/*
* _bt_restore_page -- re-enter all the index tuples on a page
*
* The page is freshly init'd, and *from (length len) is a copy of what
* had been its upper part (pd_upper to pd_special). We assume that the
* tuples had been added to the page in item-number order, and therefore
* the one with highest item number appears first (lowest on the page).
*/
static void
_bt_restore_page(Page page, char *from, int len)
{
IndexTupleData itupdata;
Size itemsz;
char *end = from + len;
Item items[MaxIndexTuplesPerPage];
uint16 itemsizes[MaxIndexTuplesPerPage];
int i;
int nitems;
/*
* To get the items back in the original order, we add them to the page in
* reverse. To figure out where one tuple ends and another begins, we
* have to scan them in forward order first.
*/
i = 0;
while (from < end)
{
/* Need to copy tuple header due to alignment considerations */
memcpy(&itupdata, from, sizeof(IndexTupleData));
itemsz = IndexTupleDSize(itupdata);
itemsz = MAXALIGN(itemsz);
items[i] = (Item) from;
itemsizes[i] = itemsz;
i++;
from += itemsz;
}
nitems = i;
for (i = nitems - 1; i >= 0; i--)
{
if (PageAddItem(page, items[i], itemsizes[i], nitems - i,
false, false) == InvalidOffsetNumber)
elog(PANIC, "_bt_restore_page: cannot add item to page");
from += itemsz;
}
}
/*
* _hash_squeezebucket(rel, bucket)
*
* Try to squeeze the tuples onto pages occurring earlier in the
* bucket chain in an attempt to free overflow pages. When we start
* the "squeezing", the page from which we start taking tuples (the
* "read" page) is the last bucket in the bucket chain and the page
* onto which we start squeezing tuples (the "write" page) is the
* first page in the bucket chain. The read page works backward and
* the write page works forward; the procedure terminates when the
* read page and write page are the same page.
*
* At completion of this procedure, it is guaranteed that all pages in
* the bucket are nonempty, unless the bucket is totally empty (in
* which case all overflow pages will be freed). The original implementation
* required that to be true on entry as well, but it's a lot easier for
* callers to leave empty overflow pages and let this guy clean it up.
*
* Caller must hold exclusive lock on the target bucket. This allows
* us to safely lock multiple pages in the bucket.
*
* Since this function is invoked in VACUUM, we provide an access strategy
* parameter that controls fetches of the bucket pages.
*/
void
_hash_squeezebucket(Relation rel,
Bucket bucket,
BlockNumber bucket_blkno,
BufferAccessStrategy bstrategy)
{
BlockNumber wblkno;
BlockNumber rblkno;
Buffer wbuf;
Buffer rbuf;
Page wpage;
Page rpage;
HashPageOpaque wopaque;
HashPageOpaque ropaque;
bool wbuf_dirty;
/*
* start squeezing into the base bucket page.
*/
wblkno = bucket_blkno;
wbuf = _hash_getbuf_with_strategy(rel,
wblkno,
HASH_WRITE,
LH_BUCKET_PAGE,
bstrategy);
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
/*
* if there aren't any overflow pages, there's nothing to squeeze.
*/
if (!BlockNumberIsValid(wopaque->hasho_nextblkno))
{
_hash_relbuf(rel, wbuf);
return;
}
/*
* Find the last page in the bucket chain by starting at the base bucket
* page and working forward. Note: we assume that a hash bucket chain is
* usually smaller than the buffer ring being used by VACUUM, else using
* the access strategy here would be counterproductive.
*/
rbuf = InvalidBuffer;
ropaque = wopaque;
do
{
rblkno = ropaque->hasho_nextblkno;
if (rbuf != InvalidBuffer)
_hash_relbuf(rel, rbuf);
rbuf = _hash_getbuf_with_strategy(rel,
rblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
} while (BlockNumberIsValid(ropaque->hasho_nextblkno));
/*
* squeeze the tuples.
