Reduce size of critial section during vacuum full, critical

sections now isn't nested. All user-defined functions now is
called outside critsections. Small improvements in WAL
protocol.

TODO: improve XLOG replay

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 1.134 2006/05/10 23:18:38 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/gist/gist.c,v 1.135 2006/05/17 16:34:59 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -347,7 +347,7 @@ gistplacetopage(GISTInsertState *state, GISTSTATE *giststate)
 		 * Form index tuples vector to split:
 		 * remove old tuple if t's needed and add new tuples to vector
 		 */
-		itvec = gistextractbuffer(state->stack->buffer, &tlen);
+		itvec = gistextractpage(state->stack->page, &tlen);
 		if ( !is_leaf ) {
 			/* on inner page we should remove old tuple */
 			int pos = state->stack->childoffnum - FirstOffsetNumber;
@@ -501,7 +501,7 @@ gistplacetopage(GISTInsertState *state, GISTSTATE *giststate)
 			}
 
 			rdata = formUpdateRdata(state->r->rd_node, state->stack->buffer,
-									offs, noffs, false,
+									offs, noffs, 
 									state->itup, state->ituplen,
 									&(state->key));
 
@@ -1157,7 +1157,7 @@ gistnewroot(Relation r, Buffer buffer, IndexTuple *itup, int len, ItemPointer ke
 		XLogRecData *rdata;
 
 		rdata = formUpdateRdata(r->rd_node, buffer,
-								NULL, 0, false,
+								NULL, 0,
 								itup, len, key);
 
 		recptr = XLogInsert(RM_GIST_ID, XLOG_GIST_NEW_ROOT, rdata);

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 1.11 2006/05/10 09:19:54 teodor Exp $
+ *			$PostgreSQL: pgsql/src/backend/access/gist/gistutil.c,v 1.12 2006/05/17 16:34:59 teodor Exp $
  *-------------------------------------------------------------------------
  */
 #include "postgres.h"
@@ -112,18 +112,17 @@ gistfitpage(IndexTuple *itvec, int len) {
  * Read buffer into itup vector
  */
 IndexTuple *
-gistextractbuffer(Buffer buffer, int *len /* out */ )
+gistextractpage(Page page, int *len /* out */ )
 {
 	OffsetNumber i,
 				maxoff;
 	IndexTuple *itvec;
-	Page		p = (Page) BufferGetPage(buffer);
 
-	maxoff = PageGetMaxOffsetNumber(p);
+	maxoff = PageGetMaxOffsetNumber(page);
 	*len = maxoff;
 	itvec = palloc(sizeof(IndexTuple) * maxoff);
 	for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
-		itvec[i - FirstOffsetNumber] = (IndexTuple) PageGetItem(p, PageGetItemId(p, i));
+		itvec[i - FirstOffsetNumber] = (IndexTuple) PageGetItem(page, PageGetItemId(page, i));
 
 	return itvec;
 }

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 1.16 2006/05/10 09:19:54 teodor Exp $
+ *			 $PostgreSQL: pgsql/src/backend/access/gist/gistxlog.c,v 1.17 2006/05/17 16:34:59 teodor Exp $
  *-------------------------------------------------------------------------
  */
 #include "postgres.h"
@@ -209,41 +209,33 @@ gistRedoPageUpdateRecord(XLogRecPtr lsn, XLogRecord *record, bool isnewroot)
 		return;
 	}
 
-	if (xlrec.data->isemptypage)
+	if (isnewroot)
+		GISTInitBuffer(buffer, 0);
+	else if (xlrec.data->ntodelete)
 	{
-		while (!PageIsEmpty(page))
-			PageIndexTupleDelete(page, FirstOffsetNumber);
+		int			i;
 
