/*
* Return whether brin_doupdate can do a samepage update.
*/
bool
brin_can_do_samepage_update(Buffer buffer, Size origsz, Size newsz)
{
return
((newsz <= origsz) ||
PageGetExactFreeSpace(BufferGetPage(buffer)) >= (newsz - origsz));
}
/*
* Update lastUsedPages cache when done modifying a page.
*
* We update the appropriate cache entry if it already contained this page
* (its freeSpace is likely obsolete), or if this page has more space than
* whatever we had cached.
*/
void
SpGistSetLastUsedPage(Relation index, Buffer buffer)
{
SpGistCache *cache = spgGetCache(index);
SpGistLastUsedPage *lup;
int freeSpace;
Page page = BufferGetPage(buffer);
BlockNumber blkno = BufferGetBlockNumber(buffer);
int flags;
/* Never enter the root page in cache, though */
if (blkno == SPGIST_HEAD_BLKNO)
return;
if (SpGistPageIsLeaf(page))
flags = GBUF_LEAF;
else
flags = GBUF_INNER_PARITY(blkno);
lup = GET_LUP(cache, flags);
freeSpace = PageGetExactFreeSpace(page);
if (lup->blkno == InvalidBlockNumber || lup->blkno == blkno ||
lup->freeSpace < freeSpace)
{
lup->blkno = blkno;
lup->freeSpace = freeSpace;
}
}
/*
* Update lastUsedPages cache when done modifying a page.
*
* We update the appropriate cache entry if it already contained this page
* (its freeSpace is likely obsolete), or if this page has more space than
* whatever we had cached.
*/
void
SpGistSetLastUsedPage(Relation index, Buffer buffer)
{
SpGistCache *cache = spgGetCache(index);
SpGistLastUsedPage *lup;
int freeSpace;
Page page = BufferGetPage(buffer);
BlockNumber blkno = BufferGetBlockNumber(buffer);
int flags;
/* Never enter fixed pages (root pages) in cache, though */
if (SpGistBlockIsFixed(blkno))
return;
if (SpGistPageIsLeaf(page))
flags = GBUF_LEAF;
else
flags = GBUF_INNER_PARITY(blkno);
if (SpGistPageStoresNulls(page))
flags |= GBUF_NULLS;
lup = GET_LUP(cache, flags);
freeSpace = PageGetExactFreeSpace(page);
if (lup->blkno == InvalidBlockNumber || lup->blkno == blkno ||
lup->freeSpace < freeSpace)
{
lup->blkno = blkno;
lup->freeSpace = freeSpace;
}
}
/*
* Allocate and initialize a new buffer of the type and parity specified by
* flags. The returned buffer is already pinned and exclusive-locked.
*
* When requesting an inner page, if we get one with the wrong parity,
* we just release the buffer and try again. We will get a different page
* because GetFreeIndexPage will have marked the page used in FSM. The page
* is entered in our local lastUsedPages cache, so there's some hope of
* making use of it later in this session, but otherwise we rely on VACUUM
* to eventually re-enter the page in FSM, making it available for recycling.
* Note that such a page does not get marked dirty here, so unless it's used
* fairly soon, the buffer will just get discarded and the page will remain
* as it was on disk.
*
* When we return a buffer to the caller, the page is *not* entered into
* the lastUsedPages cache; we expect the caller will do so after it's taken
* whatever space it will use. This is because after the caller has used up
* some space, the page might have less space than whatever was cached already
* so we'd rather not trash the old cache entry.
