/*
* Initialize metapage for bloom index.
*/
void
BloomInitMetapage(Relation index)
{
Page metaPage;
Buffer metaBuffer;
BloomMetaPageData *metadata;
GenericXLogState *state;
/*
* Make a new buffer, since it first buffer it should be associated with
* block number 0 (BLOOM_METAPAGE_BLKNO).
*/
metaBuffer = BloomNewBuffer(index);
Assert(BufferGetBlockNumber(metaBuffer) == BLOOM_METAPAGE_BLKNO);
/* Initialize bloom index options */
if (!index->rd_options)
index->rd_options = palloc0(sizeof(BloomOptions));
adjustBloomOptions((BloomOptions *) index->rd_options);
/* Initialize contents of meta page */
state = GenericXLogStart(index);
metaPage = GenericXLogRegisterBuffer(state, metaBuffer, GENERIC_XLOG_FULL_IMAGE);
BloomInitPage(metaPage, BLOOM_META);
metadata = BloomPageGetMeta(metaPage);
memset(metadata, 0, sizeof(BloomMetaPageData));
metadata->magickNumber = BLOOM_MAGICK_NUMBER;
metadata->opts = *((BloomOptions *) index->rd_options);
((PageHeader) metaPage)->pd_lower += sizeof(BloomMetaPageData);
GenericXLogFinish(state);
UnlockReleaseBuffer(metaBuffer);
}
/*
* Fill in metapage for bloom index.
*/
void
BloomFillMetapage(Relation index, Page metaPage)
{
BloomOptions *opts;
BloomMetaPageData *metadata;
/*
* Choose the index's options. If reloptions have been assigned, use
* those, otherwise create default options.
*/
opts = (BloomOptions *) index->rd_options;
if (!opts)
opts = makeDefaultBloomOptions();
/*
* Initialize contents of meta page, including a copy of the options,
* which are now frozen for the life of the index.
*/
BloomInitPage(metaPage, BLOOM_META);
metadata = BloomPageGetMeta(metaPage);
memset(metadata, 0, sizeof(BloomMetaPageData));
metadata->magickNumber = BLOOM_MAGICK_NUMBER;
metadata->opts = *opts;
((PageHeader) metaPage)->pd_lower += sizeof(BloomMetaPageData);
/* If this fails, probably FreeBlockNumberArray size calc is wrong: */
Assert(((PageHeader) metaPage)->pd_lower <= ((PageHeader) metaPage)->pd_upper);
}
void
BloomInitMetabuffer(Buffer b, Relation index)
{
BloomMetaPageData *metadata;
Page page = BufferGetPage(b);
BloomInitPage(page, BLOOM_META, BufferGetPageSize(b));
metadata = BloomPageGetMeta(page);
memset(metadata, 0, sizeof(BloomMetaPageData));
metadata->magickNumber = BLOOM_MAGICK_NUMBER;
metadata->opts = *makeDefaultBloomOptions((BloomOptions*)index->rd_options);
}
/*
* Fill BloomState structure for particular index.
*/
void
initBloomState(BloomState *state, Relation index)
{
int i;
state->nColumns = index->rd_att->natts;
/* Initialize hash function for each attribute */
for (i = 0; i < index->rd_att->natts; i++)
{
fmgr_info_copy(&(state->hashFn[i]),
index_getprocinfo(index, i + 1, BLOOM_HASH_PROC),
CurrentMemoryContext);
state->collations[i] = index->rd_indcollation[i];
}
/* Initialize amcache if needed with options from metapage */
if (!index->rd_amcache)
{
Buffer buffer;
Page page;
BloomMetaPageData *meta;
BloomOptions *opts;
opts = MemoryContextAlloc(index->rd_indexcxt, sizeof(BloomOptions));
buffer = ReadBuffer(index, BLOOM_METAPAGE_BLKNO);
LockBuffer(buffer, BUFFER_LOCK_SHARE);
page = BufferGetPage(buffer);
if (!BloomPageIsMeta(page))
elog(ERROR, "Relation is not a bloom index");
meta = BloomPageGetMeta(BufferGetPage(buffer));
if (meta->magickNumber != BLOOM_MAGICK_NUMBER)
elog(ERROR, "Relation is not a bloom index");
*opts = meta->opts;
UnlockReleaseBuffer(buffer);
index->rd_amcache = (void *) opts;
}
memcpy(&state->opts, index->rd_amcache, sizeof(state->opts));
state->sizeOfBloomTuple = BLOOMTUPLEHDRSZ +
sizeof(BloomSignatureWord) * state->opts.bloomLength;
}
void
initBloomState(BloomState *state, Relation index)
{
int i;
state->nColumns = index->rd_att->natts;
for (i = 0; i < index->rd_att->natts; i++)
{
fmgr_info_copy(&(state->hashFn[i]),
index_getprocinfo(index, i + 1, BLOOM_HASH_PROC),
CurrentMemoryContext);
}
if (!index->rd_amcache)
{
Buffer buffer;
BloomMetaPageData *meta;
BloomOptions *opts;
opts = MemoryContextAlloc(index->rd_indexcxt, sizeof(BloomOptions));
buffer = ReadBuffer(index, BLOOM_METAPAGE_BLKNO);
LockBuffer(buffer, BUFFER_LOCK_SHARE);
if (!BloomPageIsMeta(BufferGetPage(buffer)))
elog(ERROR,"Relation is not a bloom index");
meta = BloomPageGetMeta(BufferGetPage(buffer));
if (meta->magickNumber != BLOOM_MAGICK_NUMBER)
elog(ERROR,"Relation is not a bloom index");
*opts = meta->opts;
UnlockReleaseBuffer(buffer);
index->rd_amcache = (void*)opts;
}
state->opts = (BloomOptions*)index->rd_amcache;
state->sizeOfBloomTuple = BLOOMTUPLEHDRSZ + sizeof(SignType) * state->opts->bloomLength;
}
/*
* 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.