-		if (xlrec.data->blkno == GIST_ROOT_BLKNO)
-			GistPageSetLeaf(page);
-		else
-			GistPageSetDeleted(page);
+		for (i = 0; i < xlrec.data->ntodelete; i++)
+			PageIndexTupleDelete(page, xlrec.todelete[i]);
+		if (GistPageIsLeaf(page))
+			GistMarkTuplesDeleted(page);
 	}
-	else
-	{
-		if (isnewroot)
-			GISTInitBuffer(buffer, 0);
-		else if (xlrec.data->ntodelete)
-		{
-			int			i;
 
-			for (i = 0; i < xlrec.data->ntodelete; i++)
-				PageIndexTupleDelete(page, xlrec.todelete[i]);
-			if (GistPageIsLeaf(page))
-				GistMarkTuplesDeleted(page);
-		}
+	/* add tuples */
+	if (xlrec.len > 0)
+		gistfillbuffer(reln, page, xlrec.itup, xlrec.len, InvalidOffsetNumber);
 
-		/* add tuples */
-		if (xlrec.len > 0)
-			gistfillbuffer(reln, page, xlrec.itup, xlrec.len, InvalidOffsetNumber);
+	/*
+	 * special case: leafpage, nothing to insert, nothing to delete, then
+	 * vacuum marks page
+	 */
+	if (GistPageIsLeaf(page) && xlrec.len == 0 && xlrec.data->ntodelete == 0)
+		GistClearTuplesDeleted(page);
 
-		/*
-		 * special case: leafpage, nothing to insert, nothing to delete, then
-		 * vacuum marks page
-		 */
-		if (GistPageIsLeaf(page) && xlrec.len == 0 && xlrec.data->ntodelete == 0)
-			GistClearTuplesDeleted(page);
-	}
+	if ( !GistPageIsLeaf(page) && PageGetMaxOffsetNumber(page) == InvalidOffsetNumber && xldata->blkno == GIST_ROOT_BLKNO )
+		/* all links on non-leaf root page was deleted by vacuum full,
+		   so root page becomes a leaf */
+		GistPageSetLeaf(page);
 
 	GistPageGetOpaque(page)->rightlink = InvalidBlockNumber;
 	PageSetLSN(page, lsn);
@@ -252,6 +244,29 @@ gistRedoPageUpdateRecord(XLogRecPtr lsn, XLogRecord *record, bool isnewroot)
 	UnlockReleaseBuffer(buffer);
 }
 
+static void
+gistRedoPageDeleteRecord(XLogRecPtr lsn, XLogRecord *record)
+{
+	gistxlogPageDelete *xldata = (gistxlogPageDelete *) XLogRecGetData(record);
+	Relation	reln;
+	Buffer		buffer;
+	Page		page;
+
+	reln = XLogOpenRelation(xldata->node);
+	buffer = XLogReadBuffer(reln, xldata->blkno, false);
+	if (!BufferIsValid(buffer))
+		return;
+
+	GISTInitBuffer( buffer, 0 );
+	page = (Page) BufferGetPage(buffer);
+	GistPageSetDeleted(page);
+
+	PageSetLSN(page, lsn);
+	PageSetTLI(page, ThisTimeLineID);
+	MarkBufferDirty(buffer);
+	UnlockReleaseBuffer(buffer);
+}
+
 static void
 decodePageSplitRecord(PageSplitRecord *decoded, XLogRecord *record)
 {
@@ -382,6 +397,9 @@ gist_redo(XLogRecPtr lsn, XLogRecord *record)
 		case XLOG_GIST_PAGE_UPDATE:
 			gistRedoPageUpdateRecord(lsn, record, false);
 			break;
+		case XLOG_GIST_PAGE_DELETE:
+			gistRedoPageDeleteRecord(lsn, record);
+			break;
 		case XLOG_GIST_NEW_ROOT:
 			gistRedoPageUpdateRecord(lsn, record, true);
 			break;
@@ -405,8 +423,10 @@ gist_redo(XLogRecPtr lsn, XLogRecord *record)
 static void
 out_target(StringInfo buf, RelFileNode node, ItemPointerData key)
 {
-	appendStringInfo(buf, "rel %u/%u/%u; tid %u/%u",
-			node.spcNode, node.dbNode, node.relNode,
+	appendStringInfo(buf, "rel %u/%u/%u",
+			node.spcNode, node.dbNode, node.relNode);
+	if ( ItemPointerIsValid( &key ) )
+		appendStringInfo(buf, "; tid %u/%u",
 			ItemPointerGetBlockNumber(&key),
 			ItemPointerGetOffsetNumber(&key));
 }
@@ -418,6 +438,14 @@ out_gistxlogPageUpdate(StringInfo buf, gistxlogPageUpdate *xlrec)
 	appendStringInfo(buf, "; block number %u", xlrec->blkno);
 }
 
