/*
* Insert a new item to a page.
*
* Returns true if the insertion was finished. On false, the page was split and
* the parent needs to be updated. (a root split returns true as it doesn't
* need any further action by the caller to complete)
*
* When inserting a downlink to a internal page, 'childbuf' contains the
* child page that was split. Its GIN_INCOMPLETE_SPLIT flag will be cleared
* atomically with the insert. Also, the existing item at the given location
* is updated to point to 'updateblkno'.
*
* stack->buffer is locked on entry, and is kept locked.
*/
static bool
ginPlaceToPage(GinBtree btree, GinBtreeStack *stack,
void *insertdata, BlockNumber updateblkno,
Buffer childbuf, GinStatsData *buildStats)
{
Page page = BufferGetPage(stack->buffer);
XLogRecData *payloadrdata;
bool fit;
uint16 xlflags = 0;
Page childpage = NULL;
if (GinPageIsData(page))
xlflags |= GIN_INSERT_ISDATA;
if (GinPageIsLeaf(page))
{
xlflags |= GIN_INSERT_ISLEAF;
Assert(!BufferIsValid(childbuf));
Assert(updateblkno == InvalidBlockNumber);
}
else
{
Assert(BufferIsValid(childbuf));
Assert(updateblkno != InvalidBlockNumber);
childpage = BufferGetPage(childbuf);
}
/*
* Try to put the incoming tuple on the page. If it doesn't fit,
* placeToPage method will return false and leave the page unmodified, and
* we'll have to split the page.
*/
START_CRIT_SECTION();
fit = btree->placeToPage(btree, stack->buffer, stack->off,
insertdata, updateblkno,
&payloadrdata);
if (fit)
{
MarkBufferDirty(stack->buffer);
/* An insert to an internal page finishes the split of the child. */
if (childbuf != InvalidBuffer)
{
GinPageGetOpaque(childpage)->flags &= ~GIN_INCOMPLETE_SPLIT;
MarkBufferDirty(childbuf);
}
if (RelationNeedsWAL(btree->index))
{
XLogRecPtr recptr;
XLogRecData rdata[3];
ginxlogInsert xlrec;
BlockIdData childblknos[2];
xlrec.node = btree->index->rd_node;
xlrec.blkno = BufferGetBlockNumber(stack->buffer);
xlrec.offset = stack->off;
xlrec.flags = xlflags;
rdata[0].buffer = InvalidBuffer;
rdata[0].data = (char *) &xlrec;
rdata[0].len = sizeof(ginxlogInsert);
/*
* Log information about child if this was an insertion of a
* downlink.
*/
if (childbuf != InvalidBuffer)
{
rdata[0].next = &rdata[1];
BlockIdSet(&childblknos[0], BufferGetBlockNumber(childbuf));
BlockIdSet(&childblknos[1], GinPageGetOpaque(childpage)->rightlink);
rdata[1].buffer = InvalidBuffer;
rdata[1].data = (char *) childblknos;
rdata[1].len = sizeof(BlockIdData) * 2;
rdata[1].next = &rdata[2];
rdata[2].buffer = childbuf;
rdata[2].buffer_std = false;
rdata[2].data = NULL;
rdata[2].len = 0;
rdata[2].next = payloadrdata;
}
else
rdata[0].next = payloadrdata;
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_INSERT, rdata);
PageSetLSN(page, recptr);
if (childbuf != InvalidBuffer)
PageSetLSN(childpage, recptr);
}
END_CRIT_SECTION();
return true;
}
else
{
/* Didn't fit, have to split */
Buffer rbuffer;
Page newlpage;
BlockNumber savedRightLink;
Page rpage;
XLogRecData rdata[2];
ginxlogSplit data;
Buffer lbuffer = InvalidBuffer;
Page newrootpg = NULL;
END_CRIT_SECTION();
rbuffer = GinNewBuffer(btree->index);
/* During index build, count the new page */
if (buildStats)
{
if (btree->isData)
buildStats->nDataPages++;
else
buildStats->nEntryPages++;
}
savedRightLink = GinPageGetOpaque(page)->rightlink;
/*
* newlpage is a pointer to memory page, it is not associated with a
* buffer. stack->buffer is not touched yet.
