/*
* Try to find parent for current stack position. Returns correct parent and
* child's offset in stack->parent. The root page is never released, to
* to prevent conflict with vacuum process.
*/
static void
ginFindParents(GinBtree btree, GinBtreeStack *stack)
{
Page page;
Buffer buffer;
BlockNumber blkno,
leftmostBlkno;
OffsetNumber offset;
GinBtreeStack *root;
GinBtreeStack *ptr;
/*
* Unwind the stack all the way up to the root, leaving only the root
* item.
*
* Be careful not to release the pin on the root page! The pin on root
* page is required to lock out concurrent vacuums on the tree.
*/
root = stack->parent;
while (root->parent)
{
ReleaseBuffer(root->buffer);
root = root->parent;
}
Assert(root->blkno == btree->rootBlkno);
Assert(BufferGetBlockNumber(root->buffer) == btree->rootBlkno);
root->off = InvalidOffsetNumber;
blkno = root->blkno;
buffer = root->buffer;
offset = InvalidOffsetNumber;
ptr = (GinBtreeStack *) palloc(sizeof(GinBtreeStack));
for (;;)
{
LockBuffer(buffer, GIN_EXCLUSIVE);
page = BufferGetPage(buffer);
if (GinPageIsLeaf(page))
elog(ERROR, "Lost path");
if (GinPageIsIncompleteSplit(page))
{
Assert(blkno != btree->rootBlkno);
ptr->blkno = blkno;
ptr->buffer = buffer;
/*
* parent may be wrong, but if so, the ginFinishSplit call will
* recurse to call ginFindParents again to fix it.
*/
ptr->parent = root;
ptr->off = InvalidOffsetNumber;
ginFinishSplit(btree, ptr, false, NULL);
}
leftmostBlkno = btree->getLeftMostChild(btree, page);
while ((offset = btree->findChildPtr(btree, page, stack->blkno, InvalidOffsetNumber)) == InvalidOffsetNumber)
{
blkno = GinPageGetOpaque(page)->rightlink;
if (blkno == InvalidBlockNumber)
{
UnlockReleaseBuffer(buffer);
break;
}
buffer = ginStepRight(buffer, btree->index, GIN_EXCLUSIVE);
page = BufferGetPage(buffer);
/* finish any incomplete splits, as above */
if (GinPageIsIncompleteSplit(page))
{
Assert(blkno != btree->rootBlkno);
ptr->blkno = blkno;
ptr->buffer = buffer;
ptr->parent = root;
ptr->off = InvalidOffsetNumber;
ginFinishSplit(btree, ptr, false, NULL);
}
}
if (blkno != InvalidBlockNumber)
{
ptr->blkno = blkno;
ptr->buffer = buffer;
ptr->parent = root; /* it may be wrong, but in next call we will
* correct */
ptr->off = offset;
stack->parent = ptr;
return;
}
/* Descend down to next level */
blkno = leftmostBlkno;
buffer = ReadBuffer(btree->index, blkno);
}
}
/*
* Recursive guts of FreeSpaceMapVacuum
*
* Examine the FSM page indicated by addr, as well as its children, updating
* upper-level nodes that cover the heap block range from start to end-1.
* (It's okay if end is beyond the actual end of the map.)
* Return the maximum freespace value on this page.
*
* If addr is past the end of the FSM, set *eof_p to true and return 0.
*
* This traverses the tree in depth-first order. The tree is stored
* physically in depth-first order, so this should be pretty I/O efficient.
*/
static uint8
fsm_vacuum_page(Relation rel, FSMAddress addr,
BlockNumber start, BlockNumber end,
bool *eof_p)
{
Buffer buf;
Page page;
uint8 max_avail;
/* Read the page if it exists, or return EOF */
buf = fsm_readbuf(rel, addr, false);
if (!BufferIsValid(buf))
{
*eof_p = true;
return 0;
}
else
*eof_p = false;
page = BufferGetPage(buf);
/*
* If we're above the bottom level, recurse into children, and fix the
* information stored about them at this level.
*/
if (addr.level > FSM_BOTTOM_LEVEL)
{
FSMAddress fsm_start,
fsm_end;
uint16 fsm_start_slot,
fsm_end_slot;
int slot,
start_slot,
end_slot;
bool eof = false;
/*
* Compute the range of slots we need to update on this page, given
* the requested range of heap blocks to consider. The first slot to
* update is the one covering the "start" block, and the last slot is
* the one covering "end - 1". (Some of this work will be duplicated
* in each recursive call, but it's cheap enough to not worry about.)
*/
fsm_start = fsm_get_location(start, &fsm_start_slot);
fsm_end = fsm_get_location(end - 1, &fsm_end_slot);
while (fsm_start.level < addr.level)
{
fsm_start = fsm_get_parent(fsm_start, &fsm_start_slot);
fsm_end = fsm_get_parent(fsm_end, &fsm_end_slot);
}
Assert(fsm_start.level == addr.level);
if (fsm_start.logpageno == addr.logpageno)
start_slot = fsm_start_slot;
else if (fsm_start.logpageno > addr.logpageno)
start_slot = SlotsPerFSMPage; /* shouldn't get here... */
else
start_slot = 0;
if (fsm_end.logpageno == addr.logpageno)
end_slot = fsm_end_slot;
else if (fsm_end.logpageno > addr.logpageno)
end_slot = SlotsPerFSMPage - 1;
else
end_slot = -1; /* shouldn't get here... */
for (slot = start_slot; slot <= end_slot; slot++)
{
int child_avail;
CHECK_FOR_INTERRUPTS();
/* After we hit end-of-file, just clear the rest of the slots */
if (!eof)
child_avail = fsm_vacuum_page(rel, fsm_get_child(addr, slot),
start, end,
&eof);
else
child_avail = 0;
/* Update information about the child */
if (fsm_get_avail(page, slot) != child_avail)
{
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
fsm_set_avail(page, slot, child_avail);
MarkBufferDirtyHint(buf, false);
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
}
}
}
/* Now get the maximum value on the page, to return to caller */
max_avail = fsm_get_max_avail(page);
/*
* Reset the next slot pointer. This encourages the use of low-numbered
* pages, increasing the chances that a later vacuum can truncate the
* relation. We don't bother with a lock here, nor with marking the page
* dirty if it wasn't already, since this is just a hint.
*/
((FSMPage) PageGetContents(page))->fp_next_slot = 0;
ReleaseBuffer(buf);
return max_avail;
}
/*
* visibilitymap_truncate - truncate the visibility map
*
* The caller must hold AccessExclusiveLock on the relation, to ensure that
* other backends receive the smgr invalidation event that this function sends
* before they access the VM again.
*
* nheapblocks is the new size of the heap.
*/
void
visibilitymap_truncate(Relation rel, BlockNumber nheapblocks)
{
BlockNumber newnblocks;
/* last remaining block, byte, and bit */
BlockNumber truncBlock = HEAPBLK_TO_MAPBLOCK(nheapblocks);
uint32 truncByte = HEAPBLK_TO_MAPBYTE(nheapblocks);
uint8 truncBit = HEAPBLK_TO_MAPBIT(nheapblocks);
#ifdef TRACE_VISIBILITYMAP
elog(DEBUG1, "vm_truncate %s %d", RelationGetRelationName(rel), nheapblocks);
#endif
RelationOpenSmgr(rel);
/*
* If no visibility map has been created yet for this relation, there's
* nothing to truncate.
*/
if (!smgrexists(rel->rd_smgr, VISIBILITYMAP_FORKNUM))
return;
/*
* Unless the new size is exactly at a visibility map page boundary, the
* tail bits in the last remaining map page, representing truncated heap
* blocks, need to be cleared. This is not only tidy, but also necessary
* because we don't get a chance to clear the bits if the heap is extended
* again.
*/
if (truncByte != 0 || truncBit != 0)
{
Buffer mapBuffer;
Page page;
char *map;
newnblocks = truncBlock + 1;
mapBuffer = vm_readbuf(rel, truncBlock, false);
if (!BufferIsValid(mapBuffer))
{
/* nothing to do, the file was already smaller */
return;
}
page = BufferGetPage(mapBuffer);
map = PageGetContents(page);
LockBuffer(mapBuffer, BUFFER_LOCK_EXCLUSIVE);
/* Clear out the unwanted bytes. */
MemSet(&map[truncByte + 1], 0, MAPSIZE - (truncByte + 1));
/*
* Mask out the unwanted bits of the last remaining byte.
*
* ((1 << 0) - 1) = 00000000 ((1 << 1) - 1) = 00000001 ... ((1 << 6) -
* 1) = 00111111 ((1 << 7) - 1) = 01111111
*/
map[truncByte] &= (1 << truncBit) - 1;
MarkBufferDirty(mapBuffer);
UnlockReleaseBuffer(mapBuffer);
}
else
newnblocks = truncBlock;
if (smgrnblocks(rel->rd_smgr, VISIBILITYMAP_FORKNUM) <= newnblocks)
{
/* nothing to do, the file was already smaller than requested size */
return;
}
/* Truncate the unused VM pages, and send smgr inval message */
smgrtruncate(rel->rd_smgr, VISIBILITYMAP_FORKNUM, newnblocks);
/*
* We might as well update the local smgr_vm_nblocks setting. smgrtruncate
* sent an smgr cache inval message, which will cause other backends to
* invalidate their copy of smgr_vm_nblocks, and this one too at the next
* command boundary. But this ensures it isn't outright wrong until then.
*/
if (rel->rd_smgr)
rel->rd_smgr->smgr_vm_nblocks = newnblocks;
}
/*
* Helper function to perform deletion of index entries from a bucket.
*
* This function expects that the caller has acquired a cleanup lock on the
* primary bucket page, and will return with a write lock again held on the
* primary bucket page. The lock won't necessarily be held continuously,
* though, because we'll release it when visiting overflow pages.
*
* It would be very bad if this function cleaned a page while some other
* backend was in the midst of scanning it, because hashgettuple assumes
* that the next valid TID will be greater than or equal to the current
* valid TID. There can't be any concurrent scans in progress when we first
* enter this function because of the cleanup lock we hold on the primary
* bucket page, but as soon as we release that lock, there might be. We
* handle that by conspiring to prevent those scans from passing our cleanup
* scan. To do that, we lock the next page in the bucket chain before
* releasing the lock on the previous page. (This type of lock chaining is
* not ideal, so we might want to look for a better solution at some point.)
*
* We need to retain a pin on the primary bucket to ensure that no concurrent
* split can start.
*/
void
hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
BlockNumber bucket_blkno, BufferAccessStrategy bstrategy,
uint32 maxbucket, uint32 highmask, uint32 lowmask,
double *tuples_removed, double *num_index_tuples,
bool split_cleanup,
IndexBulkDeleteCallback callback, void *callback_state)
{
BlockNumber blkno;
Buffer buf;
Bucket new_bucket PG_USED_FOR_ASSERTS_ONLY = InvalidBucket;
bool bucket_dirty = false;
blkno = bucket_blkno;
buf = bucket_buf;
if (split_cleanup)
new_bucket = _hash_get_newbucket_from_oldbucket(rel, cur_bucket,
lowmask, maxbucket);
/* Scan each page in bucket */
for (;;)
{
HashPageOpaque opaque;
OffsetNumber offno;
OffsetNumber maxoffno;
Buffer next_buf;
Page page;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
bool retain_pin = false;
bool clear_dead_marking = false;
vacuum_delay_point();
page = BufferGetPage(buf);
opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
for (offno = FirstOffsetNumber;
offno <= maxoffno;
offno = OffsetNumberNext(offno))
{
ItemPointer htup;
IndexTuple itup;
Bucket bucket;
bool kill_tuple = false;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offno));
htup = &(itup->t_tid);
/*
* To remove the dead tuples, we strictly want to rely on results
* of callback function. refer btvacuumpage for detailed reason.
*/
if (callback && callback(htup, callback_state))
{
kill_tuple = true;
if (tuples_removed)
*tuples_removed += 1;
}
else if (split_cleanup)
{
/* delete the tuples that are moved by split. */
bucket = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup),
maxbucket,
highmask,
lowmask);
/* mark the item for deletion */
if (bucket != cur_bucket)
{
/*
* We expect tuples to either belong to curent bucket or
* new_bucket. This is ensured because we don't allow
* further splits from bucket that contains garbage. See
* comments in _hash_expandtable.
*/
Assert(bucket == new_bucket);
kill_tuple = true;
}
}
if (kill_tuple)
{
/* mark the item for deletion */
deletable[ndeletable++] = offno;
}
else
{
/* we're keeping it, so count it */
if (num_index_tuples)
*num_index_tuples += 1;
}
}
/* retain the pin on primary bucket page till end of bucket scan */
if (blkno == bucket_blkno)
retain_pin = true;
else
retain_pin = false;
blkno = opaque->hasho_nextblkno;
/*
* Apply deletions, advance to next page and write page if needed.
*/
if (ndeletable > 0)
{
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
PageIndexMultiDelete(page, deletable, ndeletable);
bucket_dirty = true;
/*
* Let us mark the page as clean if vacuum removes the DEAD tuples
* from an index page. We do this by clearing LH_PAGE_HAS_DEAD_TUPLES
* flag.
*/
if (tuples_removed && *tuples_removed > 0 &&
opaque->hasho_flag & LH_PAGE_HAS_DEAD_TUPLES)
{
opaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
clear_dead_marking = true;
}
MarkBufferDirty(buf);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
xl_hash_delete xlrec;
XLogRecPtr recptr;
xlrec.clear_dead_marking = clear_dead_marking;
xlrec.is_primary_bucket_page = (buf == bucket_buf) ? true : false;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHashDelete);
/*
* bucket buffer needs to be registered to ensure that we can
* acquire a cleanup lock on it during replay.
*/
if (!xlrec.is_primary_bucket_page)
XLogRegisterBuffer(0, bucket_buf, REGBUF_STANDARD | REGBUF_NO_IMAGE);
XLogRegisterBuffer(1, buf, REGBUF_STANDARD);
XLogRegisterBufData(1, (char *) deletable,
ndeletable * sizeof(OffsetNumber));
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_DELETE);
PageSetLSN(BufferGetPage(buf), recptr);
}
END_CRIT_SECTION();
}
/* bail out if there are no more pages to scan. */
if (!BlockNumberIsValid(blkno))
break;
next_buf = _hash_getbuf_with_strategy(rel, blkno, HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
/*
* release the lock on previous page after acquiring the lock on next
* page
*/
if (retain_pin)
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, buf);
buf = next_buf;
}
/*
* lock the bucket page to clear the garbage flag and squeeze the bucket.
* if the current buffer is same as bucket buffer, then we already have
* lock on bucket page.
*/
if (buf != bucket_buf)
{
_hash_relbuf(rel, buf);
LockBuffer(bucket_buf, BUFFER_LOCK_EXCLUSIVE);
}
/*
* Clear the garbage flag from bucket after deleting the tuples that are
* moved by split. We purposefully clear the flag before squeeze bucket,
* so that after restart, vacuum shouldn't again try to delete the moved
* by split tuples.
*/
if (split_cleanup)
{
HashPageOpaque bucket_opaque;
Page page;
page = BufferGetPage(bucket_buf);
bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
bucket_opaque->hasho_flag &= ~LH_BUCKET_NEEDS_SPLIT_CLEANUP;
MarkBufferDirty(bucket_buf);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
XLogRecPtr recptr;
XLogBeginInsert();
XLogRegisterBuffer(0, bucket_buf, REGBUF_STANDARD);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_SPLIT_CLEANUP);
PageSetLSN(page, recptr);
}
END_CRIT_SECTION();
}
/*
* If we have deleted anything, try to compact free space. For squeezing
* the bucket, we must have a cleanup lock, else it can impact the
* ordering of tuples for a scan that has started before it.
*/
if (bucket_dirty && IsBufferCleanupOK(bucket_buf))
_hash_squeezebucket(rel, cur_bucket, bucket_blkno, bucket_buf,
bstrategy);
else
LockBuffer(bucket_buf, BUFFER_LOCK_UNLOCK);
}
/*
* Read a FSM page.
*
* If the page doesn't exist, InvalidBuffer is returned, or if 'extend' is
* true, the FSM file is extended.
*/
static Buffer
fsm_readbuf(Relation rel, FSMAddress addr, bool extend)
{
BlockNumber blkno = fsm_logical_to_physical(addr);
Buffer buf;
RelationOpenSmgr(rel);
/*
* If we haven't cached the size of the FSM yet, check it first. Also
* recheck if the requested block seems to be past end, since our cached
* value might be stale. (We send smgr inval messages on truncation, but
* not on extension.)