*/
static Buffer
allocNewBuffer(Relation index, int flags)
{
SpGistCache *cache = spgGetCache(index);
uint16 pageflags = 0;
if (GBUF_REQ_LEAF(flags))
pageflags |= SPGIST_LEAF;
if (GBUF_REQ_NULLS(flags))
pageflags |= SPGIST_NULLS;
for (;;)
{
Buffer buffer;
buffer = SpGistNewBuffer(index);
SpGistInitBuffer(buffer, pageflags);
if (pageflags & SPGIST_LEAF)
{
/* Leaf pages have no parity concerns, so just use it */
return buffer;
}
else
{
BlockNumber blkno = BufferGetBlockNumber(buffer);
int blkFlags = GBUF_INNER_PARITY(blkno);
if ((flags & GBUF_PARITY_MASK) == blkFlags)
{
/* Page has right parity, use it */
return buffer;
}
else
{
/* Page has wrong parity, record it in cache and try again */
if (pageflags & SPGIST_NULLS)
blkFlags |= GBUF_NULLS;
cache->lastUsedPages.cachedPage[blkFlags].blkno = blkno;
cache->lastUsedPages.cachedPage[blkFlags].freeSpace =
PageGetExactFreeSpace(BufferGetPage(buffer));
UnlockReleaseBuffer(buffer);
}
}
}
}
/* -------------------------------------------------
* GetHashPageStatis()
*
* Collect statistics of single hash page
* -------------------------------------------------
*/
static void
GetHashPageStats(Page page, HashIndexStat *stats)
{
OffsetNumber maxoff = PageGetMaxOffsetNumber(page);
int off;
/* count live and dead tuples, and free space */
for (off = FirstOffsetNumber; off <= maxoff; off++)
{
ItemId id = PageGetItemId(page, off);
if (!ItemIdIsDead(id))
stats->live_items++;
else
stats->dead_items++;
}
stats->free_space += PageGetExactFreeSpace(page);
}
/*
* Allocate and initialize a new buffer of the type and parity specified by
* flags. The returned buffer is already pinned and exclusive-locked.
*
* When requesting an inner page, if we get one with the wrong parity,
* we just release the buffer and try again. We will get a different page
* because GetFreeIndexPage will have marked the page used in FSM. The page
* is entered in our local lastUsedPages cache, so there's some hope of
* making use of it later in this session, but otherwise we rely on VACUUM
* to eventually re-enter the page in FSM, making it available for recycling.
* Note that such a page does not get marked dirty here, so unless it's used
* fairly soon, the buffer will just get discarded and the page will remain
* as it was on disk.
*
* When we return a buffer to the caller, the page is *not* entered into
* the lastUsedPages cache; we expect the caller will do so after it's taken
* whatever space it will use. This is because after the caller has used up
* some space, the page might have less space than whatever was cached already
* so we'd rather not trash the old cache entry.
*/
static Buffer
allocNewBuffer(Relation index, int flags)
{
SpGistCache *cache = spgGetCache(index);
for (;;)
{
Buffer buffer;
buffer = SpGistNewBuffer(index);
SpGistInitBuffer(buffer, (flags & GBUF_LEAF) ? SPGIST_LEAF : 0);
if (flags & GBUF_LEAF)
{
/* Leaf pages have no parity concerns, so just use it */
return buffer;
}
else
{
BlockNumber blkno = BufferGetBlockNumber(buffer);
int blkParity = blkno % 3;
if ((flags & GBUF_PARITY_MASK) == blkParity)
{
/* Page has right parity, use it */
return buffer;
}
else
{
/* Page has wrong parity, record it in cache and try again */
cache->lastUsedPages.innerPage[blkParity].blkno = blkno;
cache->lastUsedPages.innerPage[blkParity].freeSpace =
PageGetExactFreeSpace(BufferGetPage(buffer));
UnlockReleaseBuffer(buffer);
}
}
}
}
/*
* Build a pending-list page from the given array of tuples, and write it out.
*
* Returns amount of free space left on the page.
