gist/gistutil.c,v 1.26 2008/05/12 00:00:44 alvherre Exp $
+ *			$PostgreSQL: pgsql/src/backend/access/gist/gistutil.c,v 1.27 2008/06/12 09:12:29 heikki Exp $
 *-------------------------------------------------------------------------
 */
 #include "postgres.h"
@@ -27,11 +27,10 @@ static Datum attrS[INDEX_MAX_KEYS];
 static bool isnullS[INDEX_MAX_KEYS];

 /*
- * Write itup vector to page, has no control of free space
+ * Write itup vector to page, has no control of free space.
 */
-OffsetNumber
-gistfillbuffer(Relation r, Page page, IndexTuple *itup,
-			   int len, OffsetNumber off)
+void
+gistfillbuffer(Page page, IndexTuple *itup, int len, OffsetNumber off)
 {
 	OffsetNumber l = InvalidOffsetNumber;
 	int			i;
@@ -42,14 +41,13 @@ gistfillbuffer(Relation r, Page page, IndexTuple *itup,

 	for (i = 0; i < len; i++)
 	{
-		l = PageAddItem(page, (Item) itup[i], IndexTupleSize(itup[i]),
-						off, false, false);
+		Size sz = IndexTupleSize(itup[i]);
+		l = PageAddItem(page, (Item) itup[i], sz, off, false, false);
 		if (l == InvalidOffsetNumber)
-			elog(ERROR, "failed to add item to index page in \"%s\"",
-				 RelationGetRelationName(r));
+			elog(ERROR, "failed to add item to GiST index page, item %d out of %d, size %d bytes",
+				 i, len, sz);
 		off++;
 	}
-	return l;
 }

 /*
--- a/src/backend/access/gist/gistvacuum.c
+++ b/src/backend/access/gist/gistvacuum.c
@@ -8,7 +8,7 @@
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/gist/gistvacuum.c,v 1.35 2008/05/12 00:00:44 alvherre Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/gist/gistvacuum.c,v 1.36 2008/06/12 09:12:30 heikki Exp $
 *
 *-------------------------------------------------------------------------
 */
@@ -403,7 +403,7 @@ gistVacuumUpdate(GistVacuum *gv, BlockNumber blkno, bool needunion)
 			}
 			else
 				/* enough free space */
-				gistfillbuffer(gv->index, tempPage, addon, curlenaddon, InvalidOffsetNumber);
+				gistfillbuffer(tempPage, addon, curlenaddon, InvalidOffsetNumber);
 		}
 	}

--- a/src/backend/access/gist/gistxlog.c
+++ b/src/backend/access/gist/gistxlog.c
@@ -8,7 +8,7 @@
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
- *			 $PostgreSQL: pgsql/src/backend/access/gist/gistxlog.c,v 1.28 2008/05/12 00:00:44 alvherre Exp $
+ *			 $PostgreSQL: pgsql/src/backend/access/gist/gistxlog.c,v 1.29 2008/06/12 09:12:30 heikki Exp $
 *-------------------------------------------------------------------------
 */
 #include "postgres.h"
@@ -189,7 +189,6 @@ gistRedoPageUpdateRecord(XLogRecPtr lsn, XLogRecord *record, bool isnewroot)
 {
 	gistxlogPageUpdate *xldata = (gistxlogPageUpdate *) XLogRecGetData(record);
 	PageUpdateRecord xlrec;
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

@@ -208,8 +207,7 @@ gistRedoPageUpdateRecord(XLogRecPtr lsn, XLogRecord *record, bool isnewroot)

 	decodePageUpdateRecord(&xlrec, record);

-	reln = XLogOpenRelation(xlrec.data->node);
-	buffer = XLogReadBuffer(reln, xlrec.data->blkno, false);
+	buffer = XLogReadBuffer(xlrec.data->node, xlrec.data->blkno, false);
 	if (!BufferIsValid(buffer))
 		return;
 	page = (Page) BufferGetPage(buffer);
@@ -234,7 +232,7 @@ gistRedoPageUpdateRecord(XLogRecPtr lsn, XLogRecord *record, bool isnewroot)

 	/* add tuples */
 	if (xlrec.len > 0)
-		gistfillbuffer(reln, page, xlrec.itup, xlrec.len, InvalidOffsetNumber);
+		gistfillbuffer(page, xlrec.itup, xlrec.len, InvalidOffsetNumber);

 	/*
 	 * special case: leafpage, nothing to insert, nothing to delete, then
@@ -262,7 +260,6 @@ static void
 gistRedoPageDeleteRecord(XLogRecPtr lsn, XLogRecord *record)
 {
 	gistxlogPageDelete *xldata = (gistxlogPageDelete *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

@@ -270,8 +267,7 @@ gistRedoPageDeleteRecord(XLogRecPtr lsn, XLogRecord *record)
 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	reln = XLogOpenRelation(xldata->node);
-	buffer = XLogReadBuffer(reln, xldata->blkno, false);
+	buffer = XLogReadBuffer(xldata->node, xldata->blkno, false);
 	if (!BufferIsValid(buffer))
 		return;

@@ -319,14 +315,12 @@ static void
 gistRedoPageSplitRecord(XLogRecPtr lsn, XLogRecord *record)
 {
 	PageSplitRecord xlrec;
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	int			i;
 	int			flags;

 	decodePageSplitRecord(&xlrec, record);
-	reln = XLogOpenRelation(xlrec.data->node);
 	flags = xlrec.data->origleaf ? F_LEAF : 0;

 	/* loop around all pages */
@@ -334,7 +328,7 @@ gistRedoPageSplitRecord(XLogRecPtr lsn, XLogRecord *record)
 	{
 		NewPage    *newpage = xlrec.page + i;

-		buffer = XLogReadBuffer(reln, newpage->header->blkno, true);
+		buffer = XLogReadBuffer(xlrec.data->node, newpage->header->blkno, true);
 		Assert(BufferIsValid(buffer));
 		page = (Page) BufferGetPage(buffer);