*/
if (rel->rd_smgr->smgr_fsm_nblocks == InvalidBlockNumber ||
blkno >= rel->rd_smgr->smgr_fsm_nblocks)
{
if (smgrexists(rel->rd_smgr, FSM_FORKNUM))
rel->rd_smgr->smgr_fsm_nblocks = smgrnblocks(rel->rd_smgr,
FSM_FORKNUM);
else
rel->rd_smgr->smgr_fsm_nblocks = 0;
}
/* Handle requests beyond EOF */
if (blkno >= rel->rd_smgr->smgr_fsm_nblocks)
{
if (extend)
fsm_extend(rel, blkno + 1);
else
return InvalidBuffer;
}
/*
* Use ZERO_ON_ERROR mode, and initialize the page if necessary. The FSM
* information is not accurate anyway, so it's better to clear corrupt
* pages than error out. Since the FSM changes are not WAL-logged, the
* so-called torn page problem on crash can lead to pages with corrupt
* headers, for example.
*
* The initialize-the-page part is trickier than it looks, because of the
* possibility of multiple backends doing this concurrently, and our
* desire to not uselessly take the buffer lock in the normal path where
* the page is OK. We must take the lock to initialize the page, so
* recheck page newness after we have the lock, in case someone else
* already did it. Also, because we initially check PageIsNew with no
* lock, it's possible to fall through and return the buffer while someone
* else is still initializing the page (i.e., we might see pd_upper as set
* but other page header fields are still zeroes). This is harmless for
* callers that will take a buffer lock themselves, but some callers
* inspect the page without any lock at all. The latter is OK only so
* long as it doesn't depend on the page header having correct contents.
* Current usage is safe because PageGetContents() does not require that.
*/
buf = ReadBufferExtended(rel, FSM_FORKNUM, blkno, RBM_ZERO_ON_ERROR, NULL);
if (PageIsNew(BufferGetPage(buf)))
{
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
if (PageIsNew(BufferGetPage(buf)))
PageInit(BufferGetPage(buf), BLCKSZ, 0);
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
}
return buf;
}
/*
* Insert all matching tuples into a bitmap.
*/
int64
blgetbitmap(IndexScanDesc scan, TIDBitmap *tbm)
{
int64 ntids = 0;
BlockNumber blkno = BLOOM_HEAD_BLKNO,
npages;
int i;
BufferAccessStrategy bas;
BloomScanOpaque so = (BloomScanOpaque) scan->opaque;
if (so->sign == NULL)
{
/* New search: have to calculate search signature */
ScanKey skey = scan->keyData;
so->sign = palloc0(sizeof(BloomSignatureWord) * so->state.opts.bloomLength);
for (i = 0; i < scan->numberOfKeys; i++)
{
/*
* Assume bloom-indexable operators to be strict, so nothing could
* be found for NULL key.
*/
if (skey->sk_flags & SK_ISNULL)
{
pfree(so->sign);
so->sign = NULL;
return 0;
}
/* Add next value to the signature */
signValue(&so->state, so->sign, skey->sk_argument,
skey->sk_attno - 1);
skey++;
}
}
/*
* We're going to read the whole index. This is why we use appropriate
* buffer access strategy.
*/
bas = GetAccessStrategy(BAS_BULKREAD);
npages = RelationGetNumberOfBlocks(scan->indexRelation);
for (blkno = BLOOM_HEAD_BLKNO; blkno < npages; blkno++)
{
Buffer buffer;
Page page;
buffer = ReadBufferExtended(scan->indexRelation, MAIN_FORKNUM,
blkno, RBM_NORMAL, bas);
LockBuffer(buffer, BUFFER_LOCK_SHARE);
page = BufferGetPage(buffer);
TestForOldSnapshot(scan->xs_snapshot, scan->indexRelation, page);
if (!PageIsNew(page) && !BloomPageIsDeleted(page))
{
OffsetNumber offset,
maxOffset = BloomPageGetMaxOffset(page);
for (offset = 1; offset <= maxOffset; offset++)
{
BloomTuple *itup = BloomPageGetTuple(&so->state, page, offset);
bool res = true;
/* Check index signature with scan signature */
for (i = 0; i < so->state.opts.bloomLength; i++)
{
if ((itup->sign[i] & so->sign[i]) != so->sign[i])
{
res = false;
break;
}
}
/* Add matching tuples to bitmap */
if (res)
{
tbm_add_tuples(tbm, &itup->heapPtr, 1, true);
ntids++;
}
}
}
UnlockReleaseBuffer(buffer);
CHECK_FOR_INTERRUPTS();
}
FreeAccessStrategy(bas);
return ntids;
}
/*
* hashgettuple() -- Get the next tuple in the scan.
*/
bool
hashgettuple(IndexScanDesc scan, ScanDirection dir)
{
HashScanOpaque so = (HashScanOpaque) scan->opaque;
Relation rel = scan->indexRelation;
Buffer buf;
Page page;
OffsetNumber offnum;
ItemPointer current;
bool res;
/* Hash indexes are always lossy since we store only the hash code */
scan->xs_recheck = true;
/*
* We hold pin but not lock on current buffer while outside the hash AM.
* Reacquire the read lock here.
*/
if (BufferIsValid(so->hashso_curbuf))
LockBuffer(so->hashso_curbuf, BUFFER_LOCK_SHARE);
/*
* If we've already initialized this scan, we can just advance it in the
* appropriate direction. If we haven't done so yet, we call a routine to
* get the first item in the scan.
*/
current = &(so->hashso_curpos);
if (ItemPointerIsValid(current))
{
/*
* An insertion into the current index page could have happened while
* we didn't have read lock on it. Re-find our position by looking
* for the TID we previously returned. (Because we hold a pin on the
* primary bucket page, no deletions or splits could have occurred;
* therefore we can expect that the TID still exists in the current
* index page, at an offset >= where we were.)
*/
OffsetNumber maxoffnum;
buf = so->hashso_curbuf;
Assert(BufferIsValid(buf));
page = BufferGetPage(buf);
/*
* We don't need test for old snapshot here as the current buffer is
* pinned, so vacuum can't clean the page.
*/
maxoffnum = PageGetMaxOffsetNumber(page);
for (offnum = ItemPointerGetOffsetNumber(current);
offnum <= maxoffnum;
offnum = OffsetNumberNext(offnum))
{
IndexTuple itup;
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, offnum));
if (ItemPointerEquals(&(so->hashso_heappos), &(itup->t_tid)))
break;
}
if (offnum > maxoffnum)
elog(ERROR, "failed to re-find scan position within index \"%s\"",
RelationGetRelationName(rel));
ItemPointerSetOffsetNumber(current, offnum);
/*
* Check to see if we should kill the previously-fetched tuple.
*/
if (scan->kill_prior_tuple)
{
/*
* Yes, so remember it for later. (We'll deal with all such
* tuples at once right after leaving the index page or at
* end of scan.) In case if caller reverses the indexscan
* direction it is quite possible that the same item might
* get entered multiple times. But, we don't detect that;
* instead, we just forget any excess entries.
*/
if (so->killedItems == NULL)
so->killedItems = palloc(MaxIndexTuplesPerPage *
sizeof(HashScanPosItem));
if (so->numKilled < MaxIndexTuplesPerPage)
{
so->killedItems[so->numKilled].heapTid = so->hashso_heappos;
so->killedItems[so->numKilled].indexOffset =
ItemPointerGetOffsetNumber(&(so->hashso_curpos));
so->numKilled++;
}
}
/*
* Now continue the scan.
*/
res = _hash_next(scan, dir);
}
else
res = _hash_first(scan, dir);
/*
* Skip killed tuples if asked to.
*/
if (scan->ignore_killed_tuples)
{
while (res)
{
offnum = ItemPointerGetOffsetNumber(current);
page = BufferGetPage(so->hashso_curbuf);
if (!ItemIdIsDead(PageGetItemId(page, offnum)))
break;
res = _hash_next(scan, dir);
}
}
/* Release read lock on current buffer, but keep it pinned */
if (BufferIsValid(so->hashso_curbuf))
LockBuffer(so->hashso_curbuf, BUFFER_LOCK_UNLOCK);
/* Return current heap TID on success */
scan->xs_ctup.t_self = so->hashso_heappos;
return res;
}
/*
* _hash_init() -- Initialize the metadata page of a hash index,
* the initial buckets, and the initial bitmap page.
*
* The initial number of buckets is dependent on num_tuples, an estimate
* of the number of tuples to be loaded into the index initially. The
* chosen number of buckets is returned.
*
* We are fairly cavalier about locking here, since we know that no one else
* could be accessing this index. In particular the rule about not holding
* multiple buffer locks is ignored.
*/
uint32
_hash_init(Relation rel, double num_tuples, ForkNumber forkNum)
{
Buffer metabuf;
Buffer buf;
Buffer bitmapbuf;
Page pg;
HashMetaPage metap;
RegProcedure procid;
int32 data_width;
int32 item_width;
int32 ffactor;
uint32 num_buckets;
uint32 i;
bool use_wal;
/* safety check */
if (RelationGetNumberOfBlocksInFork(rel, forkNum) != 0)
elog(ERROR, "cannot initialize non-empty hash index \"%s\"",
RelationGetRelationName(rel));
/*
* WAL log creation of pages if the relation is persistent, or this is the
* init fork. Init forks for unlogged relations always need to be WAL
* logged.
*/
use_wal = RelationNeedsWAL(rel) || forkNum == INIT_FORKNUM;
/*
* Determine the target fill factor (in tuples per bucket) for this index.
* The idea is to make the fill factor correspond to pages about as full
* as the user-settable fillfactor parameter says. We can compute it
* exactly since the index datatype (i.e. uint32 hash key) is fixed-width.
*/
data_width = sizeof(uint32);
item_width = MAXALIGN(sizeof(IndexTupleData)) + MAXALIGN(data_width) +
sizeof(ItemIdData); /* include the line pointer */
ffactor = RelationGetTargetPageUsage(rel, HASH_DEFAULT_FILLFACTOR) / item_width;
/* keep to a sane range */
if (ffactor < 10)
ffactor = 10;
procid = index_getprocid(rel, 1, HASHSTANDARD_PROC);
/*
* We initialize the metapage, the first N bucket pages, and the first
* bitmap page in sequence, using _hash_getnewbuf to cause smgrextend()
* calls to occur. This ensures that the smgr level has the right idea of
* the physical index length.
*
* Critical section not required, because on error the creation of the
* whole relation will be rolled back.
*/
metabuf = _hash_getnewbuf(rel, HASH_METAPAGE, forkNum);
_hash_init_metabuffer(metabuf, num_tuples, procid, ffactor, false);
MarkBufferDirty(metabuf);
pg = BufferGetPage(metabuf);
metap = HashPageGetMeta(pg);
/* XLOG stuff */
if (use_wal)
{
xl_hash_init_meta_page xlrec;
XLogRecPtr recptr;
xlrec.num_tuples = num_tuples;
xlrec.procid = metap->hashm_procid;
xlrec.ffactor = metap->hashm_ffactor;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHashInitMetaPage);
XLogRegisterBuffer(0, metabuf, REGBUF_WILL_INIT);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_INIT_META_PAGE);
PageSetLSN(BufferGetPage(metabuf), recptr);
}
num_buckets = metap->hashm_maxbucket + 1;
/*
* Release buffer lock on the metapage while we initialize buckets.
* Otherwise, we'll be in interrupt holdoff and the CHECK_FOR_INTERRUPTS
* won't accomplish anything. It's a bad idea to hold buffer locks for
* long intervals in any case, since that can block the bgwriter.
*/
LockBuffer(metabuf, BUFFER_LOCK_UNLOCK);
/*
* Initialize and WAL Log the first N buckets
*/
for (i = 0; i < num_buckets; i++)
{
BlockNumber blkno;
/* Allow interrupts, in case N is huge */
CHECK_FOR_INTERRUPTS();
blkno = BUCKET_TO_BLKNO(metap, i);
buf = _hash_getnewbuf(rel, blkno, forkNum);
_hash_initbuf(buf, metap->hashm_maxbucket, i, LH_BUCKET_PAGE, false);
MarkBufferDirty(buf);
if (use_wal)
log_newpage(&rel->rd_node,
forkNum,
blkno,
BufferGetPage(buf),
true);
_hash_relbuf(rel, buf);
}
/* Now reacquire buffer lock on metapage */
LockBuffer(metabuf, BUFFER_LOCK_EXCLUSIVE);
/*
* Initialize bitmap page
*/
bitmapbuf = _hash_getnewbuf(rel, num_buckets + 1, forkNum);
_hash_initbitmapbuffer(bitmapbuf, metap->hashm_bmsize, false);
MarkBufferDirty(bitmapbuf);
/* add the new bitmap page to the metapage's list of bitmaps */
/* metapage already has a write lock */
if (metap->hashm_nmaps >= HASH_MAX_BITMAPS)
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("out of overflow pages in hash index \"%s\"",
RelationGetRelationName(rel))));
metap->hashm_mapp[metap->hashm_nmaps] = num_buckets + 1;
metap->hashm_nmaps++;
MarkBufferDirty(metabuf);
/* XLOG stuff */
if (use_wal)
{
xl_hash_init_bitmap_page xlrec;
XLogRecPtr recptr;
xlrec.bmsize = metap->hashm_bmsize;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHashInitBitmapPage);
XLogRegisterBuffer(0, bitmapbuf, REGBUF_WILL_INIT);
/*
* This is safe only because nobody else can be modifying the index at
* this stage; it's only visible to the transaction that is creating
* it.
*/
XLogRegisterBuffer(1, metabuf, REGBUF_STANDARD);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_INIT_BITMAP_PAGE);
PageSetLSN(BufferGetPage(bitmapbuf), recptr);
PageSetLSN(BufferGetPage(metabuf), recptr);
}
/* all done */
_hash_relbuf(rel, bitmapbuf);
_hash_relbuf(rel, metabuf);
return num_buckets;
}
/*
* Attempt to expand the hash table by creating one new bucket.
*
* This will silently do nothing if we don't get cleanup lock on old or
* new bucket.
*
* Complete the pending splits and remove the tuples from old bucket,
* if there are any left over from the previous split.
*
* The caller must hold a pin, but no lock, on the metapage buffer.
* The buffer is returned in the same state.
*/
void
_hash_expandtable(Relation rel, Buffer metabuf)
{
HashMetaPage metap;
Bucket old_bucket;
Bucket new_bucket;
uint32 spare_ndx;
BlockNumber start_oblkno;
BlockNumber start_nblkno;
Buffer buf_nblkno;
Buffer buf_oblkno;
Page opage;
Page npage;
HashPageOpaque oopaque;
HashPageOpaque nopaque;
uint32 maxbucket;
uint32 highmask;
uint32 lowmask;
bool metap_update_masks = false;
bool metap_update_splitpoint = false;
restart_expand:
/*
* Write-lock the meta page. It used to be necessary to acquire a
* heavyweight lock to begin a split, but that is no longer required.
*/
LockBuffer(metabuf, BUFFER_LOCK_EXCLUSIVE);
_hash_checkpage(rel, metabuf, LH_META_PAGE);
metap = HashPageGetMeta(BufferGetPage(metabuf));
/*
* Check to see if split is still needed; someone else might have already
* done one while we waited for the lock.
*
* Make sure this stays in sync with _hash_doinsert()
*/
if (metap->hashm_ntuples <=
(double) metap->hashm_ffactor * (metap->hashm_maxbucket + 1))
goto fail;
/*
* Can't split anymore if maxbucket has reached its maximum possible
* value.
*
* Ideally we'd allow bucket numbers up to UINT_MAX-1 (no higher because
* the calculation maxbucket+1 mustn't overflow). Currently we restrict
* to half that because of overflow looping in _hash_log2() and
* insufficient space in hashm_spares[]. It's moot anyway because an
* index with 2^32 buckets would certainly overflow BlockNumber and hence
* _hash_alloc_buckets() would fail, but if we supported buckets smaller
* than a disk block then this would be an independent constraint.
*
* If you change this, see also the maximum initial number of buckets in
* _hash_init().
*/
if (metap->hashm_maxbucket >= (uint32) 0x7FFFFFFE)
goto fail;
/*
* Determine which bucket is to be split, and attempt to take cleanup lock
* on the old bucket. If we can't get the lock, give up.
*
* The cleanup lock protects us not only against other backends, but
* against our own backend as well.
*
* The cleanup lock is mainly to protect the split from concurrent
* inserts. See src/backend/access/hash/README, Lock Definitions for
* further details. Due to this locking restriction, if there is any
* pending scan, the split will give up which is not good, but harmless.
*/
new_bucket = metap->hashm_maxbucket + 1;
old_bucket = (new_bucket & metap->hashm_lowmask);
start_oblkno = BUCKET_TO_BLKNO(metap, old_bucket);
buf_oblkno = _hash_getbuf_with_condlock_cleanup(rel, start_oblkno, LH_BUCKET_PAGE);
if (!buf_oblkno)
goto fail;
opage = BufferGetPage(buf_oblkno);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
/*
* We want to finish the split from a bucket as there is no apparent
* benefit by not doing so and it will make the code complicated to finish
* the split that involves multiple buckets considering the case where new
* split also fails. We don't need to consider the new bucket for
* completing the split here as it is not possible that a re-split of new
* bucket starts when there is still a pending split from old bucket.