*/
static int32
writeListPage(Relation index, Buffer buffer,
IndexTuple *tuples, int32 ntuples, BlockNumber rightlink)
{
Page page = BufferGetPage(buffer);
int32 i,
freesize,
size = 0;
OffsetNumber l,
off;
char *workspace;
char *ptr;
/* workspace could be a local array; we use palloc for alignment */
workspace = palloc(BLCKSZ);
START_CRIT_SECTION();
GinInitBuffer(buffer, GIN_LIST);
off = FirstOffsetNumber;
ptr = workspace;
for (i = 0; i < ntuples; i++)
{
int this_size = IndexTupleSize(tuples[i]);
memcpy(ptr, tuples[i], this_size);
ptr += this_size;
size += this_size;
l = PageAddItem(page, (Item) tuples[i], this_size, off, false, false);
if (l == InvalidOffsetNumber)
elog(ERROR, "failed to add item to index page in \"%s\"",
RelationGetRelationName(index));
off++;
}
Assert(size <= BLCKSZ); /* else we overran workspace */
GinPageGetOpaque(page)->rightlink = rightlink;
/*
* tail page may contain only whole row(s) or final part of row placed on
* previous pages (a "row" here meaning all the index tuples generated for
* one heap tuple)
*/
if (rightlink == InvalidBlockNumber)
{
GinPageSetFullRow(page);
GinPageGetOpaque(page)->maxoff = 1;
}
else
{
GinPageGetOpaque(page)->maxoff = 0;
}
MarkBufferDirty(buffer);
if (RelationNeedsWAL(index))
{
XLogRecData rdata[2];
ginxlogInsertListPage data;
XLogRecPtr recptr;
data.node = index->rd_node;
data.blkno = BufferGetBlockNumber(buffer);
data.rightlink = rightlink;
data.ntuples = ntuples;
rdata[0].buffer = InvalidBuffer;
rdata[0].data = (char *) &data;
rdata[0].len = sizeof(ginxlogInsertListPage);
rdata[0].next = rdata + 1;
rdata[1].buffer = buffer;
rdata[1].buffer_std = true;
rdata[1].data = workspace;
rdata[1].len = size;
rdata[1].next = NULL;
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_INSERT_LISTPAGE, rdata);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
}
/* get free space before releasing buffer */
freesize = PageGetExactFreeSpace(page);
UnlockReleaseBuffer(buffer);
END_CRIT_SECTION();
pfree(workspace);
return freesize;
}
/*
* Write the index tuples contained in *collector into the index's
* pending list.
*
* Function guarantees that all these tuples will be inserted consecutively,
* preserving order
*/
void
ginHeapTupleFastInsert(GinState *ginstate, GinTupleCollector *collector)
{
Relation index = ginstate->index;
Buffer metabuffer;
Page metapage;
GinMetaPageData *metadata = NULL;
XLogRecData rdata[2];
Buffer buffer = InvalidBuffer;
Page page = NULL;
ginxlogUpdateMeta data;
bool separateList = false;
bool needCleanup = false;
if (collector->ntuples == 0)
return;
data.node = index->rd_node;
data.ntuples = 0;
data.newRightlink = data.prevTail = InvalidBlockNumber;
rdata[0].buffer = InvalidBuffer;
rdata[0].data = (char *) &data;
rdata[0].len = sizeof(ginxlogUpdateMeta);
rdata[0].next = NULL;
metabuffer = ReadBuffer(index, GIN_METAPAGE_BLKNO);
metapage = BufferGetPage(metabuffer);
if (collector->sumsize + collector->ntuples * sizeof(ItemIdData) > GinListPageSize)
{
/*
* Total size is greater than one page => make sublist
*/
separateList = true;
}
else
{
LockBuffer(metabuffer, GIN_EXCLUSIVE);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber ||
collector->sumsize + collector->ntuples * sizeof(ItemIdData) > metadata->tailFreeSize)
{
/*
* Pending list is empty or total size is greater than freespace
* on tail page => make sublist
*
* We unlock metabuffer to keep high concurrency
*/
separateList = true;
LockBuffer(metabuffer, GIN_UNLOCK);
}
}
if (separateList)
{
/*
* We should make sublist separately and append it to the tail
*/
GinMetaPageData sublist;
memset(&sublist, 0, sizeof(GinMetaPageData));
makeSublist(index, collector->tuples, collector->ntuples, &sublist);
/*
* metapage was unlocked, see above
*/
