static void
ginBuildCallback(Relation index, HeapTuple htup, Datum *values,
bool *isnull, bool tupleIsAlive, void *state)
{
GinBuildState *buildstate = (GinBuildState *) state;
MemoryContext oldCtx;
int i;
oldCtx = MemoryContextSwitchTo(buildstate->tmpCtx);
for (i = 0; i < buildstate->ginstate.origTupdesc->natts; i++)
ginHeapTupleBulkInsert(buildstate, (OffsetNumber) (i + 1),
values[i], isnull[i],
&htup->t_self);
/* If we've maxed out our available memory, dump everything to the index */
if (buildstate->accum.allocatedMemory >= (Size) maintenance_work_mem * 1024L)
{
ItemPointerData *list;
Datum key;
GinNullCategory category;
uint32 nlist;
OffsetNumber attnum;
ginBeginBAScan(&buildstate->accum);
while ((list = ginGetBAEntry(&buildstate->accum,
&attnum, &key, &category, &nlist)) != NULL)
{
/* there could be many entries, so be willing to abort here */
CHECK_FOR_INTERRUPTS();
ginEntryInsert(&buildstate->ginstate, attnum, key, category,
list, nlist, &buildstate->buildStats);
}
MemoryContextReset(buildstate->tmpCtx);
ginInitBA(&buildstate->accum);
}
MemoryContextSwitchTo(oldCtx);
}
IndexBuildResult *
ginbuild(Relation heap, Relation index, IndexInfo *indexInfo)
{
IndexBuildResult *result;
double reltuples;
GinBuildState buildstate;
Buffer RootBuffer,
MetaBuffer;
ItemPointerData *list;
Datum key;
GinNullCategory category;
uint32 nlist;
MemoryContext oldCtx;
OffsetNumber attnum;
if (RelationGetNumberOfBlocks(index) != 0)
elog(ERROR, "index \"%s\" already contains data",
RelationGetRelationName(index));
initGinState(&buildstate.ginstate, index);
buildstate.indtuples = 0;
memset(&buildstate.buildStats, 0, sizeof(GinStatsData));
/* initialize the meta page */
MetaBuffer = GinNewBuffer(index);
/* initialize the root page */
RootBuffer = GinNewBuffer(index);
START_CRIT_SECTION();
GinInitMetabuffer(MetaBuffer);
MarkBufferDirty(MetaBuffer);
GinInitBuffer(RootBuffer, GIN_LEAF);
MarkBufferDirty(RootBuffer);
if (RelationNeedsWAL(index))
{
XLogRecPtr recptr;
Page page;
XLogBeginInsert();
XLogRegisterBuffer(0, MetaBuffer, REGBUF_WILL_INIT | REGBUF_STANDARD);
XLogRegisterBuffer(1, RootBuffer, REGBUF_WILL_INIT);
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_CREATE_INDEX);
page = BufferGetPage(RootBuffer);
PageSetLSN(page, recptr);
page = BufferGetPage(MetaBuffer);
PageSetLSN(page, recptr);
}
UnlockReleaseBuffer(MetaBuffer);
UnlockReleaseBuffer(RootBuffer);
END_CRIT_SECTION();
/* count the root as first entry page */
buildstate.buildStats.nEntryPages++;
/*
* create a temporary memory context that is used to hold data not yet
* dumped out to the index
*/
buildstate.tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
"Gin build temporary context",
ALLOCSET_DEFAULT_SIZES);
/*
* create a temporary memory context that is used for calling
* ginExtractEntries(), and can be reset after each tuple
*/
buildstate.funcCtx = AllocSetContextCreate(CurrentMemoryContext,
"Gin build temporary context for user-defined function",
ALLOCSET_DEFAULT_SIZES);
buildstate.accum.ginstate = &buildstate.ginstate;
ginInitBA(&buildstate.accum);
/*
* Do the heap scan. We disallow sync scan here because dataPlaceToPage
* prefers to receive tuples in TID order.