*/
if (H_BUCKET_BEING_SPLIT(oopaque))
{
/*
* Copy bucket mapping info now; refer the comment in code below where
* we copy this information before calling _hash_splitbucket to see
* why this is okay.
*/
maxbucket = metap->hashm_maxbucket;
highmask = metap->hashm_highmask;
lowmask = metap->hashm_lowmask;
/*
* Release the lock on metapage and old_bucket, before completing the
* split.
*/
LockBuffer(metabuf, BUFFER_LOCK_UNLOCK);
LockBuffer(buf_oblkno, BUFFER_LOCK_UNLOCK);
_hash_finish_split(rel, metabuf, buf_oblkno, old_bucket, maxbucket,
highmask, lowmask);
/* release the pin on old buffer and retry for expand. */
_hash_dropbuf(rel, buf_oblkno);
goto restart_expand;
}
/*
* Clean the tuples remained from the previous split. This operation
* requires cleanup lock and we already have one on the old bucket, so
* let's do it. We also don't want to allow further splits from the bucket
* till the garbage of previous split is cleaned. This has two
* advantages; first, it helps in avoiding the bloat due to garbage and
* second is, during cleanup of bucket, we are always sure that the
* garbage tuples belong to most recently split bucket. On the contrary,
* if we allow cleanup of bucket after meta page is updated to indicate
* the new split and before the actual split, the cleanup operation won't
* be able to decide whether the tuple has been moved to the newly created
* bucket and ended up deleting such tuples.
*/
if (H_NEEDS_SPLIT_CLEANUP(oopaque))
{
/*
* Copy bucket mapping info now; refer to the comment in code below
* where we copy this information before calling _hash_splitbucket to
* see why this is okay.
*/
maxbucket = metap->hashm_maxbucket;
highmask = metap->hashm_highmask;
lowmask = metap->hashm_lowmask;
/* Release the metapage lock. */
LockBuffer(metabuf, BUFFER_LOCK_UNLOCK);
hashbucketcleanup(rel, old_bucket, buf_oblkno, start_oblkno, NULL,
maxbucket, highmask, lowmask, NULL, NULL, true,
NULL, NULL);
_hash_dropbuf(rel, buf_oblkno);
goto restart_expand;
}
/*
* There shouldn't be any active scan on new bucket.
*
* Note: it is safe to compute the new bucket's blkno here, even though we
* may still need to update the BUCKET_TO_BLKNO mapping. This is because
* the current value of hashm_spares[hashm_ovflpoint] correctly shows
* where we are going to put a new splitpoint's worth of buckets.
*/
start_nblkno = BUCKET_TO_BLKNO(metap, new_bucket);
/*
* If the split point is increasing we need to allocate a new batch of
* bucket pages.
*/
spare_ndx = _hash_spareindex(new_bucket + 1);
if (spare_ndx > metap->hashm_ovflpoint)
{
uint32 buckets_to_add;
Assert(spare_ndx == metap->hashm_ovflpoint + 1);
/*
* We treat allocation of buckets as a separate WAL-logged action.
* Even if we fail after this operation, won't leak bucket pages;
* rather, the next split will consume this space. In any case, even
* without failure we don't use all the space in one split operation.
*/
buckets_to_add = _hash_get_totalbuckets(spare_ndx) - new_bucket;
if (!_hash_alloc_buckets(rel, start_nblkno, buckets_to_add))
{
/* can't split due to BlockNumber overflow */
_hash_relbuf(rel, buf_oblkno);
goto fail;
}
}
/*
* Physically allocate the new bucket's primary page. We want to do this
* before changing the metapage's mapping info, in case we can't get the
* disk space. Ideally, we don't need to check for cleanup lock on new
* bucket as no other backend could find this bucket unless meta page is
* updated. However, it is good to be consistent with old bucket locking.
*/
buf_nblkno = _hash_getnewbuf(rel, start_nblkno, MAIN_FORKNUM);
if (!IsBufferCleanupOK(buf_nblkno))
{
_hash_relbuf(rel, buf_oblkno);
_hash_relbuf(rel, buf_nblkno);
goto fail;
}
/*
* Since we are scribbling on the pages in the shared buffers, establish a
* critical section. Any failure in this next code leaves us with a big
* problem: the metapage is effectively corrupt but could get written back
* to disk.
*/
START_CRIT_SECTION();
/*
* Okay to proceed with split. Update the metapage bucket mapping info.
*/
metap->hashm_maxbucket = new_bucket;
if (new_bucket > metap->hashm_highmask)
{
/* Starting a new doubling */
metap->hashm_lowmask = metap->hashm_highmask;
metap->hashm_highmask = new_bucket | metap->hashm_lowmask;
metap_update_masks = true;
}
/*
* If the split point is increasing we need to adjust the hashm_spares[]
* array and hashm_ovflpoint so that future overflow pages will be created
* beyond this new batch of bucket pages.
*/
if (spare_ndx > metap->hashm_ovflpoint)
{
metap->hashm_spares[spare_ndx] = metap->hashm_spares[metap->hashm_ovflpoint];
metap->hashm_ovflpoint = spare_ndx;
metap_update_splitpoint = true;
}
MarkBufferDirty(metabuf);
/*
* Copy bucket mapping info now; this saves re-accessing the meta page
* inside _hash_splitbucket's inner loop. Note that once we drop the
* split lock, other splits could begin, so these values might be out of
* date before _hash_splitbucket finishes. That's okay, since all it
* needs is to tell which of these two buckets to map hashkeys into.
*/
maxbucket = metap->hashm_maxbucket;
highmask = metap->hashm_highmask;
lowmask = metap->hashm_lowmask;
opage = BufferGetPage(buf_oblkno);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
/*
* Mark the old bucket to indicate that split is in progress. (At
* operation end, we will clear the split-in-progress flag.) Also, for a
* primary bucket page, hasho_prevblkno stores the number of buckets that
* existed as of the last split, so we must update that value here.
*/
oopaque->hasho_flag |= LH_BUCKET_BEING_SPLIT;
oopaque->hasho_prevblkno = maxbucket;
MarkBufferDirty(buf_oblkno);
npage = BufferGetPage(buf_nblkno);
/*
* initialize the new bucket's primary page and mark it to indicate that
* split is in progress.
*/
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
nopaque->hasho_prevblkno = maxbucket;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = new_bucket;
nopaque->hasho_flag = LH_BUCKET_PAGE | LH_BUCKET_BEING_POPULATED;
nopaque->hasho_page_id = HASHO_PAGE_ID;
MarkBufferDirty(buf_nblkno);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
xl_hash_split_allocate_page xlrec;
XLogRecPtr recptr;
xlrec.new_bucket = maxbucket;
xlrec.old_bucket_flag = oopaque->hasho_flag;
xlrec.new_bucket_flag = nopaque->hasho_flag;
xlrec.flags = 0;
XLogBeginInsert();
XLogRegisterBuffer(0, buf_oblkno, REGBUF_STANDARD);
XLogRegisterBuffer(1, buf_nblkno, REGBUF_WILL_INIT);
XLogRegisterBuffer(2, metabuf, REGBUF_STANDARD);
if (metap_update_masks)
{
xlrec.flags |= XLH_SPLIT_META_UPDATE_MASKS;
XLogRegisterBufData(2, (char *) &metap->hashm_lowmask, sizeof(uint32));
XLogRegisterBufData(2, (char *) &metap->hashm_highmask, sizeof(uint32));
}
if (metap_update_splitpoint)
{
xlrec.flags |= XLH_SPLIT_META_UPDATE_SPLITPOINT;
XLogRegisterBufData(2, (char *) &metap->hashm_ovflpoint,
sizeof(uint32));
XLogRegisterBufData(2,
(char *) &metap->hashm_spares[metap->hashm_ovflpoint],
sizeof(uint32));
}
XLogRegisterData((char *) &xlrec, SizeOfHashSplitAllocPage);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_SPLIT_ALLOCATE_PAGE);
PageSetLSN(BufferGetPage(buf_oblkno), recptr);
PageSetLSN(BufferGetPage(buf_nblkno), recptr);
PageSetLSN(BufferGetPage(metabuf), recptr);
}
END_CRIT_SECTION();
/* drop lock, but keep pin */
LockBuffer(metabuf, BUFFER_LOCK_UNLOCK);
/* Relocate records to the new bucket */
_hash_splitbucket(rel, metabuf,
old_bucket, new_bucket,
buf_oblkno, buf_nblkno, NULL,
maxbucket, highmask, lowmask);
/* all done, now release the pins on primary buckets. */
_hash_dropbuf(rel, buf_oblkno);
_hash_dropbuf(rel, buf_nblkno);
return;
/* Here if decide not to split or fail to acquire old bucket lock */
fail:
/* We didn't write the metapage, so just drop lock */
LockBuffer(metabuf, BUFFER_LOCK_UNLOCK);
}
/*
* _hash_splitbucket -- split 'obucket' into 'obucket' and 'nbucket'
*
* This routine is used to partition the tuples between old and new bucket and
* is used to finish the incomplete split operations. To finish the previously
* interrupted split operation, the caller needs to fill htab. If htab is set,
* then we skip the movement of tuples that exists in htab, otherwise NULL
* value of htab indicates movement of all the tuples that belong to the new
* bucket.
*
* We are splitting a bucket that consists of a base bucket page and zero
* or more overflow (bucket chain) pages. We must relocate tuples that
* belong in the new bucket.
*
* The caller must hold cleanup locks on both buckets to ensure that
* no one else is trying to access them (see README).
*
* The caller must hold a pin, but no lock, on the metapage buffer.
* The buffer is returned in the same state. (The metapage is only
* touched if it becomes necessary to add or remove overflow pages.)
*
* Split needs to retain pin on primary bucket pages of both old and new
* buckets till end of operation. This is to prevent vacuum from starting
* while a split is in progress.
*
* In addition, the caller must have created the new bucket's base page,
* which is passed in buffer nbuf, pinned and write-locked. The lock will be
* released here and pin must be released by the caller. (The API is set up
* this way because we must do _hash_getnewbuf() before releasing the metapage
* write lock. So instead of passing the new bucket's start block number, we
* pass an actual buffer.)
*/
static void
_hash_splitbucket(Relation rel,
Buffer metabuf,
Bucket obucket,
Bucket nbucket,
Buffer obuf,
Buffer nbuf,
HTAB *htab,
uint32 maxbucket,
uint32 highmask,
uint32 lowmask)
{
Buffer bucket_obuf;
Buffer bucket_nbuf;
Page opage;
Page npage;
HashPageOpaque oopaque;
HashPageOpaque nopaque;
OffsetNumber itup_offsets[MaxIndexTuplesPerPage];
IndexTuple itups[MaxIndexTuplesPerPage];
Size all_tups_size = 0;
int i;
uint16 nitups = 0;
bucket_obuf = obuf;
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
/*
* Partition the tuples in the old bucket between the old bucket and the
* new bucket, advancing along the old bucket's overflow bucket chain and
* adding overflow pages to the new bucket as needed. Outer loop iterates
* once per page in old bucket.
*/
for (;;)
{
BlockNumber oblkno;
OffsetNumber ooffnum;
OffsetNumber omaxoffnum;
/* Scan each tuple in old page */
omaxoffnum = PageGetMaxOffsetNumber(opage);
for (ooffnum = FirstOffsetNumber;
ooffnum <= omaxoffnum;
ooffnum = OffsetNumberNext(ooffnum))
{
IndexTuple itup;
Size itemsz;
Bucket bucket;
bool found = false;
/* skip dead tuples */
if (ItemIdIsDead(PageGetItemId(opage, ooffnum)))
continue;
/*
* Before inserting a tuple, probe the hash table containing TIDs
* of tuples belonging to new bucket, if we find a match, then
* skip that tuple, else fetch the item's hash key (conveniently
* stored in the item) and determine which bucket it now belongs
* in.
*/
itup = (IndexTuple) PageGetItem(opage,
PageGetItemId(opage, ooffnum));
if (htab)
(void) hash_search(htab, &itup->t_tid, HASH_FIND, &found);
if (found)
continue;
bucket = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup),
maxbucket, highmask, lowmask);
if (bucket == nbucket)
{
IndexTuple new_itup;
/*
* make a copy of index tuple as we have to scribble on it.
*/
new_itup = CopyIndexTuple(itup);
/*
* mark the index tuple as moved by split, such tuples are
* skipped by scan if there is split in progress for a bucket.
*/
new_itup->t_info |= INDEX_MOVED_BY_SPLIT_MASK;
/*
* insert the tuple into the new bucket. if it doesn't fit on
* the current page in the new bucket, we must allocate a new
* overflow page and place the tuple on that page instead.
*/
itemsz = IndexTupleDSize(*new_itup);
itemsz = MAXALIGN(itemsz);
if (PageGetFreeSpaceForMultipleTuples(npage, nitups + 1) < (all_tups_size + itemsz))
{
/*
* Change the shared buffer state in critical section,
* otherwise any error could make it unrecoverable.
*/
START_CRIT_SECTION();
_hash_pgaddmultitup(rel, nbuf, itups, itup_offsets, nitups);
MarkBufferDirty(nbuf);
/* log the split operation before releasing the lock */
log_split_page(rel, nbuf);
END_CRIT_SECTION();
/* drop lock, but keep pin */
LockBuffer(nbuf, BUFFER_LOCK_UNLOCK);
/* be tidy */
for (i = 0; i < nitups; i++)
pfree(itups[i]);
nitups = 0;
all_tups_size = 0;
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf, (nbuf == bucket_nbuf) ? true : false);
npage = BufferGetPage(nbuf);
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
}
itups[nitups++] = new_itup;
all_tups_size += itemsz;
}
else
{
/*
* the tuple stays on this page, so nothing to do.
*/
Assert(bucket == obucket);
}
}
oblkno = oopaque->hasho_nextblkno;
/* retain the pin on the old primary bucket */
if (obuf == bucket_obuf)
LockBuffer(obuf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, obuf);
/* Exit loop if no more overflow pages in old bucket */
if (!BlockNumberIsValid(oblkno))
{
/*
* Change the shared buffer state in critical section, otherwise
* any error could make it unrecoverable.
*/
START_CRIT_SECTION();
_hash_pgaddmultitup(rel, nbuf, itups, itup_offsets, nitups);
MarkBufferDirty(nbuf);
/* log the split operation before releasing the lock */
log_split_page(rel, nbuf);
END_CRIT_SECTION();
if (nbuf == bucket_nbuf)
LockBuffer(nbuf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, nbuf);
/* be tidy */
for (i = 0; i < nitups; i++)
pfree(itups[i]);
break;
}
/* Else, advance to next old page */
obuf = _hash_getbuf(rel, oblkno, HASH_READ, LH_OVERFLOW_PAGE);
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
}
/*
* We're at the end of the old bucket chain, so we're done partitioning
* the tuples. Mark the old and new buckets to indicate split is
* finished.
*
* To avoid deadlocks due to locking order of buckets, first lock the old
* bucket and then the new bucket.
*/
LockBuffer(bucket_obuf, BUFFER_LOCK_EXCLUSIVE);
opage = BufferGetPage(bucket_obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
LockBuffer(bucket_nbuf, BUFFER_LOCK_EXCLUSIVE);
npage = BufferGetPage(bucket_nbuf);
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
START_CRIT_SECTION();
oopaque->hasho_flag &= ~LH_BUCKET_BEING_SPLIT;
nopaque->hasho_flag &= ~LH_BUCKET_BEING_POPULATED;
/*
* After the split is finished, mark the old bucket to indicate that it
* contains deletable tuples. We will clear split-cleanup flag after
* deleting such tuples either at the end of split or at the next split
* from old bucket or at the time of vacuum.
*/
oopaque->hasho_flag |= LH_BUCKET_NEEDS_SPLIT_CLEANUP;
/*
* now write the buffers, here we don't release the locks as caller is
* responsible to release locks.
*/
MarkBufferDirty(bucket_obuf);
MarkBufferDirty(bucket_nbuf);
if (RelationNeedsWAL(rel))
{
XLogRecPtr recptr;
xl_hash_split_complete xlrec;
xlrec.old_bucket_flag = oopaque->hasho_flag;
xlrec.new_bucket_flag = nopaque->hasho_flag;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHashSplitComplete);
XLogRegisterBuffer(0, bucket_obuf, REGBUF_STANDARD);
XLogRegisterBuffer(1, bucket_nbuf, REGBUF_STANDARD);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_SPLIT_COMPLETE);
PageSetLSN(BufferGetPage(bucket_obuf), recptr);
PageSetLSN(BufferGetPage(bucket_nbuf), recptr);
}
END_CRIT_SECTION();
/*
* If possible, clean up the old bucket. We might not be able to do this
* if someone else has a pin on it, but if not then we can go ahead. This
* isn't absolutely necessary, but it reduces bloat; if we don't do it
* now, VACUUM will do it eventually, but maybe not until new overflow
* pages have been allocated. Note that there's no need to clean up the
* new bucket.