*/
IndexBulkDeleteResult *
blbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
IndexBulkDeleteCallback callback, void *callback_state)
{
Relation index = info->index;
BlockNumber blkno,
npages;
FreeBlockNumberArray notFullPage;
int countPage = 0;
BloomState state;
Buffer buffer;
Page page;
GenericXLogState *gxlogState;
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
initBloomState(&state, index);
/*
* Interate over the pages. We don't care about concurrently added pages,
* they can't contain tuples to delete.
*/
npages = RelationGetNumberOfBlocks(index);
for (blkno = BLOOM_HEAD_BLKNO; blkno < npages; blkno++)
{
BloomTuple *itup,
*itupPtr,
*itupEnd;
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
gxlogState = GenericXLogStart(index);
page = GenericXLogRegisterBuffer(gxlogState, buffer, 0);
if (BloomPageIsDeleted(page))
{
UnlockReleaseBuffer(buffer);
GenericXLogAbort(gxlogState);
CHECK_FOR_INTERRUPTS();
continue;
}
/* Iterate over the tuples */
itup = itupPtr = BloomPageGetTuple(&state, page, FirstOffsetNumber);
itupEnd = BloomPageGetTuple(&state, page,
OffsetNumberNext(BloomPageGetMaxOffset(page)));
while (itup < itupEnd)
{
/* Do we have to delete this tuple? */
if (callback(&itup->heapPtr, callback_state))
{
stats->tuples_removed += 1;
BloomPageGetOpaque(page)->maxoff--;
}
else
{
if (itupPtr != itup)
{
/*
* If we already delete something before, we have to move
* this tuple backward.
*/
memmove((Pointer) itupPtr, (Pointer) itup,
state.sizeOfBloomTuple);
}
stats->num_index_tuples++;
itupPtr = BloomPageGetNextTuple(&state, itupPtr);
}
itup = BloomPageGetNextTuple(&state, itup);
}
Assert(itupPtr == BloomPageGetTuple(&state, page,
OffsetNumberNext(BloomPageGetMaxOffset(page))));
/*
* Add page to notFullPage list if we will not mark page as deleted and
* there is a free space on it
*/
if (BloomPageGetMaxOffset(page) != 0 &&
BloomPageGetFreeSpace(&state, page) > state.sizeOfBloomTuple &&
countPage < BloomMetaBlockN)
notFullPage[countPage++] = blkno;
/* Did we delete something? */
if (itupPtr != itup)
{
/* Is it empty page now? */
if (BloomPageGetMaxOffset(page) == 0)
BloomPageSetDeleted(page);
/* Adjust pg_lower */
((PageHeader) page)->pd_lower = (Pointer) itupPtr - page;
/* Finish WAL-logging */
GenericXLogFinish(gxlogState);
}
else
{
/* Didn't change anything: abort WAL-logging */
GenericXLogAbort(gxlogState);
}
UnlockReleaseBuffer(buffer);
CHECK_FOR_INTERRUPTS();
}
if (countPage > 0)
{
BloomMetaPageData *metaData;
buffer = ReadBuffer(index, BLOOM_METAPAGE_BLKNO);
LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
gxlogState = GenericXLogStart(index);
page = GenericXLogRegisterBuffer(gxlogState, buffer, 0);
metaData = BloomPageGetMeta(page);
memcpy(metaData->notFullPage, notFullPage, sizeof(BlockNumber) * countPage);
metaData->nStart = 0;
metaData->nEnd = countPage;
GenericXLogFinish(gxlogState);
UnlockReleaseBuffer(buffer);
}
return stats;
}
/*
* Insert new tuple to the bloom index.