@@ -342,7 +336,7 @@ gistRedoPageSplitRecord(XLogRecPtr lsn, XLogRecord *record)
 		GISTInitBuffer(buffer, flags);

 		/* and fill it */
-		gistfillbuffer(reln, page, newpage->itup, newpage->header->num, FirstOffsetNumber);
+		gistfillbuffer(page, newpage->itup, newpage->header->num, FirstOffsetNumber);

 		PageSetLSN(page, lsn);
 		PageSetTLI(page, ThisTimeLineID);
@@ -361,12 +355,10 @@ static void
 gistRedoCreateIndex(XLogRecPtr lsn, XLogRecord *record)
 {
 	RelFileNode *node = (RelFileNode *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

-	reln = XLogOpenRelation(*node);
-	buffer = XLogReadBuffer(reln, GIST_ROOT_BLKNO, true);
+	buffer = XLogReadBuffer(*node, GIST_ROOT_BLKNO, true);
 	Assert(BufferIsValid(buffer));
 	page = (Page) BufferGetPage(buffer);

@@ -602,7 +594,7 @@ gistContinueInsert(gistIncompleteInsert *insert)
 				lenitup;
 	Relation	index;

-	index = XLogOpenRelation(insert->node);
+	index = CreateFakeRelcacheEntry(insert->node);

 	/*
 	 * needed vector itup never will be more than initial lenblkno+2, because
@@ -624,7 +616,7 @@ gistContinueInsert(gistIncompleteInsert *insert)
 		 * it was split root, so we should only make new root. it can't be
 		 * simple insert into root, we should replace all content of root.
 		 */
-		Buffer		buffer = XLogReadBuffer(index, GIST_ROOT_BLKNO, true);
+		Buffer		buffer = XLogReadBuffer(insert->node, GIST_ROOT_BLKNO, true);

 		gistnewroot(index, buffer, itup, lenitup, NULL);
 		UnlockReleaseBuffer(buffer);
@@ -703,7 +695,7 @@ gistContinueInsert(gistIncompleteInsert *insert)
 				LockBuffer(buffers[numbuffer], GIST_EXCLUSIVE);
 				GISTInitBuffer(buffers[numbuffer], 0);
 				pages[numbuffer] = BufferGetPage(buffers[numbuffer]);
-				gistfillbuffer(index, pages[numbuffer], itup, lenitup, FirstOffsetNumber);
+				gistfillbuffer(pages[numbuffer], itup, lenitup, FirstOffsetNumber);
 				numbuffer++;

 				if (BufferGetBlockNumber(buffers[0]) == GIST_ROOT_BLKNO)
@@ -749,7 +741,7 @@ gistContinueInsert(gistIncompleteInsert *insert)

 				for (j = 0; j < ntodelete; j++)
 					PageIndexTupleDelete(pages[0], todelete[j]);
-				gistfillbuffer(index, pages[0], itup, lenitup, InvalidOffsetNumber);
+				gistfillbuffer(pages[0], itup, lenitup, InvalidOffsetNumber);

 				rdata = formUpdateRdata(index->rd_node, buffers[0],
 										todelete, ntodelete,
@@ -794,6 +786,8 @@ gistContinueInsert(gistIncompleteInsert *insert)
 		}
 	}

+	FreeFakeRelcacheEntry(index);
+
 	ereport(LOG,
 			(errmsg("index %u/%u/%u needs VACUUM FULL or REINDEX to finish crash recovery",
 			insert->node.spcNode, insert->node.dbNode, insert->node.relNode),
--- 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 1.258 2008/06/08 22:00:47 alvherre Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/heap/heapam.c,v 1.259 2008/06/12 09:12:30 heikki Exp $
 *
 *
 * INTERFACE ROUTINES
@@ -3944,7 +3944,6 @@ static void
 heap_xlog_clean(XLogRecPtr lsn, XLogRecord *record, bool clean_move)
 {
 	xl_heap_clean *xlrec = (xl_heap_clean *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	OffsetNumber *end;
@@ -3958,8 +3957,7 @@ heap_xlog_clean(XLogRecPtr lsn, XLogRecord *record, bool clean_move)
 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	reln = XLogOpenRelation(xlrec->node);
-	buffer = XLogReadBuffer(reln, xlrec->block, false);
+	buffer = XLogReadBuffer(xlrec->node, xlrec->block, false);
 	if (!BufferIsValid(buffer))
 		return;
 	page = (Page) BufferGetPage(buffer);
@@ -3980,7 +3978,7 @@ heap_xlog_clean(XLogRecPtr lsn, XLogRecord *record, bool clean_move)
 	Assert(nunused >= 0);

 	/* Update all item pointers per the record, and repair fragmentation */
-	heap_page_prune_execute(reln, buffer,
+	heap_page_prune_execute(buffer,
 							redirected, nredirected,
 							nowdead, ndead,
 							nowunused, nunused,
@@ -4002,15 +4000,13 @@ heap_xlog_freeze(XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_heap_freeze *xlrec = (xl_heap_freeze *) XLogRecGetData(record);
 	TransactionId cutoff_xid = xlrec->cutoff_xid;
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	reln = XLogOpenRelation(xlrec->node);
-	buffer = XLogReadBuffer(reln, xlrec->block, false);
+	buffer = XLogReadBuffer(xlrec->node, xlrec->block, false);
 	if (!BufferIsValid(buffer))
 		return;
 	page = (Page) BufferGetPage(buffer);
@@ -4050,7 +4046,6 @@ static void
 heap_xlog_newpage(XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_heap_newpage *xlrec = (xl_heap_newpage *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;

@@ -4058,8 +4053,7 @@ heap_xlog_newpage(XLogRecPtr lsn, XLogRecord *record)
 	 * Note: the NEWPAGE log record is used for both heaps and indexes, so do
 	 * not do anything that assumes we are touching a heap.
 	 */
-	reln = XLogOpenRelation(xlrec->node);
-	buffer = XLogReadBuffer(reln, xlrec->blkno, true);
+	buffer = XLogReadBuffer(xlrec->node, xlrec->blkno, true);
 	Assert(BufferIsValid(buffer));
 	page = (Page) BufferGetPage(buffer);