*/
if (IsBufferCleanupOK(bucket_obuf))
{
LockBuffer(bucket_nbuf, BUFFER_LOCK_UNLOCK);
hashbucketcleanup(rel, obucket, bucket_obuf,
BufferGetBlockNumber(bucket_obuf), NULL,
maxbucket, highmask, lowmask, NULL, NULL, true,
NULL, NULL);
}
else
{
LockBuffer(bucket_nbuf, BUFFER_LOCK_UNLOCK);
LockBuffer(bucket_obuf, BUFFER_LOCK_UNLOCK);
}
}
/*
* _hash_finish_split() -- Finish the previously interrupted split operation
*
* To complete the split operation, we form the hash table of TIDs in new
* bucket which is then used by split operation to skip tuples that are
* already moved before the split operation was previously interrupted.
*
* The caller must hold a pin, but no lock, on the metapage and old bucket's
* primary page buffer. The buffers are returned in the same state. (The
* metapage is only touched if it becomes necessary to add or remove overflow
* pages.)
*/
void
_hash_finish_split(Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket,
uint32 maxbucket, uint32 highmask, uint32 lowmask)
{
HASHCTL hash_ctl;
HTAB *tidhtab;
Buffer bucket_nbuf = InvalidBuffer;
Buffer nbuf;
Page npage;
BlockNumber nblkno;
BlockNumber bucket_nblkno;
HashPageOpaque npageopaque;
Bucket nbucket;
bool found;
/* Initialize hash tables used to track TIDs */
memset(&hash_ctl, 0, sizeof(hash_ctl));
hash_ctl.keysize = sizeof(ItemPointerData);
hash_ctl.entrysize = sizeof(ItemPointerData);
hash_ctl.hcxt = CurrentMemoryContext;
tidhtab =
hash_create("bucket ctids",
256, /* arbitrary initial size */
&hash_ctl,
HASH_ELEM | HASH_BLOBS | HASH_CONTEXT);
bucket_nblkno = nblkno = _hash_get_newblock_from_oldbucket(rel, obucket);
/*
* Scan the new bucket and build hash table of TIDs
*/
for (;;)
{
OffsetNumber noffnum;
OffsetNumber nmaxoffnum;
nbuf = _hash_getbuf(rel, nblkno, HASH_READ,
LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
/* remember the primary bucket buffer to acquire cleanup lock on it. */
if (nblkno == bucket_nblkno)
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
/* Scan each tuple in new page */
nmaxoffnum = PageGetMaxOffsetNumber(npage);
for (noffnum = FirstOffsetNumber;
noffnum <= nmaxoffnum;
noffnum = OffsetNumberNext(noffnum))
{
IndexTuple itup;
/* Fetch the item's TID and insert it in hash table. */
itup = (IndexTuple) PageGetItem(npage,
PageGetItemId(npage, noffnum));
(void) hash_search(tidhtab, &itup->t_tid, HASH_ENTER, &found);
Assert(!found);
}
nblkno = npageopaque->hasho_nextblkno;
/*
* release our write lock without modifying buffer and ensure to
* retain the pin on primary bucket.
*/
if (nbuf == bucket_nbuf)
LockBuffer(nbuf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, nbuf);
/* Exit loop if no more overflow pages in new bucket */
if (!BlockNumberIsValid(nblkno))
break;
}
/*
* Conditionally get the cleanup lock on old and new buckets to perform
* the split operation. If we don't get the cleanup locks, silently give
* up and next insertion on old bucket will try again to complete the
* split.
*/
if (!ConditionalLockBufferForCleanup(obuf))
{
hash_destroy(tidhtab);
return;
}
if (!ConditionalLockBufferForCleanup(bucket_nbuf))
{
LockBuffer(obuf, BUFFER_LOCK_UNLOCK);
hash_destroy(tidhtab);
return;
}
npage = BufferGetPage(bucket_nbuf);
npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
nbucket = npageopaque->hasho_bucket;
_hash_splitbucket(rel, metabuf, obucket,
nbucket, obuf, bucket_nbuf, tidhtab,
maxbucket, highmask, lowmask);
_hash_dropbuf(rel, bucket_nbuf);
hash_destroy(tidhtab);
}
/*
* Initialize a sequence's relation with the specified tuple as content
*/
static void
fill_seq_with_data(Relation rel, HeapTuple tuple)
{
Buffer buf;
Page page;
sequence_magic *sm;
OffsetNumber offnum;
/* Initialize first page of relation with special magic number */
buf = ReadBuffer(rel, P_NEW);
Assert(BufferGetBlockNumber(buf) == 0);
page = BufferGetPage(buf);
PageInit(page, BufferGetPageSize(buf), sizeof(sequence_magic));
sm = (sequence_magic *) PageGetSpecialPointer(page);
sm->magic = SEQ_MAGIC;
/* Now insert sequence tuple */
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
/*
* Since VACUUM does not process sequences, we have to force the tuple to
* have xmin = FrozenTransactionId now. Otherwise it would become
* invisible to SELECTs after 2G transactions. It is okay to do this
* because if the current transaction aborts, no other xact will ever
* examine the sequence tuple anyway.
*/
HeapTupleHeaderSetXmin(tuple->t_data, FrozenTransactionId);
HeapTupleHeaderSetXminFrozen(tuple->t_data);
HeapTupleHeaderSetCmin(tuple->t_data, FirstCommandId);
HeapTupleHeaderSetXmax(tuple->t_data, InvalidTransactionId);
tuple->t_data->t_infomask |= HEAP_XMAX_INVALID;
ItemPointerSet(&tuple->t_data->t_ctid, 0, FirstOffsetNumber);
/* check the comment above nextval_internal()'s equivalent call. */
if (RelationNeedsWAL(rel))
GetTopTransactionId();
START_CRIT_SECTION();
MarkBufferDirty(buf);
offnum = PageAddItem(page, (Item) tuple->t_data, tuple->t_len,
InvalidOffsetNumber, false, false);
if (offnum != FirstOffsetNumber)
elog(ERROR, "failed to add sequence tuple to page");
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
xl_seq_rec xlrec;
XLogRecPtr recptr;
XLogBeginInsert();
XLogRegisterBuffer(0, buf, REGBUF_WILL_INIT);
xlrec.node = rel->rd_node;
XLogRegisterData((char *) &xlrec, sizeof(xl_seq_rec));
XLogRegisterData((char *) tuple->t_data, tuple->t_len);
recptr = XLogInsert(RM_SEQ_ID, XLOG_SEQ_LOG);
PageSetLSN(page, recptr);
}
END_CRIT_SECTION();
UnlockReleaseBuffer(buf);
}
/*
* Descend the tree to the leaf page that contains or would contain the key
* we're searching for. The key should already be filled in 'btree', in
* tree-type specific manner. If btree->fullScan is true, descends to the
* leftmost leaf page.
*
* If 'searchmode' is false, on return stack->buffer is exclusively locked,
* and the stack represents the full path to the root. Otherwise stack->buffer
* is share-locked, and stack->parent is NULL.
*/
GinBtreeStack *
ginFindLeafPage(GinBtree btree, bool searchMode)
{
GinBtreeStack *stack;
stack = (GinBtreeStack *) palloc(sizeof(GinBtreeStack));
stack->blkno = btree->rootBlkno;
stack->buffer = ReadBuffer(btree->index, btree->rootBlkno);
stack->parent = NULL;
stack->predictNumber = 1;
for (;;)
{
Page page;
BlockNumber child;
int access;
stack->off = InvalidOffsetNumber;
page = BufferGetPage(stack->buffer);
access = ginTraverseLock(stack->buffer, searchMode);
/*
* If we're going to modify the tree, finish any incomplete splits we
* encounter on the way.
*/
if (!searchMode && GinPageIsIncompleteSplit(page))
ginFinishSplit(btree, stack, false, NULL);
/*
* ok, page is correctly locked, we should check to move right ..,
* root never has a right link, so small optimization
*/
while (btree->fullScan == FALSE && stack->blkno != btree->rootBlkno &&
btree->isMoveRight(btree, page))
{
BlockNumber rightlink = GinPageGetOpaque(page)->rightlink;
if (rightlink == InvalidBlockNumber)
/* rightmost page */
break;
stack->buffer = ginStepRight(stack->buffer, btree->index, access);
stack->blkno = rightlink;
page = BufferGetPage(stack->buffer);
if (!searchMode && GinPageIsIncompleteSplit(page))
ginFinishSplit(btree, stack, false, NULL);
}
if (GinPageIsLeaf(page)) /* we found, return locked page */
return stack;
/* now we have correct buffer, try to find child */
child = btree->findChildPage(btree, stack);
LockBuffer(stack->buffer, GIN_UNLOCK);
Assert(child != InvalidBlockNumber);
Assert(stack->blkno != child);
if (searchMode)
{
/* in search mode we may forget path to leaf */
stack->blkno = child;
stack->buffer = ReleaseAndReadBuffer(stack->buffer, btree->index, stack->blkno);
}
else
{
GinBtreeStack *ptr = (GinBtreeStack *) palloc(sizeof(GinBtreeStack));
ptr->parent = stack;
stack = ptr;
stack->blkno = child;
stack->buffer = ReadBuffer(btree->index, stack->blkno);
stack->predictNumber = 1;
}
}
}
/*
* Finish a split by inserting the downlink for the new page to parent.
*
* On entry, stack->buffer is exclusively locked.
*
* If freestack is true, all the buffers are released and unlocked as we
* crawl up the tree, and 'stack' is freed. Otherwise stack->buffer is kept
* locked, and stack is unmodified, except for possibly moving right to find
* the correct parent of page.
*/
static void
ginFinishSplit(GinBtree btree, GinBtreeStack *stack, bool freestack,
GinStatsData *buildStats)
{
Page page;
bool done;
bool first = true;
/*
* freestack == false when we encounter an incompletely split page during a
* scan, while freestack == true is used in the normal scenario that a
* split is finished right after the initial insert.
*/
if (!freestack)
elog(DEBUG1, "finishing incomplete split of block %u in gin index \"%s\"",
stack->blkno, RelationGetRelationName(btree->index));
/* this loop crawls up the stack until the insertion is complete */
do
{
GinBtreeStack *parent = stack->parent;
void *insertdata;
BlockNumber updateblkno;
/* search parent to lock */
LockBuffer(parent->buffer, GIN_EXCLUSIVE);
/*
* If the parent page was incompletely split, finish that split first,
* then continue with the current one.
*
* Note: we have to finish *all* incomplete splits we encounter, even
* if we have to move right. Otherwise we might choose as the target
* a page that has no downlink in the parent, and splitting it further
* would fail.
*/
if (GinPageIsIncompleteSplit(BufferGetPage(parent->buffer)))
ginFinishSplit(btree, parent, false, buildStats);
/* move right if it's needed */
page = BufferGetPage(parent->buffer);
while ((parent->off = btree->findChildPtr(btree, page, stack->blkno, parent->off)) == InvalidOffsetNumber)
{
if (GinPageRightMost(page))
{
/*
* rightmost page, but we don't find parent, we should use
* plain search...
*/
LockBuffer(parent->buffer, GIN_UNLOCK);
ginFindParents(btree, stack);
parent = stack->parent;
Assert(parent != NULL);
break;
}
parent->buffer = ginStepRight(parent->buffer, btree->index, GIN_EXCLUSIVE);
parent->blkno = BufferGetBlockNumber(parent->buffer);
page = BufferGetPage(parent->buffer);
if (GinPageIsIncompleteSplit(BufferGetPage(parent->buffer)))
ginFinishSplit(btree, parent, false, buildStats);
}
/* insert the downlink */
insertdata = btree->prepareDownlink(btree, stack->buffer);
updateblkno = GinPageGetOpaque(BufferGetPage(stack->buffer))->rightlink;
done = ginPlaceToPage(btree, parent,
insertdata, updateblkno,
stack->buffer, buildStats);
pfree(insertdata);
/*
* If the caller requested to free the stack, unlock and release the
* child buffer now. Otherwise keep it pinned and locked, but if we
* have to recurse up the tree, we can unlock the upper pages, only
* keeping the page at the bottom of the stack locked.
*/
if (!first || freestack)
LockBuffer(stack->buffer, GIN_UNLOCK);
if (freestack)
{
ReleaseBuffer(stack->buffer);
pfree(stack);
}
stack = parent;
first = false;
} while (!done);
/* unlock the parent */
LockBuffer(stack->buffer, GIN_UNLOCK);
if (freestack)
freeGinBtreeStack(stack);
}
/* ----------------
* index_getnext - get the next heap tuple from a scan
*
* The result is the next heap tuple satisfying the scan keys and the
* snapshot, or NULL if no more matching tuples exist. On success,
* the buffer containing the heap tuple is pinned (the pin will be dropped
* at the next index_getnext or index_endscan).
*
* Note: caller must check scan->xs_recheck, and perform rechecking of the
* scan keys if required. We do not do that here because we don't have
* enough information to do it efficiently in the general case.
* ----------------
*/
HeapTuple
index_getnext(IndexScanDesc scan, ScanDirection direction)
{
HeapTuple heapTuple = &scan->xs_ctup;
ItemPointer tid = &heapTuple->t_self;
FmgrInfo *procedure;
bool all_dead = false;
SCAN_CHECKS;
GET_SCAN_PROCEDURE(amgettuple);
Assert(TransactionIdIsValid(RecentGlobalXmin));
for (;;)
{
bool got_heap_tuple;
if (scan->xs_continue_hot)
{
/*
* We are resuming scan of a HOT chain after having returned an
* earlier member. Must still hold pin on current heap page.
*/
Assert(BufferIsValid(scan->xs_cbuf));
Assert(ItemPointerGetBlockNumber(tid) ==
BufferGetBlockNumber(scan->xs_cbuf));
}
else
{
bool found;
Buffer prev_buf;
/*
* If we scanned a whole HOT chain and found only dead tuples,
* tell index AM to kill its entry for that TID. We do not do this
* when in recovery because it may violate MVCC to do so. see
* comments in RelationGetIndexScan().
*/
if (!scan->xactStartedInRecovery)
scan->kill_prior_tuple = all_dead;
/*
* The AM's gettuple proc finds the next index entry matching the
* scan keys, and puts the TID in xs_ctup.t_self (ie, *tid). It
* should also set scan->xs_recheck, though we pay no attention to
* that here.
*/
found = DatumGetBool(FunctionCall2(procedure,
PointerGetDatum(scan),
Int32GetDatum(direction)));
/* Reset kill flag immediately for safety */
scan->kill_prior_tuple = false;
/* If we're out of index entries, break out of outer loop */
if (!found)
break;
pgstat_count_index_tuples(scan->indexRelation, 1);
/* Switch to correct buffer if we don't have it already */
prev_buf = scan->xs_cbuf;
scan->xs_cbuf = ReleaseAndReadBuffer(scan->xs_cbuf,
scan->heapRelation,
ItemPointerGetBlockNumber(tid));
/*
* Prune page, but only if we weren't already on this page
*/
if (prev_buf != scan->xs_cbuf)
heap_page_prune_opt(scan->heapRelation, scan->xs_cbuf,
RecentGlobalXmin);
}
/* Obtain share-lock on the buffer so we can examine visibility */
LockBuffer(scan->xs_cbuf, BUFFER_LOCK_SHARE);
got_heap_tuple = heap_hot_search_buffer(tid, scan->heapRelation,
scan->xs_cbuf,
scan->xs_snapshot,
&scan->xs_ctup,
&all_dead,
!scan->xs_continue_hot);
LockBuffer(scan->xs_cbuf, BUFFER_LOCK_UNLOCK);
if (got_heap_tuple)
{
/*
* Only in a non-MVCC snapshot can more than one member of the
* HOT chain be visible.
*/
scan->xs_continue_hot = !IsMVCCSnapshot(scan->xs_snapshot);
pgstat_count_heap_fetch(scan->indexRelation);
return heapTuple;
}
/* Loop around to ask index AM for another TID */
scan->xs_continue_hot = false;
}
/* Release any held pin on a heap page */
if (BufferIsValid(scan->xs_cbuf))
{
ReleaseBuffer(scan->xs_cbuf);
scan->xs_cbuf = InvalidBuffer;
}
return NULL; /* failure exit */
}
/*
* Workhouse routine for doing insertion into a GiST index. Note that
* this routine assumes it is invoked in a short-lived memory context,
* so it does not bother releasing palloc'd allocations.
