ginvacuum.c

postgresql8.3.4源码,开源数据库

		me = parent;
	}
	else
	{
		if (!parent->child)
		{
			me = (DataPageDeleteStack *) palloc0(sizeof(DataPageDeleteStack));
			me->parent = parent;
			parent->child = me;
			me->leftBlkno = InvalidBlockNumber;
		}
		else
			me = parent->child;
	}

	buffer = ReadBufferWithStrategy(gvs->index, blkno, gvs->strategy);
	page = BufferGetPage(buffer);

	Assert(GinPageIsData(page));

	if (!GinPageIsLeaf(page))
	{
		OffsetNumber i;

		me->blkno = blkno;
		for (i = FirstOffsetNumber; i <= GinPageGetOpaque(page)->maxoff; i++)
		{
			PostingItem *pitem = (PostingItem *) GinDataPageGetItem(page, i);

			if (ginScanToDelete(gvs, PostingItemGetBlockNumber(pitem), FALSE, me, i))
				i--;
		}
	}

	if (GinPageGetOpaque(page)->maxoff < FirstOffsetNumber)
	{
		if (!(me->leftBlkno == InvalidBlockNumber && GinPageRightMost(page)))
		{
			/* we never delete right most branch */
			Assert(!isRoot);
			if (GinPageGetOpaque(page)->maxoff < FirstOffsetNumber)
			{
				ginDeletePage(gvs, blkno, me->leftBlkno, me->parent->blkno, myoff, me->parent->isRoot);
				meDelete = TRUE;
			}
		}
	}

	ReleaseBuffer(buffer);

	if (!meDelete)
		me->leftBlkno = blkno;

	return meDelete;
}

static void
ginVacuumPostingTree(GinVacuumState *gvs, BlockNumber rootBlkno)
{
	Buffer		rootBuffer = InvalidBuffer;
	DataPageDeleteStack root,
			   *ptr,
			   *tmp;

	if (ginVacuumPostingTreeLeaves(gvs, rootBlkno, TRUE, &rootBuffer) == FALSE)
	{
		Assert(rootBuffer == InvalidBuffer);
		return;
	}

	memset(&root, 0, sizeof(DataPageDeleteStack));
	root.leftBlkno = InvalidBlockNumber;
	root.isRoot = TRUE;

	vacuum_delay_point();

	ginScanToDelete(gvs, rootBlkno, TRUE, &root, InvalidOffsetNumber);

	ptr = root.child;
	while (ptr)
	{
		tmp = ptr->child;
		pfree(ptr);
		ptr = tmp;
	}

	UnlockReleaseBuffer(rootBuffer);
}

/*
 * returns modified page or NULL if page isn't modified.
 * Function works with original page until first change is occurred,
 * then page is copied into temporary one.
 */
static Page
ginVacuumEntryPage(GinVacuumState *gvs, Buffer buffer, BlockNumber *roots, uint32 *nroot)
{
	Page		origpage = BufferGetPage(buffer),
				tmppage;
	OffsetNumber i,
				maxoff = PageGetMaxOffsetNumber(origpage);

	tmppage = origpage;

	*nroot = 0;

	for (i = FirstOffsetNumber; i <= maxoff; i++)
	{
		IndexTuple	itup = (IndexTuple) PageGetItem(tmppage, PageGetItemId(tmppage, i));

		if (GinIsPostingTree(itup))
		{
			/*
			 * store posting tree's roots for further processing, we can't
			 * vacuum it just now due to risk of deadlocks with scans/inserts
			 */
			roots[*nroot] = GinItemPointerGetBlockNumber(&itup->t_tid);
			(*nroot)++;
		}
		else if (GinGetNPosting(itup) > 0)
		{
			/*
			 * if we already create temporary page, we will make changes in
			 * place
			 */
			ItemPointerData *cleaned = (tmppage == origpage) ? NULL : GinGetPosting(itup);
			uint32		newN = ginVacuumPostingList(gvs, GinGetPosting(itup), GinGetNPosting(itup), &cleaned);

			if (GinGetNPosting(itup) != newN)
			{
				bool		isnull;
				Datum		value;

				/*
				 * Some ItemPointers was deleted, so we should remake our
				 * tuple
				 */

				if (tmppage == origpage)
				{
					/*
					 * On first difference we create temporary page in memory
					 * and copies content in to it.
					 */
					tmppage = GinPageGetCopyPage(origpage);

					if (newN > 0)
					{
						Size		pos = ((char *) GinGetPosting(itup)) - ((char *) origpage);

						memcpy(tmppage + pos, cleaned, sizeof(ItemPointerData) * newN);
					}

					pfree(cleaned);

					/* set itup pointer to new page */
					itup = (IndexTuple) PageGetItem(tmppage, PageGetItemId(tmppage, i));
				}

				value = index_getattr(itup, FirstOffsetNumber, gvs->ginstate.tupdesc, &isnull);
				itup = GinFormTuple(&gvs->ginstate, value, GinGetPosting(itup), newN);
				PageIndexTupleDelete(tmppage, i);

				if (PageAddItem(tmppage, (Item) itup, IndexTupleSize(itup), i, false, false) != i)
					elog(ERROR, "failed to add item to index page in \"%s\"",
						 RelationGetRelationName(gvs->index));

				pfree(itup);
			}
		}
	}

	return (tmppage == origpage) ? NULL : tmppage;
}

Datum
ginbulkdelete(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	index = info->index;
	BlockNumber blkno = GIN_ROOT_BLKNO;
	GinVacuumState gvs;
	Buffer		buffer;
	BlockNumber rootOfPostingTree[BLCKSZ / (sizeof(IndexTupleData) + sizeof(ItemId))];
	uint32		nRoot;

