/*
 * Retail deletion of a single tuple.
 *
 * NB: this is no longer called externally, but is still needed by
 * gistlayerinsert().  That dependency will have to be fixed if GIST
 * is ever going to allow concurrent insertions.
 */
static void
gistdelete(Relation r, ItemPointer tid)
{
	BlockNumber blkno;
	OffsetNumber offnum;
	Buffer		buf;
	Page		page;

	/*
	 * Since GIST is not marked "amconcurrent" in pg_am, caller should
	 * have acquired exclusive lock on index relation.	We need no locking
	 * here.
	 */

	blkno = ItemPointerGetBlockNumber(tid);
	offnum = ItemPointerGetOffsetNumber(tid);

	/* adjust any scans that will be affected by this deletion */
	/* NB: this works only for scans in *this* backend! */
	gistadjscans(r, GISTOP_DEL, blkno, offnum);

	/* delete the index tuple */
	buf = ReadBuffer(r, blkno);
	page = BufferGetPage(buf);

	PageIndexTupleDelete(page, offnum);

	WriteBuffer(buf);
}

/*
 * Delete tuple on leaf page if tuples was existed and we
 * should update it, update old child blkno to new right page
 * if child split is occured
 */
static BlockNumber
entryPreparePage(GinBtree btree, Page page, OffsetNumber off)
{
	BlockNumber ret = InvalidBlockNumber;

	Assert(btree->entry);
	Assert(!GinPageIsData(page));

	if (btree->isDelete)
	{
		Assert(GinPageIsLeaf(page));
		PageIndexTupleDelete(page, off);
	}

	if (!GinPageIsLeaf(page) && btree->rightblkno != InvalidBlockNumber)
	{
		IndexTuple	itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, off));

		ItemPointerSet(&itup->t_tid, btree->rightblkno, InvalidOffsetNumber);
		ret = btree->rightblkno;
	}

	btree->rightblkno = InvalidBlockNumber;

	return ret;
}

/*
 * Delete tuple on leaf page if tuples existed and we
 * should update it, update old child blkno to new right page
 * if child split occurred
 */
static BlockNumber
entryPreparePage(RumBtree btree, Page page, OffsetNumber off)
{
	BlockNumber ret = InvalidBlockNumber;

	Assert(btree->entry);
	Assert(!RumPageIsData(page));

	if (btree->isDelete)
	{
		Assert(RumPageIsLeaf(page));
		PageIndexTupleDelete(page, off);
	}

	if (!RumPageIsLeaf(page) && btree->rightblkno != InvalidBlockNumber)
	{
		IndexTuple	itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, off));

		RumSetDownlink(itup, btree->rightblkno);
		ret = btree->rightblkno;
	}

	btree->rightblkno = InvalidBlockNumber;

	return ret;
}

/*
 * Delete tuple on leaf page if tuples existed and we
 * should update it, update old child blkno to new right page
 * if child split occurred
 */
static void
entryPreparePage(GinBtree btree, Page page, OffsetNumber off,
				 GinBtreeEntryInsertData *insertData, BlockNumber updateblkno)
{
	Assert(insertData->entry);
	Assert(!GinPageIsData(page));

	if (insertData->isDelete)
	{
		Assert(GinPageIsLeaf(page));
		PageIndexTupleDelete(page, off);
	}

	if (!GinPageIsLeaf(page) && updateblkno != InvalidBlockNumber)
	{
		IndexTuple	itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, off));

		GinSetDownlink(itup, updateblkno);
	}
}

/*
 * Add a leaf tuple, or replace an existing placeholder tuple.  This is used
 * to replay SpGistPageAddNewItem() operations.  If the offset points at an
 * existing tuple, it had better be a placeholder tuple.
 */
static void
addOrReplaceTuple(Page page, Item tuple, int size, OffsetNumber offset)
{
	if (offset <= PageGetMaxOffsetNumber(page))
	{
		SpGistDeadTuple dt = (SpGistDeadTuple) PageGetItem(page,
														   PageGetItemId(page, offset));

		if (dt->tupstate != SPGIST_PLACEHOLDER)
			elog(ERROR, "SPGiST tuple to be replaced is not a placeholder");

		Assert(SpGistPageGetOpaque(page)->nPlaceholder > 0);
		SpGistPageGetOpaque(page)->nPlaceholder--;

		PageIndexTupleDelete(page, offset);
	}

	Assert(offset <= PageGetMaxOffsetNumber(page) + 1);

	if (PageAddItem(page, tuple, size, offset, false, false) != offset)
		elog(ERROR, "failed to add item of size %u to SPGiST index page",
			 size);
}

static void
ginRedoInsertEntry(Buffer buffer, bool isLeaf, BlockNumber rightblkno, void *rdata)
{
	Page		page = BufferGetPage(buffer);
	ginxlogInsertEntry *data = (ginxlogInsertEntry *) rdata;
	OffsetNumber offset = data->offset;
	IndexTuple	itup;

	if (rightblkno != InvalidBlockNumber)
	{
		/* update link to right page after split */
		Assert(!GinPageIsLeaf(page));
		Assert(offset >= FirstOffsetNumber && offset <= PageGetMaxOffsetNumber(page));
		itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offset));
		GinSetDownlink(itup, rightblkno);
	}

	if (data->isDelete)
	{
		Assert(GinPageIsLeaf(page));
		Assert(offset >= FirstOffsetNumber && offset <= PageGetMaxOffsetNumber(page));
		PageIndexTupleDelete(page, offset);
	}

	itup = &data->tuple;

	if (PageAddItem(page, (Item) itup, IndexTupleSize(itup), offset, false, false) == InvalidOffsetNumber)
	{
		RelFileNode node;
		ForkNumber	forknum;
		BlockNumber blknum;

