/*
* 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);
}
}
/*
* _hash_splitbucket -- split 'obucket' into 'obucket' and 'nbucket'
*
* We are splitting a bucket that consists of a base bucket page and zero
* or more overflow (bucket chain) pages. We must relocate tuples that
* belong in the new bucket, and compress out any free space in the old
* bucket.
*
* The caller must hold exclusive locks on both buckets to ensure that
* no one else is trying to access them (see README).
*
* The caller must hold a pin, but no lock, on the metapage buffer.
* The buffer is returned in the same state. (The metapage is only
* touched if it becomes necessary to add or remove overflow pages.)
*/
static void
_hash_splitbucket(Relation rel,
Buffer metabuf,
Bucket obucket,
Bucket nbucket,
BlockNumber start_oblkno,
BlockNumber start_nblkno,
uint32 maxbucket,
uint32 highmask,
uint32 lowmask)
{
Bucket bucket;
Buffer obuf;
Buffer nbuf;
BlockNumber oblkno;
BlockNumber nblkno;
bool null;
Datum datum;
HashItem hitem;
HashPageOpaque oopaque;
HashPageOpaque nopaque;
IndexTuple itup;
Size itemsz;
OffsetNumber ooffnum;
OffsetNumber noffnum;
OffsetNumber omaxoffnum;
Page opage;
Page npage;
TupleDesc itupdesc = RelationGetDescr(rel);
/*
* It should be okay to simultaneously write-lock pages from each
* bucket, since no one else can be trying to acquire buffer lock
* on pages of either bucket.
*/
oblkno = start_oblkno;
nblkno = start_nblkno;
obuf = _hash_getbuf(rel, oblkno, HASH_WRITE);
nbuf = _hash_getbuf(rel, nblkno, HASH_WRITE);
opage = BufferGetPage(obuf);
npage = BufferGetPage(nbuf);
_hash_checkpage(rel, opage, LH_BUCKET_PAGE);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
/* initialize the new bucket's primary page */
_hash_pageinit(npage, BufferGetPageSize(nbuf));
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
nopaque->hasho_prevblkno = InvalidBlockNumber;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = nbucket;
nopaque->hasho_flag = LH_BUCKET_PAGE;
nopaque->hasho_filler = HASHO_FILL;
/*
* Partition the tuples in the old bucket between the old bucket and the
* new bucket, advancing along the old bucket's overflow bucket chain
* and adding overflow pages to the new bucket as needed.
*/
ooffnum = FirstOffsetNumber;
omaxoffnum = PageGetMaxOffsetNumber(opage);
for (;;)
{
/*
* at each iteration through this loop, each of these variables
* should be up-to-date: obuf opage oopaque ooffnum omaxoffnum
*/
/* check if we're at the end of the page */
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);
}
static bool
gistplacetopage(GISTInsertState *state, GISTSTATE *giststate)
{
bool is_splitted = false;
bool is_leaf = (GistPageIsLeaf(state->stack->page)) ? true : false;
MIRROREDLOCK_BUFMGR_MUST_ALREADY_BE_HELD;
/*
* if (!is_leaf) remove old key: This node's key has been modified, either
* because a child split occurred or because we needed to adjust our key
* for an insert in a child node. Therefore, remove the old version of
* this node's key.
*
* for WAL replay, in the non-split case we handle this by setting up a
* one-element todelete array; in the split case, it's handled implicitly
* because the tuple vector passed to gistSplit won't include this tuple.
*
* XXX: If we want to change fillfactors between node and leaf, fillfactor
* = (is_leaf ? state->leaf_fillfactor : state->node_fillfactor)
*/
if (gistnospace(state->stack->page, state->itup, state->ituplen,
is_leaf ? InvalidOffsetNumber : state->stack->childoffnum,
state->freespace))
{
/* no space for insertion */
IndexTuple *itvec;
int tlen;
SplitedPageLayout *dist = NULL,
*ptr;
BlockNumber rrlink = InvalidBlockNumber;
GistNSN oldnsn;
is_splitted = true;
/*
* Form index tuples vector to split: remove old tuple if t's needed
* and add new tuples to vector
*/
itvec = gistextractpage(state->stack->page, &tlen);
if (!is_leaf)
{
/* on inner page we should remove old tuple */
int pos = state->stack->childoffnum - FirstOffsetNumber;
tlen--;
if (pos != tlen)
memmove(itvec + pos, itvec + pos + 1, sizeof(IndexTuple) * (tlen - pos));
}
itvec = gistjoinvector(itvec, &tlen, state->itup, state->ituplen);
dist = gistSplit(state->r, state->stack->page, itvec, tlen, giststate);
state->itup = (IndexTuple *) palloc(sizeof(IndexTuple) * tlen);
state->ituplen = 0;
if (state->stack->blkno != GIST_ROOT_BLKNO)
{
/*
* if non-root split then we should not allocate new buffer, but
* we must create temporary page to operate
*/
dist->buffer = state->stack->buffer;
dist->page = PageGetTempPage(BufferGetPage(dist->buffer), sizeof(GISTPageOpaqueData));
/* clean all flags except F_LEAF */
GistPageGetOpaque(dist->page)->flags = (is_leaf) ? F_LEAF : 0;
}
/* make new pages and fills them */
for (ptr = dist; ptr; ptr = ptr->next)
{
int i;
char *data;
/* get new page */
if (ptr->buffer == InvalidBuffer)
{
ptr->buffer = gistNewBuffer(state->r);
GISTInitBuffer(ptr->buffer, (is_leaf) ? F_LEAF : 0);
ptr->page = BufferGetPage(ptr->buffer);
}
ptr->block.blkno = BufferGetBlockNumber(ptr->buffer);
/*
* fill page, we can do it because all these pages are new
* (ie not linked in tree or masked by temp page
*/
data = (char *) (ptr->list);
for (i = 0; i < ptr->block.num; i++)
{
if (PageAddItem(ptr->page, (Item) data, IndexTupleSize((IndexTuple) data), i + FirstOffsetNumber, LP_USED) == InvalidOffsetNumber)
elog(ERROR, "failed to add item to index page in \"%s\"", RelationGetRelationName(state->r));
data += IndexTupleSize((IndexTuple) data);
}
/* set up ItemPointer and remember it for parent */
ItemPointerSetBlockNumber(&(ptr->itup->t_tid), ptr->block.blkno);
state->itup[state->ituplen] = ptr->itup;
state->ituplen++;
}
/* saves old rightlink */
if (state->stack->blkno != GIST_ROOT_BLKNO)
rrlink = GistPageGetOpaque(dist->page)->rightlink;
START_CRIT_SECTION();
/*
* must mark buffers dirty before XLogInsert, even though we'll still
* be changing their opaque fields below. set up right links.
*/
for (ptr = dist; ptr; ptr = ptr->next)
{
MarkBufferDirty(ptr->buffer);
GistPageGetOpaque(ptr->page)->rightlink = (ptr->next) ?
ptr->next->block.blkno : rrlink;
}
/* restore splitted non-root page */
if (state->stack->blkno != GIST_ROOT_BLKNO)
{
PageRestoreTempPage(dist->page, BufferGetPage(dist->buffer));
dist->page = BufferGetPage(dist->buffer);
}
if (!state->r->rd_istemp)
{
XLogRecPtr recptr;
XLogRecData *rdata;
rdata = formSplitRdata(state->r, state->stack->blkno,
is_leaf, &(state->key), dist);
recptr = XLogInsert(RM_GIST_ID, XLOG_GIST_PAGE_SPLIT, rdata);
for (ptr = dist; ptr; ptr = ptr->next)
{
PageSetLSN(ptr->page, recptr);
PageSetTLI(ptr->page, ThisTimeLineID);
}
}
else
{
for (ptr = dist; ptr; ptr = ptr->next)
{
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);
}
/*
* Place tuples from 'itup' to 'buffer'. If 'oldoffnum' is valid, the tuple
* at that offset is atomically removed along with inserting the new tuples.
* This is used to replace a tuple with a new one.
*
* If 'leftchildbuf' is valid, we're inserting the downlink for the page
* to the right of 'leftchildbuf', or updating the downlink for 'leftchildbuf'.
* F_FOLLOW_RIGHT flag on 'leftchildbuf' is cleared and NSN is set.
*
* If 'markfollowright' is true and the page is split, the left child is
* marked with F_FOLLOW_RIGHT flag. That is the normal case. During buffered
* index build, however, there is no concurrent access and the page splitting
* is done in a slightly simpler fashion, and false is passed.
*
* If there is not enough room on the page, it is split. All the split
* pages are kept pinned and locked and returned in *splitinfo, the caller
* is responsible for inserting the downlinks for them. However, if
* 'buffer' is the root page and it needs to be split, gistplacetopage()
* performs the split as one atomic operation, and *splitinfo is set to NIL.
* In that case, we continue to hold the root page locked, and the child
* pages are released; note that new tuple(s) are *not* on the root page
* but in one of the new child pages.
*
* If 'newblkno' is not NULL, returns the block number of page the first
* new/updated tuple was inserted to. Usually it's the given page, but could
* be its right sibling if the page was split.
*
* Returns 'true' if the page was split, 'false' otherwise.
*/
bool
gistplacetopage(Relation rel, Size freespace, GISTSTATE *giststate,
Buffer buffer,
IndexTuple *itup, int ntup, OffsetNumber oldoffnum,
BlockNumber *newblkno,
Buffer leftchildbuf,
List **splitinfo,
bool markfollowright)
{
BlockNumber blkno = BufferGetBlockNumber(buffer);
Page page = BufferGetPage(buffer);
bool is_leaf = (GistPageIsLeaf(page)) ? true : false;
XLogRecPtr recptr;
int i;
bool is_split;
/*
* Refuse to modify a page that's incompletely split. This should not
* happen because we finish any incomplete splits while we walk down the
* tree. However, it's remotely possible that another concurrent inserter
* splits a parent page, and errors out before completing the split. We
* will just throw an error in that case, and leave any split we had in
* progress unfinished too. The next insert that comes along will clean up
* the mess.
*/
if (GistFollowRight(page))
elog(ERROR, "concurrent GiST page split was incomplete");
*splitinfo = NIL;
/*
* if isupdate, remove old key: This node's key has been modified, either
* because a child split occurred or because we needed to adjust our key
* for an insert in a child node. Therefore, remove the old version of
* this node's key.