*/
void
gistdoinsert(Relation r, IndexTuple itup, Size freespace, GISTSTATE *giststate)
{
ItemId iid;
IndexTuple idxtuple;
GISTInsertStack firststack;
GISTInsertStack *stack;
GISTInsertState state;
bool xlocked = false;
memset(&state, 0, sizeof(GISTInsertState));
state.freespace = freespace;
state.r = r;
/* Start from the root */
firststack.blkno = GIST_ROOT_BLKNO;
firststack.lsn = 0;
firststack.parent = NULL;
firststack.downlinkoffnum = InvalidOffsetNumber;
state.stack = stack = &firststack;
/*
* Walk down along the path of smallest penalty, updating the parent
* pointers with the key we're inserting as we go. If we crash in the
* middle, the tree is consistent, although the possible parent updates
* were a waste.
*/
for (;;)
{
if (XLogRecPtrIsInvalid(stack->lsn))
stack->buffer = ReadBuffer(state.r, stack->blkno);
/*
* Be optimistic and grab shared lock first. Swap it for an exclusive
* lock later if we need to update the page.
*/
if (!xlocked)
{
LockBuffer(stack->buffer, GIST_SHARE);
gistcheckpage(state.r, stack->buffer);
}
stack->page = (Page) BufferGetPage(stack->buffer);
stack->lsn = PageGetLSN(stack->page);
Assert(!RelationNeedsWAL(state.r) || !XLogRecPtrIsInvalid(stack->lsn));
/*
* If this page was split but the downlink was never inserted to the
* parent because the inserting backend crashed before doing that, fix
* that now.
*/
if (GistFollowRight(stack->page))
{
if (!xlocked)
{
LockBuffer(stack->buffer, GIST_UNLOCK);
LockBuffer(stack->buffer, GIST_EXCLUSIVE);
xlocked = true;
/* someone might've completed the split when we unlocked */
if (!GistFollowRight(stack->page))
continue;
}
gistfixsplit(&state, giststate);
UnlockReleaseBuffer(stack->buffer);
xlocked = false;
state.stack = stack = stack->parent;
continue;
}
if (stack->blkno != GIST_ROOT_BLKNO &&
stack->parent->lsn < GistPageGetNSN(stack->page))
{
/*
* Concurrent split detected. There's no guarantee that the
* downlink for this page is consistent with the tuple we're
* inserting anymore, so go back to parent and rechoose the best
* child.
*/
UnlockReleaseBuffer(stack->buffer);
xlocked = false;
state.stack = stack = stack->parent;
continue;
}
if (!GistPageIsLeaf(stack->page))
{
/*
* This is an internal page so continue to walk down the tree.
* Find the child node that has the minimum insertion penalty.
*/
BlockNumber childblkno;
IndexTuple newtup;
GISTInsertStack *item;
OffsetNumber downlinkoffnum;
downlinkoffnum = gistchoose(state.r, stack->page, itup, giststate);
iid = PageGetItemId(stack->page, downlinkoffnum);
idxtuple = (IndexTuple) PageGetItem(stack->page, iid);
childblkno = ItemPointerGetBlockNumber(&(idxtuple->t_tid));
/*
* Check that it's not a leftover invalid tuple from pre-9.1
*/
if (GistTupleIsInvalid(idxtuple))
ereport(ERROR,
(errmsg("index \"%s\" contains an inner tuple marked as invalid",
RelationGetRelationName(r)),
errdetail("This is caused by an incomplete page split at crash recovery before upgrading to PostgreSQL 9.1."),
errhint("Please REINDEX it.")));
/*
* Check that the key representing the target child node is
* consistent with the key we're inserting. Update it if it's not.
*/
newtup = gistgetadjusted(state.r, idxtuple, itup, giststate);
if (newtup)
{
/*
* Swap shared lock for an exclusive one. Beware, the page may
* change while we unlock/lock the page...
*/
if (!xlocked)
{
LockBuffer(stack->buffer, GIST_UNLOCK);
LockBuffer(stack->buffer, GIST_EXCLUSIVE);
xlocked = true;
stack->page = (Page) BufferGetPage(stack->buffer);
if (PageGetLSN(stack->page) != stack->lsn)
{
/* the page was changed while we unlocked it, retry */
continue;
}
}
/*
* Update the tuple.
*
* We still hold the lock after gistinserttuple(), but it
* might have to split the page to make the updated tuple fit.
* In that case the updated tuple might migrate to the other
* half of the split, so we have to go back to the parent and
* descend back to the half that's a better fit for the new
* tuple.
*/
if (gistinserttuple(&state, stack, giststate, newtup,
downlinkoffnum))
{
/*
* If this was a root split, the root page continues to be
* the parent and the updated tuple went to one of the
* child pages, so we just need to retry from the root
* page.
*/
if (stack->blkno != GIST_ROOT_BLKNO)
{
UnlockReleaseBuffer(stack->buffer);
xlocked = false;
state.stack = stack = stack->parent;
}
continue;
}
}
LockBuffer(stack->buffer, GIST_UNLOCK);
xlocked = false;
/* descend to the chosen child */
item = (GISTInsertStack *) palloc0(sizeof(GISTInsertStack));
item->blkno = childblkno;
item->parent = stack;
item->downlinkoffnum = downlinkoffnum;
state.stack = stack = item;
}
else
{
/*
* Leaf page. Insert the new key. We've already updated all the
* parents on the way down, but we might have to split the page if
* it doesn't fit. gistinserthere() will take care of that.
*/
/*
* Swap shared lock for an exclusive one. Be careful, the page may
* change while we unlock/lock the page...
*/
if (!xlocked)
{
LockBuffer(stack->buffer, GIST_UNLOCK);
LockBuffer(stack->buffer, GIST_EXCLUSIVE);
xlocked = true;
stack->page = (Page) BufferGetPage(stack->buffer);
stack->lsn = PageGetLSN(stack->page);
if (stack->blkno == GIST_ROOT_BLKNO)
{
/*
* the only page that can become inner instead of leaf is
* the root page, so for root we should recheck it
*/
if (!GistPageIsLeaf(stack->page))
{
/*
* very rare situation: during unlock/lock index with
* number of pages = 1 was increased
*/
LockBuffer(stack->buffer, GIST_UNLOCK);
xlocked = false;
continue;
}
/*
* we don't need to check root split, because checking
* leaf/inner is enough to recognize split for root
*/
}
else if (GistFollowRight(stack->page) ||
stack->parent->lsn < GistPageGetNSN(stack->page))
{
/*
* The page was split while we momentarily unlocked the
* page. Go back to parent.
*/
UnlockReleaseBuffer(stack->buffer);
xlocked = false;
state.stack = stack = stack->parent;
continue;
}
}
/* now state.stack->(page, buffer and blkno) points to leaf page */
gistinserttuple(&state, stack, giststate, itup,
InvalidOffsetNumber);
LockBuffer(stack->buffer, GIST_UNLOCK);
/* Release any pins we might still hold before exiting */
for (; stack; stack = stack->parent)
ReleaseBuffer(stack->buffer);
break;
}
}
}
/*
* 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;
}
/*
* Traverse the tree to find path from root page to specified "child" block.
*
* returns a new insertion stack, starting from the parent of "child", up
* to the root. *downlinkoffnum is set to the offset of the downlink in the
* direct parent of child.
*
* To prevent deadlocks, this should lock only one page at a time.
*/
static GISTInsertStack *
gistFindPath(Relation r, BlockNumber child, OffsetNumber *downlinkoffnum)
{
Page page;
Buffer buffer;
OffsetNumber i,
maxoff;
ItemId iid;
IndexTuple idxtuple;
List *fifo;
GISTInsertStack *top,
*ptr;
BlockNumber blkno;
top = (GISTInsertStack *) palloc0(sizeof(GISTInsertStack));
top->blkno = GIST_ROOT_BLKNO;
top->downlinkoffnum = InvalidOffsetNumber;
fifo = list_make1(top);
while (fifo != NIL)
{
/* Get next page to visit */
top = linitial(fifo);
fifo = list_delete_first(fifo);
buffer = ReadBuffer(r, top->blkno);
LockBuffer(buffer, GIST_SHARE);
gistcheckpage(r, buffer);
page = (Page) BufferGetPage(buffer);
if (GistPageIsLeaf(page))
{
/*
* Because we scan the index top-down, all the rest of the pages
* in the queue must be leaf pages as well.
*/
UnlockReleaseBuffer(buffer);
break;
}
top->lsn = PageGetLSN(page);
/*
* If F_FOLLOW_RIGHT is set, the page to the right doesn't have a
* downlink. This should not normally happen..
*/
if (GistFollowRight(page))
elog(ERROR, "concurrent GiST page split was incomplete");
if (top->parent && top->parent->lsn < GistPageGetNSN(page) &&
GistPageGetOpaque(page)->rightlink != InvalidBlockNumber /* sanity check */ )
{
/*
* Page was split while we looked elsewhere. We didn't see the
* downlink to the right page when we scanned the parent, so add
* it to the queue now.
*
* Put the right page ahead of the queue, so that we visit it
* next. That's important, because if this is the lowest internal
* level, just above leaves, we might already have queued up some
* leaf pages, and we assume that there can't be any non-leaf
* pages behind leaf pages.
*/
ptr = (GISTInsertStack *) palloc0(sizeof(GISTInsertStack));
ptr->blkno = GistPageGetOpaque(page)->rightlink;
ptr->downlinkoffnum = InvalidOffsetNumber;
ptr->parent = top->parent;
fifo = lcons(ptr, fifo);
}
maxoff = PageGetMaxOffsetNumber(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(page, i);
idxtuple = (IndexTuple) PageGetItem(page, iid);
blkno = ItemPointerGetBlockNumber(&(idxtuple->t_tid));
if (blkno == child)
{
/* Found it! */
UnlockReleaseBuffer(buffer);
*downlinkoffnum = i;
return top;
}
else
{
/* Append this child to the list of pages to visit later */
ptr = (GISTInsertStack *) palloc0(sizeof(GISTInsertStack));
ptr->blkno = blkno;
ptr->downlinkoffnum = i;
ptr->parent = top;
fifo = lappend(fifo, ptr);
}
}
UnlockReleaseBuffer(buffer);
}
elog(ERROR, "failed to re-find parent of a page in index \"%s\", block %u",
RelationGetRelationName(r), child);
return NULL; /* keep compiler quiet */
}
/*
* visibilitymap_set - set a bit on a previously pinned page
*
* recptr is the LSN of the XLOG record we're replaying, if we're in recovery,
* or InvalidXLogRecPtr in normal running. The page LSN is advanced to the
* one provided; in normal running, we generate a new XLOG record and set the
* page LSN to that value. cutoff_xid is the largest xmin on the page being
* marked all-visible; it is needed for Hot Standby, and can be
* InvalidTransactionId if the page contains no tuples.
*
* Caller is expected to set the heap page's PD_ALL_VISIBLE bit before calling
* this function. Except in recovery, caller should also pass the heap
* buffer. When checksums are enabled and we're not in recovery, we must add
* the heap buffer to the WAL chain to protect it from being torn.
*
* You must pass a buffer containing the correct map page to this function.
* Call visibilitymap_pin first to pin the right one. This function doesn't do
* any I/O.
*/
void
visibilitymap_set(Relation rel, BlockNumber heapBlk, Buffer heapBuf,
XLogRecPtr recptr, Buffer vmBuf, TransactionId cutoff_xid)
{
BlockNumber mapBlock = HEAPBLK_TO_MAPBLOCK(heapBlk);
uint32 mapByte = HEAPBLK_TO_MAPBYTE(heapBlk);
uint8 mapBit = HEAPBLK_TO_MAPBIT(heapBlk);
Page page;
char *map;
#ifdef TRACE_VISIBILITYMAP
elog(DEBUG1, "vm_set %s %d", RelationGetRelationName(rel), heapBlk);
#endif
Assert(InRecovery || XLogRecPtrIsInvalid(recptr));
Assert(InRecovery || BufferIsValid(heapBuf));
/* Check that we have the right heap page pinned, if present */
if (BufferIsValid(heapBuf) && BufferGetBlockNumber(heapBuf) != heapBlk)
elog(ERROR, "wrong heap buffer passed to visibilitymap_set");
/* Check that we have the right VM page pinned */
if (!BufferIsValid(vmBuf) || BufferGetBlockNumber(vmBuf) != mapBlock)
elog(ERROR, "wrong VM buffer passed to visibilitymap_set");
page = BufferGetPage(vmBuf);
map = PageGetContents(page);
LockBuffer(vmBuf, BUFFER_LOCK_EXCLUSIVE);
if (!(map[mapByte] & (1 << mapBit)))
{
START_CRIT_SECTION();
map[mapByte] |= (1 << mapBit);
MarkBufferDirty(vmBuf);
if (RelationNeedsWAL(rel))
{
if (XLogRecPtrIsInvalid(recptr))
{
Assert(!InRecovery);
recptr = log_heap_visible(rel->rd_node, heapBuf, vmBuf,
cutoff_xid);
/*
* If data checksums are enabled (or wal_log_hints=on), we
* need to protect the heap page from being torn.
*/
if (XLogHintBitIsNeeded())
{
Page heapPage = BufferGetPage(heapBuf);
/* caller is expected to set PD_ALL_VISIBLE first */
Assert(PageIsAllVisible(heapPage));
PageSetLSN(heapPage, recptr);
}
}
PageSetLSN(page, recptr);
}
END_CRIT_SECTION();
}
LockBuffer(vmBuf, BUFFER_LOCK_UNLOCK);
}
/*
* Updates the stack so that child->parent is the correct parent of the
* child. child->parent must be exclusively locked on entry, and will
* remain so at exit, but it might not be the same page anymore.
*/
static void
gistFindCorrectParent(Relation r, GISTInsertStack *child)
{
GISTInsertStack *parent = child->parent;
gistcheckpage(r, parent->buffer);
parent->page = (Page) BufferGetPage(parent->buffer);
/* here we don't need to distinguish between split and page update */
if (child->downlinkoffnum == InvalidOffsetNumber ||
parent->lsn != PageGetLSN(parent->page))
{
/* parent is changed, look child in right links until found */
OffsetNumber i,
maxoff;
ItemId iid;
IndexTuple idxtuple;
GISTInsertStack *ptr;
while (true)
{
maxoff = PageGetMaxOffsetNumber(parent->page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(parent->page, i);
idxtuple = (IndexTuple) PageGetItem(parent->page, iid);
if (ItemPointerGetBlockNumber(&(idxtuple->t_tid)) == child->blkno)
{
/* yes!!, found */
child->downlinkoffnum = i;
return;
}
}
parent->blkno = GistPageGetOpaque(parent->page)->rightlink;
UnlockReleaseBuffer(parent->buffer);
if (parent->blkno == InvalidBlockNumber)
{
/*
* End of chain and still didn't find parent. It's a very-very
* rare situation when root splited.
*/
break;
}
parent->buffer = ReadBuffer(r, parent->blkno);
LockBuffer(parent->buffer, GIST_EXCLUSIVE);
gistcheckpage(r, parent->buffer);
parent->page = (Page) BufferGetPage(parent->buffer);
}
/*
* awful!!, we need search tree to find parent ... , but before we
* should release all old parent
*/
ptr = child->parent->parent; /* child->parent already released
* above */
while (ptr)
{
ReleaseBuffer(ptr->buffer);
ptr = ptr->parent;
}
/* ok, find new path */
ptr = parent = gistFindPath(r, child->blkno, &child->downlinkoffnum);
/* read all buffers as expected by caller */
/* note we don't lock them or gistcheckpage them here! */
while (ptr)
{
ptr->buffer = ReadBuffer(r, ptr->blkno);
ptr->page = (Page) BufferGetPage(ptr->buffer);
ptr = ptr->parent;
}
/* install new chain of parents to stack */
child->parent = parent;
/* make recursive call to normal processing */
LockBuffer(child->parent->buffer, GIST_EXCLUSIVE);
gistFindCorrectParent(r, child);
}
return;
}
/*
* 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.
*
* This function also deletes the tuples that are moved by split to other
* bucket.
*
* Result: a palloc'd struct containing statistical info for VACUUM displays.
*/
IndexBulkDeleteResult *
hashbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
IndexBulkDeleteCallback callback, void *callback_state)
{
Relation rel = info->index;
double tuples_removed;
double num_index_tuples;
double orig_ntuples;
Bucket orig_maxbucket;
Bucket cur_maxbucket;
Bucket cur_bucket;
Buffer metabuf = InvalidBuffer;
HashMetaPage metap;
HashMetaPage cachedmetap;
tuples_removed = 0;
num_index_tuples = 0;
/*
* We need a copy of the metapage so that we can use its hashm_spares[]
* values to compute bucket page addresses, but a cached copy should be
* good enough. (If not, we'll detect that further down and refresh the
* cache as necessary.)