		BufferGetTag(buffer, &node, &forknum, &blknum);
		elog(ERROR, "failed to add item to index page in %u/%u/%u",
			 node.spcNode, node.dbNode, node.relNode);
	}
}

//.........这里部分代码省略.........
		if (ooffnum > omaxoffnum)
		{
			/* at end of page, but check for an(other) overflow page */
			oblkno = oopaque->hasho_nextblkno;
			if (!BlockNumberIsValid(oblkno))
				break;
			/*
			 * we ran out of tuples on this particular page, but we
			 * have more overflow pages; advance to next page.
			 */
			_hash_wrtbuf(rel, obuf);

			obuf = _hash_getbuf(rel, oblkno, HASH_WRITE);
			opage = BufferGetPage(obuf);
			_hash_checkpage(rel, opage, LH_OVERFLOW_PAGE);
			oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
			ooffnum = FirstOffsetNumber;
			omaxoffnum = PageGetMaxOffsetNumber(opage);
			continue;
		}

		/*
		 * Re-hash the tuple to determine which bucket it now belongs in.
		 *
		 * It is annoying to call the hash function while holding locks,
		 * but releasing and relocking the page for each tuple is unappealing
		 * too.
		 */
		hitem = (HashItem) PageGetItem(opage, PageGetItemId(opage, ooffnum));
		itup = &(hitem->hash_itup);
		datum = index_getattr(itup, 1, itupdesc, &null);
		Assert(!null);

		bucket = _hash_hashkey2bucket(_hash_datum2hashkey(rel, datum),
									  maxbucket, highmask, lowmask);

		if (bucket == nbucket)
		{
			/*
			 * insert the tuple into the new bucket.  if it doesn't fit on
			 * the current page in the new bucket, we must allocate a new
			 * overflow page and place the tuple on that page instead.
			 */
			itemsz = IndexTupleDSize(hitem->hash_itup)
				+ (sizeof(HashItemData) - sizeof(IndexTupleData));

			itemsz = MAXALIGN(itemsz);

			if (PageGetFreeSpace(npage) < itemsz)
			{
				/* write out nbuf and drop lock, but keep pin */
				_hash_chgbufaccess(rel, nbuf, HASH_WRITE, HASH_NOLOCK);
				/* chain to a new overflow page */
				nbuf = _hash_addovflpage(rel, metabuf, nbuf);
				npage = BufferGetPage(nbuf);
				_hash_checkpage(rel, npage, LH_OVERFLOW_PAGE);
				/* we don't need nopaque within the loop */
			}

			noffnum = OffsetNumberNext(PageGetMaxOffsetNumber(npage));
			if (PageAddItem(npage, (Item) hitem, itemsz, noffnum, LP_USED)
				== InvalidOffsetNumber)
				elog(ERROR, "failed to add index item to \"%s\"",
					 RelationGetRelationName(rel));

			/*
			 * now delete the tuple from the old bucket.  after this
			 * section of code, 'ooffnum' will actually point to the
			 * ItemId to which we would point if we had advanced it before
			 * the deletion (PageIndexTupleDelete repacks the ItemId
			 * array).	this also means that 'omaxoffnum' is exactly one
			 * less than it used to be, so we really can just decrement it
			 * instead of calling PageGetMaxOffsetNumber.
			 */
			PageIndexTupleDelete(opage, ooffnum);
			omaxoffnum = OffsetNumberPrev(omaxoffnum);
		}
		else
		{
			/*
			 * the tuple stays on this page.  we didn't move anything, so
			 * we didn't delete anything and therefore we don't have to
			 * change 'omaxoffnum'.
			 */
			Assert(bucket == obucket);
			ooffnum = OffsetNumberNext(ooffnum);
		}
	}

	/*
	 * We're at the end of the old bucket chain, so we're done partitioning
	 * the tuples.  Before quitting, call _hash_squeezebucket to ensure the
	 * tuples remaining in the old bucket (including the overflow pages) are
	 * packed as tightly as possible.  The new bucket is already tight.
	 */
	_hash_wrtbuf(rel, obuf);
	_hash_wrtbuf(rel, nbuf);

	_hash_squeezebucket(rel, obucket, start_oblkno);
}

/*
 * Bulk deletion of all index entries pointing to a set of heap tuples.
 * The set of target tuples is specified via a callback routine that tells
 * whether any given heap tuple (identified by ItemPointer) is being deleted.
 *
 * Result: a palloc'd struct containing statistical info for VACUUM displays.
 */
Datum
rtbulkdelete(PG_FUNCTION_ARGS)
{
	Relation	rel = (Relation) PG_GETARG_POINTER(0);
	IndexBulkDeleteCallback callback = (IndexBulkDeleteCallback) PG_GETARG_POINTER(1);
	void	   *callback_state = (void *) PG_GETARG_POINTER(2);
	IndexBulkDeleteResult *result;
	BlockNumber num_pages;
	double		tuples_removed;
	double		num_index_tuples;
	IndexScanDesc iscan;

	tuples_removed = 0;
	num_index_tuples = 0;

	/*
	 * Since rtree is not marked "amconcurrent" in pg_am, caller should have
	 * acquired exclusive lock on index relation.  We need no locking here.
	 */