*
* for WAL replay, in the non-split case we handle this by setting up a
* one-element todelete array; in the split case, it's handled implicitly
* because the tuple vector passed to gistSplit won't include this tuple.
*/
is_split = gistnospace(page, itup, ntup, oldoffnum, freespace);
if (is_split)
{
/* no space for insertion */
IndexTuple *itvec;
int tlen;
SplitedPageLayout *dist = NULL,
*ptr;
BlockNumber oldrlink = InvalidBlockNumber;
GistNSN oldnsn = 0;
SplitedPageLayout rootpg;
bool is_rootsplit;
is_rootsplit = (blkno == GIST_ROOT_BLKNO);
/*
* Form index tuples vector to split. If we're replacing an old tuple,
* remove the old version from the vector.
*/
itvec = gistextractpage(page, &tlen);
if (OffsetNumberIsValid(oldoffnum))
{
/* on inner page we should remove old tuple */
int pos = oldoffnum - FirstOffsetNumber;
tlen--;
if (pos != tlen)
memmove(itvec + pos, itvec + pos + 1, sizeof(IndexTuple) * (tlen - pos));
}
itvec = gistjoinvector(itvec, &tlen, itup, ntup);
dist = gistSplit(rel, page, itvec, tlen, giststate);
/*
* Set up pages to work with. Allocate new buffers for all but the
* leftmost page. The original page becomes the new leftmost page, and
* is just replaced with the new contents.
*
* For a root-split, allocate new buffers for all child pages, the
* original page is overwritten with new root page containing
* downlinks to the new child pages.
*/
ptr = dist;
if (!is_rootsplit)
{
/* save old rightlink and NSN */
oldrlink = GistPageGetOpaque(page)->rightlink;
oldnsn = GistPageGetNSN(page);
dist->buffer = buffer;
dist->block.blkno = BufferGetBlockNumber(buffer);
dist->page = PageGetTempPageCopySpecial(BufferGetPage(buffer));
/* clean all flags except F_LEAF */
GistPageGetOpaque(dist->page)->flags = (is_leaf) ? F_LEAF : 0;
ptr = ptr->next;
}
for (; ptr; ptr = ptr->next)
{
/* Allocate new page */
ptr->buffer = gistNewBuffer(rel);
GISTInitBuffer(ptr->buffer, (is_leaf) ? F_LEAF : 0);
ptr->page = BufferGetPage(ptr->buffer);
ptr->block.blkno = BufferGetBlockNumber(ptr->buffer);
}
/*
* Now that we know which blocks the new pages go to, set up downlink
* tuples to point to them.
*/
for (ptr = dist; ptr; ptr = ptr->next)
{
ItemPointerSetBlockNumber(&(ptr->itup->t_tid), ptr->block.blkno);
GistTupleSetValid(ptr->itup);
}
/*
* If this is a root split, we construct the new root page with the
* downlinks here directly, instead of requiring the caller to insert
* them. Add the new root page to the list along with the child pages.
*/
if (is_rootsplit)
{
IndexTuple *downlinks;
int ndownlinks = 0;
int i;
rootpg.buffer = buffer;
rootpg.page = PageGetTempPageCopySpecial(BufferGetPage(rootpg.buffer));
GistPageGetOpaque(rootpg.page)->flags = 0;
/* Prepare a vector of all the downlinks */
for (ptr = dist; ptr; ptr = ptr->next)
ndownlinks++;
downlinks = palloc(sizeof(IndexTuple) * ndownlinks);
for (i = 0, ptr = dist; ptr; ptr = ptr->next)
downlinks[i++] = ptr->itup;
rootpg.block.blkno = GIST_ROOT_BLKNO;
rootpg.block.num = ndownlinks;
rootpg.list = gistfillitupvec(downlinks, ndownlinks,
&(rootpg.lenlist));
rootpg.itup = NULL;
rootpg.next = dist;
dist = &rootpg;
}
else
{
/* Prepare split-info to be returned to caller */
for (ptr = dist; ptr; ptr = ptr->next)
{
GISTPageSplitInfo *si = palloc(sizeof(GISTPageSplitInfo));
si->buf = ptr->buffer;
si->downlink = ptr->itup;
*splitinfo = lappend(*splitinfo, si);
}
}
/*
* Fill all pages. All the pages are new, ie. freshly allocated empty
* pages, or a temporary copy of the old page.
*/
for (ptr = dist; ptr; ptr = ptr->next)
{
char *data = (char *) (ptr->list);
for (i = 0; i < ptr->block.num; i++)
{
IndexTuple thistup = (IndexTuple) data;
if (PageAddItem(ptr->page, (Item) data, IndexTupleSize(thistup), i + FirstOffsetNumber, false, false) == InvalidOffsetNumber)
elog(ERROR, "failed to add item to index page in \"%s\"", RelationGetRelationName(rel));
/*
* If this is the first inserted/updated tuple, let the caller
* know which page it landed on.
*/
if (newblkno && ItemPointerEquals(&thistup->t_tid, &(*itup)->t_tid))
*newblkno = ptr->block.blkno;
data += IndexTupleSize(thistup);
}
/* Set up rightlinks */
if (ptr->next && ptr->block.blkno != GIST_ROOT_BLKNO)
GistPageGetOpaque(ptr->page)->rightlink =
ptr->next->block.blkno;
else
GistPageGetOpaque(ptr->page)->rightlink = oldrlink;
/*
* Mark the all but the right-most page with the follow-right
* flag. It will be cleared as soon as the downlink is inserted
* into the parent, but this ensures that if we error out before
* that, the index is still consistent. (in buffering build mode,
* any error will abort the index build anyway, so this is not
* needed.)
*/
if (ptr->next && !is_rootsplit && markfollowright)
GistMarkFollowRight(ptr->page);
else
GistClearFollowRight(ptr->page);
/*
* Copy the NSN of the original page to all pages. The
* F_FOLLOW_RIGHT flags ensure that scans will follow the
* rightlinks until the downlinks are inserted.
*/
GistPageSetNSN(ptr->page, oldnsn);
}
START_CRIT_SECTION();
/*
* Must mark buffers dirty before XLogInsert, even though we'll still
* be changing their opaque fields below.
*/
for (ptr = dist; ptr; ptr = ptr->next)
MarkBufferDirty(ptr->buffer);
if (BufferIsValid(leftchildbuf))
MarkBufferDirty(leftchildbuf);
/*
* The first page in the chain was a temporary working copy meant to
* replace the old page. Copy it over the old page.
*/
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;
rdata = formUpdateRdata(rel->rd_node, buffer,
todelete, ntodelete,
NULL, 0,
NULL);
xlinfo = (gistxlogPageUpdate *) rdata->next->data;
recptr = XLogInsert(RM_GIST_ID, XLOG_GIST_PAGE_UPDATE, rdata);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
pfree(xlinfo);
pfree(rdata);
}
else
PageSetLSN(page, XLogRecPtrForTemp);
END_CRIT_SECTION();
}
}
else
{
/* check for split proceeded after look at parent */
pushStackIfSplited(page, stack);
maxoff = PageGetMaxOffsetNumber(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(page, i);
idxtuple = (IndexTuple) PageGetItem(page, iid);
ptr = (GistBDItem *) palloc(sizeof(GistBDItem));
ptr->blkno = ItemPointerGetBlockNumber(&(idxtuple->t_tid));
ptr->parentlsn = PageGetLSN(page);
ptr->next = stack->next;
stack->next = ptr;
if (GistTupleIsInvalid(idxtuple))
stats->needFullVacuum = true;
}
}
UnlockReleaseBuffer(buffer);
ptr = stack->next;
pfree(stack);
stack = ptr;
vacuum_delay_point();
}
PG_RETURN_POINTER(stats);
}
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++;
else
SpGistPageGetOpaque(page)->nRedirection++;
PageSetLSN(page, lsn);
PageSetTLI(page, ThisTimeLineID);
MarkBufferDirty(buffer);
}
UnlockReleaseBuffer(buffer);
}
}
/*
* Update parent downlink. Since parent could be in either of the
* previous two buffers, it's a bit tricky to determine which BKP bit
* applies.
*/
if (xldata->blknoParent == xldata->blkno)
bbi = 0;
else if (xldata->blknoParent == xldata->blknoNew)
bbi = 1;
else
bbi = 2;
if (!(record->xl_info & XLR_SET_BKP_BLOCK(bbi)))
{
buffer = XLogReadBuffer(xldata->node, xldata->blknoParent, false);
if (BufferIsValid(buffer))
{
page = BufferGetPage(buffer);
if (!XLByteLE(lsn, PageGetLSN(page)))
{
SpGistInnerTuple innerTuple;
innerTuple = (SpGistInnerTuple) PageGetItem(page,
PageGetItemId(page, xldata->offnumParent));
spgUpdateNodeLink(innerTuple, xldata->nodeI,
xldata->blknoNew, xldata->offnumNew);
PageSetLSN(page, lsn);
PageSetTLI(page, ThisTimeLineID);
MarkBufferDirty(buffer);
}
UnlockReleaseBuffer(buffer);
}
}
}
}
/*
* 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)
{
/*
* 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);
MarkBufferDirty(buffer);
UnlockReleaseBuffer(buffer);
}
/*
* 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)
elog(ERROR, "incorrect index offsets supplied");
if (totallen > (Size) (pd_special - pd_lower))
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted item lengths: total %u, available space %u",
(unsigned int) totallen, pd_special - pd_lower)));
/* sort itemIdSortData array into decreasing itemoff order */
qsort((char *) itemidbase, nused, sizeof(itemIdSortData),
itemoffcompare);
/* compactify page and install new itemids */
upper = pd_special;
for (i = 0, itemidptr = itemidbase; i < nused; i++, itemidptr++)
{
lp = PageGetItemId(page, itemidptr->offsetindex + 1);
upper -= itemidptr->alignedlen;
memmove((char *) page + upper,
(char *) page + itemidptr->itemoff,
itemidptr->alignedlen);
*lp = itemidptr->olditemid;
lp->lp_off = upper;
}
phdr->pd_lower = SizeOfPageHeaderData + nused * sizeof(ItemIdData);
phdr->pd_upper = upper;
pfree(itemidbase);
}
/*
* 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;
/* don't increment offno, instead decrement maxoffno */
maxoffno = OffsetNumberPrev(maxoffno);
tuples_removed += 1;
}
else
{
offno = OffsetNumberNext(offno);
num_index_tuples += 1;
}
}
/*
* Write page if needed, advance to next page.