*/
newlpage = btree->splitPage(btree, stack->buffer, rbuffer, stack->off,
insertdata, updateblkno,
&payloadrdata);
data.node = btree->index->rd_node;
data.rblkno = BufferGetBlockNumber(rbuffer);
data.flags = xlflags;
if (childbuf != InvalidBuffer)
{
Page childpage = BufferGetPage(childbuf);
GinPageGetOpaque(childpage)->flags &= ~GIN_INCOMPLETE_SPLIT;
data.leftChildBlkno = BufferGetBlockNumber(childbuf);
data.rightChildBlkno = GinPageGetOpaque(childpage)->rightlink;
}
else
data.leftChildBlkno = data.rightChildBlkno = InvalidBlockNumber;
rdata[0].buffer = InvalidBuffer;
rdata[0].data = (char *) &data;
rdata[0].len = sizeof(ginxlogSplit);
if (childbuf != InvalidBuffer)
{
rdata[0].next = &rdata[1];
rdata[1].buffer = childbuf;
rdata[1].buffer_std = false;
rdata[1].data = NULL;
rdata[1].len = 0;
rdata[1].next = payloadrdata;
}
else
rdata[0].next = payloadrdata;
rpage = BufferGetPage(rbuffer);
if (stack->parent == NULL)
{
/*
* split root, so we need to allocate new left page and place
* pointer on root to left and right page
*/
lbuffer = GinNewBuffer(btree->index);
/* During index build, count the newly-added root page */
if (buildStats)
{
if (btree->isData)
buildStats->nDataPages++;
else
buildStats->nEntryPages++;
}
/*
* root never has a right-link, so we borrow the rrlink field to
* store the root block number.
*/
data.rrlink = BufferGetBlockNumber(stack->buffer);
data.lblkno = BufferGetBlockNumber(lbuffer);
data.flags |= GIN_SPLIT_ROOT;
GinPageGetOpaque(rpage)->rightlink = InvalidBlockNumber;
GinPageGetOpaque(newlpage)->rightlink = BufferGetBlockNumber(rbuffer);
/*
* Construct a new root page containing downlinks to the new left
* and right pages. (do this in a temporary copy first rather
* than overwriting the original page directly, so that we can still
* abort gracefully if this fails.)
*/
newrootpg = PageGetTempPage(rpage);
GinInitPage(newrootpg, GinPageGetOpaque(newlpage)->flags & ~GIN_LEAF, BLCKSZ);
btree->fillRoot(btree, newrootpg,
BufferGetBlockNumber(lbuffer), newlpage,
BufferGetBlockNumber(rbuffer), rpage);
}
else
{
/* split non-root page */
data.rrlink = savedRightLink;
data.lblkno = BufferGetBlockNumber(stack->buffer);
GinPageGetOpaque(rpage)->rightlink = savedRightLink;
GinPageGetOpaque(newlpage)->flags |= GIN_INCOMPLETE_SPLIT;
GinPageGetOpaque(newlpage)->rightlink = BufferGetBlockNumber(rbuffer);
}
/*
* Ok, we have the new contents of the left page in a temporary copy
* now (newlpage), and the newly-allocated right block has been filled
* in. The original page is still unchanged.
*
* If this is a root split, we also have a temporary page containing
* the new contents of the root. Copy the new left page to a
* newly-allocated block, and initialize the (original) root page the
* new copy. Otherwise, copy over the temporary copy of the new left
* page over the old left page.
*/
START_CRIT_SECTION();
MarkBufferDirty(rbuffer);
if (stack->parent == NULL)
{
PageRestoreTempPage(newlpage, BufferGetPage(lbuffer));
MarkBufferDirty(lbuffer);
newlpage = newrootpg;
}
PageRestoreTempPage(newlpage, BufferGetPage(stack->buffer));
MarkBufferDirty(stack->buffer);
/* write WAL record */
if (RelationNeedsWAL(btree->index))
{
XLogRecPtr recptr;
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_SPLIT, rdata);
PageSetLSN(BufferGetPage(stack->buffer), recptr);
PageSetLSN(rpage, recptr);
if (stack->parent == NULL)
PageSetLSN(BufferGetPage(lbuffer), recptr);
}
END_CRIT_SECTION();
/*
* We can release the lock on the right page now, but keep the
* original buffer locked.