	/*
	 * XXX generic implementation --- should be improved!
	 */

	/* walk through the entire index */
	iscan = index_beginscan(NULL, rel, SnapshotAny, 0, NULL);
	/* including killed tuples */
	iscan->ignore_killed_tuples = false;

	while (index_getnext_indexitem(iscan, ForwardScanDirection))
	{
		vacuum_delay_point();

		if (callback(&iscan->xs_ctup.t_self, callback_state))
		{
			ItemPointerData indextup = iscan->currentItemData;
			BlockNumber blkno;
			OffsetNumber offnum;
			Buffer		buf;
			Page		page;

			blkno = ItemPointerGetBlockNumber(&indextup);
			offnum = ItemPointerGetOffsetNumber(&indextup);

			/* adjust any scans that will be affected by this deletion */
			/* (namely, my own scan) */
			rtadjscans(rel, RTOP_DEL, blkno, offnum);

			/* delete the index tuple */
			buf = ReadBuffer(rel, blkno);
			page = BufferGetPage(buf);

			PageIndexTupleDelete(page, offnum);

			WriteBuffer(buf);

			tuples_removed += 1;
		}
		else
			num_index_tuples += 1;
	}

	index_endscan(iscan);

	/* return statistics */
	num_pages = RelationGetNumberOfBlocks(rel);

	result = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
	result->num_pages = num_pages;
	result->num_index_tuples = num_index_tuples;
	result->tuples_removed = tuples_removed;

	PG_RETURN_POINTER(result);
}

//.........这里部分代码省略.........
				PageSetLSN(ptr->page, XLogRecPtrForTemp);
			}
		}

		/* set up NSN */
		oldnsn = GistPageGetOpaque(dist->page)->nsn;
		if (state->stack->blkno == GIST_ROOT_BLKNO)
			/* if root split we should put initial value */
			oldnsn = PageGetLSN(dist->page);

		for (ptr = dist; ptr; ptr = ptr->next)
		{
			/* only for last set oldnsn */
			GistPageGetOpaque(ptr->page)->nsn = (ptr->next) ?
				PageGetLSN(ptr->page) : oldnsn;
		}

		/*
		 * release buffers, if it was a root split then release all buffers
		 * because we create all buffers
		 */
		ptr = (state->stack->blkno == GIST_ROOT_BLKNO) ? dist : dist->next;
		for (; ptr; ptr = ptr->next)
			UnlockReleaseBuffer(ptr->buffer);

		if (state->stack->blkno == GIST_ROOT_BLKNO)
		{
			gistnewroot(state->r, state->stack->buffer, state->itup, state->ituplen, &(state->key));
			state->needInsertComplete = false;
		}

		END_CRIT_SECTION();
	}
	else
	{
		/* enough space */
		START_CRIT_SECTION();

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

		MarkBufferDirty(state->stack->buffer);

		if (!state->r->rd_istemp)
		{
			OffsetNumber noffs = 0,
						offs[1];
			XLogRecPtr	recptr;
			XLogRecData *rdata;

			if (!is_leaf)
			{
				/* only on inner page we should delete previous version */
				offs[0] = state->stack->childoffnum;
				noffs = 1;
			}

			rdata = formUpdateRdata(state->r, state->stack->buffer,
									offs, noffs,
									state->itup, state->ituplen,
									&(state->key));

			recptr = XLogInsert(RM_GIST_ID, XLOG_GIST_PAGE_UPDATE, rdata);
			PageSetLSN(state->stack->page, recptr);
			PageSetTLI(state->stack->page, ThisTimeLineID);
		}
		else
			PageSetLSN(state->stack->page, XLogRecPtrForTemp);

		if (state->stack->blkno == GIST_ROOT_BLKNO)
			state->needInsertComplete = false;

		END_CRIT_SECTION();

		if (state->ituplen > 1)
		{						/* previous is_splitted==true */

			/*
			 * child was splited, so we must form union for insertion in
			 * parent
			 */
			IndexTuple	newtup = gistunion(state->r, state->itup, state->ituplen, giststate);

			ItemPointerSetBlockNumber(&(newtup->t_tid), state->stack->blkno);
			state->itup[0] = newtup;
			state->ituplen = 1;
		}
		else if (is_leaf)
		{
			/*
			 * itup[0] store key to adjust parent, we set it to valid to
			 * correct check by GistTupleIsInvalid macro in gistgetadjusted()
			 */
			ItemPointerSetBlockNumber(&(state->itup[0]->t_tid), state->stack->blkno);
			GistTupleSetValid(state->itup[0]);
		}
	}
	return is_splitted;
}

/*
 * redo any page update (except page split)
 */
static void
gistRedoPageUpdateRecord(XLogRecPtr lsn, XLogRecord *record)
{
	char	   *begin = XLogRecGetData(record);
	gistxlogPageUpdate *xldata = (gistxlogPageUpdate *) begin;
	Buffer		buffer;
	Page		page;
	char	   *data;

	if (BlockNumberIsValid(xldata->leftchild))
		gistRedoClearFollowRight(xldata->node, lsn, xldata->leftchild);

	/* nothing more to do if page was backed up (and no info to do it with) */
	if (record->xl_info & XLR_BKP_BLOCK_1)
		return;

	buffer = XLogReadBuffer(xldata->node, xldata->blkno, false);
	if (!BufferIsValid(buffer))
		return;
	page = (Page) BufferGetPage(buffer);

	if (XLByteLE(lsn, PageGetLSN(page)))
	{
		UnlockReleaseBuffer(buffer);
		return;
	}

	data = begin + sizeof(gistxlogPageUpdate);