*/
bool
blinsert(Relation index, Datum *values, bool *isnull,
ItemPointer ht_ctid, Relation heapRel,
IndexUniqueCheck checkUnique,
IndexInfo *indexInfo)
{
BloomState blstate;
BloomTuple *itup;
MemoryContext oldCtx;
MemoryContext insertCtx;
BloomMetaPageData *metaData;
Buffer buffer,
metaBuffer;
Page page,
metaPage;
BlockNumber blkno = InvalidBlockNumber;
OffsetNumber nStart;
GenericXLogState *state;
insertCtx = AllocSetContextCreate(CurrentMemoryContext,
"Bloom insert temporary context",
ALLOCSET_DEFAULT_SIZES);
oldCtx = MemoryContextSwitchTo(insertCtx);
initBloomState(&blstate, index);
itup = BloomFormTuple(&blstate, ht_ctid, values, isnull);
/*
* At first, try to insert new tuple to the first page in notFullPage
* array. If successful, we don't need to modify the meta page.
*/
metaBuffer = ReadBuffer(index, BLOOM_METAPAGE_BLKNO);
LockBuffer(metaBuffer, BUFFER_LOCK_SHARE);
metaData = BloomPageGetMeta(BufferGetPage(metaBuffer));
if (metaData->nEnd > metaData->nStart)
{
Page page;
blkno = metaData->notFullPage[metaData->nStart];
Assert(blkno != InvalidBlockNumber);
/* Don't hold metabuffer lock while doing insert */
LockBuffer(metaBuffer, BUFFER_LOCK_UNLOCK);
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
state = GenericXLogStart(index);
page = GenericXLogRegisterBuffer(state, buffer, 0);
/*
* We might have found a page that was recently deleted by VACUUM. If
* so, we can reuse it, but we must reinitialize it.
*/
if (PageIsNew(page) || BloomPageIsDeleted(page))
BloomInitPage(page, 0);
if (BloomPageAddItem(&blstate, page, itup))
{
/* Success! Apply the change, clean up, and exit */
GenericXLogFinish(state);
UnlockReleaseBuffer(buffer);
ReleaseBuffer(metaBuffer);
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(insertCtx);
return false;
}
/* Didn't fit, must try other pages */
GenericXLogAbort(state);
UnlockReleaseBuffer(buffer);
}
else
{
/* No entries in notFullPage */
LockBuffer(metaBuffer, BUFFER_LOCK_UNLOCK);
}
/*
* Try other pages in notFullPage array. We will have to change nStart in
* metapage. Thus, grab exclusive lock on metapage.
*/
LockBuffer(metaBuffer, BUFFER_LOCK_EXCLUSIVE);
/* nStart might have changed while we didn't have lock */
nStart = metaData->nStart;
/* Skip first page if we already tried it above */
if (nStart < metaData->nEnd &&
blkno == metaData->notFullPage[nStart])
nStart++;
/*
* This loop iterates for each page we try from the notFullPage array, and
* will also initialize a GenericXLogState for the fallback case of having
* to allocate a new page.
*/
for (;;)
{
state = GenericXLogStart(index);
/* get modifiable copy of metapage */
metaPage = GenericXLogRegisterBuffer(state, metaBuffer, 0);
metaData = BloomPageGetMeta(metaPage);
if (nStart >= metaData->nEnd)
break; /* no more entries in notFullPage array */
blkno = metaData->notFullPage[nStart];
Assert(blkno != InvalidBlockNumber);
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
page = GenericXLogRegisterBuffer(state, buffer, 0);
/* Basically same logic as above */
if (PageIsNew(page) || BloomPageIsDeleted(page))
BloomInitPage(page, 0);
if (BloomPageAddItem(&blstate, page, itup))
{
/* Success! Apply the changes, clean up, and exit */
metaData->nStart = nStart;
GenericXLogFinish(state);
UnlockReleaseBuffer(buffer);
UnlockReleaseBuffer(metaBuffer);
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(insertCtx);
return false;
}
/* Didn't fit, must try other pages */
GenericXLogAbort(state);
UnlockReleaseBuffer(buffer);
nStart++;
}
/*
* Didn't find place to insert in notFullPage array. Allocate new page.
* (XXX is it good to do this while holding ex-lock on the metapage??)