*/
wbuf_dirty = false;
for (;;)
{
OffsetNumber roffnum;
OffsetNumber maxroffnum;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
/* Scan each tuple in "read" page */
maxroffnum = PageGetMaxOffsetNumber(rpage);
for (roffnum = FirstOffsetNumber;
roffnum <= maxroffnum;
roffnum = OffsetNumberNext(roffnum))
{
IndexTuple itup;
Size itemsz;
itup = (IndexTuple) PageGetItem(rpage,
PageGetItemId(rpage, roffnum));
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
/*
* Walk up the bucket chain, looking for a page big enough for
* this item. Exit if we reach the read page.
*/
while (PageGetFreeSpace(wpage) < itemsz)
{
Assert(!PageIsEmpty(wpage));
wblkno = wopaque->hasho_nextblkno;
Assert(BlockNumberIsValid(wblkno));
if (wbuf_dirty)
_hash_wrtbuf(rel, wbuf);
else
_hash_relbuf(rel, wbuf);
/* nothing more to do if we reached the read page */
if (rblkno == wblkno)
{
if (ndeletable > 0)
{
/* Delete tuples we already moved off read page */
PageIndexMultiDelete(rpage, deletable, ndeletable);
_hash_wrtbuf(rel, rbuf);
}
else
_hash_relbuf(rel, rbuf);
return;
}
wbuf = _hash_getbuf_with_strategy(rel,
wblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
Assert(wopaque->hasho_bucket == bucket);
wbuf_dirty = false;
}
/*
* we have found room so insert on the "write" page, being careful
* to preserve hashkey ordering. (If we insert many tuples into
* the same "write" page it would be worth qsort'ing instead of
* doing repeated _hash_pgaddtup.)
*/
(void) _hash_pgaddtup(rel, wbuf, itemsz, itup);
wbuf_dirty = true;
/* remember tuple for deletion from "read" page */
deletable[ndeletable++] = roffnum;
}
/*
* If we reach here, there are no live tuples on the "read" page ---
* it was empty when we got to it, or we moved them all. So we can
* just free the page without bothering with deleting tuples
* individually. Then advance to the previous "read" page.
*
* Tricky point here: if our read and write pages are adjacent in the
* bucket chain, our write lock on wbuf will conflict with
* _hash_freeovflpage's attempt to update the sibling links of the
* removed page. However, in that case we are done anyway, so we can
* simply drop the write lock before calling _hash_freeovflpage.
*/
rblkno = ropaque->hasho_prevblkno;
Assert(BlockNumberIsValid(rblkno));
/* are we freeing the page adjacent to wbuf? */
if (rblkno == wblkno)
{
/* yes, so release wbuf lock first */
if (wbuf_dirty)
_hash_wrtbuf(rel, wbuf);
else
_hash_relbuf(rel, wbuf);
/* free this overflow page (releases rbuf) */
_hash_freeovflpage(rel, rbuf, bstrategy);
/* done */
return;
}
/* free this overflow page, then get the previous one */
_hash_freeovflpage(rel, rbuf, bstrategy);
rbuf = _hash_getbuf_with_strategy(rel,
rblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
}
/* NOTREACHED */
}
/*----------
* Add an item to a disk page from the sort output.
*
* We must be careful to observe the page layout conventions of nbtsearch.c:
* - rightmost pages start data items at P_HIKEY instead of at P_FIRSTKEY.
* - on non-leaf pages, the key portion of the first item need not be
* stored, we should store only the link.
*
* A leaf page being built looks like:
*
* +----------------+---------------------------------+
* | PageHeaderData | linp0 linp1 linp2 ... |
* +-----------+----+---------------------------------+
* | ... linpN | |
* +-----------+--------------------------------------+
* | ^ last |
* | |
* +-------------+------------------------------------+
* | | itemN ... |
* +-------------+------------------+-----------------+
* | ... item3 item2 item1 | "special space" |
* +--------------------------------+-----------------+
*
* Contrast this with the diagram in bufpage.h; note the mismatch
* between linps and items. This is because we reserve linp0 as a
* placeholder for the pointer to the "high key" item; when we have
* filled up the page, we will set linp0 to point to itemN and clear
* linpN. On the other hand, if we find this is the last (rightmost)
* page, we leave the items alone and slide the linp array over.