@@ -4076,7 +4070,6 @@ static void
 heap_xlog_delete(XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_heap_delete *xlrec = (xl_heap_delete *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	OffsetNumber offnum;
@@ -4086,8 +4079,7 @@ heap_xlog_delete(XLogRecPtr lsn, XLogRecord *record)
 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	reln = XLogOpenRelation(xlrec->target.node);
-	buffer = XLogReadBuffer(reln,
+	buffer = XLogReadBuffer(xlrec->target.node,
 							ItemPointerGetBlockNumber(&(xlrec->target.tid)),
 							false);
 	if (!BufferIsValid(buffer))
@@ -4133,7 +4125,6 @@ static void
 heap_xlog_insert(XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_heap_insert *xlrec = (xl_heap_insert *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	OffsetNumber offnum;
@@ -4149,11 +4140,9 @@ heap_xlog_insert(XLogRecPtr lsn, XLogRecord *record)
 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	reln = XLogOpenRelation(xlrec->target.node);
-
 	if (record->xl_info & XLOG_HEAP_INIT_PAGE)
 	{
-		buffer = XLogReadBuffer(reln,
+		buffer = XLogReadBuffer(xlrec->target.node,
 							 ItemPointerGetBlockNumber(&(xlrec->target.tid)),
 								true);
 		Assert(BufferIsValid(buffer));
@@ -4163,7 +4152,7 @@ heap_xlog_insert(XLogRecPtr lsn, XLogRecord *record)
 	}
 	else
 	{
-		buffer = XLogReadBuffer(reln,
+		buffer = XLogReadBuffer(xlrec->target.node,
 							 ItemPointerGetBlockNumber(&(xlrec->target.tid)),
 								false);
 		if (!BufferIsValid(buffer))
@@ -4216,7 +4205,6 @@ static void
 heap_xlog_update(XLogRecPtr lsn, XLogRecord *record, bool move, bool hot_update)
 {
 	xl_heap_update *xlrec = (xl_heap_update *) XLogRecGetData(record);
-	Relation	reln = XLogOpenRelation(xlrec->target.node);
 	Buffer		buffer;
 	bool		samepage = (ItemPointerGetBlockNumber(&(xlrec->newtid)) ==
 							ItemPointerGetBlockNumber(&(xlrec->target.tid)));
@@ -4242,7 +4230,7 @@ heap_xlog_update(XLogRecPtr lsn, XLogRecord *record, bool move, bool hot_update)

 	/* Deal with old tuple version */

-	buffer = XLogReadBuffer(reln,
+	buffer = XLogReadBuffer(xlrec->target.node,
 							ItemPointerGetBlockNumber(&(xlrec->target.tid)),
 							false);
 	if (!BufferIsValid(buffer))
@@ -4317,7 +4305,7 @@ newt:;

 	if (record->xl_info & XLOG_HEAP_INIT_PAGE)
 	{
-		buffer = XLogReadBuffer(reln,
+		buffer = XLogReadBuffer(xlrec->target.node,
 								ItemPointerGetBlockNumber(&(xlrec->newtid)),
 								true);
 		Assert(BufferIsValid(buffer));
@@ -4327,7 +4315,7 @@ newt:;
 	}
 	else
 	{
-		buffer = XLogReadBuffer(reln,
+		buffer = XLogReadBuffer(xlrec->target.node,
 								ItemPointerGetBlockNumber(&(xlrec->newtid)),
 								false);
 		if (!BufferIsValid(buffer))
@@ -4399,7 +4387,6 @@ static void
 heap_xlog_lock(XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_heap_lock *xlrec = (xl_heap_lock *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	OffsetNumber offnum;
@@ -4409,8 +4396,7 @@ heap_xlog_lock(XLogRecPtr lsn, XLogRecord *record)
 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	reln = XLogOpenRelation(xlrec->target.node);
-	buffer = XLogReadBuffer(reln,
+	buffer = XLogReadBuffer(xlrec->target.node,
 							ItemPointerGetBlockNumber(&(xlrec->target.tid)),
 							false);
 	if (!BufferIsValid(buffer))
@@ -4458,7 +4444,6 @@ static void
 heap_xlog_inplace(XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_heap_inplace *xlrec = (xl_heap_inplace *) XLogRecGetData(record);
-	Relation	reln = XLogOpenRelation(xlrec->target.node);
 	Buffer		buffer;
 	Page		page;
 	OffsetNumber offnum;
@@ -4470,7 +4455,7 @@ heap_xlog_inplace(XLogRecPtr lsn, XLogRecord *record)
 	if (record->xl_info & XLR_BKP_BLOCK_1)
 		return;

-	buffer = XLogReadBuffer(reln,
+	buffer = XLogReadBuffer(xlrec->target.node,
 							ItemPointerGetBlockNumber(&(xlrec->target.tid)),
 							false);
 	if (!BufferIsValid(buffer))
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -8,7 +8,7 @@
 *
 *
 * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/heap/pruneheap.c,v 1.13 2008/06/08 22:00:47 alvherre Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/heap/pruneheap.c,v 1.14 2008/06/12 09:12:30 heikki Exp $
 *
 *-------------------------------------------------------------------------
 */
@@ -225,7 +225,7 @@ heap_page_prune(Relation relation, Buffer buffer, TransactionId OldestXmin,
 		 * and update the page's hint bit about whether it has free line
 		 * pointers.
 		 */
-		heap_page_prune_execute(relation, buffer,
+		heap_page_prune_execute(buffer,
 								prstate.redirected, prstate.nredirected,
 								prstate.nowdead, prstate.ndead,
 								prstate.nowunused, prstate.nunused,
@@ -696,7 +696,7 @@ heap_prune_record_unused(PruneState *prstate, OffsetNumber offnum)
 * arguments are identical to those of log_heap_clean().
 */
 void
-heap_page_prune_execute(Relation reln, Buffer buffer,
+heap_page_prune_execute(Buffer buffer,
 						OffsetNumber *redirected, int nredirected,
 						OffsetNumber *nowdead, int ndead,
 						OffsetNumber *nowunused, int nunused,
--- a/src/backend/access/nbtree/nbtxlog.c
+++ b/src/backend/access/nbtree/nbtxlog.c
@@ -8,7 +8,7 @@
 * Portions Copyright (c) 1994, Regents of the University of California
 *
 * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/nbtree/nbtxlog.c,v 1.51 2008/05/12 00:00:45 alvherre Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/nbtree/nbtxlog.c,v 1.52 2008/06/12 09:12:30 heikki Exp $
 *
 *-------------------------------------------------------------------------
 */
@@ -150,7 +150,7 @@ _bt_restore_page(Page page, char *from, int len)
 }