*/
cachedmetap = _hash_getcachedmetap(rel, &metabuf, false);
Assert(cachedmetap != NULL);
orig_maxbucket = cachedmetap->hashm_maxbucket;
orig_ntuples = cachedmetap->hashm_ntuples;
/* Scan the buckets that we know exist */
cur_bucket = 0;
cur_maxbucket = orig_maxbucket;
loop_top:
while (cur_bucket <= cur_maxbucket)
{
BlockNumber bucket_blkno;
BlockNumber blkno;
Buffer bucket_buf;
Buffer buf;
HashPageOpaque bucket_opaque;
Page page;
bool split_cleanup = false;
/* Get address of bucket's start page */
bucket_blkno = BUCKET_TO_BLKNO(cachedmetap, cur_bucket);
blkno = bucket_blkno;
/*
* We need to acquire a cleanup lock on the primary bucket page to out
* wait concurrent scans before deleting the dead tuples.
*/
buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL, info->strategy);
LockBufferForCleanup(buf);
_hash_checkpage(rel, buf, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/*
* If the bucket contains tuples that are moved by split, then we need
* to delete such tuples. We can't delete such tuples if the split
* operation on bucket is not finished as those are needed by scans.
*/
if (!H_BUCKET_BEING_SPLIT(bucket_opaque) &&
H_NEEDS_SPLIT_CLEANUP(bucket_opaque))
{
split_cleanup = true;
/*
* This bucket might have been split since we last held a lock on
* the metapage. If so, hashm_maxbucket, hashm_highmask and
* hashm_lowmask might be old enough to cause us to fail to remove
* tuples left behind by the most recent split. To prevent that,
* now that the primary page of the target bucket has been locked
* (and thus can't be further split), check whether we need to
* update our cached metapage data.
*
* NB: The check for InvalidBlockNumber is only needed for
* on-disk compatibility with indexes created before we started
* storing hashm_maxbucket in the primary page's hasho_prevblkno.
*/
if (bucket_opaque->hasho_prevblkno != InvalidBlockNumber &&
bucket_opaque->hasho_prevblkno > cachedmetap->hashm_maxbucket)
{
cachedmetap = _hash_getcachedmetap(rel, &metabuf, true);
Assert(cachedmetap != NULL);
}
}
bucket_buf = buf;
hashbucketcleanup(rel, cur_bucket, bucket_buf, blkno, info->strategy,
cachedmetap->hashm_maxbucket,
cachedmetap->hashm_highmask,
cachedmetap->hashm_lowmask, &tuples_removed,
&num_index_tuples, split_cleanup,
callback, callback_state);
_hash_dropbuf(rel, bucket_buf);
/* Advance to next bucket */
cur_bucket++;
}
if (BufferIsInvalid(metabuf))
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_NOLOCK, LH_META_PAGE);
/* Write-lock metapage and check for split since we started */
LockBuffer(metabuf, BUFFER_LOCK_EXCLUSIVE);
metap = HashPageGetMeta(BufferGetPage(metabuf));
if (cur_maxbucket != metap->hashm_maxbucket)
{
/* There's been a split, so process the additional bucket(s) */
LockBuffer(metabuf, BUFFER_LOCK_UNLOCK);
cachedmetap = _hash_getcachedmetap(rel, &metabuf, true);
Assert(cachedmetap != NULL);
cur_maxbucket = cachedmetap->hashm_maxbucket;
goto loop_top;
}
/* Okay, we're really done. Update tuple count in metapage. */
START_CRIT_SECTION();
if (orig_maxbucket == metap->hashm_maxbucket &&
orig_ntuples == metap->hashm_ntuples)
{
/*
* No one has split or inserted anything since start of scan, so
* believe our count as gospel.
*/
metap->hashm_ntuples = num_index_tuples;
}
else
{
/*
* Otherwise, our count is untrustworthy since we may have
* double-scanned tuples in split buckets. Proceed by dead-reckoning.
* (Note: we still return estimated_count = false, because using this
* count is better than not updating reltuples at all.)
*/
if (metap->hashm_ntuples > tuples_removed)
metap->hashm_ntuples -= tuples_removed;
else
metap->hashm_ntuples = 0;
num_index_tuples = metap->hashm_ntuples;
}
MarkBufferDirty(metabuf);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
xl_hash_update_meta_page xlrec;
XLogRecPtr recptr;
xlrec.ntuples = metap->hashm_ntuples;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, sizeof(SizeOfHashUpdateMetaPage));
XLogRegisterBuffer(0, metabuf, REGBUF_STANDARD);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_UPDATE_META_PAGE);
PageSetLSN(BufferGetPage(metabuf), recptr);
}
END_CRIT_SECTION();
_hash_relbuf(rel, metabuf);
/* return statistics */
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
stats->estimated_count = false;
stats->num_index_tuples = num_index_tuples;
stats->tuples_removed += tuples_removed;
/* hashvacuumcleanup will fill in num_pages */
return stats;
}
/*
* Insert an index tuple into the index relation. The revmap is updated to
* mark the range containing the given page as pointing to the inserted entry.
* A WAL record is written.
*
* The buffer, if valid, is first checked for free space to insert the new
* entry; if there isn't enough, a new buffer is obtained and pinned. No
* buffer lock must be held on entry, no buffer lock is held on exit.
*
* Return value is the offset number where the tuple was inserted.
*/
OffsetNumber
brin_doinsert(Relation idxrel, BlockNumber pagesPerRange,
BrinRevmap *revmap, Buffer *buffer, BlockNumber heapBlk,
BrinTuple *tup, Size itemsz)
{
Page page;
BlockNumber blk;
OffsetNumber off;
Buffer revmapbuf;
ItemPointerData tid;
bool extended;
Assert(itemsz == MAXALIGN(itemsz));
/* If the item is oversized, don't even bother. */
if (itemsz > BrinMaxItemSize)
{
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index row size %lu exceeds maximum %lu for index \"%s\"",
(unsigned long) itemsz,
(unsigned long) BrinMaxItemSize,
RelationGetRelationName(idxrel))));
return InvalidOffsetNumber; /* keep compiler quiet */
}
/* Make sure the revmap is long enough to contain the entry we need */
brinRevmapExtend(revmap, heapBlk);
/*
* Acquire lock on buffer supplied by caller, if any. If it doesn't have
* enough space, unpin it to obtain a new one below.
*/
if (BufferIsValid(*buffer))
{
/*
* It's possible that another backend (or ourselves!) extended the
* revmap over the page we held a pin on, so we cannot assume that
* it's still a regular page.
*/
LockBuffer(*buffer, BUFFER_LOCK_EXCLUSIVE);
if (br_page_get_freespace(BufferGetPage(*buffer)) < itemsz)
{
UnlockReleaseBuffer(*buffer);
*buffer = InvalidBuffer;
}
}
/*
* If we still don't have a usable buffer, have brin_getinsertbuffer
* obtain one for us.
*/
if (!BufferIsValid(*buffer))
{
do
*buffer = brin_getinsertbuffer(idxrel, InvalidBuffer, itemsz, &extended);
while (!BufferIsValid(*buffer));
}
else
extended = false;
/* Now obtain lock on revmap buffer */
revmapbuf = brinLockRevmapPageForUpdate(revmap, heapBlk);
page = BufferGetPage(*buffer);
blk = BufferGetBlockNumber(*buffer);
/* Execute the actual insertion */
START_CRIT_SECTION();
if (extended)
brin_page_init(BufferGetPage(*buffer), BRIN_PAGETYPE_REGULAR);
off = PageAddItem(page, (Item) tup, itemsz, InvalidOffsetNumber,
false, false);
if (off == InvalidOffsetNumber)
elog(ERROR, "could not insert new index tuple to page");
MarkBufferDirty(*buffer);
BRIN_elog((DEBUG2, "inserted tuple (%u,%u) for range starting at %u",
blk, off, heapBlk));
ItemPointerSet(&tid, blk, off);
brinSetHeapBlockItemptr(revmapbuf, pagesPerRange, heapBlk, tid);
MarkBufferDirty(revmapbuf);
/* XLOG stuff */
if (RelationNeedsWAL(idxrel))
{
xl_brin_insert xlrec;
XLogRecPtr recptr;
uint8 info;
info = XLOG_BRIN_INSERT | (extended ? XLOG_BRIN_INIT_PAGE : 0);
xlrec.heapBlk = heapBlk;
xlrec.pagesPerRange = pagesPerRange;
xlrec.offnum = off;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfBrinInsert);
XLogRegisterBuffer(0, *buffer, REGBUF_STANDARD | (extended ? REGBUF_WILL_INIT : 0));
XLogRegisterBufData(0, (char *) tup, itemsz);
XLogRegisterBuffer(1, revmapbuf, 0);
recptr = XLogInsert(RM_BRIN_ID, info);
PageSetLSN(page, recptr);
PageSetLSN(BufferGetPage(revmapbuf), recptr);
}
END_CRIT_SECTION();
/* Tuple is firmly on buffer; we can release our locks */
LockBuffer(*buffer, BUFFER_LOCK_UNLOCK);
LockBuffer(revmapbuf, BUFFER_LOCK_UNLOCK);
if (extended)
FreeSpaceMapVacuum(idxrel);
return off;
}
/*
* FreeSpaceMapTruncateRel - adjust for truncation of a relation.
*
* The caller must hold AccessExclusiveLock on the relation, to ensure that
* other backends receive the smgr invalidation event that this function sends
* before they access the FSM again.
*
* nblocks is the new size of the heap.
*/
void
FreeSpaceMapTruncateRel(Relation rel, BlockNumber nblocks)
{
BlockNumber new_nfsmblocks;
FSMAddress first_removed_address;
uint16 first_removed_slot;
Buffer buf;
RelationOpenSmgr(rel);
/*
* If no FSM has been created yet for this relation, there's nothing to
* truncate.
*/
if (!smgrexists(rel->rd_smgr, FSM_FORKNUM))
return;
/* Get the location in the FSM of the first removed heap block */
first_removed_address = fsm_get_location(nblocks, &first_removed_slot);
/*
* Zero out the tail of the last remaining FSM page. If the slot
* representing the first removed heap block is at a page boundary, as the
* first slot on the FSM page that first_removed_address points to, we can
* just truncate that page altogether.
*/
if (first_removed_slot > 0)
{
buf = fsm_readbuf(rel, first_removed_address, false);
if (!BufferIsValid(buf))
return; /* nothing to do; the FSM was already smaller */
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
/* NO EREPORT(ERROR) from here till changes are logged */
START_CRIT_SECTION();
fsm_truncate_avail(BufferGetPage(buf), first_removed_slot);
/*
* Truncation of a relation is WAL-logged at a higher-level, and we
* will be called at WAL replay. But if checksums are enabled, we need
* to still write a WAL record to protect against a torn page, if the
* page is flushed to disk before the truncation WAL record. We cannot
* use MarkBufferDirtyHint here, because that will not dirty the page
* during recovery.
*/
MarkBufferDirty(buf);
if (!InRecovery && RelationNeedsWAL(rel) && XLogHintBitIsNeeded())
log_newpage_buffer(buf, false);
END_CRIT_SECTION();
UnlockReleaseBuffer(buf);
new_nfsmblocks = fsm_logical_to_physical(first_removed_address) + 1;
}
else
{
new_nfsmblocks = fsm_logical_to_physical(first_removed_address);
if (smgrnblocks(rel->rd_smgr, FSM_FORKNUM) <= new_nfsmblocks)
return; /* nothing to do; the FSM was already smaller */
}
/* Truncate the unused FSM pages, and send smgr inval message */
smgrtruncate(rel->rd_smgr, FSM_FORKNUM, new_nfsmblocks);
/*
* We might as well update the local smgr_fsm_nblocks setting.
* smgrtruncate sent an smgr cache inval message, which will cause other
* backends to invalidate their copy of smgr_fsm_nblocks, and this one too
* at the next command boundary. But this ensures it isn't outright wrong
* until then.
*/
if (rel->rd_smgr)
rel->rd_smgr->smgr_fsm_nblocks = new_nfsmblocks;
/*
* Update upper-level FSM pages to account for the truncation. This is
* important because the just-truncated pages were likely marked as
* all-free, and would be preferentially selected.
*/
FreeSpaceMapVacuumRange(rel, nblocks, InvalidBlockNumber);
}
/*
* Update tuple origtup (size origsz), located in offset oldoff of buffer
* oldbuf, to newtup (size newsz) as summary tuple for the page range starting
* at heapBlk. oldbuf must not be locked on entry, and is not locked at exit.
*
* If samepage is true, attempt to put the new tuple in the same page, but if
* there's no room, use some other one.
*
* If the update is successful, return true; the revmap is updated to point to
* the new tuple. If the update is not done for whatever reason, return false.
* Caller may retry the update if this happens.
*/
bool
brin_doupdate(Relation idxrel, BlockNumber pagesPerRange,
BrinRevmap *revmap, BlockNumber heapBlk,
Buffer oldbuf, OffsetNumber oldoff,
const BrinTuple *origtup, Size origsz,
const BrinTuple *newtup, Size newsz,
bool samepage)
{
Page oldpage;
ItemId oldlp;
BrinTuple *oldtup;
Size oldsz;
Buffer newbuf;
bool extended;
Assert(newsz == MAXALIGN(newsz));
/* If the item is oversized, don't bother. */
if (newsz > BrinMaxItemSize)
{
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index row size %lu exceeds maximum %lu for index \"%s\"",
(unsigned long) newsz,
(unsigned long) BrinMaxItemSize,
RelationGetRelationName(idxrel))));
return false; /* keep compiler quiet */
}
/* make sure the revmap is long enough to contain the entry we need */
brinRevmapExtend(revmap, heapBlk);
if (!samepage)
{
/* need a page on which to put the item */
newbuf = brin_getinsertbuffer(idxrel, oldbuf, newsz, &extended);
if (!BufferIsValid(newbuf))
{
Assert(!extended);
return false;
}
/*
* Note: it's possible (though unlikely) that the returned newbuf is
* the same as oldbuf, if brin_getinsertbuffer determined that the old
* buffer does in fact have enough space.
*/
if (newbuf == oldbuf)
{
Assert(!extended);
newbuf = InvalidBuffer;
}
}
else
{
LockBuffer(oldbuf, BUFFER_LOCK_EXCLUSIVE);
newbuf = InvalidBuffer;
extended = false;
}
oldpage = BufferGetPage(oldbuf);
oldlp = PageGetItemId(oldpage, oldoff);
/*
* Check that the old tuple wasn't updated concurrently: it might have
* moved someplace else entirely ...
*/
if (!ItemIdIsNormal(oldlp))
{
LockBuffer(oldbuf, BUFFER_LOCK_UNLOCK);
/*
* If this happens, and the new buffer was obtained by extending the
* relation, then we need to ensure we don't leave it uninitialized or
* forget about it.
*/
if (BufferIsValid(newbuf))
{
if (extended)
brin_initialize_empty_new_buffer(idxrel, newbuf);
UnlockReleaseBuffer(newbuf);
if (extended)
FreeSpaceMapVacuum(idxrel);
}
return false;
}
oldsz = ItemIdGetLength(oldlp);
oldtup = (BrinTuple *) PageGetItem(oldpage, oldlp);
/*
* ... or it might have been updated in place to different contents.
*/
if (!brin_tuples_equal(oldtup, oldsz, origtup, origsz))
{
LockBuffer(oldbuf, BUFFER_LOCK_UNLOCK);
if (BufferIsValid(newbuf))
{
if (extended)
brin_initialize_empty_new_buffer(idxrel, newbuf);
UnlockReleaseBuffer(newbuf);
if (extended)
FreeSpaceMapVacuum(idxrel);
}
return false;
}
/*
* Great, the old tuple is intact. We can proceed with the update.
*
* If there's enough room in the old page for the new tuple, replace it.
*
* Note that there might now be enough space on the page even though the
* caller told us there isn't, if a concurrent update moved another tuple
* elsewhere or replaced a tuple with a smaller one.
*/
if (((BrinPageFlags(oldpage) & BRIN_EVACUATE_PAGE) == 0) &&
brin_can_do_samepage_update(oldbuf, origsz, newsz))
{
if (BufferIsValid(newbuf))
{
/* as above */
if (extended)
brin_initialize_empty_new_buffer(idxrel, newbuf);
UnlockReleaseBuffer(newbuf);
}
START_CRIT_SECTION();
PageIndexDeleteNoCompact(oldpage, &oldoff, 1);
if (PageAddItem(oldpage, (Item) newtup, newsz, oldoff, true,
false) == InvalidOffsetNumber)
elog(ERROR, "failed to add BRIN tuple");
MarkBufferDirty(oldbuf);
/* XLOG stuff */
if (RelationNeedsWAL(idxrel))
{
xl_brin_samepage_update xlrec;
XLogRecPtr recptr;
uint8 info = XLOG_BRIN_SAMEPAGE_UPDATE;
xlrec.offnum = oldoff;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfBrinSamepageUpdate);
XLogRegisterBuffer(0, oldbuf, REGBUF_STANDARD);
XLogRegisterBufData(0, (char *) newtup, newsz);
recptr = XLogInsert(RM_BRIN_ID, info);
PageSetLSN(oldpage, recptr);
}
END_CRIT_SECTION();
LockBuffer(oldbuf, BUFFER_LOCK_UNLOCK);
if (extended)
FreeSpaceMapVacuum(idxrel);
return true;
}
else if (newbuf == InvalidBuffer)
{
/*
* Not enough space, but caller said that there was. Tell them to
* start over.