*/
blkno = opaque->hasho_nextblkno;
if (page_dirty)
_hash_wrtbuf(rel, buf);
else
_hash_relbuf(rel, buf);
}
/* If we deleted anything, try to compact free space */
if (bucket_dirty)
_hash_squeezebucket(rel, cur_bucket, bucket_blkno);
/* Release bucket lock */
_hash_droplock(rel, bucket_blkno, HASH_EXCLUSIVE);
/* Advance to next bucket */
cur_bucket++;
}
/* Write-lock metapage and check for split since we started */
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_WRITE);
_hash_checkpage(rel, metabuf, LH_META_PAGE);
metap = (HashMetaPage) BufferGetPage(metabuf);
if (cur_maxbucket != metap->hashm_maxbucket)
{
/* There's been a split, so process the additional bucket(s) */
cur_maxbucket = metap->hashm_maxbucket;
memcpy(&local_metapage, metap, sizeof(local_metapage));
_hash_relbuf(rel, metabuf);
goto loop_top;
}
/* Okay, we're really done. Update tuple count in metapage. */
if (orig_maxbucket == metap->hashm_maxbucket &&
orig_ntuples == metap->hashm_ntuples)
{
/*
* No one has split or inserted anything since start of scan, so
* believe our count as gospel.
*/
metap->hashm_ntuples = num_index_tuples;
}
else
{
/*
* Otherwise, our count is untrustworthy since we may have
* double-scanned tuples in split buckets. Proceed by dead-reckoning.
*/
if (metap->hashm_ntuples > tuples_removed)
metap->hashm_ntuples -= tuples_removed;
else
metap->hashm_ntuples = 0;
num_index_tuples = metap->hashm_ntuples;
}
_hash_wrtbuf(rel, metabuf);
/* return statistics */
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
stats->num_index_tuples = num_index_tuples;
stats->tuples_removed += tuples_removed;
/* hashvacuumcleanup will fill in num_pages */
PG_RETURN_POINTER(stats);
}
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(&leafTupleHdr, leafTuple, sizeof(SpGistLeafTupleData));
/*
* In normal operation we would have both current and parent pages locked
* simultaneously; but in WAL replay it should be safe to update the leaf
* page before updating the parent.
*/
if (xldata->newPage)
{
buffer = XLogReadBuffer(xldata->node, xldata->blknoLeaf, true);
SpGistInitBuffer(buffer,
SPGIST_LEAF | (xldata->storesNulls ? SPGIST_NULLS : 0));
action = BLK_NEEDS_REDO;
}
else
action = XLogReadBufferForRedo(lsn, record, 0,
xldata->node, xldata->blknoLeaf,
&buffer);
if (action == BLK_NEEDS_REDO)
{
page = BufferGetPage(buffer);
/* insert new tuple */
if (xldata->offnumLeaf != xldata->offnumHeadLeaf)
{
/* normal cases, tuple was added by SpGistPageAddNewItem */
addOrReplaceTuple(page, (Item) leafTuple, leafTupleHdr.size,
xldata->offnumLeaf);
/* update head tuple's chain link if needed */
if (xldata->offnumHeadLeaf != InvalidOffsetNumber)
{
SpGistLeafTuple head;
head = (SpGistLeafTuple) PageGetItem(page,
PageGetItemId(page, xldata->offnumHeadLeaf));
Assert(head->nextOffset == leafTupleHdr.nextOffset);
head->nextOffset = xldata->offnumLeaf;
}
}
else
{
/* replacing a DEAD tuple */
PageIndexTupleDelete(page, xldata->offnumLeaf);
if (PageAddItem(page, (Item) leafTuple, leafTupleHdr.size,
xldata->offnumLeaf, false, false) != xldata->offnumLeaf)
elog(ERROR, "failed to add item of size %u to SPGiST index page",
leafTupleHdr.size);
}
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
}
if (BufferIsValid(buffer))
UnlockReleaseBuffer(buffer);
/* update parent downlink if necessary */
if (xldata->blknoParent != InvalidBlockNumber)
{
if (XLogReadBufferForRedo(lsn, record, 1,
xldata->node, xldata->blknoParent,
&buffer) == BLK_NEEDS_REDO)
{
SpGistInnerTuple tuple;
page = BufferGetPage(buffer);
tuple = (SpGistInnerTuple) PageGetItem(page,
PageGetItemId(page, xldata->offnumParent));
spgUpdateNodeLink(tuple, xldata->nodeI,
xldata->blknoLeaf, xldata->offnumLeaf);
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
}
if (BufferIsValid(buffer))
UnlockReleaseBuffer(buffer);
}
}
static void
btree_xlog_mark_page_halfdead(uint8 info, XLogReaderState *record)
{
XLogRecPtr lsn = record->EndRecPtr;
xl_btree_mark_page_halfdead *xlrec = (xl_btree_mark_page_halfdead *) XLogRecGetData(record);
Buffer buffer;
Page page;
BTPageOpaque pageop;
IndexTupleData trunctuple;
/*
* In normal operation, we would lock all the pages this WAL record
* touches before changing any of them. In WAL replay, it should be okay
* to lock just one page at a time, since no concurrent index updates can
* be happening, and readers should not care whether they arrive at the
* target page or not (since it's surely empty).
*/
/* parent page */
if (XLogReadBufferForRedo(record, 1, &buffer) == BLK_NEEDS_REDO)
{
OffsetNumber poffset;
ItemId itemid;
IndexTuple itup;
OffsetNumber nextoffset;
BlockNumber rightsib;
page = (Page) BufferGetPage(buffer);
pageop = (BTPageOpaque) PageGetSpecialPointer(page);
poffset = xlrec->poffset;
nextoffset = OffsetNumberNext(poffset);
itemid = PageGetItemId(page, nextoffset);
itup = (IndexTuple) PageGetItem(page, itemid);
rightsib = ItemPointerGetBlockNumber(&itup->t_tid);
itemid = PageGetItemId(page, poffset);
itup = (IndexTuple) PageGetItem(page, itemid);
ItemPointerSet(&(itup->t_tid), rightsib, P_HIKEY);
nextoffset = OffsetNumberNext(poffset);
PageIndexTupleDelete(page, nextoffset);
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
}
if (BufferIsValid(buffer))
UnlockReleaseBuffer(buffer);
/* Rewrite the leaf page as a halfdead page */
buffer = XLogInitBufferForRedo(record, 0);
page = (Page) BufferGetPage(buffer);
_bt_pageinit(page, BufferGetPageSize(buffer));
pageop = (BTPageOpaque) PageGetSpecialPointer(page);
pageop->btpo_prev = xlrec->leftblk;
pageop->btpo_next = xlrec->rightblk;
pageop->btpo.level = 0;
pageop->btpo_flags = BTP_HALF_DEAD | BTP_LEAF;
pageop->btpo_cycleid = 0;
/*
* Construct a dummy hikey item that points to the next parent to be
* deleted (if any).
*/
MemSet(&trunctuple, 0, sizeof(IndexTupleData));
trunctuple.t_info = sizeof(IndexTupleData);
if (xlrec->topparent != InvalidBlockNumber)
ItemPointerSet(&trunctuple.t_tid, xlrec->topparent, P_HIKEY);
else
ItemPointerSetInvalid(&trunctuple.t_tid);
if (PageAddItem(page, (Item) &trunctuple, sizeof(IndexTupleData), P_HIKEY,
false, false) == InvalidOffsetNumber)
elog(ERROR, "could not add dummy high key to half-dead page");
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
UnlockReleaseBuffer(buffer);
}
static ArrayTuple
gistVacuumUpdate(GistVacuum *gv, BlockNumber blkno, bool needunion)
{
ArrayTuple res = {NULL, 0, false};
Buffer buffer;
Page page,
tempPage = NULL;
OffsetNumber i,
maxoff;
ItemId iid;
int lenaddon = 4,
curlenaddon = 0,
nOffToDelete = 0,
nBlkToDelete = 0;
IndexTuple idxtuple,
*addon = NULL;
bool needwrite = false;
OffsetNumber offToDelete[MaxOffsetNumber];
BlockNumber blkToDelete[MaxOffsetNumber];
ItemPointerData *completed = NULL;
int ncompleted = 0,
lencompleted = 16;
vacuum_delay_point();
buffer = ReadBufferWithStrategy(gv->index, blkno, gv->strategy);
LockBuffer(buffer, GIST_EXCLUSIVE);
gistcheckpage(gv->index, buffer);
page = (Page) BufferGetPage(buffer);
maxoff = PageGetMaxOffsetNumber(page);
if (GistPageIsLeaf(page))
{
if (GistTuplesDeleted(page))
needunion = needwrite = true;
}
else
{
completed = (ItemPointerData *) palloc(sizeof(ItemPointerData) * lencompleted);
addon = (IndexTuple *) palloc(sizeof(IndexTuple) * lenaddon);
/* get copy of page to work */
tempPage = GistPageGetCopyPage(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
ArrayTuple chldtuple;
bool needchildunion;
iid = PageGetItemId(tempPage, i);
idxtuple = (IndexTuple) PageGetItem(tempPage, iid);
needchildunion = (GistTupleIsInvalid(idxtuple)) ? true : false;
if (needchildunion)
elog(DEBUG2, "gistVacuumUpdate: need union for block %u",
ItemPointerGetBlockNumber(&(idxtuple->t_tid)));
chldtuple = gistVacuumUpdate(gv, ItemPointerGetBlockNumber(&(idxtuple->t_tid)),
needchildunion);
if (chldtuple.ituplen || chldtuple.emptypage)
{
/* update tuple or/and inserts new */
if (chldtuple.emptypage)
blkToDelete[nBlkToDelete++] = ItemPointerGetBlockNumber(&(idxtuple->t_tid));
offToDelete[nOffToDelete++] = i;
PageIndexTupleDelete(tempPage, i);
i--;
maxoff--;
needwrite = needunion = true;
if (chldtuple.ituplen)
{
Assert(chldtuple.emptypage == false);
while (curlenaddon + chldtuple.ituplen >= lenaddon)
{
lenaddon *= 2;
addon = (IndexTuple *) repalloc(addon, sizeof(IndexTuple) * lenaddon);
}
memcpy(addon + curlenaddon, chldtuple.itup, chldtuple.ituplen * sizeof(IndexTuple));
curlenaddon += chldtuple.ituplen;
if (chldtuple.ituplen > 1)
{
/*
* child was split, so we need mark completion
* insert(split)
*/
int j;
while (ncompleted + chldtuple.ituplen > lencompleted)
{
lencompleted *= 2;
completed = (ItemPointerData *) repalloc(completed, sizeof(ItemPointerData) * lencompleted);
}
for (j = 0; j < chldtuple.ituplen; j++)
{
ItemPointerCopy(&(chldtuple.itup[j]->t_tid), completed + ncompleted);
ncompleted++;
}
}
pfree(chldtuple.itup);
}
}
}
Assert(maxoff == PageGetMaxOffsetNumber(tempPage));
if (curlenaddon)
{
/* insert updated tuples */
if (gistnospace(tempPage, addon, curlenaddon, InvalidOffsetNumber, 0))
{
/* there is no space on page to insert tuples */
res = vacuumSplitPage(gv, tempPage, buffer, addon, curlenaddon);
tempPage = NULL; /* vacuumSplitPage() free tempPage */
needwrite = needunion = false; /* gistSplit already forms
* unions and writes pages */
}
else
/* enough free space */
gistfillbuffer(gv->index, tempPage, addon, curlenaddon, InvalidOffsetNumber);
}
}
/*
* If page is empty, we should remove pointer to it before deleting page
* (except root)
*/
if (blkno != GIST_ROOT_BLKNO && (PageIsEmpty(page) || (tempPage && PageIsEmpty(tempPage))))
{
/*
* New version of page is empty, so leave it unchanged, upper call
* will mark our page as deleted. In case of page split we never will
* be here...