*/
reltuples = IndexBuildHeapScan(heap, index, indexInfo, false,
ginBuildCallback, (void *) &buildstate, NULL);
/* dump remaining entries to the index */
oldCtx = MemoryContextSwitchTo(buildstate.tmpCtx);
ginBeginBAScan(&buildstate.accum);
while ((list = ginGetBAEntry(&buildstate.accum,
&attnum, &key, &category, &nlist)) != NULL)
{
/* there could be many entries, so be willing to abort here */
CHECK_FOR_INTERRUPTS();
ginEntryInsert(&buildstate.ginstate, attnum, key, category,
list, nlist, &buildstate.buildStats);
}
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(buildstate.funcCtx);
MemoryContextDelete(buildstate.tmpCtx);
/*
* Update metapage stats
*/
buildstate.buildStats.nTotalPages = RelationGetNumberOfBlocks(index);
ginUpdateStats(index, &buildstate.buildStats);
/*
* Return statistics
*/
result = (IndexBuildResult *) palloc(sizeof(IndexBuildResult));
result->heap_tuples = reltuples;
result->index_tuples = buildstate.indtuples;
return result;
}
/*
* Move tuples from pending pages into regular GIN structure.
*
* This can be called concurrently by multiple backends, so it must cope.
* On first glance it looks completely not concurrent-safe and not crash-safe
* either. The reason it's okay is that multiple insertion of the same entry
* is detected and treated as a no-op by gininsert.c. If we crash after
* posting entries to the main index and before removing them from the
* pending list, it's okay because when we redo the posting later on, nothing
* bad will happen. Likewise, if two backends simultaneously try to post
* a pending entry into the main index, one will succeed and one will do
* nothing. We try to notice when someone else is a little bit ahead of
* us in the process, but that's just to avoid wasting cycles. Only the
* action of removing a page from the pending list really needs exclusive
* lock.
*
* vac_delay indicates that ginInsertCleanup is called from vacuum process,
* so call vacuum_delay_point() periodically.
* If stats isn't null, we count deleted pending pages into the counts.
*/
void
ginInsertCleanup(GinState *ginstate,
bool vac_delay, IndexBulkDeleteResult *stats)
{
Relation index = ginstate->index;
Buffer metabuffer,
buffer;
Page metapage,
page;
GinMetaPageData *metadata;
MemoryContext opCtx,
oldCtx;
BuildAccumulator accum;
KeyArray datums;
BlockNumber blkno;
metabuffer = ReadBuffer(index, GIN_METAPAGE_BLKNO);
LockBuffer(metabuffer, GIN_SHARE);
metapage = BufferGetPage(metabuffer);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber)
{
/* Nothing to do */
UnlockReleaseBuffer(metabuffer);
return;
}
/*
* Read and lock head of pending list
*/
blkno = metadata->head;
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* Initialize. All temporary space will be in opCtx
*/
opCtx = AllocSetContextCreate(CurrentMemoryContext,
"GIN insert cleanup temporary context",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
oldCtx = MemoryContextSwitchTo(opCtx);
initKeyArray(&datums, 128);
ginInitBA(&accum);
accum.ginstate = ginstate;
/*
* At the top of this loop, we have pin and lock on the current page of
* the pending list. However, we'll release that before exiting the loop.
* Note we also have pin but not lock on the metapage.
*/
for (;;)
{
if (GinPageIsDeleted(page))
{
/* another cleanup process is running concurrently */
UnlockReleaseBuffer(buffer);
break;
}
/*
* read page's datums into accum
*/
processPendingPage(&accum, &datums, page, FirstOffsetNumber);
if (vac_delay)
vacuum_delay_point();
/*
* Is it time to flush memory to disk? Flush if we are at the end of
* the pending list, or if we have a full row and memory is getting
* full.
*
* XXX using up maintenance_work_mem here is probably unreasonably
* much, since vacuum might already be using that much.
*/
if (GinPageGetOpaque(page)->rightlink == InvalidBlockNumber ||
(GinPageHasFullRow(page) &&
(accum.allocatedMemory >= maintenance_work_mem * 1024L)))
{
ItemPointerData *list;
uint32 nlist;
Datum key;
GinNullCategory category;
OffsetNumber maxoff,
attnum;
/*
* Unlock current page to increase performance. Changes of page
* will be checked later by comparing maxoff after completion of
* memory flush.