 static void
-_bt_restore_meta(Relation reln, XLogRecPtr lsn,
+_bt_restore_meta(RelFileNode rnode, XLogRecPtr lsn,
 				 BlockNumber root, uint32 level,
 				 BlockNumber fastroot, uint32 fastlevel)
 {
@@ -159,7 +159,7 @@ _bt_restore_meta(Relation reln, XLogRecPtr lsn,
 	BTMetaPageData *md;
 	BTPageOpaque pageop;

-	metabuf = XLogReadBuffer(reln, BTREE_METAPAGE, true);
+	metabuf = XLogReadBuffer(rnode, BTREE_METAPAGE, true);
 	Assert(BufferIsValid(metabuf));
 	metapg = BufferGetPage(metabuf);

@@ -194,7 +194,6 @@ btree_xlog_insert(bool isleaf, bool ismeta,
 				  XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_btree_insert *xlrec = (xl_btree_insert *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	char	   *datapos;
@@ -220,11 +219,9 @@ btree_xlog_insert(bool isleaf, bool ismeta,
 	if ((record->xl_info & XLR_BKP_BLOCK_1) && !ismeta && isleaf)
 		return;					/* nothing to do */

-	reln = XLogOpenRelation(xlrec->target.node);
-
 	if (!(record->xl_info & XLR_BKP_BLOCK_1))
 	{
-		buffer = XLogReadBuffer(reln,
+		buffer = XLogReadBuffer(xlrec->target.node,
 							 ItemPointerGetBlockNumber(&(xlrec->target.tid)),
 								false);
 		if (BufferIsValid(buffer))
@@ -251,7 +248,7 @@ btree_xlog_insert(bool isleaf, bool ismeta,
 	}

 	if (ismeta)
-		_bt_restore_meta(reln, lsn,
+		_bt_restore_meta(xlrec->target.node, lsn,
 						 md.root, md.level,
 						 md.fastroot, md.fastlevel);

@@ -265,7 +262,6 @@ btree_xlog_split(bool onleft, bool isroot,
 				 XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_btree_split *xlrec = (xl_btree_split *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		rbuf;
 	Page		rpage;
 	BTPageOpaque ropaque;
@@ -277,8 +273,6 @@ btree_xlog_split(bool onleft, bool isroot,
 	Item		left_hikey = NULL;
 	Size		left_hikeysz = 0;

-	reln = XLogOpenRelation(xlrec->node);
-
 	datapos = (char *) xlrec + SizeOfBtreeSplit;
 	datalen = record->xl_len - SizeOfBtreeSplit;

@@ -328,7 +322,7 @@ btree_xlog_split(bool onleft, bool isroot,
 	}

 	/* Reconstruct right (new) sibling from scratch */
-	rbuf = XLogReadBuffer(reln, xlrec->rightsib, true);
+	rbuf = XLogReadBuffer(xlrec->node, xlrec->rightsib, true);
 	Assert(BufferIsValid(rbuf));
 	rpage = (Page) BufferGetPage(rbuf);

@@ -369,7 +363,7 @@ btree_xlog_split(bool onleft, bool isroot,
 	 */
 	if (!(record->xl_info & XLR_BKP_BLOCK_1))
 	{
-		Buffer		lbuf = XLogReadBuffer(reln, xlrec->leftsib, false);
+		Buffer		lbuf = XLogReadBuffer(xlrec->node, xlrec->leftsib, false);

 		if (BufferIsValid(lbuf))
 		{
@@ -439,7 +433,7 @@ btree_xlog_split(bool onleft, bool isroot,
 	/* Fix left-link of the page to the right of the new right sibling */
 	if (xlrec->rnext != P_NONE && !(record->xl_info & XLR_BKP_BLOCK_2))
 	{
-		Buffer		buffer = XLogReadBuffer(reln, xlrec->rnext, false);
+		Buffer		buffer = XLogReadBuffer(xlrec->node, xlrec->rnext, false);

 		if (BufferIsValid(buffer))
 		{
@@ -468,7 +462,6 @@ static void
 btree_xlog_delete(XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_btree_delete *xlrec;
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	BTPageOpaque opaque;
@@ -477,8 +470,7 @@ btree_xlog_delete(XLogRecPtr lsn, XLogRecord *record)
 		return;

 	xlrec = (xl_btree_delete *) XLogRecGetData(record);
-	reln = XLogOpenRelation(xlrec->node);
-	buffer = XLogReadBuffer(reln, xlrec->block, false);
+	buffer = XLogReadBuffer(xlrec->node, xlrec->block, false);
 	if (!BufferIsValid(buffer))
 		return;
 	page = (Page) BufferGetPage(buffer);
@@ -517,7 +509,6 @@ static void
 btree_xlog_delete_page(uint8 info, XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_btree_delete_page *xlrec = (xl_btree_delete_page *) XLogRecGetData(record);
-	Relation	reln;
 	BlockNumber parent;
 	BlockNumber target;
 	BlockNumber leftsib;
@@ -526,7 +517,6 @@ btree_xlog_delete_page(uint8 info, XLogRecPtr lsn, XLogRecord *record)
 	Page		page;
 	BTPageOpaque pageop;