*/
LockBuffer(oldbuf, BUFFER_LOCK_UNLOCK);
return false;
}
else
{
/*
* Not enough free space on the oldpage. Put the new tuple on the new
* page, and update the revmap.
*/
Page newpage = BufferGetPage(newbuf);
Buffer revmapbuf;
ItemPointerData newtid;
OffsetNumber newoff;
BlockNumber newblk = InvalidBlockNumber;
Size freespace = 0;
revmapbuf = brinLockRevmapPageForUpdate(revmap, heapBlk);
START_CRIT_SECTION();
/*
* We need to initialize the page if it's newly obtained. Note we
* will WAL-log the initialization as part of the update, so we don't
* need to do that here.
*/
if (extended)
brin_page_init(BufferGetPage(newbuf), BRIN_PAGETYPE_REGULAR);
PageIndexDeleteNoCompact(oldpage, &oldoff, 1);
newoff = PageAddItem(newpage, (Item) newtup, newsz,
InvalidOffsetNumber, false, false);
if (newoff == InvalidOffsetNumber)
elog(ERROR, "failed to add BRIN tuple to new page");
MarkBufferDirty(oldbuf);
MarkBufferDirty(newbuf);
/* needed to update FSM below */
if (extended)
{
newblk = BufferGetBlockNumber(newbuf);
freespace = br_page_get_freespace(newpage);
}
ItemPointerSet(&newtid, BufferGetBlockNumber(newbuf), newoff);
brinSetHeapBlockItemptr(revmapbuf, pagesPerRange, heapBlk, newtid);
MarkBufferDirty(revmapbuf);
/* XLOG stuff */
if (RelationNeedsWAL(idxrel))
{
xl_brin_update xlrec;
XLogRecPtr recptr;
uint8 info;
info = XLOG_BRIN_UPDATE | (extended ? XLOG_BRIN_INIT_PAGE : 0);
xlrec.insert.offnum = newoff;
xlrec.insert.heapBlk = heapBlk;
xlrec.insert.pagesPerRange = pagesPerRange;
xlrec.oldOffnum = oldoff;
XLogBeginInsert();
/* new page */
XLogRegisterData((char *) &xlrec, SizeOfBrinUpdate);
XLogRegisterBuffer(0, newbuf, REGBUF_STANDARD | (extended ? REGBUF_WILL_INIT : 0));
XLogRegisterBufData(0, (char *) newtup, newsz);
/* revmap page */
XLogRegisterBuffer(1, revmapbuf, REGBUF_STANDARD);
/* old page */
XLogRegisterBuffer(2, oldbuf, REGBUF_STANDARD);
recptr = XLogInsert(RM_BRIN_ID, info);
PageSetLSN(oldpage, recptr);
PageSetLSN(newpage, recptr);
PageSetLSN(BufferGetPage(revmapbuf), recptr);
}
END_CRIT_SECTION();
LockBuffer(revmapbuf, BUFFER_LOCK_UNLOCK);
LockBuffer(oldbuf, BUFFER_LOCK_UNLOCK);
UnlockReleaseBuffer(newbuf);
if (extended)
{
Assert(BlockNumberIsValid(newblk));
RecordPageWithFreeSpace(idxrel, newblk, freespace);
FreeSpaceMapVacuum(idxrel);
}
return true;
}
}
/*
* Search the tree for a heap page with at least min_cat of free space
*/
static BlockNumber
fsm_search(Relation rel, uint8 min_cat)
{
int restarts = 0;
FSMAddress addr = FSM_ROOT_ADDRESS;
for (;;)
{
int slot;
Buffer buf;
uint8 max_avail = 0;
/* Read the FSM page. */
buf = fsm_readbuf(rel, addr, false);
/* Search within the page */
if (BufferIsValid(buf))
{
LockBuffer(buf, BUFFER_LOCK_SHARE);
slot = fsm_search_avail(buf, min_cat,
(addr.level == FSM_BOTTOM_LEVEL),
false);
if (slot == -1)
max_avail = fsm_get_max_avail(BufferGetPage(buf));
UnlockReleaseBuffer(buf);
}
else
slot = -1;
if (slot != -1)
{
/*
* Descend the tree, or return the found block if we're at the
* bottom.
*/
if (addr.level == FSM_BOTTOM_LEVEL)
return fsm_get_heap_blk(addr, slot);
addr = fsm_get_child(addr, slot);
}
else if (addr.level == FSM_ROOT_LEVEL)
{
/*
* At the root, failure means there's no page with enough free
* space in the FSM. Give up.
*/
return InvalidBlockNumber;
}
else
{
uint16 parentslot;
FSMAddress parent;
/*
* At lower level, failure can happen if the value in the upper-
* level node didn't reflect the value on the lower page. Update
* the upper node, to avoid falling into the same trap again, and
* start over.
*
* There's a race condition here, if another backend updates this
* page right after we release it, and gets the lock on the parent
* page before us. We'll then update the parent page with the now
* stale information we had. It's OK, because it should happen
* rarely, and will be fixed by the next vacuum.
*/
parent = fsm_get_parent(addr, &parentslot);
fsm_set_and_search(rel, parent, parentslot, max_avail, 0);
/*
* If the upper pages are badly out of date, we might need to loop
* quite a few times, updating them as we go. Any inconsistencies
* should eventually be corrected and the loop should end. Looping
* indefinitely is nevertheless scary, so provide an emergency
* valve.
*/
if (restarts++ > 10000)
return InvalidBlockNumber;
/* Start search all over from the root */
addr = FSM_ROOT_ADDRESS;
}
}
}
/*
* Return a pinned and exclusively locked buffer which can be used to insert an
* index item of size itemsz (caller must ensure not to request sizes
* impossible to fulfill). If oldbuf is a valid buffer, it is also locked (in
* an order determined to avoid deadlocks.)
*
* If we find that the old page is no longer a regular index page (because
* of a revmap extension), the old buffer is unlocked and we return
* InvalidBuffer.
*
* If there's no existing page with enough free space to accommodate the new
* item, the relation is extended. If this happens, *extended is set to true,
* and it is the caller's responsibility to initialize the page (and WAL-log
* that fact) prior to use.
*
* Note that in some corner cases it is possible for this routine to extend the
* relation and then not return the buffer. It is this routine's
* responsibility to WAL-log the page initialization and to record the page in
* FSM if that happens. Such a buffer may later be reused by this routine.
*/
static Buffer
brin_getinsertbuffer(Relation irel, Buffer oldbuf, Size itemsz,
bool *extended)
{
BlockNumber oldblk;
BlockNumber newblk;
Page page;
int freespace;
/* callers must have checked */
Assert(itemsz <= BrinMaxItemSize);
*extended = false;
if (BufferIsValid(oldbuf))
oldblk = BufferGetBlockNumber(oldbuf);
else
oldblk = InvalidBlockNumber;
/*
* Loop until we find a page with sufficient free space. By the time we
* return to caller out of this loop, both buffers are valid and locked;
* if we have to restart here, neither buffer is locked and buf is not a
* pinned buffer.
*/
newblk = RelationGetTargetBlock(irel);
if (newblk == InvalidBlockNumber)
newblk = GetPageWithFreeSpace(irel, itemsz);
for (;;)
{
Buffer buf;
bool extensionLockHeld = false;
CHECK_FOR_INTERRUPTS();
if (newblk == InvalidBlockNumber)
{
/*
* There's not enough free space in any existing index page,
* according to the FSM: extend the relation to obtain a shiny new
* page.
*/
if (!RELATION_IS_LOCAL(irel))
{
LockRelationForExtension(irel, ExclusiveLock);
extensionLockHeld = true;
}
buf = ReadBuffer(irel, P_NEW);
newblk = BufferGetBlockNumber(buf);
*extended = true;
BRIN_elog((DEBUG2, "brin_getinsertbuffer: extending to page %u",
BufferGetBlockNumber(buf)));
}
else if (newblk == oldblk)
{
/*
* There's an odd corner-case here where the FSM is out-of-date,
* and gave us the old page.
*/
buf = oldbuf;
}
else
{
buf = ReadBuffer(irel, newblk);
}
/*
* We lock the old buffer first, if it's earlier than the new one; but
* before we do, we need to check that it hasn't been turned into a
* revmap page concurrently; if we detect that it happened, give up
* and tell caller to start over.
*/
if (BufferIsValid(oldbuf) && oldblk < newblk)
{
LockBuffer(oldbuf, BUFFER_LOCK_EXCLUSIVE);
if (!BRIN_IS_REGULAR_PAGE(BufferGetPage(oldbuf)))
{
LockBuffer(oldbuf, BUFFER_LOCK_UNLOCK);
/*
* It is possible that the new page was obtained from
* extending the relation. In that case, we must be sure to
* record it in the FSM before leaving, because otherwise the
* space would be lost forever. However, we cannot let an
* uninitialized page get in the FSM, so we need to initialize
* it first.
*/
if (*extended)
{
brin_initialize_empty_new_buffer(irel, buf);
/* shouldn't matter, but don't confuse caller */
*extended = false;
}
if (extensionLockHeld)
UnlockRelationForExtension(irel, ExclusiveLock);
ReleaseBuffer(buf);
return InvalidBuffer;
}
}
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
if (extensionLockHeld)
UnlockRelationForExtension(irel, ExclusiveLock);
page = BufferGetPage(buf);
/*
* We have a new buffer to insert into. Check that the new page has
* enough free space, and return it if it does; otherwise start over.
* Note that we allow for the FSM to be out of date here, and in that
* case we update it and move on.
*
* (br_page_get_freespace also checks that the FSM didn't hand us a
* page that has since been repurposed for the revmap.)
*/
freespace = *extended ?
BrinMaxItemSize : br_page_get_freespace(page);
if (freespace >= itemsz)
{
RelationSetTargetBlock(irel, BufferGetBlockNumber(buf));
/*
* Since the target block specification can get lost on cache
* invalidations, make sure we update the more permanent FSM with
* data about it before going away.
*/
if (*extended)
RecordPageWithFreeSpace(irel, BufferGetBlockNumber(buf),
freespace);
/*
* Lock the old buffer if not locked already. Note that in this
* case we know for sure it's a regular page: it's later than the
* new page we just got, which is not a revmap page, and revmap
* pages are always consecutive.
*/
if (BufferIsValid(oldbuf) && oldblk > newblk)
{
LockBuffer(oldbuf, BUFFER_LOCK_EXCLUSIVE);
Assert(BRIN_IS_REGULAR_PAGE(BufferGetPage(oldbuf)));
}
return buf;
}
/* This page is no good. */
/*
* If an entirely new page does not contain enough free space for the
* new item, then surely that item is oversized. Complain loudly; but
* first make sure we initialize the page and record it as free, for
* next time.
*/
if (*extended)
{
brin_initialize_empty_new_buffer(irel, buf);
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index row size %lu exceeds maximum %lu for index \"%s\"",
(unsigned long) itemsz,
(unsigned long) freespace,
RelationGetRelationName(irel))));
return InvalidBuffer; /* keep compiler quiet */
}
if (newblk != oldblk)
UnlockReleaseBuffer(buf);
if (BufferIsValid(oldbuf) && oldblk <= newblk)
LockBuffer(oldbuf, BUFFER_LOCK_UNLOCK);
newblk = RecordAndGetPageWithFreeSpace(irel, newblk, freespace, itemsz);
}
}
/*
* Optionally prune and repair fragmentation in the specified page.
*
* This is an opportunistic function. It will perform housekeeping
* only if the page heuristically looks like a candidate for pruning and we
* can acquire buffer cleanup lock without blocking.
*
* Note: this is called quite often. It's important that it fall out quickly
* if there's not any use in pruning.
*
* Caller must have pin on the buffer, and must *not* have a lock on it.
*
* OldestXmin is the cutoff XID used to distinguish whether tuples are DEAD
* or RECENTLY_DEAD (see HeapTupleSatisfiesVacuum).
*/
void
heap_page_prune_opt(Relation relation, Buffer buffer)
{
Page page = BufferGetPage(buffer);
Size minfree;
TransactionId OldestXmin;
/*
* We can't write WAL in recovery mode, so there's no point trying to
* clean the page. The master will likely issue a cleaning WAL record soon
* anyway, so this is no particular loss.
*/
if (RecoveryInProgress())
return;
/*
* Use the appropriate xmin horizon for this relation. If it's a proper
* catalog relation or a user defined, additional, catalog relation, we
* need to use the horizon that includes slots, otherwise the data-only
* horizon can be used. Note that the toast relation of user defined
* relations are *not* considered catalog relations.
*
* It is OK to apply the old snapshot limit before acquiring the cleanup
* lock because the worst that can happen is that we are not quite as
* aggressive about the cleanup (by however many transaction IDs are
* consumed between this point and acquiring the lock). This allows us to
* save significant overhead in the case where the page is found not to be
* prunable.
*/
if (IsCatalogRelation(relation) ||
RelationIsAccessibleInLogicalDecoding(relation))
OldestXmin = RecentGlobalXmin;
else
OldestXmin =
TransactionIdLimitedForOldSnapshots(RecentGlobalDataXmin,
relation);
Assert(TransactionIdIsValid(OldestXmin));
/*
* Let's see if we really need pruning.
*
* Forget it if page is not hinted to contain something prunable that's
* older than OldestXmin.
*/
if (!PageIsPrunable(page, OldestXmin))
return;
/*
* We prune when a previous UPDATE failed to find enough space on the page
* for a new tuple version, or when free space falls below the relation's
* fill-factor target (but not less than 10%).
*
* Checking free space here is questionable since we aren't holding any
* lock on the buffer; in the worst case we could get a bogus answer. It's
* unlikely to be *seriously* wrong, though, since reading either pd_lower
* or pd_upper is probably atomic. Avoiding taking a lock seems more
* important than sometimes getting a wrong answer in what is after all
* just a heuristic estimate.
*/
minfree = RelationGetTargetPageFreeSpace(relation,
HEAP_DEFAULT_FILLFACTOR);
minfree = Max(minfree, BLCKSZ / 10);
if (PageIsFull(page) || PageGetHeapFreeSpace(page) < minfree)
{
/* OK, try to get exclusive buffer lock */
if (!ConditionalLockBufferForCleanup(buffer))
return;
/*
* Now that we have buffer lock, get accurate information about the
* page's free space, and recheck the heuristic about whether to
* prune. (We needn't recheck PageIsPrunable, since no one else could
* have pruned while we hold pin.)
*/
if (PageIsFull(page) || PageGetHeapFreeSpace(page) < minfree)
{
TransactionId ignore = InvalidTransactionId; /* return value not
* needed */
/* OK to prune */
(void) heap_page_prune(relation, buffer, OldestXmin, true, &ignore);
}
/* And release buffer lock */
LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
}
}
/*
* btvacuumpage --- VACUUM one page
*
* This processes a single page for btvacuumscan(). In some cases we
* must go back and re-examine previously-scanned pages; this routine
* recurses when necessary to handle that case.
*
* blkno is the page to process. orig_blkno is the highest block number
* reached by the outer btvacuumscan loop (the same as blkno, unless we
* are recursing to re-examine a previous page).
*/
static void
btvacuumpage(BTVacState *vstate, BlockNumber blkno, BlockNumber orig_blkno)
{
IndexVacuumInfo *info = vstate->info;
IndexBulkDeleteResult *stats = vstate->stats;
IndexBulkDeleteCallback callback = vstate->callback;
void *callback_state = vstate->callback_state;
Relation rel = info->index;
bool delete_now;
BlockNumber recurse_to;
Buffer buf;
Page page;
BTPageOpaque opaque;
restart:
delete_now = false;
recurse_to = P_NONE;
/* call vacuum_delay_point while not holding any buffer lock */
vacuum_delay_point();
/*
* We can't use _bt_getbuf() here because it always applies
* _bt_checkpage(), which will barf on an all-zero page. We want to
* recycle all-zero pages, not fail. Also, we want to use a nondefault
* buffer access strategy.
*/
buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL,
info->strategy);
LockBuffer(buf, BT_READ);
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (!PageIsNew(page))
_bt_checkpage(rel, buf);
/*
* If we are recursing, the only case we want to do anything with is a
* live leaf page having the current vacuum cycle ID. Any other state
* implies we already saw the page (eg, deleted it as being empty).