LockBuffer(metabuffer, GIN_EXCLUSIVE);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber)
{
/*
* Main list is empty, so just insert sublist as main list
*/
START_CRIT_SECTION();
metadata->head = sublist.head;
metadata->tail = sublist.tail;
metadata->tailFreeSize = sublist.tailFreeSize;
metadata->nPendingPages = sublist.nPendingPages;
metadata->nPendingHeapTuples = sublist.nPendingHeapTuples;
}
else
{
/*
* Merge lists
*/
data.prevTail = metadata->tail;
data.newRightlink = sublist.head;
buffer = ReadBuffer(index, metadata->tail);
LockBuffer(buffer, GIN_EXCLUSIVE);
page = BufferGetPage(buffer);
rdata[0].next = rdata + 1;
rdata[1].buffer = buffer;
rdata[1].buffer_std = true;
rdata[1].data = NULL;
rdata[1].len = 0;
rdata[1].next = NULL;
Assert(GinPageGetOpaque(page)->rightlink == InvalidBlockNumber);
START_CRIT_SECTION();
GinPageGetOpaque(page)->rightlink = sublist.head;
MarkBufferDirty(buffer);
metadata->tail = sublist.tail;
metadata->tailFreeSize = sublist.tailFreeSize;
metadata->nPendingPages += sublist.nPendingPages;
metadata->nPendingHeapTuples += sublist.nPendingHeapTuples;
}
}
else
{
/*
* Insert into tail page. Metapage is already locked
*/
OffsetNumber l,
off;
int i,
tupsize;
char *ptr;
buffer = ReadBuffer(index, metadata->tail);
LockBuffer(buffer, GIN_EXCLUSIVE);
page = BufferGetPage(buffer);
off = (PageIsEmpty(page)) ? FirstOffsetNumber :
OffsetNumberNext(PageGetMaxOffsetNumber(page));
rdata[0].next = rdata + 1;
rdata[1].buffer = buffer;
rdata[1].buffer_std = true;
ptr = rdata[1].data = (char *) palloc(collector->sumsize);
rdata[1].len = collector->sumsize;
rdata[1].next = NULL;
data.ntuples = collector->ntuples;
START_CRIT_SECTION();
/*
* Increase counter of heap tuples
*/
Assert(GinPageGetOpaque(page)->maxoff <= metadata->nPendingHeapTuples);
GinPageGetOpaque(page)->maxoff++;
metadata->nPendingHeapTuples++;
for (i = 0; i < collector->ntuples; i++)
{
tupsize = IndexTupleSize(collector->tuples[i]);
l = PageAddItem(page, (Item) collector->tuples[i], tupsize, off, false, false);
if (l == InvalidOffsetNumber)
elog(ERROR, "failed to add item to index page in \"%s\"",
RelationGetRelationName(index));
memcpy(ptr, collector->tuples[i], tupsize);
ptr += tupsize;
off++;
}
Assert((ptr - rdata[1].data) <= collector->sumsize);
metadata->tailFreeSize = PageGetExactFreeSpace(page);
MarkBufferDirty(buffer);
}
/*
* Write metabuffer, make xlog entry
*/
MarkBufferDirty(metabuffer);
if (RelationNeedsWAL(index))
{
XLogRecPtr recptr;
memcpy(&data.metadata, metadata, sizeof(GinMetaPageData));
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_UPDATE_META_PAGE, rdata);
PageSetLSN(metapage, recptr);
PageSetTLI(metapage, ThisTimeLineID);
if (buffer != InvalidBuffer)
{
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
}
}
if (buffer != InvalidBuffer)
UnlockReleaseBuffer(buffer);
/*
* Force pending list cleanup when it becomes too long. And,
* ginInsertCleanup could take significant amount of time, so we prefer to
* call it when it can do all the work in a single collection cycle. In
* non-vacuum mode, it shouldn't require maintenance_work_mem, so fire it
* while pending list is still small enough to fit into work_mem.
*
* ginInsertCleanup() should not be called inside our CRIT_SECTION.
*/
if (metadata->nPendingPages * GIN_PAGE_FREESIZE > work_mem * 1024L)
needCleanup = true;
UnlockReleaseBuffer(metabuffer);
END_CRIT_SECTION();
if (needCleanup)
ginInsertCleanup(ginstate, false, NULL);
}
/*
* Get a buffer of the type and parity specified by flags, having at least
* as much free space as indicated by needSpace. We use the lastUsedPages
* cache to assign the same buffer previously requested when possible.
* The returned buffer is already pinned and exclusive-locked.
*
* *isNew is set true if the page was initialized here, false if it was
* already valid.