*
* If page was empty it can't become non-empty during processing
*/
res.emptypage = true;
UnlockReleaseBuffer(buffer);
}
else
{
/* write page and remove its childs if it need */
START_CRIT_SECTION();
if (tempPage && needwrite)
{
PageRestoreTempPage(tempPage, page);
tempPage = NULL;
}
/* Empty index */
if (PageIsEmpty(page) && blkno == GIST_ROOT_BLKNO)
{
needwrite = true;
GistPageSetLeaf(page);
}
if (needwrite)
{
MarkBufferDirty(buffer);
GistClearTuplesDeleted(page);
if (!gv->index->rd_istemp)
{
XLogRecData *rdata;
XLogRecPtr recptr;
char *xlinfo;
rdata = formUpdateRdata(gv->index->rd_node, buffer,
offToDelete, nOffToDelete,
addon, curlenaddon, NULL);
xlinfo = rdata->next->data;
recptr = XLogInsert(RM_GIST_ID, XLOG_GIST_PAGE_UPDATE, rdata);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
pfree(xlinfo);
pfree(rdata);
}
else
PageSetLSN(page, XLogRecPtrForTemp);
}
END_CRIT_SECTION();
if (needunion && !PageIsEmpty(page))
{
res.itup = (IndexTuple *) palloc(sizeof(IndexTuple));
res.ituplen = 1;
res.itup[0] = PageMakeUnionKey(gv, buffer);
}
UnlockReleaseBuffer(buffer);
/* delete empty children, now we havn't any links to pointed subtrees */
for (i = 0; i < nBlkToDelete; i++)
gistDeleteSubtree(gv, blkToDelete[i]);
if (ncompleted && !gv->index->rd_istemp)
gistxlogInsertCompletion(gv->index->rd_node, completed, ncompleted);
}
for (i = 0; i < curlenaddon; i++)
pfree(addon[i]);
if (addon)
pfree(addon);
if (completed)
pfree(completed);
if (tempPage)
pfree(tempPage);
return res;
}
/*
* 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] = GinGetDownlink(itup);
(*nroot)++;
}
else if (GinGetNPosting(itup) > 0)
{
int nitems;
ItemPointer items_orig;
bool free_items_orig;
ItemPointer items;
/* Get list of item pointers from the tuple. */
if (GinItupIsCompressed(itup))
{
items_orig = ginPostingListDecode((GinPostingList *) GinGetPosting(itup), &nitems);
free_items_orig = true;
}
else
{
items_orig = (ItemPointer) GinGetPosting(itup);
nitems = GinGetNPosting(itup);
free_items_orig = false;
}
/* Remove any items from the list that need to be vacuumed. */
items = ginVacuumItemPointers(gvs, items_orig, nitems, &nitems);
if (free_items_orig)
pfree(items_orig);
/* If any item pointers were removed, recreate the tuple. */
if (items)
{
OffsetNumber attnum;
Datum key;
GinNullCategory category;
GinPostingList *plist;
int plistsize;
if (nitems > 0)
{
plist = ginCompressPostingList(items, nitems, GinMaxItemSize, NULL);
plistsize = SizeOfGinPostingList(plist);
}
else
{
plist = NULL;
plistsize = 0;
}
/*
* if we already created a temporary page, make changes in
* place
*/
if (tmppage == origpage)
{
/*
* On first difference, create a temporary copy of the
* page and copy the tuple's posting list to it.
*/
tmppage = PageGetTempPageCopy(origpage);
/* set itup pointer to new page */
itup = (IndexTuple) PageGetItem(tmppage, PageGetItemId(tmppage, i));
}
attnum = gintuple_get_attrnum(&gvs->ginstate, itup);
key = gintuple_get_key(&gvs->ginstate, itup, &category);
itup = GinFormTuple(&gvs->ginstate, attnum, key, category,
(char *) plist, plistsize,
nitems, true);
if (plist)
pfree(plist);
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);
pfree(items);
}
}
}
return (tmppage == origpage) ? NULL : tmppage;
}
/*
* 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;
itemIdSortData itemidbase[MaxIndexTuplesPerPage];
ItemIdData newitemids[MaxIndexTuplesPerPage];
itemIdSort itemidptr;
ItemId lp;
int nline,
nused;
Size totallen;
Size size;
unsigned offset;
int nextitm;
OffsetNumber offnum;
Assert(nitems <= MaxIndexTuplesPerPage);
/*
* 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);
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, length = %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->alignedlen = MAXALIGN(size);
totallen += itemidptr->alignedlen;
newitemids[nused] = *lp;
itemidptr++;
nused++;
}
}
/* this will catch invalid or out-of-order itemnos[] */
if (nextitm != nitems)
elog(ERROR, "incorrect index offsets supplied");
if (totallen > (Size) (pd_special - pd_lower))
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("corrupted item lengths: total %u, available space %u",
(unsigned int) totallen, pd_special - pd_lower)));
/*
* Looks good. Overwrite the line pointers with the copy, from which we've
* removed all the unused items.
*/
memcpy(phdr->pd_linp, newitemids, nused * sizeof(ItemIdData));
phdr->pd_lower = SizeOfPageHeaderData + nused * sizeof(ItemIdData);
/* and compactify the tuple data */
compactify_tuples(itemidbase, nused, page);
}
/*
* _hash_squeezebucket(rel, bucket)
*
* Try to squeeze the tuples onto pages occurring earlier in the
* bucket chain in an attempt to free overflow pages. When we start
* the "squeezing", the page from which we start taking tuples (the
* "read" page) is the last bucket in the bucket chain and the page
* onto which we start squeezing tuples (the "write" page) is the
* first page in the bucket chain. The read page works backward and
* the write page works forward; the procedure terminates when the
* read page and write page are the same page.
*
* At completion of this procedure, it is guaranteed that all pages in
* the bucket are nonempty, unless the bucket is totally empty (in
* which case all overflow pages will be freed). The original implementation
* required that to be true on entry as well, but it's a lot easier for
* callers to leave empty overflow pages and let this guy clean it up.
*
* Caller must hold exclusive lock on the target bucket. This allows
* us to safely lock multiple pages in the bucket.
*/
void
_hash_squeezebucket(Relation rel,
Bucket bucket,
BlockNumber bucket_blkno)
{
Buffer wbuf;
Buffer rbuf = 0;
BlockNumber wblkno;
BlockNumber rblkno;
Page wpage;
Page rpage;
HashPageOpaque wopaque;
HashPageOpaque ropaque;
OffsetNumber woffnum;
OffsetNumber roffnum;
IndexTuple itup;
Size itemsz;
/*
* start squeezing into the base bucket page.
*/
wblkno = bucket_blkno;
wbuf = _hash_getbuf(rel, wblkno, HASH_WRITE);
_hash_checkpage(rel, wbuf, LH_BUCKET_PAGE);
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
/*
* if there aren't any overflow pages, there's nothing to squeeze.
*/
if (!BlockNumberIsValid(wopaque->hasho_nextblkno))
{
_hash_relbuf(rel, wbuf);
return;
}
/*
* find the last page in the bucket chain by starting at the base bucket
* page and working forward.
*/
ropaque = wopaque;
do
{
rblkno = ropaque->hasho_nextblkno;
if (ropaque != wopaque)
_hash_relbuf(rel, rbuf);
rbuf = _hash_getbuf(rel, rblkno, HASH_WRITE);
_hash_checkpage(rel, rbuf, LH_OVERFLOW_PAGE);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
} while (BlockNumberIsValid(ropaque->hasho_nextblkno));
/*
* squeeze the tuples.
*/
roffnum = FirstOffsetNumber;
for (;;)
{
/* this test is needed in case page is empty on entry */
if (roffnum <= PageGetMaxOffsetNumber(rpage))
{
itup = (IndexTuple) PageGetItem(rpage,
PageGetItemId(rpage, roffnum));
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
/*
* Walk up the bucket chain, looking for a page big enough for
* this item. Exit if we reach the read page.
*/
while (PageGetFreeSpace(wpage) < itemsz)
{
Assert(!PageIsEmpty(wpage));
wblkno = wopaque->hasho_nextblkno;
Assert(BlockNumberIsValid(wblkno));
_hash_wrtbuf(rel, wbuf);
if (rblkno == wblkno)
{
/* wbuf is already released */
_hash_wrtbuf(rel, rbuf);
return;
}
wbuf = _hash_getbuf(rel, wblkno, HASH_WRITE);
_hash_checkpage(rel, wbuf, LH_OVERFLOW_PAGE);
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
Assert(wopaque->hasho_bucket == bucket);
}
/*
* we have found room so insert on the "write" page.