	/* Delete old tuples */
	if (xldata->ntodelete > 0)
	{
		int			i;
		OffsetNumber *todelete = (OffsetNumber *) data;

		data += sizeof(OffsetNumber) * xldata->ntodelete;

		for (i = 0; i < xldata->ntodelete; i++)
			PageIndexTupleDelete(page, todelete[i]);
		if (GistPageIsLeaf(page))
			GistMarkTuplesDeleted(page);
	}

	/* add tuples */
	if (data - begin < record->xl_len)
	{
		OffsetNumber off = (PageIsEmpty(page)) ? FirstOffsetNumber :
		OffsetNumberNext(PageGetMaxOffsetNumber(page));

		while (data - begin < record->xl_len)
		{
			IndexTuple	itup = (IndexTuple) data;
			Size		sz = IndexTupleSize(itup);
			OffsetNumber l;

			data += sz;

			l = PageAddItem(page, (Item) itup, sz, off, false, false);
			if (l == InvalidOffsetNumber)
				elog(ERROR, "failed to add item to GiST index page, size %d bytes",
					 (int) sz);
			off++;
		}
	}
	else
	{
		/*
		 * special case: leafpage, nothing to insert, nothing to delete, then
		 * vacuum marks page
		 */
		if (GistPageIsLeaf(page) && xldata->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);
	PageSetTLI(page, ThisTimeLineID);
	MarkBufferDirty(buffer);
	UnlockReleaseBuffer(buffer);
}

static void
spgRedoSplitTuple(XLogRecPtr lsn, XLogRecord *record)
{
	char	   *ptr = XLogRecGetData(record);
	spgxlogSplitTuple *xldata = (spgxlogSplitTuple *) ptr;
	char	   *prefixTuple;
	SpGistInnerTupleData prefixTupleHdr;
	char	   *postfixTuple;
	SpGistInnerTupleData postfixTupleHdr;
	Buffer		buffer;
	Page		page;

	ptr += sizeof(spgxlogSplitTuple);
	prefixTuple = ptr;
	/* the prefix tuple is unaligned, so make a copy to access its header */
	memcpy(&prefixTupleHdr, prefixTuple, sizeof(SpGistInnerTupleData));
	ptr += prefixTupleHdr.size;
	postfixTuple = ptr;
	/* postfix tuple is also unaligned */
	memcpy(&postfixTupleHdr, postfixTuple, sizeof(SpGistInnerTupleData));

	/*
	 * In normal operation we would have both pages locked simultaneously; but
	 * in WAL replay it should be safe to update them one at a time, as long
	 * as we do it in the right order.
	 */

	/* insert postfix tuple first to avoid dangling link */
	if (xldata->blknoPostfix != xldata->blknoPrefix)
	{
		XLogRedoAction action;

		if (xldata->newPage)
		{
			buffer = XLogReadBuffer(xldata->node, xldata->blknoPostfix, true);
			/* SplitTuple is not used for nulls pages */
			SpGistInitBuffer(buffer, 0);
			action = BLK_NEEDS_REDO;
		}
		else
			action = XLogReadBufferForRedo(lsn, record, 1,
										   xldata->node, xldata->blknoPostfix,
										   &buffer);

		if (action == BLK_NEEDS_REDO)
		{
			page = BufferGetPage(buffer);

			addOrReplaceTuple(page, (Item) postfixTuple,
							  postfixTupleHdr.size, xldata->offnumPostfix);

			PageSetLSN(page, lsn);
			MarkBufferDirty(buffer);
		}
		if (BufferIsValid(buffer))
			UnlockReleaseBuffer(buffer);
	}

	/* now handle the original page */
	if (XLogReadBufferForRedo(lsn, record, 0, xldata->node, xldata->blknoPrefix,
							  &buffer) == BLK_NEEDS_REDO)
	{
		page = BufferGetPage(buffer);
		PageIndexTupleDelete(page, xldata->offnumPrefix);
		if (PageAddItem(page, (Item) prefixTuple, prefixTupleHdr.size,
				 xldata->offnumPrefix, false, false) != xldata->offnumPrefix)
			elog(ERROR, "failed to add item of size %u to SPGiST index page",
				 prefixTupleHdr.size);

		if (xldata->blknoPostfix == xldata->blknoPrefix)
			addOrReplaceTuple(page, (Item) postfixTuple, postfixTupleHdr.size,
							  xldata->offnumPostfix);

		PageSetLSN(page, lsn);
		MarkBufferDirty(buffer);
	}
	if (BufferIsValid(buffer))
		UnlockReleaseBuffer(buffer);
}

//.........这里部分代码省略.........
		 */
		PageRestoreTempPage(dist->page, BufferGetPage(dist->buffer));
		dist->page = BufferGetPage(dist->buffer);

		/* Write the WAL record */
		if (RelationNeedsWAL(rel))
			recptr = gistXLogSplit(rel->rd_node, blkno, is_leaf,
								   dist, oldrlink, oldnsn, leftchildbuf,
								   markfollowright);
		else
			recptr = gistGetFakeLSN(rel);

		for (ptr = dist; ptr; ptr = ptr->next)
		{
			PageSetLSN(ptr->page, recptr);
		}

		/*
		 * Return the new child buffers to the caller.
		 *
		 * If this was a root split, we've already inserted the downlink
		 * pointers, in the form of a new root page. Therefore we can release
		 * all the new buffers, and keep just the root page locked.
		 */
		if (is_rootsplit)
		{
			for (ptr = dist->next; ptr; ptr = ptr->next)
				UnlockReleaseBuffer(ptr->buffer);
		}
	}
	else
	{
		/*
		 * Enough space. We also get here if ntuples==0.
		 */
		START_CRIT_SECTION();

		if (OffsetNumberIsValid(oldoffnum))
			PageIndexTupleDelete(page, oldoffnum);
		gistfillbuffer(page, itup, ntup, InvalidOffsetNumber);