*/
maxoff = PageGetMaxOffsetNumber(page);
LockBuffer(buffer, GIN_UNLOCK);
/*
* Moving collected data into regular structure can take
* significant amount of time - so, run it without locking pending
* list.
*/
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
{
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
if (vac_delay)
vacuum_delay_point();
}
/*
* Lock the whole list to remove pages
*/
LockBuffer(metabuffer, GIN_EXCLUSIVE);
LockBuffer(buffer, GIN_SHARE);
if (GinPageIsDeleted(page))
{
/* another cleanup process is running concurrently */
UnlockReleaseBuffer(buffer);
LockBuffer(metabuffer, GIN_UNLOCK);
break;
}
/*
* While we left the page unlocked, more stuff might have gotten
* added to it. If so, process those entries immediately. There
* shouldn't be very many, so we don't worry about the fact that
* we're doing this with exclusive lock. Insertion algorithm
* guarantees that inserted row(s) will not continue on next page.
* NOTE: intentionally no vacuum_delay_point in this loop.
*/
if (PageGetMaxOffsetNumber(page) != maxoff)
{
ginInitBA(&accum);
processPendingPage(&accum, &datums, page, maxoff + 1);
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
}
/*
* Remember next page - it will become the new list head
*/
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer); /* shiftList will do exclusive
* locking */
/*
* remove readed pages from pending list, at this point all
* content of readed pages is in regular structure
*/
if (shiftList(index, metabuffer, blkno, stats))
{
/* another cleanup process is running concurrently */
LockBuffer(metabuffer, GIN_UNLOCK);
break;
}
Assert(blkno == metadata->head);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* if we removed the whole pending list just exit
*/
if (blkno == InvalidBlockNumber)
break;
/*
* release memory used so far and reinit state
*/
MemoryContextReset(opCtx);
initKeyArray(&datums, datums.maxvalues);
ginInitBA(&accum);
}
else
{
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer);
}
/*
* Read next page in pending list
*/
CHECK_FOR_INTERRUPTS();
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
}
ReleaseBuffer(metabuffer);
/* Clean up temporary space */
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(opCtx);
}
Datum
ginbuild(PG_FUNCTION_ARGS)
{
Relation heap = (Relation) PG_GETARG_POINTER(0);
Relation index = (Relation) PG_GETARG_POINTER(1);
IndexInfo *indexInfo = (IndexInfo *) PG_GETARG_POINTER(2);
IndexBuildResult *result;
double reltuples;
GinBuildState buildstate;
Buffer RootBuffer,
MetaBuffer;
ItemPointerData *list;
Datum key;
GinNullCategory category;
uint32 nlist;
MemoryContext oldCtx;
OffsetNumber attnum;
if (RelationGetNumberOfBlocks(index) != 0)
elog(ERROR, "index \"%s\" already contains data",
RelationGetRelationName(index));
initGinState(&buildstate.ginstate, index);
buildstate.indtuples = 0;
memset(&buildstate.buildStats, 0, sizeof(GinStatsData));
/* initialize the meta page */
MetaBuffer = GinNewBuffer(index);
/* initialize the root page */
RootBuffer = GinNewBuffer(index);
START_CRIT_SECTION();
GinInitMetabuffer(MetaBuffer);
MarkBufferDirty(MetaBuffer);
GinInitBuffer(RootBuffer, GIN_LEAF);
MarkBufferDirty(RootBuffer);
if (RelationNeedsWAL(index))
{
XLogRecPtr recptr;
XLogRecData rdata;
Page page;
rdata.buffer = InvalidBuffer;
rdata.data = (char *) &(index->rd_node);
rdata.len = sizeof(RelFileNode);
rdata.next = NULL;
recptr = XLogInsert(RM_GIN_ID, XLOG_GIN_CREATE_INDEX, &rdata);
page = BufferGetPage(RootBuffer);
PageSetLSN(page, recptr);
page = BufferGetPage(MetaBuffer);
PageSetLSN(page, recptr);
}
UnlockReleaseBuffer(MetaBuffer);
UnlockReleaseBuffer(RootBuffer);
END_CRIT_SECTION();
/* count the root as first entry page */
buildstate.buildStats.nEntryPages++;
/*
* create a temporary memory context that is reset once for each tuple
* inserted into the index
*/
buildstate.tmpCtx = AllocSetContextCreate(CurrentMemoryContext,
"Gin build temporary context",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
buildstate.funcCtx = AllocSetContextCreate(buildstate.tmpCtx,
"Gin build temporary context for user-defined function",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
buildstate.accum.ginstate = &buildstate.ginstate;
ginInitBA(&buildstate.accum);
/*
* Do the heap scan. We disallow sync scan here because dataPlaceToPage
* prefers to receive tuples in TID order.