*/
woffnum = OffsetNumberNext(PageGetMaxOffsetNumber(wpage));
if (PageAddItem(wpage, (Item) itup, itemsz, woffnum, LP_USED)
== InvalidOffsetNumber)
elog(ERROR, "failed to add index item to \"%s\"",
RelationGetRelationName(rel));
/*
* delete the tuple from the "read" page. PageIndexTupleDelete
* repacks the ItemId array, so 'roffnum' will be "advanced" to
* the "next" ItemId.
*/
PageIndexTupleDelete(rpage, roffnum);
}
/*
* if the "read" page is now empty because of the deletion (or because
* it was empty when we got to it), free it.
*
* Tricky point here: if our read and write pages are adjacent in the
* bucket chain, our write lock on wbuf will conflict with
* _hash_freeovflpage's attempt to update the sibling links of the
* removed page. However, in that case we are done anyway, so we can
* simply drop the write lock before calling _hash_freeovflpage.
*/
if (PageIsEmpty(rpage))
{
rblkno = ropaque->hasho_prevblkno;
Assert(BlockNumberIsValid(rblkno));
/* are we freeing the page adjacent to wbuf? */
if (rblkno == wblkno)
{
/* yes, so release wbuf lock first */
_hash_wrtbuf(rel, wbuf);
/* free this overflow page (releases rbuf) */
_hash_freeovflpage(rel, rbuf);
/* done */
return;
}
/* free this overflow page, then get the previous one */
_hash_freeovflpage(rel, rbuf);
rbuf = _hash_getbuf(rel, rblkno, HASH_WRITE);
_hash_checkpage(rel, rbuf, LH_OVERFLOW_PAGE);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
roffnum = FirstOffsetNumber;
}
}
/* NOTREACHED */
}
static void
spgRedoAddLeaf(XLogRecPtr lsn, XLogRecord *record)
{
char *ptr = XLogRecGetData(record);
spgxlogAddLeaf *xldata = (spgxlogAddLeaf *) ptr;
SpGistLeafTuple leafTuple;
Buffer buffer;
Page page;
/* we assume this is adequately aligned */
ptr += sizeof(spgxlogAddLeaf);
leafTuple = (SpGistLeafTuple) ptr;
if (!(record->xl_info & XLR_BKP_BLOCK_1))
{
buffer = XLogReadBuffer(xldata->node, xldata->blknoLeaf,
xldata->newPage);
if (BufferIsValid(buffer))
{
page = BufferGetPage(buffer);
if (xldata->newPage)
SpGistInitBuffer(buffer, SPGIST_LEAF);
if (!XLByteLE(lsn, PageGetLSN(page)))
{
/* insert new tuple */
if (xldata->offnumLeaf != xldata->offnumHeadLeaf)
{
/* normal cases, tuple was added by SpGistPageAddNewItem */
addOrReplaceTuple(page, (Item) leafTuple, leafTuple->size,
xldata->offnumLeaf);
/* update head tuple's chain link if needed */
if (xldata->offnumHeadLeaf != InvalidOffsetNumber)
{
SpGistLeafTuple head;
head = (SpGistLeafTuple) PageGetItem(page,
PageGetItemId(page, xldata->offnumHeadLeaf));
Assert(head->nextOffset == leafTuple->nextOffset);
head->nextOffset = xldata->offnumLeaf;
}
}
else
{
/* replacing a DEAD tuple */
PageIndexTupleDelete(page, xldata->offnumLeaf);
if (PageAddItem(page,
(Item) leafTuple, leafTuple->size,
xldata->offnumLeaf, false, false) != xldata->offnumLeaf)
elog(ERROR, "failed to add item of size %u to SPGiST index page",
leafTuple->size);
}
PageSetLSN(page, lsn);
PageSetTLI(page, ThisTimeLineID);
MarkBufferDirty(buffer);
}
UnlockReleaseBuffer(buffer);
}
}
/* update parent downlink if necessary */
if (xldata->blknoParent != InvalidBlockNumber &&
!(record->xl_info & XLR_BKP_BLOCK_2))
{
buffer = XLogReadBuffer(xldata->node, xldata->blknoParent, false);
if (BufferIsValid(buffer))
{
page = BufferGetPage(buffer);
if (!XLByteLE(lsn, PageGetLSN(page)))
{
SpGistInnerTuple tuple;
tuple = (SpGistInnerTuple) PageGetItem(page,
PageGetItemId(page, xldata->offnumParent));
spgUpdateNodeLink(tuple, xldata->nodeI,
xldata->blknoLeaf, xldata->offnumLeaf);
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;
char *innerTuple;
SpGistInnerTupleData innerTupleHdr;
SpGistState state;
Buffer buffer;
Page page;
int bbi;
XLogRedoAction action;
ptr += sizeof(spgxlogAddNode);
innerTuple = ptr;
/* the tuple is unaligned, so make a copy to access its header */
memcpy(&innerTupleHdr, innerTuple, sizeof(SpGistInnerTupleData));
fillFakeState(&state, xldata->stateSrc);
if (xldata->blknoNew == InvalidBlockNumber)
{
/* update in place */
Assert(xldata->blknoParent == InvalidBlockNumber);
if (XLogReadBufferForRedo(lsn, record, 0, xldata->node, xldata->blkno,
&buffer) == BLK_NEEDS_REDO)
{
page = BufferGetPage(buffer);
PageIndexTupleDelete(page, xldata->offnum);
if (PageAddItem(page, (Item) innerTuple, innerTupleHdr.size,
xldata->offnum, false, false) != xldata->offnum)
elog(ERROR, "failed to add item of size %u to SPGiST index page",
innerTupleHdr.size);
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
}
if (BufferIsValid(buffer))
UnlockReleaseBuffer(buffer);
}
else
{
/*
* In normal operation we would have all three pages (source, dest,
* and parent) 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.
*
* The logic here depends on the assumption that blkno != blknoNew,
* else we can't tell which BKP bit goes with which page, and the LSN
* checks could go wrong too.
*/
Assert(xldata->blkno != xldata->blknoNew);
/* Install new tuple first so redirect is valid */
if (xldata->newPage)
{
buffer = XLogReadBuffer(xldata->node, xldata->blknoNew, true);
/* AddNode is not used for nulls pages */
SpGistInitBuffer(buffer, 0);
action = BLK_NEEDS_REDO;
}
else
action = XLogReadBufferForRedo(lsn, record, 1,
xldata->node, xldata->blknoNew,
&buffer);
if (action == BLK_NEEDS_REDO)
{
page = BufferGetPage(buffer);
addOrReplaceTuple(page, (Item) innerTuple,
innerTupleHdr.size, xldata->offnumNew);
/*
* If parent is in this same page, don't advance LSN; doing so
* would fool us into not applying the parent downlink update
* below. We'll update the LSN when we fix the parent downlink.
*/
if (xldata->blknoParent != xldata->blknoNew)
{
PageSetLSN(page, lsn);
}
MarkBufferDirty(buffer);
}
if (BufferIsValid(buffer))
UnlockReleaseBuffer(buffer);
/* Delete old tuple, replacing it with redirect or placeholder tuple */
if (XLogReadBufferForRedo(lsn, record, 0, xldata->node, xldata->blkno,
&buffer) == BLK_NEEDS_REDO)
{
SpGistDeadTuple dt;
page = BufferGetPage(buffer);
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++;
else
SpGistPageGetOpaque(page)->nRedirection++;
/*
* If parent is in this same page, don't advance LSN; doing so
* would fool us into not applying the parent downlink update
* below. We'll update the LSN when we fix the parent downlink.
*/
if (xldata->blknoParent != xldata->blkno)
{
PageSetLSN(page, lsn);
}
MarkBufferDirty(buffer);
}
if (BufferIsValid(buffer))
UnlockReleaseBuffer(buffer);
/*
* Update parent downlink. Since parent could be in either of the
* previous two buffers, it's a bit tricky to determine which BKP bit
* applies.
*/
if (xldata->blknoParent == xldata->blkno)
bbi = 0;
else if (xldata->blknoParent == xldata->blknoNew)
bbi = 1;
else
bbi = 2;
if (record->xl_info & XLR_BKP_BLOCK(bbi))
{
if (bbi == 2) /* else we already did it */
(void) RestoreBackupBlock(lsn, record, bbi, false, false);
action = BLK_RESTORED;
buffer = InvalidBuffer;
}
else
{
action = XLogReadBufferForRedo(lsn, record, bbi, xldata->node,
xldata->blknoParent, &buffer);
Assert(action != BLK_RESTORED);
}
if (action == BLK_NEEDS_REDO)
{
SpGistInnerTuple innerTuple;
page = BufferGetPage(buffer);
innerTuple = (SpGistInnerTuple) PageGetItem(page,
PageGetItemId(page, xldata->offnumParent));
spgUpdateNodeLink(innerTuple, xldata->nodeI,
xldata->blknoNew, xldata->offnumNew);
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
}
if (BufferIsValid(buffer))
UnlockReleaseBuffer(buffer);
}
}
/*
* 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] = GinGetDownlink(itup);
(*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)
{
OffsetNumber attnum;
Datum key;
GinNullCategory category;
/*
* 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 = PageGetTempPageCopy(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));
}
attnum = gintuple_get_attrnum(&gvs->ginstate, itup);
key = gintuple_get_key(&gvs->ginstate, itup, &category);
itup = GinFormTuple(&gvs->ginstate, attnum, key, category,
GinGetPosting(itup), newN, true);
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;
}
/*
* redo any page update (except page split)
*/
static void
gistRedoPageUpdateRecord(XLogReaderState *record)
{
XLogRecPtr lsn = record->EndRecPtr;
gistxlogPageUpdate *xldata = (gistxlogPageUpdate *) XLogRecGetData(record);
Buffer buffer;
Page page;
if (XLogReadBufferForRedo(record, 0, &buffer) == BLK_NEEDS_REDO)
{
char *begin;
char *data;
Size datalen;
int ninserted = 0;
data = begin = XLogRecGetBlockData(record, 0, &datalen);
page = (Page) BufferGetPage(buffer);
/* 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 < datalen)
{
OffsetNumber off = (PageIsEmpty(page)) ? FirstOffsetNumber :
OffsetNumberNext(PageGetMaxOffsetNumber(page));
while (data - begin < datalen)
{
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++;
ninserted++;
}
}
Assert(ninserted == xldata->ntoinsert);
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
}
/*
* Fix follow-right data on left child page
*
* This must be done while still holding the lock on the target page. Note
* that even if the target page no longer exists, we still attempt to
* replay the change on the child page.