		MarkBufferDirty(buffer);

		if (BufferIsValid(leftchildbuf))
			MarkBufferDirty(leftchildbuf);

		if (RelationNeedsWAL(rel))
		{
			OffsetNumber ndeloffs = 0,
						deloffs[1];

			if (OffsetNumberIsValid(oldoffnum))
			{
				deloffs[0] = oldoffnum;
				ndeloffs = 1;
			}

			recptr = gistXLogUpdate(rel->rd_node, buffer,
									deloffs, ndeloffs, itup, ntup,
									leftchildbuf);

			PageSetLSN(page, recptr);
		}
		else
		{
			recptr = gistGetFakeLSN(rel);
			PageSetLSN(page, recptr);
		}

		if (newblkno)
			*newblkno = blkno;
	}

	/*
	 * If we inserted the downlink for a child page, set NSN and clear
	 * F_FOLLOW_RIGHT flag on the left child, so that concurrent scans know to
	 * follow the rightlink if and only if they looked at the parent page
	 * before we inserted the downlink.
	 *
	 * Note that we do this *after* writing the WAL record. That means that
	 * the possible full page image in the WAL record does not include these
	 * changes, and they must be replayed even if the page is restored from
	 * the full page image. There's a chicken-and-egg problem: if we updated
	 * the child pages first, we wouldn't know the recptr of the WAL record
	 * we're about to write.
	 */
	if (BufferIsValid(leftchildbuf))
	{
		Page		leftpg = BufferGetPage(leftchildbuf);

		GistPageSetNSN(leftpg, recptr);
		GistClearFollowRight(leftpg);

		PageSetLSN(leftpg, recptr);
	}

	END_CRIT_SECTION();

	return is_split;
}

/*
 * Bulk deletion of all index entries pointing to a set of heap tuples and
 * check invalid tuples after crash recovery.
 * The set of target tuples is specified via a callback routine that tells
 * whether any given heap tuple (identified by ItemPointer) is being deleted.
 *
 * Result: a palloc'd struct containing statistical info for VACUUM displays.
 */
Datum
gistbulkdelete(PG_FUNCTION_ARGS)
{
	IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
	GistBulkDeleteResult *stats = (GistBulkDeleteResult *) PG_GETARG_POINTER(1);
	IndexBulkDeleteCallback callback = (IndexBulkDeleteCallback) PG_GETARG_POINTER(2);
	void	   *callback_state = (void *) PG_GETARG_POINTER(3);
	Relation	rel = info->index;
	GistBDItem *stack,
			   *ptr;

	/* first time through? */
	if (stats == NULL)
		stats = (GistBulkDeleteResult *) palloc0(sizeof(GistBulkDeleteResult));
	/* we'll re-count the tuples each time */
	stats->std.num_index_tuples = 0;

	stack = (GistBDItem *) palloc0(sizeof(GistBDItem));
	stack->blkno = GIST_ROOT_BLKNO;

	while (stack)
	{
		Buffer		buffer = ReadBufferWithStrategy(rel, stack->blkno, info->strategy);
		Page		page;
		OffsetNumber i,
					maxoff;
		IndexTuple	idxtuple;
		ItemId		iid;

		LockBuffer(buffer, GIST_SHARE);
		gistcheckpage(rel, buffer);
		page = (Page) BufferGetPage(buffer);

		if (GistPageIsLeaf(page))
		{
			OffsetNumber todelete[MaxOffsetNumber];
			int			ntodelete = 0;

			LockBuffer(buffer, GIST_UNLOCK);
			LockBuffer(buffer, GIST_EXCLUSIVE);

			page = (Page) BufferGetPage(buffer);
			if (stack->blkno == GIST_ROOT_BLKNO && !GistPageIsLeaf(page))
			{
				/* only the root can become non-leaf during relock */
				UnlockReleaseBuffer(buffer);
				/* one more check */
				continue;
			}

			/*
			 * check for split proceeded after look at parent, we should check
			 * it after relock
			 */
			pushStackIfSplited(page, stack);

			/*
			 * Remove deletable tuples from page
			 */

			maxoff = PageGetMaxOffsetNumber(page);

			for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
			{
				iid = PageGetItemId(page, i);
				idxtuple = (IndexTuple) PageGetItem(page, iid);

				if (callback(&(idxtuple->t_tid), callback_state))
				{
					todelete[ntodelete] = i - ntodelete;
					ntodelete++;
					stats->std.tuples_removed += 1;
				}
				else
					stats->std.num_index_tuples += 1;
			}

			if (ntodelete)
			{
				START_CRIT_SECTION();

				MarkBufferDirty(buffer);

				for (i = 0; i < ntodelete; i++)
					PageIndexTupleDelete(page, todelete[i]);
				GistMarkTuplesDeleted(page);

				if (!rel->rd_istemp)
				{
					XLogRecData *rdata;
					XLogRecPtr	recptr;
					gistxlogPageUpdate *xlinfo;
//.........这里部分代码省略.........

static void
spgRedoSplitTuple(XLogRecPtr lsn, XLogRecord *record)
{
	char	   *ptr = XLogRecGetData(record);
	spgxlogSplitTuple *xldata = (spgxlogSplitTuple *) ptr;
	SpGistInnerTuple prefixTuple;
	SpGistInnerTuple postfixTuple;
	Buffer		buffer;
	Page		page;

	/* we assume this is adequately aligned */
	ptr += sizeof(spgxlogSplitTuple);
	prefixTuple = (SpGistInnerTuple) ptr;
	ptr += prefixTuple->size;
	postfixTuple = (SpGistInnerTuple) ptr;