-	reln = XLogOpenRelation(xlrec->target.node);
 	parent = ItemPointerGetBlockNumber(&(xlrec->target.tid));
 	target = xlrec->deadblk;
 	leftsib = xlrec->leftblk;
@@ -535,7 +525,7 @@ btree_xlog_delete_page(uint8 info, XLogRecPtr lsn, XLogRecord *record)
 	/* parent page */
 	if (!(record->xl_info & XLR_BKP_BLOCK_1))
 	{
-		buffer = XLogReadBuffer(reln, parent, false);
+		buffer = XLogReadBuffer(xlrec->target.node, parent, false);
 		if (BufferIsValid(buffer))
 		{
 			page = (Page) BufferGetPage(buffer);
@@ -581,7 +571,7 @@ btree_xlog_delete_page(uint8 info, XLogRecPtr lsn, XLogRecord *record)
 	/* Fix left-link of right sibling */
 	if (!(record->xl_info & XLR_BKP_BLOCK_2))
 	{
-		buffer = XLogReadBuffer(reln, rightsib, false);
+		buffer = XLogReadBuffer(xlrec->target.node, rightsib, false);
 		if (BufferIsValid(buffer))
 		{
 			page = (Page) BufferGetPage(buffer);
@@ -607,7 +597,7 @@ btree_xlog_delete_page(uint8 info, XLogRecPtr lsn, XLogRecord *record)
 	{
 		if (leftsib != P_NONE)
 		{
-			buffer = XLogReadBuffer(reln, leftsib, false);
+			buffer = XLogReadBuffer(xlrec->target.node, leftsib, false);
 			if (BufferIsValid(buffer))
 			{
 				page = (Page) BufferGetPage(buffer);
@@ -630,7 +620,7 @@ btree_xlog_delete_page(uint8 info, XLogRecPtr lsn, XLogRecord *record)
 	}

 	/* Rewrite target page as empty deleted page */
-	buffer = XLogReadBuffer(reln, target, true);
+	buffer = XLogReadBuffer(xlrec->target.node, target, true);
 	Assert(BufferIsValid(buffer));
 	page = (Page) BufferGetPage(buffer);

@@ -655,7 +645,7 @@ btree_xlog_delete_page(uint8 info, XLogRecPtr lsn, XLogRecord *record)

 		memcpy(&md, (char *) xlrec + SizeOfBtreeDeletePage,
 			   sizeof(xl_btree_metadata));
-		_bt_restore_meta(reln, lsn,
+		_bt_restore_meta(xlrec->target.node, lsn,
 						 md.root, md.level,
 						 md.fastroot, md.fastlevel);
 	}
@@ -672,14 +662,12 @@ static void
 btree_xlog_newroot(XLogRecPtr lsn, XLogRecord *record)
 {
 	xl_btree_newroot *xlrec = (xl_btree_newroot *) XLogRecGetData(record);
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	BTPageOpaque pageop;
 	BlockNumber downlink = 0;

-	reln = XLogOpenRelation(xlrec->node);
-	buffer = XLogReadBuffer(reln, xlrec->rootblk, true);
+	buffer = XLogReadBuffer(xlrec->node, xlrec->rootblk, true);
 	Assert(BufferIsValid(buffer));
 	page = (Page) BufferGetPage(buffer);

@@ -711,7 +699,7 @@ btree_xlog_newroot(XLogRecPtr lsn, XLogRecord *record)
 	MarkBufferDirty(buffer);
 	UnlockReleaseBuffer(buffer);

-	_bt_restore_meta(reln, lsn,
+	_bt_restore_meta(xlrec->node, lsn,
 					 xlrec->rootblk, xlrec->level,
 					 xlrec->rootblk, xlrec->level);

@@ -904,9 +892,7 @@ btree_xlog_cleanup(void)
 	foreach(l, incomplete_actions)
 	{
 		bt_incomplete_action *action = (bt_incomplete_action *) lfirst(l);
-		Relation	reln;

-		reln = XLogOpenRelation(action->node);
 		if (action->is_split)
 		{
 			/* finish an incomplete split */
@@ -917,14 +903,15 @@ btree_xlog_cleanup(void)
 			BTPageOpaque lpageop,
 						rpageop;
 			bool		is_only;
+			Relation	reln;

-			lbuf = XLogReadBuffer(reln, action->leftblk, false);
+			lbuf = XLogReadBuffer(action->node, action->leftblk, false);
 			/* failure is impossible because we wrote this page earlier */
 			if (!BufferIsValid(lbuf))
 				elog(PANIC, "btree_xlog_cleanup: left block unfound");
 			lpage = (Page) BufferGetPage(lbuf);
 			lpageop = (BTPageOpaque) PageGetSpecialPointer(lpage);
-			rbuf = XLogReadBuffer(reln, action->rightblk, false);
+			rbuf = XLogReadBuffer(action->node, action->rightblk, false);
 			/* failure is impossible because we wrote this page earlier */
 			if (!BufferIsValid(rbuf))
 				elog(PANIC, "btree_xlog_cleanup: right block unfound");
@@ -934,18 +921,26 @@ btree_xlog_cleanup(void)
 			/* if the pages are all of their level, it's a only-page split */
 			is_only = P_LEFTMOST(lpageop) && P_RIGHTMOST(rpageop);

+			reln = CreateFakeRelcacheEntry(action->node);
 			_bt_insert_parent(reln, lbuf, rbuf, NULL,
 							  action->is_root, is_only);
+			FreeFakeRelcacheEntry(reln);
 		}
 		else
 		{
 			/* finish an incomplete deletion (of a half-dead page) */
 			Buffer		buf;