*/
if (XLogRecHasBlockRef(record, 1))
gistRedoClearFollowRight(record, 1);
if (BufferIsValid(buffer))
UnlockReleaseBuffer(buffer);
}
/*
* _bt_pagedel() -- Delete a page from the b-tree, if legal to do so.
*
* This action unlinks the page from the b-tree structure, removing all
* pointers leading to it --- but not touching its own left and right links.
* The page cannot be physically reclaimed right away, since other processes
* may currently be trying to follow links leading to the page; they have to
* be allowed to use its right-link to recover. See nbtree/README.
*
* On entry, the target buffer must be pinned and locked (either read or write
* lock is OK). This lock and pin will be dropped before exiting.
*
* The "stack" argument can be a search stack leading (approximately) to the
* target page, or NULL --- outside callers typically pass NULL since they
* have not done such a search, but internal recursion cases pass the stack
* to avoid duplicated search effort.
*
* Returns the number of pages successfully deleted (zero if page cannot
* be deleted now; could be more than one if parent pages were deleted too).
*
* NOTE: this leaks memory. Rather than trying to clean up everything
* carefully, it's better to run it in a temp context that can be reset
* frequently.
*/
int
_bt_pagedel(Relation rel, Buffer buf, BTStack stack, bool vacuum_full)
{
int result;
BlockNumber target,
leftsib,
rightsib,
parent;
OffsetNumber poffset,
maxoff;
uint32 targetlevel,
ilevel;
ItemId itemid;
IndexTuple targetkey,
itup;
ScanKey itup_scankey;
Buffer lbuf,
rbuf,
pbuf;
bool parent_half_dead;
bool parent_one_child;
bool rightsib_empty;
Buffer metabuf = InvalidBuffer;
Page metapg = NULL;
BTMetaPageData *metad = NULL;
Page page;
BTPageOpaque opaque;
MIRROREDLOCK_BUFMGR_MUST_ALREADY_BE_HELD;
// Fetch gp_persistent_relation_node information that will be added to XLOG record.
RelationFetchGpRelationNodeForXLog(rel);
/*
* We can never delete rightmost pages nor root pages. While at it, check
* that page is not already deleted and is empty.
*/
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (P_RIGHTMOST(opaque) || P_ISROOT(opaque) || P_ISDELETED(opaque) ||
P_FIRSTDATAKEY(opaque) <= PageGetMaxOffsetNumber(page))
{
/* Should never fail to delete a half-dead page */
Assert(!P_ISHALFDEAD(opaque));
_bt_relbuf(rel, buf);
return 0;
}
/*
* Save info about page, including a copy of its high key (it must have
* one, being non-rightmost).
*/
target = BufferGetBlockNumber(buf);
targetlevel = opaque->btpo.level;
leftsib = opaque->btpo_prev;
itemid = PageGetItemId(page, P_HIKEY);
targetkey = CopyIndexTuple((IndexTuple) PageGetItem(page, itemid));
/*
* To avoid deadlocks, we'd better drop the target page lock before going
* further.
*/
_bt_relbuf(rel, buf);
/*
* We need an approximate pointer to the page's parent page. We use the
* standard search mechanism to search for the page's high key; this will
* give us a link to either the current parent or someplace to its left
* (if there are multiple equal high keys). In recursion cases, the
* caller already generated a search stack and we can just re-use that
* work.
*/
if (stack == NULL)
{
if (!InRecovery)
{
/* we need an insertion scan key to do our search, so build one */
itup_scankey = _bt_mkscankey(rel, targetkey);
/* find the leftmost leaf page containing this key */
stack = _bt_search(rel, rel->rd_rel->relnatts, itup_scankey, false,
&lbuf, BT_READ);
/* don't need a pin on that either */
_bt_relbuf(rel, lbuf);
/*
* If we are trying to delete an interior page, _bt_search did
* more than we needed. Locate the stack item pointing to our
* parent level.
*/
ilevel = 0;
for (;;)
{
if (stack == NULL)
elog(ERROR, "not enough stack items");
if (ilevel == targetlevel)
break;
stack = stack->bts_parent;
ilevel++;
}
}
else
{
/*
* During WAL recovery, we can't use _bt_search (for one reason,
* it might invoke user-defined comparison functions that expect
* facilities not available in recovery mode). Instead, just set
* up a dummy stack pointing to the left end of the parent tree
* level, from which _bt_getstackbuf will walk right to the parent
* page. Painful, but we don't care too much about performance in
* this scenario.
*/
pbuf = _bt_get_endpoint(rel, targetlevel + 1, false);
stack = (BTStack) palloc(sizeof(BTStackData));
stack->bts_blkno = BufferGetBlockNumber(pbuf);
stack->bts_offset = InvalidOffsetNumber;
/* bts_btentry will be initialized below */
stack->bts_parent = NULL;
_bt_relbuf(rel, pbuf);
}
}
/*
* We cannot delete a page that is the rightmost child of its immediate
* parent, unless it is the only child --- in which case the parent has to
* be deleted too, and the same condition applies recursively to it. We
* have to check this condition all the way up before trying to delete. We
* don't need to re-test when deleting a non-leaf page, though.
*/
if (targetlevel == 0 &&
!_bt_parent_deletion_safe(rel, target, stack))
return 0;
/*
* We have to lock the pages we need to modify in the standard order:
* moving right, then up. Else we will deadlock against other writers.
*
* So, we need to find and write-lock the current left sibling of the
* target page. The sibling that was current a moment ago could have
* split, so we may have to move right. This search could fail if either
* the sibling or the target page was deleted by someone else meanwhile;
* if so, give up. (Right now, that should never happen, since page
* deletion is only done in VACUUM and there shouldn't be multiple VACUUMs
* concurrently on the same table.)
*/
if (leftsib != P_NONE)
{
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
while (P_ISDELETED(opaque) || opaque->btpo_next != target)
{
/* step right one page */
leftsib = opaque->btpo_next;
_bt_relbuf(rel, lbuf);
if (leftsib == P_NONE)
{
elog(LOG, "no left sibling (concurrent deletion?) in \"%s\"",
RelationGetRelationName(rel));
return 0;
}
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
}
}
else
lbuf = InvalidBuffer;
/*
* Next write-lock the target page itself. It should be okay to take just
* a write lock not a superexclusive lock, since no scans would stop on an
* empty page.
*/
buf = _bt_getbuf(rel, target, BT_WRITE);
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
/*
* Check page is still empty etc, else abandon deletion. The empty check
* is necessary since someone else might have inserted into it while we
* didn't have it locked; the others are just for paranoia's sake.
*/
if (P_RIGHTMOST(opaque) || P_ISROOT(opaque) || P_ISDELETED(opaque) ||
P_FIRSTDATAKEY(opaque) <= PageGetMaxOffsetNumber(page))
{
_bt_relbuf(rel, buf);
if (BufferIsValid(lbuf))
_bt_relbuf(rel, lbuf);
return 0;
}
if (opaque->btpo_prev != leftsib)
elog(ERROR, "left link changed unexpectedly in block %u of index \"%s\"",
target, RelationGetRelationName(rel));
/*
* And next write-lock the (current) right sibling.
*/
rightsib = opaque->btpo_next;
rbuf = _bt_getbuf(rel, rightsib, BT_WRITE);
page = BufferGetPage(rbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (opaque->btpo_prev != target)
elog(ERROR, "right sibling's left-link doesn't match: "
"block %u links to %u instead of expected %u in index \"%s\"",
rightsib, opaque->btpo_prev, target,
RelationGetRelationName(rel));
/*
* Next find and write-lock the current parent of the target page. This is
* essentially the same as the corresponding step of splitting.
*/
ItemPointerSet(&(stack->bts_btentry.t_tid), target, P_HIKEY);
pbuf = _bt_getstackbuf(rel, stack, BT_WRITE);
if (pbuf == InvalidBuffer)
elog(ERROR, "failed to re-find parent key in index \"%s\" for deletion target page %u",
RelationGetRelationName(rel), target);
parent = stack->bts_blkno;
poffset = stack->bts_offset;
/*
* If the target is the rightmost child of its parent, then we can't
* delete, unless it's also the only child --- in which case the parent
* changes to half-dead status. The "can't delete" case should have been
* detected by _bt_parent_deletion_safe, so complain if we see it now.
*/
page = BufferGetPage(pbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
maxoff = PageGetMaxOffsetNumber(page);
parent_half_dead = false;
parent_one_child = false;
if (poffset >= maxoff)
{
if (poffset == P_FIRSTDATAKEY(opaque))
parent_half_dead = true;
else
elog(ERROR, "failed to delete rightmost child %u of block %u in index \"%s\"",
target, parent, RelationGetRelationName(rel));
}
else
{
/* Will there be exactly one child left in this parent? */
if (OffsetNumberNext(P_FIRSTDATAKEY(opaque)) == maxoff)
parent_one_child = true;
}
/*
* If we are deleting the next-to-last page on the target's level, then
* the rightsib is a candidate to become the new fast root. (In theory, it
* might be possible to push the fast root even further down, but the odds
* of doing so are slim, and the locking considerations daunting.)
*
* We don't support handling this in the case where the parent is becoming
* half-dead, even though it theoretically could occur.
*
* We can safely acquire a lock on the metapage here --- see comments for
* _bt_newroot().
*/
if (leftsib == P_NONE && !parent_half_dead)
{
page = BufferGetPage(rbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo.level == targetlevel);
if (P_RIGHTMOST(opaque))
{
/* rightsib will be the only one left on the level */
metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_WRITE);
metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg);
/*
* The expected case here is btm_fastlevel == targetlevel+1; if
* the fastlevel is <= targetlevel, something is wrong, and we
* choose to overwrite it to fix it.