*/
buffer = BloomNewBuffer(index);
page = GenericXLogRegisterBuffer(state, buffer, GENERIC_XLOG_FULL_IMAGE);
BloomInitPage(page, 0);
if (!BloomPageAddItem(&blstate, page, itup))
{
/* We shouldn't be here since we're inserting to an empty page */
elog(ERROR, "could not add new bloom tuple to empty page");
}
/* Reset notFullPage array to contain just this new page */
metaData->nStart = 0;
metaData->nEnd = 1;
metaData->notFullPage[0] = BufferGetBlockNumber(buffer);
/* Apply the changes, clean up, and exit */
GenericXLogFinish(state);
UnlockReleaseBuffer(buffer);
UnlockReleaseBuffer(metaBuffer);
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(insertCtx);
return false;
}
/*
* 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.
*/
IndexBulkDeleteResult *
blbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
IndexBulkDeleteCallback callback, void *callback_state)
{
Relation index = info->index;
BlockNumber blkno,
npages;
FreeBlockNumberArray notFullPage;
int countPage = 0;
BloomState state;
Buffer buffer;
Page page;
BloomMetaPageData *metaData;
GenericXLogState *gxlogState;
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
initBloomState(&state, index);
/*
* Interate over the pages. We don't care about concurrently added pages,
* they can't contain tuples to delete.
*/
npages = RelationGetNumberOfBlocks(index);
for (blkno = BLOOM_HEAD_BLKNO; blkno < npages; blkno++)
{
BloomTuple *itup,
*itupPtr,
*itupEnd;
vacuum_delay_point();
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
gxlogState = GenericXLogStart(index);
page = GenericXLogRegisterBuffer(gxlogState, buffer, 0);
/* Ignore empty/deleted pages until blvacuumcleanup() */
if (PageIsNew(page) || BloomPageIsDeleted(page))
{
UnlockReleaseBuffer(buffer);
GenericXLogAbort(gxlogState);
continue;
}
/*
* Iterate over the tuples. itup points to current tuple being
* scanned, itupPtr points to where to save next non-deleted tuple.
*/
itup = itupPtr = BloomPageGetTuple(&state, page, FirstOffsetNumber);
itupEnd = BloomPageGetTuple(&state, page,
OffsetNumberNext(BloomPageGetMaxOffset(page)));
while (itup < itupEnd)
{
/* Do we have to delete this tuple? */
if (callback(&itup->heapPtr, callback_state))
{
/* Yes; adjust count of tuples that will be left on page */
BloomPageGetOpaque(page)->maxoff--;
stats->tuples_removed += 1;
}
else
{
/* No; copy it to itupPtr++, but skip copy if not needed */
if (itupPtr != itup)
memmove((Pointer) itupPtr, (Pointer) itup,
state.sizeOfBloomTuple);
itupPtr = BloomPageGetNextTuple(&state, itupPtr);
}
itup = BloomPageGetNextTuple(&state, itup);
}
/* Assert that we counted correctly */
Assert(itupPtr == BloomPageGetTuple(&state, page,
OffsetNumberNext(BloomPageGetMaxOffset(page))));
/*
* Add page to new notFullPage list if we will not mark page as
* deleted and there is free space on it
*/
if (BloomPageGetMaxOffset(page) != 0 &&
BloomPageGetFreeSpace(&state, page) >= state.sizeOfBloomTuple &&
countPage < BloomMetaBlockN)
notFullPage[countPage++] = blkno;
/* Did we delete something? */
if (itupPtr != itup)
{
/* Is it empty page now? */
if (BloomPageGetMaxOffset(page) == 0)
BloomPageSetDeleted(page);
/* Adjust pg_lower */
((PageHeader) page)->pd_lower = (Pointer) itupPtr - page;
/* Finish WAL-logging */
GenericXLogFinish(gxlogState);
}
else
{
/* Didn't change anything: abort WAL-logging */
GenericXLogAbort(gxlogState);
}
UnlockReleaseBuffer(buffer);
}
/*
* Update the metapage's notFullPage list with whatever we found. Our
* info could already be out of date at this point, but blinsert() will
* cope if so.
*/
buffer = ReadBuffer(index, BLOOM_METAPAGE_BLKNO);
LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
gxlogState = GenericXLogStart(index);
page = GenericXLogRegisterBuffer(gxlogState, buffer, 0);
metaData = BloomPageGetMeta(page);
memcpy(metaData->notFullPage, notFullPage, sizeof(BlockNumber) * countPage);
metaData->nStart = 0;
metaData->nEnd = countPage;
GenericXLogFinish(gxlogState);
UnlockReleaseBuffer(buffer);
return stats;
}