*/
UnlockReleaseBuffer(rbuffer);
if (stack->parent == NULL)
UnlockReleaseBuffer(lbuffer);
/*
* If we split the root, we're done. Otherwise the split is not
* complete until the downlink for the new page has been inserted to
* the parent.
*/
if (stack->parent == NULL)
return true;
else
return false;
}
}
/*
* rtdosplit -- split a page in the tree.
*
* rtpicksplit does the interesting work of choosing the split.
* This routine just does the bit-pushing.
*/
static void
rtdosplit(Relation r,
Buffer buffer,
RTSTACK *stack,
IndexTuple itup,
RTSTATE *rtstate)
{
Page p;
Buffer leftbuf,
rightbuf;
Page left,
right;
ItemId itemid;
IndexTuple item;
IndexTuple ltup,
rtup;
OffsetNumber maxoff;
OffsetNumber i;
OffsetNumber leftoff,
rightoff;
BlockNumber lbknum,
rbknum;
BlockNumber bufblock;
RTreePageOpaque opaque;
bool *isnull;
SPLITVEC v;
OffsetNumber *spl_left,
*spl_right;
TupleDesc tupDesc;
int n;
OffsetNumber newitemoff;
p = (Page) BufferGetPage(buffer);
opaque = (RTreePageOpaque) PageGetSpecialPointer(p);
rtpicksplit(r, p, &v, itup, rtstate);
/*
* The root of the tree is the first block in the relation. If we're
* about to split the root, we need to do some hocus-pocus to enforce this
* guarantee.
*/
if (BufferGetBlockNumber(buffer) == P_ROOT)
{
leftbuf = ReadBuffer(r, P_NEW);
RTInitBuffer(leftbuf, opaque->flags);
lbknum = BufferGetBlockNumber(leftbuf);
left = (Page) BufferGetPage(leftbuf);
}
else
{
leftbuf = buffer;
IncrBufferRefCount(buffer);
lbknum = BufferGetBlockNumber(buffer);
left = (Page) PageGetTempPage(p, sizeof(RTreePageOpaqueData));
}
rightbuf = ReadBuffer(r, P_NEW);
RTInitBuffer(rightbuf, opaque->flags);
rbknum = BufferGetBlockNumber(rightbuf);
right = (Page) BufferGetPage(rightbuf);
spl_left = v.spl_left;
spl_right = v.spl_right;
leftoff = rightoff = FirstOffsetNumber;
maxoff = PageGetMaxOffsetNumber(p);
newitemoff = OffsetNumberNext(maxoff);
/*
* spl_left contains a list of the offset numbers of the tuples that will
* go to the left page. For each offset number, get the tuple item, then
* add the item to the left page. Similarly for the right side.
*/
/* fill left node */
for (n = 0; n < v.spl_nleft; n++)
{
i = *spl_left;
if (i == newitemoff)
item = itup;
else
{
itemid = PageGetItemId(p, i);
item = (IndexTuple) PageGetItem(p, itemid);
}
if (PageAddItem(left, (Item) item, IndexTupleSize(item),
leftoff, LP_USED) == InvalidOffsetNumber)
elog(ERROR, "failed to add index item to \"%s\"",
RelationGetRelationName(r));
leftoff = OffsetNumberNext(leftoff);
spl_left++; /* advance in left split vector */
}
/* fill right node */
for (n = 0; n < v.spl_nright; n++)
{
i = *spl_right;
if (i == newitemoff)
item = itup;
else
{
itemid = PageGetItemId(p, i);
item = (IndexTuple) PageGetItem(p, itemid);
}
if (PageAddItem(right, (Item) item, IndexTupleSize(item),
rightoff, LP_USED) == InvalidOffsetNumber)
elog(ERROR, "failed to add index item to \"%s\"",
RelationGetRelationName(r));
rightoff = OffsetNumberNext(rightoff);
spl_right++; /* advance in right split vector */
}
/* Make sure we consumed all of the split vectors, and release 'em */
Assert(*spl_left == InvalidOffsetNumber);
Assert(*spl_right == InvalidOffsetNumber);
pfree(v.spl_left);
pfree(v.spl_right);
if ((bufblock = BufferGetBlockNumber(buffer)) != P_ROOT)
PageRestoreTempPage(left, p);
WriteBuffer(leftbuf);
WriteBuffer(rightbuf);
/*
* Okay, the page is split. We have three things left to do:
*
* 1) Adjust any active scans on this index to cope with changes we
* introduced in its structure by splitting this page.