*/
if (metad->btm_fastlevel > targetlevel + 1)
{
/* no update wanted */
_bt_relbuf(rel, metabuf);
metabuf = InvalidBuffer;
}
}
}
/*
* Check that the parent-page index items we're about to delete/overwrite
* contain what we expect. This can fail if the index has become
* corrupt for some reason. We want to throw any error before entering
* the critical section --- otherwise it'd be a PANIC.
*
* The test on the target item is just an Assert because _bt_getstackbuf
* should have guaranteed it has the expected contents. The test on the
* next-child downlink is known to sometimes fail in the field, though.
*/
page = BufferGetPage(pbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
#ifdef USE_ASSERT_CHECKING
itemid = PageGetItemId(page, poffset);
itup = (IndexTuple) PageGetItem(page, itemid);
Assert(ItemPointerGetBlockNumber(&(itup->t_tid)) == target);
#endif
if (!parent_half_dead)
{
OffsetNumber nextoffset;
nextoffset = OffsetNumberNext(poffset);
itemid = PageGetItemId(page, nextoffset);
itup = (IndexTuple) PageGetItem(page, itemid);
if (ItemPointerGetBlockNumber(&(itup->t_tid)) != rightsib)
elog(ERROR, "right sibling %u of block %u is not next child %u of block %u in index \"%s\"",
rightsib, target, ItemPointerGetBlockNumber(&(itup->t_tid)),
parent, RelationGetRelationName(rel));
}
/*
* Here we begin doing the deletion.
*/
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
/*
* Update parent. The normal case is a tad tricky because we want to
* delete the target's downlink and the *following* key. Easiest way is
* to copy the right sibling's downlink over the target downlink, and then
* delete the following item.
*/
if (parent_half_dead)
{
PageIndexTupleDelete(page, poffset);
opaque->btpo_flags |= BTP_HALF_DEAD;
}
else
{
OffsetNumber nextoffset;
itemid = PageGetItemId(page, poffset);
itup = (IndexTuple) PageGetItem(page, itemid);
ItemPointerSet(&(itup->t_tid), rightsib, P_HIKEY);
nextoffset = OffsetNumberNext(poffset);
PageIndexTupleDelete(page, nextoffset);
}
/*
* Update siblings' side-links. Note the target page's side-links will
* continue to point to the siblings. Asserts here are just rechecking
* things we already verified above.
*/
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo_next == target);
opaque->btpo_next = rightsib;
}
page = BufferGetPage(rbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo_prev == target);
opaque->btpo_prev = leftsib;
rightsib_empty = (P_FIRSTDATAKEY(opaque) > PageGetMaxOffsetNumber(page));
/*
* Mark the page itself deleted. It can be recycled when all current
* transactions are gone; or immediately if we're doing VACUUM FULL.
*/
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
opaque->btpo_flags &= ~BTP_HALF_DEAD;
opaque->btpo_flags |= BTP_DELETED;
opaque->btpo.xact =
vacuum_full ? FrozenTransactionId : ReadNewTransactionId();
/* And update the metapage, if needed */
if (BufferIsValid(metabuf))
{
metad->btm_fastroot = rightsib;
metad->btm_fastlevel = targetlevel;
MarkBufferDirty(metabuf);
}
/* Must mark buffers dirty before XLogInsert */
MarkBufferDirty(pbuf);
MarkBufferDirty(rbuf);
MarkBufferDirty(buf);
if (BufferIsValid(lbuf))
MarkBufferDirty(lbuf);
/* XLOG stuff */
if (!rel->rd_istemp)
{
xl_btree_delete_page xlrec;
xl_btree_metadata xlmeta;
uint8 xlinfo;
XLogRecPtr recptr;
XLogRecData rdata[5];
XLogRecData *nextrdata;
xl_btreetid_set(&(xlrec.target), rel, parent, poffset);
xlrec.deadblk = target;
xlrec.leftblk = leftsib;
xlrec.rightblk = rightsib;
rdata[0].data = (char *) &xlrec;
rdata[0].len = SizeOfBtreeDeletePage;
rdata[0].buffer = InvalidBuffer;
rdata[0].next = nextrdata = &(rdata[1]);
if (BufferIsValid(metabuf))
{
xlmeta.root = metad->btm_root;
xlmeta.level = metad->btm_level;
xlmeta.fastroot = metad->btm_fastroot;
xlmeta.fastlevel = metad->btm_fastlevel;
nextrdata->data = (char *) &xlmeta;
nextrdata->len = sizeof(xl_btree_metadata);
nextrdata->buffer = InvalidBuffer;
nextrdata->next = nextrdata + 1;
nextrdata++;
xlinfo = XLOG_BTREE_DELETE_PAGE_META;
}
else if (parent_half_dead)
xlinfo = XLOG_BTREE_DELETE_PAGE_HALF;
else
xlinfo = XLOG_BTREE_DELETE_PAGE;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->next = nextrdata + 1;
nextrdata->buffer = pbuf;
nextrdata->buffer_std = true;
nextrdata++;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->buffer = rbuf;
nextrdata->buffer_std = true;
nextrdata->next = NULL;
if (BufferIsValid(lbuf))
{
nextrdata->next = nextrdata + 1;
nextrdata++;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->buffer = lbuf;
nextrdata->buffer_std = true;
nextrdata->next = NULL;
}
recptr = XLogInsert(RM_BTREE_ID, xlinfo, rdata);
if (BufferIsValid(metabuf))
{
PageSetLSN(metapg, recptr);
PageSetTLI(metapg, ThisTimeLineID);
}
page = BufferGetPage(pbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
page = BufferGetPage(rbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
page = BufferGetPage(buf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
}
}
END_CRIT_SECTION();
/* release metapage; send out relcache inval if metapage changed */
if (BufferIsValid(metabuf))
{
CacheInvalidateRelcache(rel);
_bt_relbuf(rel, metabuf);
}
/* can always release leftsib immediately */
if (BufferIsValid(lbuf))
_bt_relbuf(rel, lbuf);
/*
* If parent became half dead, recurse to delete it. Otherwise, if right
* sibling is empty and is now the last child of the parent, recurse to
* try to delete it. (These cases cannot apply at the same time, though
* the second case might itself recurse to the first.)
*
* When recursing to parent, we hold the lock on the target page until
* done. This delays any insertions into the keyspace that was just
* effectively reassigned to the parent's right sibling. If we allowed
* that, and there were enough such insertions before we finish deleting
* the parent, page splits within that keyspace could lead to inserting
* out-of-order keys into the grandparent level. It is thought that that
* wouldn't have any serious consequences, but it still seems like a
* pretty bad idea.
*/
if (parent_half_dead)
{
/* recursive call will release pbuf */
_bt_relbuf(rel, rbuf);
result = _bt_pagedel(rel, pbuf, stack->bts_parent, vacuum_full) + 1;
_bt_relbuf(rel, buf);
}
else if (parent_one_child && rightsib_empty)
{
_bt_relbuf(rel, pbuf);
_bt_relbuf(rel, buf);
/* recursive call will release rbuf */
result = _bt_pagedel(rel, rbuf, stack, vacuum_full) + 1;
}
else
{
_bt_relbuf(rel, pbuf);
_bt_relbuf(rel, buf);
_bt_relbuf(rel, rbuf);
result = 1;
}
return result;
}
/*
* _bt_pagedel() -- Delete a page from the b-tree.
*
* This action unlinks the page from the b-tree structure, removing all
* pointers leading to it --- but not touching its own left and right links.
* The page cannot be physically reclaimed right away, since other processes
* may currently be trying to follow links leading to the page; they have to
* be allowed to use its right-link to recover. See nbtree/README.
*
* On entry, the target buffer must be pinned and read-locked. This lock and
* pin will be dropped before exiting.
*
* Returns the number of pages successfully deleted (zero on failure; could
* be more than one if parent blocks were deleted).
*
* NOTE: this leaks memory. Rather than trying to clean up everything
* carefully, it's better to run it in a temp context that can be reset
* frequently.
*/
int
_bt_pagedel(Relation rel, Buffer buf, bool vacuum_full)
{
BlockNumber target,
leftsib,
rightsib,
parent;
OffsetNumber poffset,
maxoff;
uint32 targetlevel,
ilevel;
ItemId itemid;
BTItem targetkey,
btitem;
ScanKey itup_scankey;
BTStack stack;
Buffer lbuf,
rbuf,
pbuf;
bool parent_half_dead;
bool parent_one_child;
bool rightsib_empty;
Buffer metabuf = InvalidBuffer;
Page metapg = NULL;
BTMetaPageData *metad = NULL;
Page page;
BTPageOpaque opaque;
/*
* We can never delete rightmost pages nor root pages. While at it, check
* that page is not already deleted and is empty.
*/
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (P_RIGHTMOST(opaque) || P_ISROOT(opaque) || P_ISDELETED(opaque) ||
P_FIRSTDATAKEY(opaque) <= PageGetMaxOffsetNumber(page))
{
_bt_relbuf(rel, buf);
return 0;
}
/*
* Save info about page, including a copy of its high key (it must have
* one, being non-rightmost).
*/
target = BufferGetBlockNumber(buf);
targetlevel = opaque->btpo.level;
leftsib = opaque->btpo_prev;
itemid = PageGetItemId(page, P_HIKEY);
targetkey = CopyBTItem((BTItem) PageGetItem(page, itemid));
/*
* We need to get an approximate pointer to the page's parent page. Use
* the standard search mechanism to search for the page's high key; this
* will give us a link to either the current parent or someplace to its
* left (if there are multiple equal high keys). To avoid deadlocks, we'd
* better drop the target page lock first.
*/
_bt_relbuf(rel, buf);
/* we need a scan key to do our search, so build one */
itup_scankey = _bt_mkscankey(rel, &(targetkey->bti_itup));
/* find the leftmost leaf page containing this key */
stack = _bt_search(rel, rel->rd_rel->relnatts, itup_scankey, false,
&lbuf, BT_READ);
/* don't need a pin on that either */
_bt_relbuf(rel, lbuf);
/*
* If we are trying to delete an interior page, _bt_search did more than
* we needed. Locate the stack item pointing to our parent level.