*/
if (blkno != orig_blkno)
{
if (_bt_page_recyclable(page) ||
P_IGNORE(opaque) ||
!P_ISLEAF(opaque) ||
opaque->btpo_cycleid != vstate->cycleid)
{
_bt_relbuf(rel, buf);
return;
}
}
/* Page is valid, see what to do with it */
if (_bt_page_recyclable(page))
{
/* Okay to recycle this page */
RecordFreeIndexPage(rel, blkno);
vstate->totFreePages++;
stats->pages_deleted++;
}
else if (P_ISDELETED(opaque))
{
/* Already deleted, but can't recycle yet */
stats->pages_deleted++;
}
else if (P_ISHALFDEAD(opaque))
{
/* Half-dead, try to delete */
delete_now = true;
}
else if (P_ISLEAF(opaque))
{
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable;
OffsetNumber offnum,
minoff,
maxoff;
/*
* Trade in the initial read lock for a super-exclusive write lock on
* this page. We must get such a lock on every leaf page over the
* course of the vacuum scan, whether or not it actually contains any
* deletable tuples --- see nbtree/README.
*/
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockBufferForCleanup(buf);
/*
* Remember highest leaf page number we've taken cleanup lock on; see
* notes in btvacuumscan
*/
if (blkno > vstate->lastBlockLocked)
vstate->lastBlockLocked = blkno;
/*
* Check whether we need to recurse back to earlier pages. What we
* are concerned about is a page split that happened since we started
* the vacuum scan. If the split moved some tuples to a lower page
* then we might have missed 'em. If so, set up for tail recursion.
* (Must do this before possibly clearing btpo_cycleid below!)
*/
if (vstate->cycleid != 0 &&
opaque->btpo_cycleid == vstate->cycleid &&
!(opaque->btpo_flags & BTP_SPLIT_END) &&
!P_RIGHTMOST(opaque) &&
opaque->btpo_next < orig_blkno)
recurse_to = opaque->btpo_next;
/*
* Scan over all items to see which ones need deleted according to the
* callback function.
*/
ndeletable = 0;
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
if (callback)
{
for (offnum = minoff;
offnum <= maxoff;
offnum = OffsetNumberNext(offnum))
{
IndexTuple itup;
ItemPointer htup;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offnum));
htup = &(itup->t_tid);
/*
* During Hot Standby we currently assume that
* XLOG_BTREE_VACUUM records do not produce conflicts. That is
* only true as long as the callback function depends only
* upon whether the index tuple refers to heap tuples removed
* in the initial heap scan. When vacuum starts it derives a
* value of OldestXmin. Backends taking later snapshots could
* have a RecentGlobalXmin with a later xid than the vacuum's
* OldestXmin, so it is possible that row versions deleted
* after OldestXmin could be marked as killed by other
* backends. The callback function *could* look at the index
* tuple state in isolation and decide to delete the index
* tuple, though currently it does not. If it ever did, we
* would need to reconsider whether XLOG_BTREE_VACUUM records
* should cause conflicts. If they did cause conflicts they
* would be fairly harsh conflicts, since we haven't yet
* worked out a way to pass a useful value for
* latestRemovedXid on the XLOG_BTREE_VACUUM records. This
* applies to *any* type of index that marks index tuples as
* killed.
*/
if (callback(htup, callback_state))
deletable[ndeletable++] = offnum;
}
}
/*
* Apply any needed deletes. We issue just one _bt_delitems_vacuum()
* call per page, so as to minimize WAL traffic.
*/
if (ndeletable > 0)
{
/*
* Notice that the issued XLOG_BTREE_VACUUM WAL record includes an
* instruction to the replay code to get cleanup lock on all pages
* between the previous lastBlockVacuumed and this page. This
* ensures that WAL replay locks all leaf pages at some point.
*
* Since we can visit leaf pages out-of-order when recursing,
* replay might end up locking such pages an extra time, but it
* doesn't seem worth the amount of bookkeeping it'd take to avoid
* that.
*/
_bt_delitems_vacuum(rel, buf, deletable, ndeletable,
vstate->lastBlockVacuumed);
/*
* Remember highest leaf page number we've issued a
* XLOG_BTREE_VACUUM WAL record for.
*/
if (blkno > vstate->lastBlockVacuumed)
vstate->lastBlockVacuumed = blkno;
stats->tuples_removed += ndeletable;
/* must recompute maxoff */
maxoff = PageGetMaxOffsetNumber(page);
}
else
{
/*
* If the page has been split during this vacuum cycle, it seems
* worth expending a write to clear btpo_cycleid even if we don't
* have any deletions to do. (If we do, _bt_delitems_vacuum takes
* care of this.) This ensures we won't process the page again.
*
* We treat this like a hint-bit update because there's no need to
* WAL-log it.
*/
if (vstate->cycleid != 0 &&
opaque->btpo_cycleid == vstate->cycleid)
{
opaque->btpo_cycleid = 0;
MarkBufferDirtyHint(buf, true);
}
}
/*
* If it's now empty, try to delete; else count the live tuples. We
* don't delete when recursing, though, to avoid putting entries into
* freePages out-of-order (doesn't seem worth any extra code to handle
* the case).
*/
if (minoff > maxoff)
delete_now = (blkno == orig_blkno);
else
stats->num_index_tuples += maxoff - minoff + 1;
}
if (delete_now)
{
MemoryContext oldcontext;
int ndel;
/* Run pagedel in a temp context to avoid memory leakage */
MemoryContextReset(vstate->pagedelcontext);
oldcontext = MemoryContextSwitchTo(vstate->pagedelcontext);
ndel = _bt_pagedel(rel, buf);
/* count only this page, else may double-count parent */
if (ndel)
stats->pages_deleted++;
MemoryContextSwitchTo(oldcontext);
/* pagedel released buffer, so we shouldn't */
}
else
_bt_relbuf(rel, buf);
/*
* This is really tail recursion, but if the compiler is too stupid to
* optimize it as such, we'd eat an uncomfortably large amount of stack
* space per recursion level (due to the deletable[] array). A failure is
* improbable since the number of levels isn't likely to be large ... but
* just in case, let's hand-optimize into a loop.
*/
if (recurse_to != P_NONE)
{
blkno = recurse_to;
goto restart;
}
}
/* ----------------
* index_getnext - get the next heap tuple from a scan
*
* The result is the next heap tuple satisfying the scan keys and the
* snapshot, or NULL if no more matching tuples exist. On success,
* the buffer containing the heap tuple is pinned (the pin will be dropped
* at the next index_getnext or index_endscan).
*
* Note: caller must check scan->xs_recheck, and perform rechecking of the
* scan keys if required. We do not do that here because we don't have
* enough information to do it efficiently in the general case.
* ----------------
*/
HeapTuple
index_getnext(IndexScanDesc scan, ScanDirection direction)
{
HeapTuple heapTuple = &scan->xs_ctup;
ItemPointer tid = &heapTuple->t_self;
FmgrInfo *procedure;
SCAN_CHECKS;
GET_SCAN_PROCEDURE(amgettuple);
Assert(TransactionIdIsValid(RecentGlobalXmin));
/*
* We always reset xs_hot_dead; if we are here then either we are just
* starting the scan, or we previously returned a visible tuple, and in
* either case it's inappropriate to kill the prior index entry.
*/
scan->xs_hot_dead = false;
for (;;)
{
OffsetNumber offnum;
bool at_chain_start;
Page dp;
if (scan->xs_next_hot != InvalidOffsetNumber)
{
/*
* We are resuming scan of a HOT chain after having returned an
* earlier member. Must still hold pin on current heap page.
*/
Assert(BufferIsValid(scan->xs_cbuf));
Assert(ItemPointerGetBlockNumber(tid) ==
BufferGetBlockNumber(scan->xs_cbuf));
Assert(TransactionIdIsValid(scan->xs_prev_xmax));
offnum = scan->xs_next_hot;
at_chain_start = false;
scan->xs_next_hot = InvalidOffsetNumber;
}
else
{
bool found;
Buffer prev_buf;
/*
* If we scanned a whole HOT chain and found only dead tuples,
* tell index AM to kill its entry for that TID. We do not do this
* when in recovery because it may violate MVCC to do so. see
* comments in RelationGetIndexScan().
*/
if (!scan->xactStartedInRecovery)
scan->kill_prior_tuple = scan->xs_hot_dead;
/*
* The AM's gettuple proc finds the next index entry matching the
* scan keys, and puts the TID in xs_ctup.t_self (ie, *tid). It
* should also set scan->xs_recheck, though we pay no attention to
* that here.
*/
found = DatumGetBool(FunctionCall2(procedure,
PointerGetDatum(scan),
Int32GetDatum(direction)));
/* Reset kill flag immediately for safety */
scan->kill_prior_tuple = false;
/* If we're out of index entries, break out of outer loop */
if (!found)
break;
pgstat_count_index_tuples(scan->indexRelation, 1);
/* Switch to correct buffer if we don't have it already */
prev_buf = scan->xs_cbuf;
scan->xs_cbuf = ReleaseAndReadBuffer(scan->xs_cbuf,
scan->heapRelation,
ItemPointerGetBlockNumber(tid));
/*
* Prune page, but only if we weren't already on this page
*/
if (prev_buf != scan->xs_cbuf)
heap_page_prune_opt(scan->heapRelation, scan->xs_cbuf,
RecentGlobalXmin);
/* Prepare to scan HOT chain starting at index-referenced offnum */
offnum = ItemPointerGetOffsetNumber(tid);
at_chain_start = true;
/* We don't know what the first tuple's xmin should be */
scan->xs_prev_xmax = InvalidTransactionId;
/* Initialize flag to detect if all entries are dead */
scan->xs_hot_dead = true;
}
/* Obtain share-lock on the buffer so we can examine visibility */
LockBuffer(scan->xs_cbuf, BUFFER_LOCK_SHARE);
dp = (Page) BufferGetPage(scan->xs_cbuf);
/* Scan through possible multiple members of HOT-chain */
for (;;)
{
ItemId lp;
ItemPointer ctid;
bool valid;
/* check for bogus TID */
if (offnum < FirstOffsetNumber ||
offnum > PageGetMaxOffsetNumber(dp))
break;
lp = PageGetItemId(dp, offnum);
/* check for unused, dead, or redirected items */
if (!ItemIdIsNormal(lp))
{
/* We should only see a redirect at start of chain */
if (ItemIdIsRedirected(lp) && at_chain_start)
{
/* Follow the redirect */
offnum = ItemIdGetRedirect(lp);
at_chain_start = false;
continue;
}
/* else must be end of chain */
break;
}
/*
* We must initialize all of *heapTuple (ie, scan->xs_ctup) since
* it is returned to the executor on success.
*/
heapTuple->t_data = (HeapTupleHeader) PageGetItem(dp, lp);
heapTuple->t_len = ItemIdGetLength(lp);
ItemPointerSetOffsetNumber(tid, offnum);
heapTuple->t_tableOid = RelationGetRelid(scan->heapRelation);
ctid = &heapTuple->t_data->t_ctid;
/*
* Shouldn't see a HEAP_ONLY tuple at chain start. (This test
* should be unnecessary, since the chain root can't be removed
* while we have pin on the index entry, but let's make it
* anyway.)
*/
if (at_chain_start && HeapTupleIsHeapOnly(heapTuple))
break;
/*
* The xmin should match the previous xmax value, else chain is
* broken. (Note: this test is not optional because it protects
* us against the case where the prior chain member's xmax aborted
* since we looked at it.)
*/
if (TransactionIdIsValid(scan->xs_prev_xmax) &&
!TransactionIdEquals(scan->xs_prev_xmax,
HeapTupleHeaderGetXmin(heapTuple->t_data)))
break;
/* If it's visible per the snapshot, we must return it */
valid = HeapTupleSatisfiesVisibility(heapTuple, scan->xs_snapshot,
scan->xs_cbuf);
CheckForSerializableConflictOut(valid, scan->heapRelation,
heapTuple, scan->xs_cbuf);
if (valid)
{
/*
* If the snapshot is MVCC, we know that it could accept at
* most one member of the HOT chain, so we can skip examining
* any more members. Otherwise, check for continuation of the
* HOT-chain, and set state for next time.
*/
if (IsMVCCSnapshot(scan->xs_snapshot)
&& !IsolationIsSerializable())
scan->xs_next_hot = InvalidOffsetNumber;
else if (HeapTupleIsHotUpdated(heapTuple))
{
Assert(ItemPointerGetBlockNumber(ctid) ==
ItemPointerGetBlockNumber(tid));
scan->xs_next_hot = ItemPointerGetOffsetNumber(ctid);
scan->xs_prev_xmax = HeapTupleHeaderGetXmax(heapTuple->t_data);
}
else
scan->xs_next_hot = InvalidOffsetNumber;
PredicateLockTuple(scan->heapRelation, heapTuple);
LockBuffer(scan->xs_cbuf, BUFFER_LOCK_UNLOCK);
pgstat_count_heap_fetch(scan->indexRelation);
return heapTuple;
}
/*
* If we can't see it, maybe no one else can either. Check to see
* if the tuple is dead to all transactions. If we find that all
* the tuples in the HOT chain are dead, we'll signal the index AM
* to not return that TID on future indexscans.
*/
if (scan->xs_hot_dead &&
HeapTupleSatisfiesVacuum(heapTuple->t_data, RecentGlobalXmin,
scan->xs_cbuf) != HEAPTUPLE_DEAD)
scan->xs_hot_dead = false;
/*
* Check to see if HOT chain continues past this tuple; if so
* fetch the next offnum (we don't bother storing it into
* xs_next_hot, but must store xs_prev_xmax), and loop around.
*/
if (HeapTupleIsHotUpdated(heapTuple))
{
Assert(ItemPointerGetBlockNumber(ctid) ==
ItemPointerGetBlockNumber(tid));
offnum = ItemPointerGetOffsetNumber(ctid);
at_chain_start = false;
scan->xs_prev_xmax = HeapTupleHeaderGetXmax(heapTuple->t_data);
}
else
break; /* end of chain */
} /* loop over a single HOT chain */
LockBuffer(scan->xs_cbuf, BUFFER_LOCK_UNLOCK);
/* Loop around to ask index AM for another TID */
scan->xs_next_hot = InvalidOffsetNumber;
}
/* Release any held pin on a heap page */
if (BufferIsValid(scan->xs_cbuf))
{
ReleaseBuffer(scan->xs_cbuf);
scan->xs_cbuf = InvalidBuffer;
}
return NULL; /* failure exit */
}
/*
* Optionally prune and repair fragmentation in the specified page.
*
* This is an opportunistic function. It will perform housekeeping
* only if the page heuristically looks like a candidate for pruning and we
* can acquire buffer cleanup lock without blocking.
*
* Note: this is called quite often. It's important that it fall out quickly
* if there's not any use in pruning.
*
* Caller must have pin on the buffer, and must *not* have a lock on it.
*
* OldestXmin is the cutoff XID used to distinguish whether tuples are DEAD
* or RECENTLY_DEAD (see HeapTupleSatisfiesVacuum).
*/
void
heap_page_prune_opt(Relation relation, Buffer buffer, TransactionId OldestXmin)
{
Page page = BufferGetPage(buffer);
Size minfree;
/*
* Let's see if we really need pruning.
*
* Forget it if page is not hinted to contain something prunable that's
* older than OldestXmin.
*/
if (!PageIsPrunable(page, OldestXmin))
return;
/*
* We can't write WAL in recovery mode, so there's no point trying to
* clean the page. The master will likely issue a cleaning WAL record soon
* anyway, so this is no particular loss.
*/
if (RecoveryInProgress())
return;
/*
* We prune when a previous UPDATE failed to find enough space on the page
* for a new tuple version, or when free space falls below the relation's
* fill-factor target (but not less than 10%).
*
* Checking free space here is questionable since we aren't holding any
* lock on the buffer; in the worst case we could get a bogus answer. It's
* unlikely to be *seriously* wrong, though, since reading either pd_lower
* or pd_upper is probably atomic. Avoiding taking a lock seems more
* important than sometimes getting a wrong answer in what is after all
* just a heuristic estimate.
*/
minfree = RelationGetTargetPageFreeSpace(relation,
HEAP_DEFAULT_FILLFACTOR);
minfree = Max(minfree, BLCKSZ / 10);
if (PageIsFull(page) || PageGetHeapFreeSpace(page) < minfree)
{
/* OK, try to get exclusive buffer lock */
if (!ConditionalLockBufferForCleanup(buffer))
return;
/*
* Now that we have buffer lock, get accurate information about the
* page's free space, and recheck the heuristic about whether to
* prune. (We needn't recheck PageIsPrunable, since no one else could
* have pruned while we hold pin.)
*/
if (PageIsFull(page) || PageGetHeapFreeSpace(page) < minfree)
{
TransactionId ignore = InvalidTransactionId; /* return value not
* needed */
/* OK to prune */
(void) heap_page_prune(relation, buffer, OldestXmin, true, &ignore);
}
/* And release buffer lock */
LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
}
}