*
* 'last' pointer indicates the last offset added to the page.
*----------
*/
static void
_bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup)
{
Page npage;
BlockNumber nblkno;
OffsetNumber last_off;
Size pgspc;
Size itupsz;
/*
* This is a handy place to check for cancel interrupts during the btree
* load phase of index creation.
*/
CHECK_FOR_INTERRUPTS();
npage = state->btps_page;
nblkno = state->btps_blkno;
last_off = state->btps_lastoff;
pgspc = PageGetFreeSpace(npage);
itupsz = IndexTupleDSize(*itup);
itupsz = MAXALIGN(itupsz);
/*
* Check whether the item can fit on a btree page at all. (Eventually, we
* ought to try to apply TOAST methods if not.) We actually need to be
* able to fit three items on every page, so restrict any one item to 1/3
* the per-page available space. Note that at this point, itupsz doesn't
* include the ItemId.
*
* NOTE: similar code appears in _bt_insertonpg() to defend against
* oversize items being inserted into an already-existing index. But
* during creation of an index, we don't go through there.
*/
if (itupsz > BTMaxItemSize(npage))
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index row size %zu exceeds maximum %zu for index \"%s\"",
itupsz, BTMaxItemSize(npage),
RelationGetRelationName(wstate->index)),
errhint("Values larger than 1/3 of a buffer page cannot be indexed.\n"
"Consider a function index of an MD5 hash of the value, "
"or use full text indexing."),
errtableconstraint(wstate->heap,
RelationGetRelationName(wstate->index))));
/*
* Check to see if page is "full". It's definitely full if the item won't
* fit. Otherwise, compare to the target freespace derived from the
* fillfactor. However, we must put at least two items on each page, so
* disregard fillfactor if we don't have that many.
*/
if (pgspc < itupsz || (pgspc < state->btps_full && last_off > P_FIRSTKEY))
{
/*
* Finish off the page and write it out.
*/
Page opage = npage;
BlockNumber oblkno = nblkno;
ItemId ii;
ItemId hii;
IndexTuple oitup;
/* Create new page of same level */
npage = _bt_blnewpage(state->btps_level);
/* and assign it a page position */
nblkno = wstate->btws_pages_alloced++;
/*
* We copy the last item on the page into the new page, and then
* rearrange the old page so that the 'last item' becomes its high key
* rather than a true data item. There had better be at least two
* items on the page already, else the page would be empty of useful
* data.
*/
Assert(last_off > P_FIRSTKEY);
ii = PageGetItemId(opage, last_off);
oitup = (IndexTuple) PageGetItem(opage, ii);
_bt_sortaddtup(npage, ItemIdGetLength(ii), oitup, P_FIRSTKEY);
/*
* Move 'last' into the high key position on opage
*/
hii = PageGetItemId(opage, P_HIKEY);
*hii = *ii;
ItemIdSetUnused(ii); /* redundant */
((PageHeader) opage)->pd_lower -= sizeof(ItemIdData);
/*
* Link the old page into its parent, using its minimum key. If we
* don't have a parent, we have to create one; this adds a new btree
* level.
*/
if (state->btps_next == NULL)
state->btps_next = _bt_pagestate(wstate, state->btps_level + 1);
Assert(state->btps_minkey != NULL);
ItemPointerSet(&(state->btps_minkey->t_tid), oblkno, P_HIKEY);
_bt_buildadd(wstate, state->btps_next, state->btps_minkey);
pfree(state->btps_minkey);
/*
* Save a copy of the minimum key for the new page. We have to copy
* it off the old page, not the new one, in case we are not at leaf
* level.
*/
state->btps_minkey = CopyIndexTuple(oitup);
/*
* Set the sibling links for both pages.