	/* insert postfix tuple first to avoid dangling link */
	if (xldata->blknoPostfix != xldata->blknoPrefix &&
		!(record->xl_info & XLR_BKP_BLOCK_2))
	{
		buffer = XLogReadBuffer(xldata->node, xldata->blknoPostfix,
								xldata->newPage);
		if (BufferIsValid(buffer))
		{
			page = BufferGetPage(buffer);

			if (xldata->newPage)
				SpGistInitBuffer(buffer, 0);

			if (!XLByteLE(lsn, PageGetLSN(page)))
			{
				addOrReplaceTuple(page, (Item) postfixTuple,
								  postfixTuple->size, xldata->offnumPostfix);

				PageSetLSN(page, lsn);
				PageSetTLI(page, ThisTimeLineID);
				MarkBufferDirty(buffer);
			}
			UnlockReleaseBuffer(buffer);
		}
	}

	/* now handle the original page */
	if (!(record->xl_info & XLR_BKP_BLOCK_1))
	{
		buffer = XLogReadBuffer(xldata->node, xldata->blknoPrefix, false);
		if (BufferIsValid(buffer))
		{
			page = BufferGetPage(buffer);
			if (!XLByteLE(lsn, PageGetLSN(page)))
			{
				PageIndexTupleDelete(page, xldata->offnumPrefix);
				if (PageAddItem(page, (Item) prefixTuple, prefixTuple->size,
								xldata->offnumPrefix, false, false) != xldata->offnumPrefix)
					elog(ERROR, "failed to add item of size %u to SPGiST index page",
						 prefixTuple->size);

				if (xldata->blknoPostfix == xldata->blknoPrefix)
					addOrReplaceTuple(page, (Item) postfixTuple,
									  postfixTuple->size,
									  xldata->offnumPostfix);

				PageSetLSN(page, lsn);
				PageSetTLI(page, ThisTimeLineID);
				MarkBufferDirty(buffer);
			}
			UnlockReleaseBuffer(buffer);
		}
	}
}

static void
spgRedoAddNode(XLogRecPtr lsn, XLogRecord *record)
{
	char	   *ptr = XLogRecGetData(record);
	spgxlogAddNode *xldata = (spgxlogAddNode *) ptr;
	SpGistInnerTuple innerTuple;
	SpGistState state;
	Buffer		buffer;
	Page		page;
	int			bbi;

	/* we assume this is adequately aligned */
	ptr += sizeof(spgxlogAddNode);
	innerTuple = (SpGistInnerTuple) ptr;

	fillFakeState(&state, xldata->stateSrc);

	if (xldata->blknoNew == InvalidBlockNumber)
	{
		/* update in place */
		Assert(xldata->blknoParent == InvalidBlockNumber);
		if (!(record->xl_info & XLR_BKP_BLOCK_1))
		{
			buffer = XLogReadBuffer(xldata->node, xldata->blkno, false);
			if (BufferIsValid(buffer))
			{
				page = BufferGetPage(buffer);
				if (!XLByteLE(lsn, PageGetLSN(page)))
				{
					PageIndexTupleDelete(page, xldata->offnum);
					if (PageAddItem(page, (Item) innerTuple, innerTuple->size,
									xldata->offnum,
									false, false) != xldata->offnum)
						elog(ERROR, "failed to add item of size %u to SPGiST index page",
							 innerTuple->size);

					PageSetLSN(page, lsn);
					PageSetTLI(page, ThisTimeLineID);
					MarkBufferDirty(buffer);
				}
				UnlockReleaseBuffer(buffer);
			}
		}
	}
	else
	{
		/* Install new tuple first so redirect is valid */
		if (!(record->xl_info & XLR_BKP_BLOCK_2))
		{
			buffer = XLogReadBuffer(xldata->node, xldata->blknoNew,
									xldata->newPage);
			if (BufferIsValid(buffer))
			{
				page = BufferGetPage(buffer);

				if (xldata->newPage)
					SpGistInitBuffer(buffer, 0);

				if (!XLByteLE(lsn, PageGetLSN(page)))
				{
					addOrReplaceTuple(page, (Item) innerTuple,
									  innerTuple->size, xldata->offnumNew);

					PageSetLSN(page, lsn);
					PageSetTLI(page, ThisTimeLineID);
					MarkBufferDirty(buffer);
				}
				UnlockReleaseBuffer(buffer);
			}
		}

		/* Delete old tuple, replacing it with redirect or placeholder tuple */
		if (!(record->xl_info & XLR_BKP_BLOCK_1))
		{
			buffer = XLogReadBuffer(xldata->node, xldata->blkno, false);
			if (BufferIsValid(buffer))
			{
				page = BufferGetPage(buffer);
				if (!XLByteLE(lsn, PageGetLSN(page)))
				{
					SpGistDeadTuple dt;

					if (state.isBuild)
						dt = spgFormDeadTuple(&state, SPGIST_PLACEHOLDER,
											  InvalidBlockNumber,
											  InvalidOffsetNumber);
					else
						dt = spgFormDeadTuple(&state, SPGIST_REDIRECT,
											  xldata->blknoNew,
											  xldata->offnumNew);

					PageIndexTupleDelete(page, xldata->offnum);
					if (PageAddItem(page, (Item) dt, dt->size,
									xldata->offnum,
									false, false) != xldata->offnum)
						elog(ERROR, "failed to add item of size %u to SPGiST index page",
							 dt->size);

					if (state.isBuild)
						SpGistPageGetOpaque(page)->nPlaceholder++;
//.........这里部分代码省略.........