*
* 2) "Tighten" the bounding box of the pointer to the left page in the
* parent node in the tree, if any. Since we moved a bunch of stuff off
* the left page, we expect it to get smaller. This happens in the
* internal insertion routine.
*
* 3) Insert a pointer to the right page in the parent. This may cause
* the parent to split. If it does, we need to repeat steps one and two
* for each split node in the tree.
*/
/* adjust active scans */
rtadjscans(r, RTOP_SPLIT, bufblock, FirstOffsetNumber);
tupDesc = r->rd_att;
isnull = (bool *) palloc(r->rd_rel->relnatts * sizeof(bool));
memset(isnull, false, r->rd_rel->relnatts * sizeof(bool));
ltup = index_form_tuple(tupDesc, &(v.spl_ldatum), isnull);
rtup = index_form_tuple(tupDesc, &(v.spl_rdatum), isnull);
pfree(isnull);
pfree(DatumGetPointer(v.spl_ldatum));
pfree(DatumGetPointer(v.spl_rdatum));
/* set pointers to new child pages in the internal index tuples */
ItemPointerSet(&(ltup->t_tid), lbknum, 1);
ItemPointerSet(&(rtup->t_tid), rbknum, 1);
rtintinsert(r, stack, ltup, rtup, rtstate);
pfree(ltup);
pfree(rtup);
}
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;
}
/*
* Insert a new item to a page.
*
* Returns true if the insertion was finished. On false, the page was split and
* the parent needs to be updated. (A root split returns true as it doesn't
* need any further action by the caller to complete.)
*
* When inserting a downlink to an internal page, 'childbuf' contains the
* child page that was split. Its GIN_INCOMPLETE_SPLIT flag will be cleared
* atomically with the insert. Also, the existing item at offset stack->off
* in the target page is updated to point to updateblkno.
*
* stack->buffer is locked on entry, and is kept locked.
* Likewise for childbuf, if given.
*/
static bool
ginPlaceToPage(GinBtree btree, GinBtreeStack *stack,
void *insertdata, BlockNumber updateblkno,
Buffer childbuf, GinStatsData *buildStats)
{
Page page = BufferGetPage(stack->buffer);
bool result;
GinPlaceToPageRC rc;
uint16 xlflags = 0;
Page childpage = NULL;
Page newlpage = NULL,
newrpage = NULL;
void *ptp_workspace = NULL;
XLogRecData payloadrdata[10];
MemoryContext tmpCxt;
MemoryContext oldCxt;
/*
* We do all the work of this function and its subfunctions in a temporary
* memory context. This avoids leakages and simplifies APIs, since some
* subfunctions allocate storage that has to survive until we've finished
* the WAL insertion.
*/
tmpCxt = AllocSetContextCreate(CurrentMemoryContext,
"ginPlaceToPage temporary context",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
oldCxt = MemoryContextSwitchTo(tmpCxt);
if (GinPageIsData(page))
xlflags |= GIN_INSERT_ISDATA;
if (GinPageIsLeaf(page))
{
xlflags |= GIN_INSERT_ISLEAF;
Assert(!BufferIsValid(childbuf));
Assert(updateblkno == InvalidBlockNumber);
}
else
{
Assert(BufferIsValid(childbuf));
Assert(updateblkno != InvalidBlockNumber);
childpage = BufferGetPage(childbuf);
}
/*
* See if the incoming tuple will fit on the page. beginPlaceToPage will
* decide if the page needs to be split, and will compute the split
* contents if so. See comments for beginPlaceToPage and execPlaceToPage
* functions for more details of the API here.