*/
Buffer
SpGistGetBuffer(Relation index, int flags, int needSpace, bool *isNew)
{
SpGistCache *cache = spgGetCache(index);
SpGistLastUsedPage *lup;
/* Bail out if even an empty page wouldn't meet the demand */
if (needSpace > SPGIST_PAGE_CAPACITY)
elog(ERROR, "desired SPGiST tuple size is too big");
/*
* If possible, increase the space request to include relation's
* fillfactor. This ensures that when we add unrelated tuples to a page,
* we try to keep 100-fillfactor% available for adding tuples that are
* related to the ones already on it. But fillfactor mustn't cause an
* error for requests that would otherwise be legal.
*/
needSpace += RelationGetTargetPageFreeSpace(index,
SPGIST_DEFAULT_FILLFACTOR);
needSpace = Min(needSpace, SPGIST_PAGE_CAPACITY);
/* Get the cache entry for this flags setting */
lup = GET_LUP(cache, flags);
/* If we have nothing cached, just turn it over to allocNewBuffer */
if (lup->blkno == InvalidBlockNumber)
{
*isNew = true;
return allocNewBuffer(index, flags);
}
/* fixed pages should never be in cache */
Assert(!SpGistBlockIsFixed(lup->blkno));
/* If cached freeSpace isn't enough, don't bother looking at the page */
if (lup->freeSpace >= needSpace)
{
Buffer buffer;
Page page;
buffer = ReadBuffer(index, lup->blkno);
if (!ConditionalLockBuffer(buffer))
{
/*
* buffer is locked by another process, so return a new buffer
*/
ReleaseBuffer(buffer);
*isNew = true;
return allocNewBuffer(index, flags);
}
page = BufferGetPage(buffer);
if (PageIsNew(page) || SpGistPageIsDeleted(page) || PageIsEmpty(page))
{
/* OK to initialize the page */
uint16 pageflags = 0;
if (GBUF_REQ_LEAF(flags))
pageflags |= SPGIST_LEAF;
if (GBUF_REQ_NULLS(flags))
pageflags |= SPGIST_NULLS;
SpGistInitBuffer(buffer, pageflags);
lup->freeSpace = PageGetExactFreeSpace(page) - needSpace;
*isNew = true;
return buffer;
}
/*
* Check that page is of right type and has enough space. We must
* recheck this since our cache isn't necessarily up to date.
*/
if ((GBUF_REQ_LEAF(flags) ? SpGistPageIsLeaf(page) : !SpGistPageIsLeaf(page)) &&
(GBUF_REQ_NULLS(flags) ? SpGistPageStoresNulls(page) : !SpGistPageStoresNulls(page)))
{
int freeSpace = PageGetExactFreeSpace(page);
if (freeSpace >= needSpace)
{
/* Success, update freespace info and return the buffer */
lup->freeSpace = freeSpace - needSpace;
*isNew = false;
return buffer;
}
}
/*
* fallback to allocation of new buffer
*/
UnlockReleaseBuffer(buffer);
}
/* No success with cache, so return a new buffer */
*isNew = true;
return allocNewBuffer(index, flags);
}
/*
* Add a new item to the page, replacing a PLACEHOLDER item if possible.
* Return the location it's inserted at, or InvalidOffsetNumber on failure.
*
* If startOffset isn't NULL, we start searching for placeholders at
* *startOffset, and update that to the next place to search. This is just
* an optimization for repeated insertions.
*
* If errorOK is false, we throw error when there's not enough room,
* rather than returning InvalidOffsetNumber.
*/
OffsetNumber
SpGistPageAddNewItem(SpGistState *state, Page page, Item item, Size size,
OffsetNumber *startOffset, bool errorOK)
{
SpGistPageOpaque opaque = SpGistPageGetOpaque(page);
OffsetNumber i,
maxoff,
offnum;
if (opaque->nPlaceholder > 0 &&
PageGetExactFreeSpace(page) + SGDTSIZE >= MAXALIGN(size))
{
/* Try to replace a placeholder */
maxoff = PageGetMaxOffsetNumber(page);
offnum = InvalidOffsetNumber;
for (;;)
{
if (startOffset && *startOffset != InvalidOffsetNumber)
i = *startOffset;
else
i = FirstOffsetNumber;
for (; i <= maxoff; i++)
{
SpGistDeadTuple it = (SpGistDeadTuple) PageGetItem(page,
PageGetItemId(page, i));
if (it->tupstate == SPGIST_PLACEHOLDER)
{
offnum = i;
break;
}
}
/* Done if we found a placeholder */
if (offnum != InvalidOffsetNumber)
break;
if (startOffset && *startOffset != InvalidOffsetNumber)
{
/* Hint was no good, re-search from beginning */
*startOffset = InvalidOffsetNumber;
continue;
}
/* Hmm, no placeholder found? */
opaque->nPlaceholder = 0;
break;
}
if (offnum != InvalidOffsetNumber)
{
/* Replace the placeholder tuple */
PageIndexTupleDelete(page, offnum);
offnum = PageAddItem(page, item, size, offnum, false, false);
/*
* We should not have failed given the size check at the top of
* the function, but test anyway. If we did fail, we must PANIC
* because we've already deleted the placeholder tuple, and
* there's no other way to keep the damage from getting to disk.