*/
reltuples = IndexBuildHeapScan(heap, index, indexInfo, false,
ginBuildCallback, (void *) &buildstate);
/* dump remaining entries to the index */
oldCtx = MemoryContextSwitchTo(buildstate.tmpCtx);
ginBeginBAScan(&buildstate.accum);
while ((list = ginGetBAEntry(&buildstate.accum,
&attnum, &key, &category, &nlist)) != NULL)
{
/* there could be many entries, so be willing to abort here */
CHECK_FOR_INTERRUPTS();
ginEntryInsert(&buildstate.ginstate, attnum, key, category,
list, nlist, &buildstate.buildStats);
}
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(buildstate.tmpCtx);
/*
* Update metapage stats
*/
buildstate.buildStats.nTotalPages = RelationGetNumberOfBlocks(index);
ginUpdateStats(index, &buildstate.buildStats);
/*
* Return statistics
*/
result = (IndexBuildResult *) palloc(sizeof(IndexBuildResult));
result->heap_tuples = reltuples;
result->index_tuples = buildstate.indtuples;
PG_RETURN_POINTER(result);
}
/*
* Move tuples from pending pages into regular GIN structure.
*
* On first glance it looks completely not crash-safe. But if we crash
* after posting entries to the main index and before removing them from the
* pending list, it's okay because when we redo the posting later on, nothing
* bad will happen.
*
* fill_fsm indicates that ginInsertCleanup should add deleted pages
* to FSM otherwise caller is responsible to put deleted pages into
* FSM.
*
* If stats isn't null, we count deleted pending pages into the counts.
*/
void
ginInsertCleanup(GinState *ginstate, bool full_clean,
bool fill_fsm, IndexBulkDeleteResult *stats)
{
Relation index = ginstate->index;
Buffer metabuffer,
buffer;
Page metapage,
page;
GinMetaPageData *metadata;
MemoryContext opCtx,
oldCtx;
BuildAccumulator accum;
KeyArray datums;
BlockNumber blkno,
blknoFinish;
bool cleanupFinish = false;
bool fsm_vac = false;
Size workMemory;
bool inVacuum = (stats == NULL);
/*
* We would like to prevent concurrent cleanup process. For that we will
* lock metapage in exclusive mode using LockPage() call. Nobody other
* will use that lock for metapage, so we keep possibility of concurrent
* insertion into pending list
*/
if (inVacuum)
{
/*
* We are called from [auto]vacuum/analyze or gin_clean_pending_list()
* and we would like to wait concurrent cleanup to finish.
*/
LockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
workMemory =
(IsAutoVacuumWorkerProcess() && autovacuum_work_mem != -1) ?
autovacuum_work_mem : maintenance_work_mem;
}
else
{
/*
* We are called from regular insert and if we see concurrent cleanup
* just exit in hope that concurrent process will clean up pending
* list.
*/
if (!ConditionalLockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock))
return;
workMemory = work_mem;
}
metabuffer = ReadBuffer(index, GIN_METAPAGE_BLKNO);
LockBuffer(metabuffer, GIN_SHARE);
metapage = BufferGetPage(metabuffer);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber)
{
/* Nothing to do */
UnlockReleaseBuffer(metabuffer);
UnlockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
return;
}
/*
* Remember a tail page to prevent infinite cleanup if other backends add
* new tuples faster than we can cleanup.