*/
rc = btree->beginPlaceToPage(btree, stack->buffer, stack,
insertdata, updateblkno,
&ptp_workspace,
&newlpage, &newrpage,
payloadrdata);
if (rc == GPTP_NO_WORK)
{
/* Nothing to do */
result = true;
}
else if (rc == GPTP_INSERT)
{
/* It will fit, perform the insertion */
START_CRIT_SECTION();
/* Perform the page update, and set up WAL data about it */
btree->execPlaceToPage(btree, stack->buffer, stack,
insertdata, updateblkno,
ptp_workspace, payloadrdata);
MarkBufferDirty(stack->buffer);
/* An insert to an internal page finishes the split of the child. */
if (BufferIsValid(childbuf))
{
GinPageGetOpaque(childpage)->flags &= ~GIN_INCOMPLETE_SPLIT;
MarkBufferDirty(childbuf);
}
if (RelationNeedsWAL(btree->index))
{
XLogRecPtr recptr;
XLogRecData rdata[3];
ginxlogInsert xlrec;
BlockIdData childblknos[2];
xlrec.node = btree->index->rd_node;
xlrec.blkno = BufferGetBlockNumber(stack->buffer);
xlrec.flags = xlflags;
rdata[0].buffer = InvalidBuffer;
rdata[0].data = (char *) &xlrec;
rdata[0].len = sizeof(ginxlogInsert);
/*
* Log information about child if this was an insertion of a
* downlink.
*/
if (BufferIsValid(childbuf))
{
rdata[0].next = &rdata[1];
BlockIdSet(&childblknos[0], BufferGetBlockNumber(childbuf));
BlockIdSet(&childblknos[1], GinPageGetOpaque(childpage)->rightlink);
rdata[1].buffer = InvalidBuffer;
rdata[1].data = (char *) childblknos;
rdata[1].len = sizeof(BlockIdData) * 2;
rdata[1].next = &rdata[2];
rdata[2].buffer = childbuf;
rdata[2].buffer_std = true;
rdata[2].data = NULL;
rdata[2].len = 0;
rdata[2].next = payloadrdata;
}
else
rdata[0].next = payloadrdata;
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_INSERT, rdata);
PageSetLSN(page, recptr);
if (BufferIsValid(childbuf))
PageSetLSN(childpage, recptr);
}
END_CRIT_SECTION();
/* Insertion is complete. */
result = true;
}
else if (rc == GPTP_SPLIT)
{
/*
* Didn't fit, need to split. The split has been computed in newlpage
* and newrpage, which are pointers to palloc'd pages, not associated
* with buffers. stack->buffer is not touched yet.
*/
Buffer rbuffer;
BlockNumber savedRightLink;
ginxlogSplit data;
Buffer lbuffer = InvalidBuffer;
Page newrootpg = NULL;
/* Get a new index page to become the right page */
rbuffer = GinNewBuffer(btree->index);
/* During index build, count the new page */
if (buildStats)
{
if (btree->isData)
buildStats->nDataPages++;
else
buildStats->nEntryPages++;
}
savedRightLink = GinPageGetOpaque(page)->rightlink;
/* Begin setting up WAL record (which we might not use) */
data.node = btree->index->rd_node;
data.rblkno = BufferGetBlockNumber(rbuffer);
data.flags = xlflags;
if (BufferIsValid(childbuf))
{
data.leftChildBlkno = BufferGetBlockNumber(childbuf);
data.rightChildBlkno = GinPageGetOpaque(childpage)->rightlink;
}
else
data.leftChildBlkno = data.rightChildBlkno = InvalidBlockNumber;
if (stack->parent == NULL)
{
/*
* splitting the root, so we need to allocate new left page and
* place pointers to left and right page on root page.