*/
if (offnum != InvalidOffsetNumber)
{
Assert(opaque->nPlaceholder > 0);
opaque->nPlaceholder--;
if (startOffset)
*startOffset = offnum + 1;
}
else
elog(PANIC, "failed to add item of size %u to SPGiST index page",
(int) size);
return offnum;
}
}
/* No luck in replacing a placeholder, so just add it to the page */
offnum = PageAddItem(page, item, size,
InvalidOffsetNumber, false, false);
if (offnum == InvalidOffsetNumber && !errorOK)
elog(ERROR, "failed to add item of size %u to SPGiST index page",
(int) size);
return offnum;
}
/*
* Write the index tuples contained in *collector into the index's
* pending list.
*
* Function guarantees that all these tuples will be inserted consecutively,
* preserving order
*/
void
ginHeapTupleFastInsert(GinState *ginstate, GinTupleCollector *collector)
{
Relation index = ginstate->index;
Buffer metabuffer;
Page metapage;
GinMetaPageData *metadata = NULL;
Buffer buffer = InvalidBuffer;
Page page = NULL;
ginxlogUpdateMeta data;
bool separateList = false;
bool needCleanup = false;
int cleanupSize;
bool needWal;
if (collector->ntuples == 0)
return;
needWal = RelationNeedsWAL(index);
data.node = index->rd_node;
data.ntuples = 0;
data.newRightlink = data.prevTail = InvalidBlockNumber;
metabuffer = ReadBuffer(index, GIN_METAPAGE_BLKNO);
metapage = BufferGetPage(metabuffer);
if (collector->sumsize + collector->ntuples * sizeof(ItemIdData) > GinListPageSize)
{
/*
* Total size is greater than one page => make sublist
*/
separateList = true;
}
else
{
LockBuffer(metabuffer, GIN_EXCLUSIVE);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber ||
collector->sumsize + collector->ntuples * sizeof(ItemIdData) > metadata->tailFreeSize)
{
/*
* Pending list is empty or total size is greater than freespace
* on tail page => make sublist
*
* We unlock metabuffer to keep high concurrency
*/
separateList = true;
LockBuffer(metabuffer, GIN_UNLOCK);
}
}
if (separateList)
{
/*
* We should make sublist separately and append it to the tail
*/
GinMetaPageData sublist;
memset(&sublist, 0, sizeof(GinMetaPageData));
makeSublist(index, collector->tuples, collector->ntuples, &sublist);
if (needWal)
XLogBeginInsert();
/*
* metapage was unlocked, see above
*/
LockBuffer(metabuffer, GIN_EXCLUSIVE);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber)
{
/*
* Main list is empty, so just insert sublist as main list
*/
START_CRIT_SECTION();
metadata->head = sublist.head;
metadata->tail = sublist.tail;
metadata->tailFreeSize = sublist.tailFreeSize;
metadata->nPendingPages = sublist.nPendingPages;
metadata->nPendingHeapTuples = sublist.nPendingHeapTuples;
}
else
{
/*
* Merge lists
*/
data.prevTail = metadata->tail;
data.newRightlink = sublist.head;
buffer = ReadBuffer(index, metadata->tail);
LockBuffer(buffer, GIN_EXCLUSIVE);
page = BufferGetPage(buffer);
Assert(GinPageGetOpaque(page)->rightlink == InvalidBlockNumber);
START_CRIT_SECTION();
GinPageGetOpaque(page)->rightlink = sublist.head;
MarkBufferDirty(buffer);
metadata->tail = sublist.tail;
metadata->tailFreeSize = sublist.tailFreeSize;
metadata->nPendingPages += sublist.nPendingPages;