+static void
+out_gistxlogPageDelete(StringInfo buf, gistxlogPageDelete *xlrec)
+{
+	appendStringInfo(buf, "page_delete: rel %u/%u/%u; blkno %u",
+			xlrec->node.spcNode, xlrec->node.dbNode, xlrec->node.relNode,
+			xlrec->blkno);
+}
+
 static void
 out_gistxlogPageSplit(StringInfo buf, gistxlogPageSplit *xlrec)
 {
@@ -438,6 +466,9 @@ gist_desc(StringInfo buf, uint8 xl_info, char *rec)
 			appendStringInfo(buf, "page_update: ");
 			out_gistxlogPageUpdate(buf, (gistxlogPageUpdate *) rec);
 			break;
+		case XLOG_GIST_PAGE_DELETE:
+			out_gistxlogPageDelete(buf, (gistxlogPageDelete *) rec);
+			break;
 		case XLOG_GIST_NEW_ROOT:
 			appendStringInfo(buf, "new_root: ");
 			out_target(buf, ((gistxlogPageUpdate *) rec)->node, ((gistxlogPageUpdate *) rec)->key);
@@ -643,7 +674,7 @@ gistContinueInsert(gistIncompleteInsert *insert)
 					 * we split root, just copy tuples from old root to new
 					 * page
 					 */
-					parentitup = gistextractbuffer(buffers[numbuffer - 1],
+					parentitup = gistextractpage(pages[numbuffer - 1],
 												   &pituplen);
 
 					/* sanity check */
@@ -796,7 +827,7 @@ formSplitRdata(RelFileNode node, BlockNumber blkno, bool page_is_leaf,
  */
 XLogRecData *
 formUpdateRdata(RelFileNode node, Buffer buffer,
-				OffsetNumber *todelete, int ntodelete, bool emptypage,
+				OffsetNumber *todelete, int ntodelete,
 				IndexTuple *itup, int ituplen, ItemPointer key)
 {
 	XLogRecData *rdata;
@@ -804,35 +835,37 @@ formUpdateRdata(RelFileNode node, Buffer buffer,
 	int			cur,
 				i;
 
-	/* ugly wart in API: emptypage causes us to ignore other inputs */
-	if (emptypage)
-		ntodelete = ituplen = 0;
-
-	rdata = (XLogRecData *) palloc(sizeof(XLogRecData) * (2 + ituplen));
+	rdata = (XLogRecData *) palloc(sizeof(XLogRecData) * (3 + ituplen));
 	xlrec = (gistxlogPageUpdate *) palloc(sizeof(gistxlogPageUpdate));
 
 	xlrec->node = node;
 	xlrec->blkno = BufferGetBlockNumber(buffer);
 	xlrec->ntodelete = ntodelete;
-	xlrec->isemptypage = emptypage;
+
 	if (key)
 		xlrec->key = *key;
 	else
 		ItemPointerSetInvalid(&(xlrec->key));
 
-	rdata[0].data = (char *) xlrec;
-	rdata[0].len = sizeof(gistxlogPageUpdate);
-	rdata[0].buffer = InvalidBuffer;
+	rdata[0].buffer = buffer;
+	rdata[0].buffer_std = true;
+	rdata[0].data = NULL;
+	rdata[0].len = 0;
 	rdata[0].next = &(rdata[1]);
 