-			buf = XLogReadBuffer(reln, action->delblk, false);
+			buf = XLogReadBuffer(action->node, action->delblk, false);
 			if (BufferIsValid(buf))
+			{
+				Relation reln;
+
+				reln = CreateFakeRelcacheEntry(action->node);
 				if (_bt_pagedel(reln, buf, NULL, true) == 0)
 					elog(PANIC, "btree_xlog_cleanup: _bt_pagdel failed");
+				FreeFakeRelcacheEntry(reln);
+			}
 		}
 	}
 	incomplete_actions = NIL;
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -7,7 +7,7 @@
 * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
 * Portions Copyright (c) 1994, Regents of the University of California
 *
- * $PostgreSQL: pgsql/src/backend/access/transam/xlog.c,v 1.313 2008/06/08 22:00:47 alvherre Exp $
+ * $PostgreSQL: pgsql/src/backend/access/transam/xlog.c,v 1.314 2008/06/12 09:12:30 heikki Exp $
 *
 *-------------------------------------------------------------------------
 */
@@ -2840,7 +2840,6 @@ CleanupBackupHistory(void)
 static void
 RestoreBkpBlocks(XLogRecord *record, XLogRecPtr lsn)
 {
-	Relation	reln;
 	Buffer		buffer;
 	Page		page;
 	BkpBlock	bkpb;
@@ -2856,8 +2855,7 @@ RestoreBkpBlocks(XLogRecord *record, XLogRecPtr lsn)
 		memcpy(&bkpb, blk, sizeof(BkpBlock));
 		blk += sizeof(BkpBlock);

-		reln = XLogOpenRelation(bkpb.node);
-		buffer = XLogReadBuffer(reln, bkpb.block, true);
+		buffer = XLogReadBuffer(bkpb.node, bkpb.block, true);
 		Assert(BufferIsValid(buffer));
 		page = (Page) BufferGetPage(buffer);

@@ -5064,9 +5062,7 @@ StartupXLOG(void)
 								BACKUP_LABEL_FILE, BACKUP_LABEL_OLD)));
 		}

-		/* Start up the recovery environment */
-		XLogInitRelationCache();
-
+		/* Initialize resource managers */
 		for (rmid = 0; rmid <= RM_MAX_ID; rmid++)
 		{
 			if (RmgrTable[rmid].rm_startup != NULL)
@@ -5330,11 +5326,6 @@ StartupXLOG(void)
 		 * allows some extra error checking in xlog_redo.
 		 */
 		CreateCheckPoint(CHECKPOINT_IS_SHUTDOWN | CHECKPOINT_IMMEDIATE);
-
-		/*
-		 * Close down recovery environment
-		 */
-		XLogCloseRelationCache();
 	}

 	/*
--- a/src/backend/access/transam/xlogutils.c
+++ b/src/backend/access/transam/xlogutils.c
@@ -11,7 +11,7 @@
 * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
 * Portions Copyright (c) 1994, Regents of the University of California
 *
- * $PostgreSQL: pgsql/src/backend/access/transam/xlogutils.c,v 1.54 2008/06/08 22:00:47 alvherre Exp $
+ * $PostgreSQL: pgsql/src/backend/access/transam/xlogutils.c,v 1.55 2008/06/12 09:12:30 heikki Exp $
 *
 *-------------------------------------------------------------------------
 */
@@ -190,6 +190,9 @@ XLogCheckInvalidPages(void)

 	if (foundone)
 		elog(PANIC, "WAL contains references to invalid pages");
+
+	hash_destroy(invalid_page_tab);
+	invalid_page_tab = NULL;
 }


@@ -218,27 +221,40 @@ XLogCheckInvalidPages(void)
 * at the end of WAL replay.)
 */
 Buffer
-XLogReadBuffer(Relation reln, BlockNumber blkno, bool init)
+XLogReadBuffer(RelFileNode rnode, BlockNumber blkno, bool init)
 {
-	BlockNumber lastblock = RelationGetNumberOfBlocks(reln);
+	BlockNumber lastblock;
 	Buffer		buffer;
+	SMgrRelation smgr;

 	Assert(blkno != P_NEW);

+	/* Open the relation at smgr level */
+	smgr = smgropen(rnode);
+
+	/*
+	 * Create the target file if it doesn't already exist.  This lets us cope
+	 * if the replay sequence contains writes to a relation that is later
+	 * deleted.  (The original coding of this routine would instead suppress
+	 * the writes, but that seems like it risks losing valuable data if the
+	 * filesystem loses an inode during a crash.  Better to write the data
+	 * until we are actually told to delete the file.)
+	 */
+	smgrcreate(smgr, false, true);
+
+	lastblock = smgrnblocks(smgr);
+
 	if (blkno < lastblock)
 	{
 		/* page exists in file */
-		if (init)
-			buffer = ReadOrZeroBuffer(reln, blkno);
-		else
-			buffer = ReadBuffer(reln, blkno);
+		buffer = ReadBufferWithoutRelcache(rnode, false, blkno, init);
 	}
 	else
 	{
 		/* hm, page doesn't exist in file */
 		if (!init)
 		{
-			log_invalid_page(reln->rd_node, blkno, false);
+			log_invalid_page(rnode, blkno, false);
 			return InvalidBuffer;
 		}
 		/* OK to extend the file */
@@ -249,7 +265,7 @@ XLogReadBuffer(Relation reln, BlockNumber blkno, bool init)
 		{
 			if (buffer != InvalidBuffer)
 				ReleaseBuffer(buffer);
-			buffer = ReadBuffer(reln, P_NEW);
+			buffer = ReadBufferWithoutRelcache(rnode, false, P_NEW, false);
 			lastblock++;
 		}
 		Assert(BufferGetBlockNumber(buffer) == blkno);
@@ -265,7 +281,7 @@ XLogReadBuffer(Relation reln, BlockNumber blkno, bool init)
 		if (PageIsNew((PageHeader) page))
 		{
 			UnlockReleaseBuffer(buffer);
-			log_invalid_page(reln->rd_node, blkno, true);
+			log_invalid_page(rnode, blkno, true);
 			return InvalidBuffer;
 		}
 	}
@@ -275,226 +291,81 @@ XLogReadBuffer(Relation reln, BlockNumber blkno, bool init)