*/
lbuffer = GinNewBuffer(btree->index);
/* During index build, count the new left page */
if (buildStats)
{
if (btree->isData)
buildStats->nDataPages++;
else
buildStats->nEntryPages++;
}
/*
* root never has a right-link, so we borrow the rrlink field to
* store the root block number.
*/
data.rrlink = BufferGetBlockNumber(stack->buffer);
data.lblkno = BufferGetBlockNumber(lbuffer);
data.flags |= GIN_SPLIT_ROOT;
GinPageGetOpaque(newrpage)->rightlink = InvalidBlockNumber;
GinPageGetOpaque(newlpage)->rightlink = BufferGetBlockNumber(rbuffer);
/*
* Construct a new root page containing downlinks to the new left
* and right pages. (Do this in a temporary copy rather than
* overwriting the original page directly, since we're not in the
* critical section yet.)
*/
newrootpg = PageGetTempPage(newrpage);
GinInitPage(newrootpg, GinPageGetOpaque(newlpage)->flags & ~(GIN_LEAF | GIN_COMPRESSED), BLCKSZ);
btree->fillRoot(btree, newrootpg,
BufferGetBlockNumber(lbuffer), newlpage,
BufferGetBlockNumber(rbuffer), newrpage);
}
else
{
/* splitting a non-root page */
data.rrlink = savedRightLink;
data.lblkno = BufferGetBlockNumber(stack->buffer);
GinPageGetOpaque(newrpage)->rightlink = savedRightLink;
GinPageGetOpaque(newlpage)->flags |= GIN_INCOMPLETE_SPLIT;
GinPageGetOpaque(newlpage)->rightlink = BufferGetBlockNumber(rbuffer);
}
/*
* OK, we have the new contents of the left page in a temporary copy
* now (newlpage), and likewise for the new contents of the
* newly-allocated right block. The original page is still unchanged.
*
* If this is a root split, we also have a temporary page containing
* the new contents of the root.
*/
START_CRIT_SECTION();
MarkBufferDirty(rbuffer);
MarkBufferDirty(stack->buffer);
/*
* Restore the temporary copies over the real buffers.
*/
if (stack->parent == NULL)
{
/* Splitting the root, three pages to update */
MarkBufferDirty(lbuffer);
memcpy(page, newrootpg, BLCKSZ);
memcpy(BufferGetPage(lbuffer), newlpage, BLCKSZ);
memcpy(BufferGetPage(rbuffer), newrpage, BLCKSZ);
}
else
{
/* Normal split, only two pages to update */
memcpy(page, newlpage, BLCKSZ);
memcpy(BufferGetPage(rbuffer), newrpage, BLCKSZ);
}
/* We also clear childbuf's INCOMPLETE_SPLIT flag, if passed */
if (BufferIsValid(childbuf))
{
GinPageGetOpaque(childpage)->flags &= ~GIN_INCOMPLETE_SPLIT;
MarkBufferDirty(childbuf);
}
/* write WAL record */
if (RelationNeedsWAL(btree->index))
{
XLogRecData rdata[2];
XLogRecPtr recptr;
rdata[0].buffer = InvalidBuffer;
rdata[0].data = (char *) &data;
rdata[0].len = sizeof(ginxlogSplit);
if (BufferIsValid(childbuf))
{
rdata[0].next = &rdata[1];
rdata[1].buffer = childbuf;
rdata[1].buffer_std = true;
rdata[1].data = NULL;
rdata[1].len = 0;
rdata[1].next = payloadrdata;
}
else
rdata[0].next = payloadrdata;
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_SPLIT, rdata);
PageSetLSN(page, recptr);
PageSetLSN(BufferGetPage(rbuffer), recptr);
if (stack->parent == NULL)
PageSetLSN(BufferGetPage(lbuffer), recptr);
if (BufferIsValid(childbuf))
PageSetLSN(childpage, recptr);
}
END_CRIT_SECTION();
/*
* We can release the locks/pins on the new pages now, but keep
* stack->buffer locked. childbuf doesn't get unlocked either.