metadata->nPendingHeapTuples += sublist.nPendingHeapTuples;
if (needWal)
XLogRegisterBuffer(1, buffer, REGBUF_STANDARD);
}
}
else
{
/*
* Insert into tail page. Metapage is already locked
*/
OffsetNumber l,
off;
int i,
tupsize;
char *ptr;
char *collectordata;
buffer = ReadBuffer(index, metadata->tail);
LockBuffer(buffer, GIN_EXCLUSIVE);
page = BufferGetPage(buffer);
off = (PageIsEmpty(page)) ? FirstOffsetNumber :
OffsetNumberNext(PageGetMaxOffsetNumber(page));
collectordata = ptr = (char *) palloc(collector->sumsize);
data.ntuples = collector->ntuples;
if (needWal)
XLogBeginInsert();
START_CRIT_SECTION();
/*
* Increase counter of heap tuples
*/
Assert(GinPageGetOpaque(page)->maxoff <= metadata->nPendingHeapTuples);
GinPageGetOpaque(page)->maxoff++;
metadata->nPendingHeapTuples++;
for (i = 0; i < collector->ntuples; i++)
{
tupsize = IndexTupleSize(collector->tuples[i]);
l = PageAddItem(page, (Item) collector->tuples[i], tupsize, off, false, false);
if (l == InvalidOffsetNumber)
elog(ERROR, "failed to add item to index page in \"%s\"",
RelationGetRelationName(index));
memcpy(ptr, collector->tuples[i], tupsize);
ptr += tupsize;
off++;
}
Assert((ptr - collectordata) <= collector->sumsize);
if (needWal)
{
XLogRegisterBuffer(1, buffer, REGBUF_STANDARD);
XLogRegisterBufData(1, collectordata, collector->sumsize);
}
metadata->tailFreeSize = PageGetExactFreeSpace(page);
MarkBufferDirty(buffer);
}
/*
* Set pd_lower just past the end of the metadata. This is essential,
* because without doing so, metadata will be lost if xlog.c compresses
* the page. (We must do this here because pre-v11 versions of PG did not
* set the metapage's pd_lower correctly, so a pg_upgraded index might
* contain the wrong value.)
*/
((PageHeader) metapage)->pd_lower =
((char *) metadata + sizeof(GinMetaPageData)) - (char *) metapage;
/*
* Write metabuffer, make xlog entry
*/
MarkBufferDirty(metabuffer);
if (needWal)
{
XLogRecPtr recptr;
memcpy(&data.metadata, metadata, sizeof(GinMetaPageData));
XLogRegisterBuffer(0, metabuffer, REGBUF_WILL_INIT | REGBUF_STANDARD);
XLogRegisterData((char *) &data, sizeof(ginxlogUpdateMeta));
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_UPDATE_META_PAGE);
PageSetLSN(metapage, recptr);
if (buffer != InvalidBuffer)
{
PageSetLSN(page, recptr);
}
}
if (buffer != InvalidBuffer)
UnlockReleaseBuffer(buffer);
/*
* Force pending list cleanup when it becomes too long. And,
* ginInsertCleanup could take significant amount of time, so we prefer to
* call it when it can do all the work in a single collection cycle. In
* non-vacuum mode, it shouldn't require maintenance_work_mem, so fire it
* while pending list is still small enough to fit into
* gin_pending_list_limit.
*
* ginInsertCleanup() should not be called inside our CRIT_SECTION.
*/
cleanupSize = GinGetPendingListCleanupSize(index);
if (metadata->nPendingPages * GIN_PAGE_FREESIZE > cleanupSize * 1024L)
needCleanup = true;
UnlockReleaseBuffer(metabuffer);
END_CRIT_SECTION();
/*
* Since it could contend with concurrent cleanup process we cleanup
* pending list not forcibly.
*/
if (needCleanup)
ginInsertCleanup(ginstate, false, true, false, NULL);
}