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

	gvs.index = index;
	gvs.result = stats;
	gvs.callback = callback;
	gvs.callback_state = callback_state;
	gvs.strategy = info->strategy;
	initGinState(&gvs.ginstate, index);

	buffer = ReadBufferWithStrategy(index, blkno, info->strategy);

	/* find leaf page */
	for (;;)
	{
		Page		page = BufferGetPage(buffer);
		IndexTuple	itup;

		LockBuffer(buffer, GIN_SHARE);

		Assert(!GinPageIsData(page));

		if (GinPageIsLeaf(page))
		{
			LockBuffer(buffer, GIN_UNLOCK);
			LockBuffer(buffer, GIN_EXCLUSIVE);

			if (blkno == GIN_ROOT_BLKNO && !GinPageIsLeaf(page))
			{
				LockBuffer(buffer, GIN_UNLOCK);
				continue;		/* check it one more */
			}
			break;
		}

		Assert(PageGetMaxOffsetNumber(page) >= FirstOffsetNumber);

		itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, FirstOffsetNumber));
		blkno = GinItemPointerGetBlockNumber(&(itup)->t_tid);
		Assert(blkno != InvalidBlockNumber);

		UnlockReleaseBuffer(buffer);
		buffer = ReadBufferWithStrategy(index, blkno, info->strategy);
	}

	/* right now we found leftmost page in entry's BTree */

	for (;;)
	{
		Page		page = BufferGetPage(buffer);
		Page		resPage;
		uint32		i;

		Assert(!GinPageIsData(page));

		resPage = ginVacuumEntryPage(&gvs, buffer, rootOfPostingTree, &nRoot);

		blkno = GinPageGetOpaque(page)->rightlink;

		if (resPage)
		{
			START_CRIT_SECTION();
			PageRestoreTempPage(resPage, page);
			MarkBufferDirty(buffer);
			xlogVacuumPage(gvs.index, buffer);
			UnlockReleaseBuffer(buffer);
			END_CRIT_SECTION();
		}
		else
		{
			UnlockReleaseBuffer(buffer);
		}

		vacuum_delay_point();

		for (i = 0; i < nRoot; i++)
		{
			ginVacuumPostingTree(&gvs, rootOfPostingTree[i]);
			vacuum_delay_point();
		}

		if (blkno == InvalidBlockNumber)		/* rightmost page */
			break;

		buffer = ReadBufferWithStrategy(index, blkno, info->strategy);
		LockBuffer(buffer, GIN_EXCLUSIVE);
	}

	PG_RETURN_POINTER(gvs.result);
}

Datum
ginvacuumcleanup(PG_FUNCTION_ARGS)
{
	IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
	IndexBulkDeleteResult *stats = (IndexBulkDeleteResult *) PG_GETARG_POINTER(1);
	Relation	index = info->index;
	bool		needLock;
	BlockNumber npages,
				blkno;
	BlockNumber totFreePages,
				nFreePages,
			   *freePages,
				maxFreePages;
	BlockNumber lastBlock = GIN_ROOT_BLKNO,
				lastFilledBlock = GIN_ROOT_BLKNO;

	/* Set up all-zero stats if ginbulkdelete wasn't called */
	if (stats == NULL)
		stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));

	/*
	 * XXX we always report the heap tuple count as the number of index
	 * entries.  This is bogus if the index is partial, but it's real hard to
	 * tell how many distinct heap entries are referenced by a GIN index.
	 */
	stats->num_index_tuples = info->num_heap_tuples;

	/*
	 * If vacuum full, we already have exclusive lock on the index. Otherwise,
	 * need lock unless it's local to this backend.
	 */
	if (info->vacuum_full)
		needLock = false;
	else
		needLock = !RELATION_IS_LOCAL(index);

	if (needLock)
		LockRelationForExtension(index, ExclusiveLock);
	npages = RelationGetNumberOfBlocks(index);
	if (needLock)
		UnlockRelationForExtension(index, ExclusiveLock);

	maxFreePages = npages;
	if (maxFreePages > MaxFSMPages)
		maxFreePages = MaxFSMPages;

	totFreePages = nFreePages = 0;
	freePages = (BlockNumber *) palloc(sizeof(BlockNumber) * maxFreePages);

	for (blkno = GIN_ROOT_BLKNO + 1; blkno < npages; blkno++)
	{
		Buffer		buffer;
		Page		page;

		vacuum_delay_point();

		buffer = ReadBufferWithStrategy(index, blkno, info->strategy);
		LockBuffer(buffer, GIN_SHARE);
		page = (Page) BufferGetPage(buffer);

		if (GinPageIsDeleted(page))
		{
			if (nFreePages < maxFreePages)
				freePages[nFreePages++] = blkno;
			totFreePages++;
		}
		else
			lastFilledBlock = blkno;

		UnlockReleaseBuffer(buffer);
	}
	lastBlock = npages - 1;

	if (info->vacuum_full && nFreePages > 0)
	{
		/* try to truncate index */
		int			i;

		for (i = 0; i < nFreePages; i++)
			if (freePages[i] >= lastFilledBlock)
			{
				totFreePages = nFreePages = i;
				break;
			}

		if (lastBlock > lastFilledBlock)
			RelationTruncate(index, lastFilledBlock + 1);

		stats->pages_removed = lastBlock - lastFilledBlock;
	}

	RecordIndexFreeSpace(&index->rd_node, totFreePages, nFreePages, freePages);
	stats->pages_free = totFreePages;

	if (needLock)
		LockRelationForExtension(index, ExclusiveLock);
	stats->num_pages = RelationGetNumberOfBlocks(index);
	if (needLock)
		UnlockRelationForExtension(index, ExclusiveLock);

	PG_RETURN_POINTER(stats);
}