*/
UnlockReleaseBuffer(rbuffer);
if (stack->parent == NULL)
UnlockReleaseBuffer(lbuffer);
/*
* If we split the root, we're done. Otherwise the split is not
* complete until the downlink for the new page has been inserted to
* the parent.
*/
result = (stack->parent == NULL);
}
else
{
elog(ERROR, "invalid return code from GIN placeToPage method: %d", rc);
result = false; /* keep compiler quiet */
}
/* Clean up temp context */
MemoryContextSwitchTo(oldCxt);
MemoryContextDelete(tmpCxt);
return result;
}
/*
* gistSplit -- split a page in the tree.
*/
static IndexTuple *
gistSplit(Relation r,
Buffer buffer,
IndexTuple *itup, /* contains compressed entry */
int *len,
GISTSTATE *giststate,
InsertIndexResult *res)
{
Page p;
Buffer leftbuf,
rightbuf;
Page left,
right;
IndexTuple *lvectup,
*rvectup,
*newtup;
BlockNumber lbknum,
rbknum;
GISTPageOpaque opaque;
GIST_SPLITVEC v;
GistEntryVector *entryvec;
bool *decompvec;
int i,
j,
nlen;
int MaxGrpId = 1;
Datum datum;
bool IsNull;
p = (Page) BufferGetPage(buffer);
opaque = (GISTPageOpaque) PageGetSpecialPointer(p);
/*
* The root of the tree is the first block in the relation. If we're
* about to split the root, we need to do some hocus-pocus to enforce
* this guarantee.
*/
if (BufferGetBlockNumber(buffer) == GISTP_ROOT)
{
leftbuf = ReadBuffer(r, P_NEW);
GISTInitBuffer(leftbuf, opaque->flags);
lbknum = BufferGetBlockNumber(leftbuf);
left = (Page) BufferGetPage(leftbuf);
}
else
{
leftbuf = buffer;
IncrBufferRefCount(buffer);
lbknum = BufferGetBlockNumber(buffer);
left = (Page) PageGetTempPage(p, sizeof(GISTPageOpaqueData));
}
rightbuf = ReadBuffer(r, P_NEW);
GISTInitBuffer(rightbuf, opaque->flags);
rbknum = BufferGetBlockNumber(rightbuf);
right = (Page) BufferGetPage(rightbuf);
/* generate the item array */
entryvec = palloc(GEVHDRSZ + (*len + 1) * sizeof(GISTENTRY));
entryvec->n = *len + 1;
decompvec = (bool *) palloc((*len + 1) * sizeof(bool));
for (i = 1; i <= *len; i++)
{
datum = index_getattr(itup[i - 1], 1, giststate->tupdesc, &IsNull);
gistdentryinit(giststate, 0, &(entryvec->vector[i]),
datum, r, p, i,
ATTSIZE(datum, giststate->tupdesc, 1, IsNull), FALSE, IsNull);
if ((!isAttByVal(giststate, 0)) && entryvec->vector[i].key != datum)
decompvec[i] = TRUE;
else
decompvec[i] = FALSE;
}
/*
* now let the user-defined picksplit function set up the split
* vector; in entryvec have no null value!!