/*
 * redo any page update (except page split)
 */
static void
gistRedoPageUpdateRecord(XLogRecPtr lsn, XLogRecord *record)
{
	char	   *begin = XLogRecGetData(record);
	gistxlogPageUpdate *xldata = (gistxlogPageUpdate *) begin;
	Buffer		buffer;
	Page		page;
	char	   *data;

	/*
	 * We need to acquire and hold lock on target page while updating the left
	 * child page.	If we have a full-page image of target page, getting the
	 * lock is a side-effect of restoring that image.  Note that even if the
	 * target page no longer exists, we'll still attempt to replay the change
	 * on the child page.
	 */
	if (record->xl_info & XLR_BKP_BLOCK(0))
		buffer = RestoreBackupBlock(lsn, record, 0, false, true);
	else
		buffer = XLogReadBuffer(xldata->node, xldata->blkno, false);

	/* Fix follow-right data on left child page */
	if (BlockNumberIsValid(xldata->leftchild))
		gistRedoClearFollowRight(lsn, record, 1,
								 xldata->node, xldata->leftchild);

	/* Done if target page no longer exists */
	if (!BufferIsValid(buffer))
		return;

	/* nothing more to do if page was backed up (and no info to do it with) */
	if (record->xl_info & XLR_BKP_BLOCK(0))
	{
		UnlockReleaseBuffer(buffer);
		return;
	}

	page = (Page) BufferGetPage(buffer);

	/* nothing more to do if change already applied */
	if (lsn <= PageGetLSN(page))
	{
		UnlockReleaseBuffer(buffer);
		return;
	}

	data = begin + sizeof(gistxlogPageUpdate);

	/* Delete old tuples */
	if (xldata->ntodelete > 0)
	{
		int			i;
		OffsetNumber *todelete = (OffsetNumber *) data;

		data += sizeof(OffsetNumber) * xldata->ntodelete;

		for (i = 0; i < xldata->ntodelete; i++)
			PageIndexTupleDelete(page, todelete[i]);
		if (GistPageIsLeaf(page))
			GistMarkTuplesDeleted(page);
	}

	/* add tuples */
	if (data - begin < record->xl_len)
	{
		OffsetNumber off = (PageIsEmpty(page)) ? FirstOffsetNumber :
		OffsetNumberNext(PageGetMaxOffsetNumber(page));

		while (data - begin < record->xl_len)
		{
			IndexTuple	itup = (IndexTuple) data;
			Size		sz = IndexTupleSize(itup);
			OffsetNumber l;

			data += sz;

			l = PageAddItem(page, (Item) itup, sz, off, false, false);
			if (l == InvalidOffsetNumber)
				elog(ERROR, "failed to add item to GiST index page, size %d bytes",
					 (int) sz);
			off++;
		}
	}
	else
	{
		/*
		 * special case: leafpage, nothing to insert, nothing to delete, then
		 * vacuum marks page
		 */
		if (GistPageIsLeaf(page) && xldata->ntodelete == 0)
			GistClearTuplesDeleted(page);
	}

	if (!GistPageIsLeaf(page) &&
		PageGetMaxOffsetNumber(page) == InvalidOffsetNumber &&
		xldata->blkno == GIST_ROOT_BLKNO)
	{
//.........这里部分代码省略.........

/*
 * PageIndexMultiDelete
 *
 * This routine handles the case of deleting multiple tuples from an
 * index page at once.	It is considerably faster than a loop around
 * PageIndexTupleDelete ... however, the caller *must* supply the array
 * of item numbers to be deleted in item number order!
 */
void
PageIndexMultiDelete(Page page, OffsetNumber *itemnos, int nitems)
{
	PageHeader	phdr = (PageHeader) page;
	Offset		pd_lower = phdr->pd_lower;
	Offset		pd_upper = phdr->pd_upper;
	Offset		pd_special = phdr->pd_special;
	itemIdSort	itemidbase,
				itemidptr;
	ItemId		lp;
	int			nline,
				nused;
	int			i;
	Size		totallen;
	Offset		upper;
	Size		size;
	unsigned	offset;
	int			nextitm;
	OffsetNumber offnum;

	/*
	 * If there aren't very many items to delete, then retail
	 * PageIndexTupleDelete is the best way.  Delete the items in reverse
	 * order so we don't have to think about adjusting item numbers for
	 * previous deletions.
	 *
	 * TODO: tune the magic number here
	 */
	if (nitems <= 2)
	{
		while (--nitems >= 0)
			PageIndexTupleDelete(page, itemnos[nitems]);
		return;
	}

	/*
	 * As with PageRepairFragmentation, paranoia seems justified.
	 */
	if (pd_lower < SizeOfPageHeaderData ||
		pd_lower > pd_upper ||
		pd_upper > pd_special ||
		pd_special > BLCKSZ ||
		pd_special != MAXALIGN(pd_special))
		ereport(ERROR,
				(errcode(ERRCODE_DATA_CORRUPTED),
				 errmsg("corrupted page pointers: lower = %u, upper = %u, special = %u",
						pd_lower, pd_upper, pd_special)));