*/
ilevel = 0;
for (;;)
{
if (stack == NULL)
elog(ERROR, "not enough stack items");
if (ilevel == targetlevel)
break;
stack = stack->bts_parent;
ilevel++;
}
/*
* We have to lock the pages we need to modify in the standard order:
* moving right, then up. Else we will deadlock against other writers.
*
* So, we need to find and write-lock the current left sibling of the
* target page. The sibling that was current a moment ago could have
* split, so we may have to move right. This search could fail if either
* the sibling or the target page was deleted by someone else meanwhile;
* if so, give up. (Right now, that should never happen, since page
* deletion is only done in VACUUM and there shouldn't be multiple VACUUMs
* concurrently on the same table.)
*/
if (leftsib != P_NONE)
{
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
while (P_ISDELETED(opaque) || opaque->btpo_next != target)
{
/* step right one page */
leftsib = opaque->btpo_next;
_bt_relbuf(rel, lbuf);
if (leftsib == P_NONE)
{
elog(LOG, "no left sibling (concurrent deletion?) in \"%s\"",
RelationGetRelationName(rel));
return 0;
}
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
}
}
else
lbuf = InvalidBuffer;
/*
* Next write-lock the target page itself. It should be okay to take just
* a write lock not a superexclusive lock, since no scans would stop on an
* empty page.
*/
buf = _bt_getbuf(rel, target, BT_WRITE);
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
/*
* Check page is still empty etc, else abandon deletion. The empty check
* is necessary since someone else might have inserted into it while we
* didn't have it locked; the others are just for paranoia's sake.
*/
if (P_RIGHTMOST(opaque) || P_ISROOT(opaque) || P_ISDELETED(opaque) ||
P_FIRSTDATAKEY(opaque) <= PageGetMaxOffsetNumber(page))
{
_bt_relbuf(rel, buf);
if (BufferIsValid(lbuf))
_bt_relbuf(rel, lbuf);
return 0;
}
if (opaque->btpo_prev != leftsib)
elog(ERROR, "left link changed unexpectedly in block %u of \"%s\"",
target, RelationGetRelationName(rel));
/*
* And next write-lock the (current) right sibling.
*/
rightsib = opaque->btpo_next;
rbuf = _bt_getbuf(rel, rightsib, BT_WRITE);
/*
* Next find and write-lock the current parent of the target page. This is
* essentially the same as the corresponding step of splitting. However,
* it's possible for the search to fail (for reasons explained in README).
* If that happens, we recover by searching the whole parent level, which
* is a tad inefficient but doesn't happen often enough to be a problem.
*/
ItemPointerSet(&(stack->bts_btitem.bti_itup.t_tid),
target, P_HIKEY);
pbuf = _bt_getstackbuf(rel, stack, BT_WRITE);
if (pbuf == InvalidBuffer)
{
/* Find the leftmost page in the parent level */
pbuf = _bt_get_endpoint(rel, opaque->btpo.level + 1, false);
stack->bts_blkno = BufferGetBlockNumber(pbuf);
stack->bts_offset = InvalidOffsetNumber;
_bt_relbuf(rel, pbuf);
/* and repeat search from there */
pbuf = _bt_getstackbuf(rel, stack, BT_WRITE);
if (pbuf == InvalidBuffer)
elog(ERROR, "failed to re-find parent key in \"%s\" for deletion target page %u",
RelationGetRelationName(rel), target);
}
parent = stack->bts_blkno;
poffset = stack->bts_offset;
/*
* If the target is the rightmost child of its parent, then we can't
* delete, unless it's also the only child --- in which case the parent
* changes to half-dead status.
*/
page = BufferGetPage(pbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
maxoff = PageGetMaxOffsetNumber(page);
parent_half_dead = false;
parent_one_child = false;
if (poffset >= maxoff)
{
if (poffset == P_FIRSTDATAKEY(opaque))
parent_half_dead = true;
else
{
_bt_relbuf(rel, pbuf);
_bt_relbuf(rel, rbuf);
_bt_relbuf(rel, buf);
if (BufferIsValid(lbuf))
_bt_relbuf(rel, lbuf);
return 0;
}
}
else
{
/* Will there be exactly one child left in this parent? */
if (OffsetNumberNext(P_FIRSTDATAKEY(opaque)) == maxoff)
parent_one_child = true;
}
/*
* If we are deleting the next-to-last page on the target's level, then
* the rightsib is a candidate to become the new fast root. (In theory, it
* might be possible to push the fast root even further down, but the odds
* of doing so are slim, and the locking considerations daunting.)
*
* We can safely acquire a lock on the metapage here --- see comments for
* _bt_newroot().
*/
if (leftsib == P_NONE)
{
page = BufferGetPage(rbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo.level == targetlevel);
if (P_RIGHTMOST(opaque))
{
/* rightsib will be the only one left on the level */
metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_WRITE);
metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg);
/*
* The expected case here is btm_fastlevel == targetlevel+1; if
* the fastlevel is <= targetlevel, something is wrong, and we
* choose to overwrite it to fix it.
*/
if (metad->btm_fastlevel > targetlevel + 1)
{
/* no update wanted */
_bt_relbuf(rel, metabuf);
metabuf = InvalidBuffer;
}
}
}
/*
* Here we begin doing the deletion.
*/
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
/*
* Update parent. The normal case is a tad tricky because we want to
* delete the target's downlink and the *following* key. Easiest way is
* to copy the right sibling's downlink over the target downlink, and then
* delete the following item.
*/
page = BufferGetPage(pbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (parent_half_dead)
{
PageIndexTupleDelete(page, poffset);
opaque->btpo_flags |= BTP_HALF_DEAD;
}
else
{
OffsetNumber nextoffset;
itemid = PageGetItemId(page, poffset);
btitem = (BTItem) PageGetItem(page, itemid);
Assert(ItemPointerGetBlockNumber(&(btitem->bti_itup.t_tid)) == target);
ItemPointerSet(&(btitem->bti_itup.t_tid), rightsib, P_HIKEY);
nextoffset = OffsetNumberNext(poffset);
/* This part is just for double-checking */
itemid = PageGetItemId(page, nextoffset);
btitem = (BTItem) PageGetItem(page, itemid);
if (ItemPointerGetBlockNumber(&(btitem->bti_itup.t_tid)) != rightsib)
elog(PANIC, "right sibling is not next child in \"%s\"",
RelationGetRelationName(rel));
PageIndexTupleDelete(page, nextoffset);
}
/*
* Update siblings' side-links. Note the target page's side-links will
* continue to point to the siblings.
*/
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo_next == target);
opaque->btpo_next = rightsib;
}
page = BufferGetPage(rbuf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->btpo_prev == target);
opaque->btpo_prev = leftsib;
rightsib_empty = (P_FIRSTDATAKEY(opaque) > PageGetMaxOffsetNumber(page));
/*
* Mark the page itself deleted. It can be recycled when all current
* transactions are gone; or immediately if we're doing VACUUM FULL.
*/
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
opaque->btpo_flags |= BTP_DELETED;
opaque->btpo.xact =
vacuum_full ? FrozenTransactionId : ReadNewTransactionId();
/* And update the metapage, if needed */
if (BufferIsValid(metabuf))
{
metad->btm_fastroot = rightsib;
metad->btm_fastlevel = targetlevel;
}
/* XLOG stuff */
if (!rel->rd_istemp)
{
xl_btree_delete_page xlrec;
xl_btree_metadata xlmeta;
uint8 xlinfo;
XLogRecPtr recptr;
XLogRecData rdata[5];
XLogRecData *nextrdata;
xlrec.target.node = rel->rd_node;
ItemPointerSet(&(xlrec.target.tid), parent, poffset);
xlrec.deadblk = target;
xlrec.leftblk = leftsib;
xlrec.rightblk = rightsib;
rdata[0].data = (char *) &xlrec;
rdata[0].len = SizeOfBtreeDeletePage;
rdata[0].buffer = InvalidBuffer;
rdata[0].next = nextrdata = &(rdata[1]);
if (BufferIsValid(metabuf))
{
xlmeta.root = metad->btm_root;
xlmeta.level = metad->btm_level;
xlmeta.fastroot = metad->btm_fastroot;
xlmeta.fastlevel = metad->btm_fastlevel;
nextrdata->data = (char *) &xlmeta;
nextrdata->len = sizeof(xl_btree_metadata);
nextrdata->buffer = InvalidBuffer;
nextrdata->next = nextrdata + 1;
nextrdata++;
xlinfo = XLOG_BTREE_DELETE_PAGE_META;
}
else
xlinfo = XLOG_BTREE_DELETE_PAGE;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->next = nextrdata + 1;
nextrdata->buffer = pbuf;
nextrdata->buffer_std = true;
nextrdata++;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->buffer = rbuf;
nextrdata->buffer_std = true;
nextrdata->next = NULL;
if (BufferIsValid(lbuf))
{
nextrdata->next = nextrdata + 1;
nextrdata++;
nextrdata->data = NULL;
nextrdata->len = 0;
nextrdata->buffer = lbuf;
nextrdata->buffer_std = true;
nextrdata->next = NULL;
}
recptr = XLogInsert(RM_BTREE_ID, xlinfo, rdata);
if (BufferIsValid(metabuf))
{
PageSetLSN(metapg, recptr);
PageSetTLI(metapg, ThisTimeLineID);
}
page = BufferGetPage(pbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
page = BufferGetPage(rbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
page = BufferGetPage(buf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
}
}
END_CRIT_SECTION();
/* Write and release buffers */
if (BufferIsValid(metabuf))
_bt_wrtbuf(rel, metabuf);
_bt_wrtbuf(rel, pbuf);
_bt_wrtbuf(rel, rbuf);
_bt_wrtbuf(rel, buf);
if (BufferIsValid(lbuf))
_bt_wrtbuf(rel, lbuf);
/*
* If parent became half dead, recurse to try to delete it. Otherwise, if
* right sibling is empty and is now the last child of the parent, recurse
* to try to delete it. (These cases cannot apply at the same time,
* though the second case might itself recurse to the first.)
*/
if (parent_half_dead)
{
buf = _bt_getbuf(rel, parent, BT_READ);
return _bt_pagedel(rel, buf, vacuum_full) + 1;
}
if (parent_one_child && rightsib_empty)
{
buf = _bt_getbuf(rel, rightsib, BT_READ);
return _bt_pagedel(rel, buf, vacuum_full) + 1;
}
return 1;
}