-	rdata[1].data = (char *) todelete;
-	rdata[1].len = MAXALIGN(sizeof(OffsetNumber) * ntodelete);
-	rdata[1].buffer = buffer;
-	rdata[1].buffer_std = true;
-	rdata[1].next = NULL;
+	rdata[1].data = (char *) xlrec;
+	rdata[1].len = sizeof(gistxlogPageUpdate);
+	rdata[1].buffer = InvalidBuffer;
+	rdata[1].next = &(rdata[2]);
+
+	rdata[2].data = (char *) todelete;
+	rdata[2].len = MAXALIGN(sizeof(OffsetNumber) * ntodelete);
+	rdata[2].buffer = buffer;
+	rdata[2].buffer_std = true;
+	rdata[2].next = NULL;
 
 	/* new tuples */
-	cur = 2;
+	cur = 3;
 	for (i = 0; i < ituplen; i++)
 	{
 		rdata[cur - 1].next = &(rdata[cur]);

src/include/access/gist_private.h

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
  * Portions Copyright (c) 1994, Regents of the University of California
  *
- * $PostgreSQL: pgsql/src/include/access/gist_private.h,v 1.13 2006/05/10 09:19:54 teodor Exp $
+ * $PostgreSQL: pgsql/src/include/access/gist_private.h,v 1.14 2006/05/17 16:34:59 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -85,20 +85,21 @@ extern const XLogRecPtr XLogRecPtrForTemp;
 #define XLOG_GIST_PAGE_SPLIT		0x30
 #define XLOG_GIST_INSERT_COMPLETE	0x40
 #define XLOG_GIST_CREATE_INDEX		0x50
+#define XLOG_GIST_PAGE_DELETE		0x60
 
 typedef struct gistxlogPageUpdate
 {
 	RelFileNode node;
 	BlockNumber blkno;
 
-	uint16		ntodelete;
-	bool		isemptypage;
-
 	/*
 	 * It used to identify completeness of insert. Sets to leaf itup
 	 */
 	ItemPointerData key;
 
+	/* number of deleted offsets */
+	uint16		ntodelete;
+
 	/*
 	 * follow: 1. todelete OffsetNumbers 2. tuples to insert
 	 */
@@ -131,6 +132,11 @@ typedef struct gistxlogInsertComplete
 	/* follows ItemPointerData key to clean */
 } gistxlogInsertComplete;
 
+typedef struct gistxlogPageDelete
+{
+	RelFileNode node;
+	BlockNumber blkno;
+} gistxlogPageDelete;
 
 /* SplitedPageLayout - gistSplit function result */
 typedef struct SplitedPageLayout
@@ -249,7 +255,7 @@ extern void gist_xlog_cleanup(void);
 extern IndexTuple gist_form_invalid_tuple(BlockNumber blkno);
 
 extern XLogRecData *formUpdateRdata(RelFileNode node, Buffer buffer,
-				OffsetNumber *todelete, int ntodelete, bool emptypage,
+				OffsetNumber *todelete, int ntodelete, 
 				IndexTuple *itup, int ituplen, ItemPointer key);
 
 extern XLogRecData *formSplitRdata(RelFileNode node,
@@ -273,7 +279,7 @@ extern void gistcheckpage(Relation rel, Buffer buf);
 extern Buffer gistNewBuffer(Relation r);
 extern OffsetNumber gistfillbuffer(Relation r, Page page, IndexTuple *itup,
 			   int len, OffsetNumber off);
-extern IndexTuple *gistextractbuffer(Buffer buffer, int *len /* out */ );
+extern IndexTuple *gistextractpage(Page page, int *len /* out */ );
 extern IndexTuple *gistjoinvector(
 			   IndexTuple *itvec, int *len,
 			   IndexTuple *additvec, int addlen);