*/
blknoFinish = metadata->tail;
/*
* Read and lock head of pending list
*/
blkno = metadata->head;
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* Initialize. All temporary space will be in opCtx
*/
opCtx = AllocSetContextCreate(CurrentMemoryContext,
"GIN insert cleanup temporary context",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
oldCtx = MemoryContextSwitchTo(opCtx);
initKeyArray(&datums, 128);
ginInitBA(&accum);
accum.ginstate = ginstate;
/*
* At the top of this loop, we have pin and lock on the current page of
* the pending list. However, we'll release that before exiting the loop.
* Note we also have pin but not lock on the metapage.
*/
for (;;)
{
Assert(!GinPageIsDeleted(page));
/*
* Are we walk through the page which as we remember was a tail when
* we start our cleanup? But if caller asks us to clean up whole
* pending list then ignore old tail, we will work until list becomes
* empty.
*/
if (blkno == blknoFinish && full_clean == false)
cleanupFinish = true;
/*
* read page's datums into accum
*/
processPendingPage(&accum, &datums, page, FirstOffsetNumber);
vacuum_delay_point();
/*
* Is it time to flush memory to disk? Flush if we are at the end of
* the pending list, or if we have a full row and memory is getting
* full.
*/
if (GinPageGetOpaque(page)->rightlink == InvalidBlockNumber ||
(GinPageHasFullRow(page) &&
(accum.allocatedMemory >= workMemory * 1024L)))
{
ItemPointerData *list;
uint32 nlist;
Datum key;
GinNullCategory category;
OffsetNumber maxoff,
attnum;
/*
* Unlock current page to increase performance. Changes of page
* will be checked later by comparing maxoff after completion of
* memory flush.
*/
maxoff = PageGetMaxOffsetNumber(page);
LockBuffer(buffer, GIN_UNLOCK);
/*
* Moving collected data into regular structure can take
* significant amount of time - so, run it without locking pending
* list.
*/
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
{
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
vacuum_delay_point();
}
/*
* Lock the whole list to remove pages
*/
LockBuffer(metabuffer, GIN_EXCLUSIVE);
LockBuffer(buffer, GIN_SHARE);
Assert(!GinPageIsDeleted(page));
/*
* While we left the page unlocked, more stuff might have gotten
* added to it. If so, process those entries immediately. There
* shouldn't be very many, so we don't worry about the fact that
* we're doing this with exclusive lock. Insertion algorithm
* guarantees that inserted row(s) will not continue on next page.
* NOTE: intentionally no vacuum_delay_point in this loop.
*/
if (PageGetMaxOffsetNumber(page) != maxoff)
{
ginInitBA(&accum);
processPendingPage(&accum, &datums, page, maxoff + 1);
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
}
/*
* Remember next page - it will become the new list head
*/
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer); /* shiftList will do exclusive
* locking */
/*
* remove read pages from pending list, at this point all content
* of read pages is in regular structure
*/
shiftList(index, metabuffer, blkno, fill_fsm, stats);
/* At this point, some pending pages have been freed up */
fsm_vac = true;
Assert(blkno == metadata->head);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* if we removed the whole pending list or we cleanup tail (which
* we remembered on start our cleanup process) then just exit
*/
if (blkno == InvalidBlockNumber || cleanupFinish)
break;
/*
* release memory used so far and reinit state
*/
MemoryContextReset(opCtx);
initKeyArray(&datums, datums.maxvalues);
ginInitBA(&accum);
}
else
{
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer);
}
/*
* Read next page in pending list
*/
vacuum_delay_point();
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
}
UnlockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
ReleaseBuffer(metabuffer);
/*
* As pending list pages can have a high churn rate, it is desirable to
* recycle them immediately to the FreeSpace Map when ordinary backends
* clean the list.
*/
if (fsm_vac && fill_fsm)
IndexFreeSpaceMapVacuum(index);
/* Clean up temporary space */
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(opCtx);
}