*/
FunctionCall2(&giststate->picksplitFn[0],
PointerGetDatum(entryvec),
PointerGetDatum(&v));
/* compatibility with old code */
if (v.spl_left[v.spl_nleft - 1] == InvalidOffsetNumber)
v.spl_left[v.spl_nleft - 1] = (OffsetNumber) *len;
if (v.spl_right[v.spl_nright - 1] == InvalidOffsetNumber)
v.spl_right[v.spl_nright - 1] = (OffsetNumber) *len;
v.spl_lattr[0] = v.spl_ldatum;
v.spl_rattr[0] = v.spl_rdatum;
v.spl_lisnull[0] = false;
v.spl_risnull[0] = false;
/*
* if index is multikey, then we must to try get smaller bounding box
* for subkey(s)
*/
if (r->rd_att->natts > 1)
{
v.spl_idgrp = (int *) palloc0(sizeof(int) * (*len + 1));
v.spl_grpflag = (char *) palloc0(sizeof(char) * (*len + 1));
v.spl_ngrp = (int *) palloc(sizeof(int) * (*len + 1));
MaxGrpId = gistfindgroup(giststate, entryvec->vector, &v);
/* form union of sub keys for each page (l,p) */
gistunionsubkey(r, giststate, itup, &v);
/*
* if possible, we insert equivalent tuples with control by
* penalty for a subkey(s)
*/
if (MaxGrpId > 1)
gistadjsubkey(r, itup, len, &v, giststate);
pfree(v.spl_idgrp);
pfree(v.spl_grpflag);
pfree(v.spl_ngrp);
}
/* clean up the entry vector: its keys need to be deleted, too */
for (i = 1; i <= *len; i++)
if (decompvec[i])
pfree(DatumGetPointer(entryvec->vector[i].key));
pfree(entryvec);
pfree(decompvec);
/* form left and right vector */
lvectup = (IndexTuple *) palloc(sizeof(IndexTuple) * v.spl_nleft);
rvectup = (IndexTuple *) palloc(sizeof(IndexTuple) * v.spl_nright);
for (i = 0; i < v.spl_nleft; i++)
lvectup[i] = itup[v.spl_left[i] - 1];
for (i = 0; i < v.spl_nright; i++)
rvectup[i] = itup[v.spl_right[i] - 1];
/* write on disk (may be need another split) */
if (gistnospace(right, rvectup, v.spl_nright))
{
nlen = v.spl_nright;
newtup = gistSplit(r, rightbuf, rvectup, &nlen, giststate,
(res && rvectup[nlen - 1] == itup[*len - 1]) ? res : NULL);
ReleaseBuffer(rightbuf);
for (j = 1; j < r->rd_att->natts; j++)
if ((!isAttByVal(giststate, j)) && !v.spl_risnull[j])
pfree(DatumGetPointer(v.spl_rattr[j]));
}
else
{
OffsetNumber l;
l = gistwritebuffer(r, right, rvectup, v.spl_nright, FirstOffsetNumber);
WriteBuffer(rightbuf);
if (res)
ItemPointerSet(&((*res)->pointerData), rbknum, l);
nlen = 1;
newtup = (IndexTuple *) palloc(sizeof(IndexTuple) * 1);
newtup[0] = gistFormTuple(giststate, r, v.spl_rattr, v.spl_rattrsize, v.spl_risnull);
ItemPointerSet(&(newtup[0]->t_tid), rbknum, 1);
}
if (gistnospace(left, lvectup, v.spl_nleft))
{
int llen = v.spl_nleft;
IndexTuple *lntup;
lntup = gistSplit(r, leftbuf, lvectup, &llen, giststate,
(res && lvectup[llen - 1] == itup[*len - 1]) ? res : NULL);
ReleaseBuffer(leftbuf);
for (j = 1; j < r->rd_att->natts; j++)
if ((!isAttByVal(giststate, j)) && !v.spl_lisnull[j])
pfree(DatumGetPointer(v.spl_lattr[j]));
newtup = gistjoinvector(newtup, &nlen, lntup, llen);
pfree(lntup);
}
else
{
OffsetNumber l;
l = gistwritebuffer(r, left, lvectup, v.spl_nleft, FirstOffsetNumber);
if (BufferGetBlockNumber(buffer) != GISTP_ROOT)
PageRestoreTempPage(left, p);
WriteBuffer(leftbuf);
if (res)
ItemPointerSet(&((*res)->pointerData), lbknum, l);
nlen += 1;
newtup = (IndexTuple *) repalloc((void *) newtup, sizeof(IndexTuple) * nlen);
newtup[nlen - 1] = gistFormTuple(giststate, r, v.spl_lattr, v.spl_lattrsize, v.spl_lisnull);
ItemPointerSet(&(newtup[nlen - 1]->t_tid), lbknum, 1);
}
/* !!! pfree */
pfree(rvectup);
pfree(lvectup);
pfree(v.spl_left);
pfree(v.spl_right);
*len = nlen;
return newtup;
}