 /*
- * Lightweight "Relation" cache --- this substitutes for the normal relcache
- * during XLOG replay.
+ * Struct actually returned by XLogFakeRelcacheEntry, though the declared
+ * return type is Relation.
 */
-
-typedef struct XLogRelDesc
+typedef struct
 {
-	RelationData reldata;
-	struct XLogRelDesc *lessRecently;
-	struct XLogRelDesc *moreRecently;
-} XLogRelDesc;
+	RelationData		reldata;	/* Note: this must be first */
+	FormData_pg_class	pgc;
+} FakeRelCacheEntryData;

-typedef struct XLogRelCacheEntry
+typedef FakeRelCacheEntryData *FakeRelCacheEntry;
+
+/*
+ * Create a fake relation cache entry for a physical relation
+ *
+ * It's often convenient to use the same functions in XLOG replay as in the
+ * main codepath, but those functions typically work with a relcache entry. 
+ * We don't have a working relation cache during XLOG replay, but this 
+ * function can be used to create a fake relcache entry instead. Only the 
+ * fields related to physical storage, like rd_rel, are initialized, so the 
+ * fake entry is only usable in low-level operations like ReadBuffer().
+ *
+ * Caller must free the returned entry with FreeFakeRelcacheEntry().
+ */
+Relation
+CreateFakeRelcacheEntry(RelFileNode rnode)
 {
-	RelFileNode rnode;
-	XLogRelDesc *rdesc;
-} XLogRelCacheEntry;
+	FakeRelCacheEntry fakeentry;
+	Relation rel;

-static HTAB *_xlrelcache;
-static XLogRelDesc *_xlrelarr = NULL;
-static Form_pg_class _xlpgcarr = NULL;
-static int	_xlast = 0;
-static int	_xlcnt = 0;
+	/* Allocate the Relation struct and all related space in one block. */
+	fakeentry = palloc0(sizeof(FakeRelCacheEntryData));
+	rel = (Relation) fakeentry;

-#define _XLOG_RELCACHESIZE	512
+	rel->rd_rel = &fakeentry->pgc;
+	rel->rd_node = rnode;

-static void
-_xl_init_rel_cache(void)
-{
-	HASHCTL		ctl;
+	/* We don't know the name of the relation; use relfilenode instead */
+	sprintf(RelationGetRelationName(rel), "%u", rnode.relNode);

-	_xlcnt = _XLOG_RELCACHESIZE;
-	_xlast = 0;
-	_xlrelarr = (XLogRelDesc *) malloc(sizeof(XLogRelDesc) * _xlcnt);
-	memset(_xlrelarr, 0, sizeof(XLogRelDesc) * _xlcnt);
-	_xlpgcarr = (Form_pg_class) malloc(sizeof(FormData_pg_class) * _xlcnt);
-	memset(_xlpgcarr, 0, sizeof(FormData_pg_class) * _xlcnt);
+	/*
+	 * We set up the lockRelId in case anything tries to lock the dummy
+	 * relation.  Note that this is fairly bogus since relNode may be
+	 * different from the relation's OID.  It shouldn't really matter
+	 * though, since we are presumably running by ourselves and can't have
+	 * any lock conflicts ...
+	 */
+	rel->rd_lockInfo.lockRelId.dbId = rnode.dbNode;
+	rel->rd_lockInfo.lockRelId.relId = rnode.relNode;

-	_xlrelarr[0].moreRecently = &(_xlrelarr[0]);
-	_xlrelarr[0].lessRecently = &(_xlrelarr[0]);
+	rel->rd_targblock = InvalidBlockNumber;
+	rel->rd_smgr = NULL;

-	memset(&ctl, 0, sizeof(ctl));
-	ctl.keysize = sizeof(RelFileNode);
-	ctl.entrysize = sizeof(XLogRelCacheEntry);
-	ctl.hash = tag_hash;
-
-	_xlrelcache = hash_create("XLOG relcache", _XLOG_RELCACHESIZE,
-							  &ctl, HASH_ELEM | HASH_FUNCTION);
-}
-
-static void
-_xl_remove_hash_entry(XLogRelDesc *rdesc)
-{
-	Form_pg_class tpgc = rdesc->reldata.rd_rel;
-	XLogRelCacheEntry *hentry;
-
-	rdesc->lessRecently->moreRecently = rdesc->moreRecently;
-	rdesc->moreRecently->lessRecently = rdesc->lessRecently;
-
-	hentry = (XLogRelCacheEntry *) hash_search(_xlrelcache,
-					  (void *) &(rdesc->reldata.rd_node), HASH_REMOVE, NULL);
-	if (hentry == NULL)
-		elog(PANIC, "_xl_remove_hash_entry: file was not found in cache");
-
-	RelationCloseSmgr(&(rdesc->reldata));
-
-	memset(rdesc, 0, sizeof(XLogRelDesc));
-	memset(tpgc, 0, sizeof(FormData_pg_class));
-	rdesc->reldata.rd_rel = tpgc;
-}
-
-static XLogRelDesc *
-_xl_new_reldesc(void)
-{
-	XLogRelDesc *res;
-
-	_xlast++;
-	if (_xlast < _xlcnt)
-	{
-		_xlrelarr[_xlast].reldata.rd_rel = &(_xlpgcarr[_xlast]);
-		return &(_xlrelarr[_xlast]);
-	}
-
-	/* reuse */
-	res = _xlrelarr[0].moreRecently;
-
-	_xl_remove_hash_entry(res);
-
-	_xlast--;
-	return res;
-}
-
-
-void
-XLogInitRelationCache(void)
-{
-	_xl_init_rel_cache();
-	invalid_page_tab = NULL;
-}
-
-void
-XLogCloseRelationCache(void)
-{
-	HASH_SEQ_STATUS status;
-	XLogRelCacheEntry *hentry;
-
-	if (!_xlrelarr)
-		return;
-
-	hash_seq_init(&status, _xlrelcache);
-
-	while ((hentry = (XLogRelCacheEntry *) hash_seq_search(&status)) != NULL)
-		_xl_remove_hash_entry(hentry->rdesc);
-
-	hash_destroy(_xlrelcache);
-
-	free(_xlrelarr);
-	free(_xlpgcarr);
-
-	_xlrelarr = NULL;
+	return rel;
 }