	/*
	 * Scan the item pointer array and build a list of just the ones we are
	 * going to keep.  Notice we do not modify the page yet, since we are
	 * still validity-checking.
	 */
	nline = PageGetMaxOffsetNumber(page);
	itemidbase = (itemIdSort) palloc(sizeof(itemIdSortData) * nline);
	itemidptr = itemidbase;
	totallen = 0;
	nused = 0;
	nextitm = 0;
	for (offnum = FirstOffsetNumber; offnum <= nline; offnum = OffsetNumberNext(offnum))
	{
		lp = PageGetItemId(page, offnum);
		Assert(ItemIdHasStorage(lp));
		size = ItemIdGetLength(lp);
		offset = ItemIdGetOffset(lp);
		if (offset < pd_upper ||
			(offset + size) > pd_special ||
			offset != MAXALIGN(offset))
			ereport(ERROR,
					(errcode(ERRCODE_DATA_CORRUPTED),
					 errmsg("corrupted item pointer: offset = %u, size = %u",
							offset, (unsigned int) size)));

		if (nextitm < nitems && offnum == itemnos[nextitm])
		{
			/* skip item to be deleted */
			nextitm++;
		}
		else
		{
			itemidptr->offsetindex = nused;		/* where it will go */
			itemidptr->itemoff = offset;
			itemidptr->olditemid = *lp;
			itemidptr->alignedlen = MAXALIGN(size);
			totallen += itemidptr->alignedlen;
			itemidptr++;
			nused++;
		}
	}

	/* this will catch invalid or out-of-order itemnos[] */
	if (nextitm != nitems)
//.........这里部分代码省略.........

/*
 * Bulk deletion of all index entries pointing to a set of heap tuples.
 * The set of target tuples is specified via a callback routine that tells
 * whether any given heap tuple (identified by ItemPointer) is being deleted.
 *
 * Result: a palloc'd struct containing statistical info for VACUUM displays.
 */
Datum
hashbulkdelete(PG_FUNCTION_ARGS)
{
	IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
	IndexBulkDeleteResult *stats = (IndexBulkDeleteResult *) PG_GETARG_POINTER(1);
	IndexBulkDeleteCallback callback = (IndexBulkDeleteCallback) PG_GETARG_POINTER(2);
	void	   *callback_state = (void *) PG_GETARG_POINTER(3);
	Relation	rel = info->index;
	double		tuples_removed;
	double		num_index_tuples;
	double		orig_ntuples;
	Bucket		orig_maxbucket;
	Bucket		cur_maxbucket;
	Bucket		cur_bucket;
	Buffer		metabuf;
	HashMetaPage metap;
	HashMetaPageData local_metapage;

	tuples_removed = 0;
	num_index_tuples = 0;

	/*
	 * Read the metapage to fetch original bucket and tuple counts.  Also, we
	 * keep a copy of the last-seen metapage so that we can use its
	 * hashm_spares[] values to compute bucket page addresses.	This is a bit
	 * hokey but perfectly safe, since the interesting entries in the spares
	 * array cannot change under us; and it beats rereading the metapage for
	 * each bucket.
	 */
	metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_READ);
	_hash_checkpage(rel, metabuf, LH_META_PAGE);
	metap = (HashMetaPage) BufferGetPage(metabuf);
	orig_maxbucket = metap->hashm_maxbucket;
	orig_ntuples = metap->hashm_ntuples;
	memcpy(&local_metapage, metap, sizeof(local_metapage));
	_hash_relbuf(rel, metabuf);

	/* Scan the buckets that we know exist */
	cur_bucket = 0;
	cur_maxbucket = orig_maxbucket;

loop_top:
	while (cur_bucket <= cur_maxbucket)
	{
		BlockNumber bucket_blkno;
		BlockNumber blkno;
		bool		bucket_dirty = false;

		/* Get address of bucket's start page */
		bucket_blkno = BUCKET_TO_BLKNO(&local_metapage, cur_bucket);

		/* Exclusive-lock the bucket so we can shrink it */
		_hash_getlock(rel, bucket_blkno, HASH_EXCLUSIVE);

		/* Shouldn't have any active scans locally, either */
		if (_hash_has_active_scan(rel, cur_bucket))
			elog(ERROR, "hash index has active scan during VACUUM");

		/* Scan each page in bucket */
		blkno = bucket_blkno;
		while (BlockNumberIsValid(blkno))
		{
			Buffer		buf;
			Page		page;
			HashPageOpaque opaque;
			OffsetNumber offno;
			OffsetNumber maxoffno;
			bool		page_dirty = false;

			vacuum_delay_point();

			buf = _hash_getbuf(rel, blkno, HASH_WRITE);
			_hash_checkpage(rel, buf, LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
			page = BufferGetPage(buf);
			opaque = (HashPageOpaque) PageGetSpecialPointer(page);
			Assert(opaque->hasho_bucket == cur_bucket);

			/* Scan each tuple in page */
			offno = FirstOffsetNumber;
			maxoffno = PageGetMaxOffsetNumber(page);
			while (offno <= maxoffno)
			{
				IndexTuple	itup;
				ItemPointer htup;

				itup = (IndexTuple) PageGetItem(page,
												PageGetItemId(page, offno));
				htup = &(itup->t_tid);
				if (callback(htup, callback_state))
				{
					/* delete the item from the page */
					PageIndexTupleDelete(page, offno);
					bucket_dirty = page_dirty = true;
//.........这里部分代码省略.........

static void
spgRedoAddLeaf(XLogRecPtr lsn, XLogRecord *record)
{
	char	   *ptr = XLogRecGetData(record);
	spgxlogAddLeaf *xldata = (spgxlogAddLeaf *) ptr;
	char	   *leafTuple;
	SpGistLeafTupleData leafTupleHdr;
	Buffer		buffer;
	Page		page;
	XLogRedoAction action;

	ptr += sizeof(spgxlogAddLeaf);
	leafTuple = ptr;
	/* the leaf tuple is unaligned, so make a copy to access its header */
	memcpy(&le