 /*
- * Open a relation during XLOG replay
- *
- * Note: this once had an API that allowed NULL return on failure, but it
- * no longer does; any failure results in elog().
+ * Free a fake relation cache entry.
 */
-Relation
-XLogOpenRelation(RelFileNode rnode)
+void
+FreeFakeRelcacheEntry(Relation fakerel)
 {
-	XLogRelDesc *res;
-	XLogRelCacheEntry *hentry;
-	bool		found;
-
-	hentry = (XLogRelCacheEntry *)
-		hash_search(_xlrelcache, (void *) &rnode, HASH_FIND, NULL);
-
-	if (hentry)
-	{
-		res = hentry->rdesc;
-
-		res->lessRecently->moreRecently = res->moreRecently;
-		res->moreRecently->lessRecently = res->lessRecently;
-	}
-	else
-	{
-		res = _xl_new_reldesc();
-
-		sprintf(RelationGetRelationName(&(res->reldata)), "%u", rnode.relNode);
-
-		res->reldata.rd_node = rnode;
-
-		/*
-		 * We set up the lockRelId in case anything tries to lock the dummy
-		 * relation.  Note that this is fairly bogus since relNode may be
-		 * different from the relation's OID.  It shouldn't really matter
-		 * though, since we are presumably running by ourselves and can't have
-		 * any lock conflicts ...
-		 */
-		res->reldata.rd_lockInfo.lockRelId.dbId = rnode.dbNode;
-		res->reldata.rd_lockInfo.lockRelId.relId = rnode.relNode;
-
-		hentry = (XLogRelCacheEntry *)
-			hash_search(_xlrelcache, (void *) &rnode, HASH_ENTER, &found);
-
-		if (found)
-			elog(PANIC, "xlog relation already present on insert into cache");
-
-		hentry->rdesc = res;
-
-		res->reldata.rd_targblock = InvalidBlockNumber;
-		res->reldata.rd_smgr = NULL;
-		RelationOpenSmgr(&(res->reldata));
-
-		/*
-		 * Create the target file if it doesn't already exist.  This lets us
-		 * cope if the replay sequence contains writes to a relation that is
-		 * later deleted.  (The original coding of this routine would instead
-		 * return NULL, causing the writes to be suppressed. But that seems
-		 * like it risks losing valuable data if the filesystem loses an inode
-		 * during a crash.	Better to write the data until we are actually
-		 * told to delete the file.)
-		 */
-		smgrcreate(res->reldata.rd_smgr, res->reldata.rd_istemp, true);
-	}
-
-	res->moreRecently = &(_xlrelarr[0]);
-	res->lessRecently = _xlrelarr[0].lessRecently;
-	_xlrelarr[0].lessRecently = res;
-	res->lessRecently->moreRecently = res;
-
-	return &(res->reldata);
+	pfree(fakerel);
 }

 /*
 * Drop a relation during XLOG replay
 *
- * This is called when the relation is about to be deleted; we need to ensure
- * that there is no dangling smgr reference in the xlog relation cache.
- *
- * Currently, we don't bother to physically remove the relation from the
- * cache, we just let it age out normally.
- *
- * This also takes care of removing any open "invalid-page" records for
- * the relation.
+ * This is called when the relation is about to be deleted; we need to remove
+ * any open "invalid-page" records for the relation.
 */
 void
 XLogDropRelation(RelFileNode rnode)
 {
-	XLogRelCacheEntry *hentry;
-
-	hentry = (XLogRelCacheEntry *)
-		hash_search(_xlrelcache, (void *) &rnode, HASH_FIND, NULL);
-
-	if (hentry)
-	{
-		XLogRelDesc *rdesc = hentry->rdesc;
-
-		RelationCloseSmgr(&(rdesc->reldata));
-	}
+	/* Tell smgr to forget about this relation as well */
+	smgrclosenode(rnode);

 	forget_invalid_pages(rnode, 0);
 }
@@ -507,18 +378,14 @@ XLogDropRelation(RelFileNode rnode)
 void
 XLogDropDatabase(Oid dbid)
 {
-	HASH_SEQ_STATUS status;
-	XLogRelCacheEntry *hentry;
-
-	hash_seq_init(&status, _xlrelcache);
-
-	while ((hentry = (XLogRelCacheEntry *) hash_seq_search(&status)) != NULL)
-	{
-		XLogRelDesc *rdesc = hentry->rdesc;
-
-		if (hentry->rnode.dbNode == dbid)
-			RelationCloseSmgr(&(rdesc->reldata));
-	}
+	/*
+	 * This is unnecessarily heavy-handed, as it will close SMgrRelation
+	 * objects for other databases as well. DROP DATABASE occurs seldom
+	 * enough that it's not worth introducing a variant of smgrclose for
+	 * just this purpose. XXX: Or should we rather leave the smgr entries
+	 * dangling?
+	 */
+	smgrcloseall();

 	forget_invalid_pages_db(dbid);
 }
@@ -526,8 +393,7 @@ XLogDropDatabase(Oid dbid)
 /*
 * Truncate a relation during XLOG replay
 *
- * We don't need to do anything to the fake relcache, but we do need to
- * clean up any open "invalid-page" records for the dropped pages.
+ * We need to clean up any open "invalid-page" records for the dropped pages.
 */
 void
 XLogTruncateRelation(RelFileNode rnode, BlockNumber nblocks)