*/
{
BTPageOpaque oopaque = (BTPageOpaque) PageGetSpecialPointer(opage);
BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(npage);
oopaque->btpo_next = nblkno;
nopaque->btpo_prev = oblkno;
nopaque->btpo_next = P_NONE; /* redundant */
}
/*
* Write out the old page. We never need to touch it again, so we can
* free the opage workspace too.
*/
_bt_blwritepage(wstate, opage, oblkno);
/*
* Reset last_off to point to new page
*/
last_off = P_FIRSTKEY;
}
/*
* If the new item is the first for its page, stash a copy for later. Note
* this will only happen for the first item on a level; on later pages,
* the first item for a page is copied from the prior page in the code
* above.
*/
if (last_off == P_HIKEY)
{
Assert(state->btps_minkey == NULL);
state->btps_minkey = CopyIndexTuple(itup);
}
/*
* Add the new item into the current page.
*/
last_off = OffsetNumberNext(last_off);
_bt_sortaddtup(npage, itupsz, itup, last_off);
state->btps_page = npage;
state->btps_blkno = nblkno;
state->btps_lastoff = last_off;
}
/*
* _hash_splitbucket -- split 'obucket' into 'obucket' and 'nbucket'
*
* 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, and compress out any free space in the old
* bucket.
*
* The caller must hold exclusive 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.)
*/
static void
_hash_splitbucket(Relation rel,
Buffer metabuf,
Bucket obucket,
Bucket nbucket,
BlockNumber start_oblkno,
BlockNumber start_nblkno,
uint32 maxbucket,
uint32 highmask,
uint32 lowmask)
{
Bucket bucket;
Buffer obuf;
Buffer nbuf;
BlockNumber oblkno;
BlockNumber nblkno;
bool null;
Datum datum;
HashItem hitem;
HashPageOpaque oopaque;
HashPageOpaque nopaque;
IndexTuple itup;
Size itemsz;
OffsetNumber ooffnum;
OffsetNumber noffnum;
OffsetNumber omaxoffnum;
Page opage;
Page npage;
TupleDesc itupdesc = RelationGetDescr(rel);
/*
* It should be okay to simultaneously write-lock pages from each
* bucket, since no one else can be trying to acquire buffer lock
* on pages of either bucket.
*/
oblkno = start_oblkno;
nblkno = start_nblkno;
obuf = _hash_getbuf(rel, oblkno, HASH_WRITE);
nbuf = _hash_getbuf(rel, nblkno, HASH_WRITE);
opage = BufferGetPage(obuf);
npage = BufferGetPage(nbuf);
_hash_checkpage(rel, opage, LH_BUCKET_PAGE);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
/* initialize the new bucket's primary page */
_hash_pageinit(npage, BufferGetPageSize(nbuf));
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
nopaque->hasho_prevblkno = InvalidBlockNumber;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = nbucket;
nopaque->hasho_flag = LH_BUCKET_PAGE;
nopaque->hasho_filler = HASHO_FILL;
/*
* 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.
*/
ooffnum = FirstOffsetNumber;
omaxoffnum = PageGetMaxOffsetNumber(opage);
for (;;)
{
/*
* at each iteration through this loop, each of these variables
* should be up-to-date: obuf opage oopaque ooffnum omaxoffnum
*/
/* check if we're at the end of the page */
if (ooffnum > omaxoffnum)
{
/* at end of page, but check for an(other) overflow page */
oblkno = oopaque->hasho_nextblkno;
if (!BlockNumberIsValid(oblkno))
break;
/*
* we ran out of tuples on this particular page, but we
* have more overflow pages; advance to next page.
*/
_hash_wrtbuf(rel, obuf);
obuf = _hash_getbuf(rel, oblkno, HASH_WRITE);
opage = BufferGetPage(obuf);
_hash_checkpage(rel, opage, LH_OVERFLOW_PAGE);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
ooffnum = FirstOffsetNumber;
omaxoffnum = PageGetMaxOffsetNumber(opage);
continue;
}
/*
* Re-hash the tuple to determine which bucket it now belongs in.
*
* It is annoying to call the hash function while holding locks,
* but releasing and relocking the page for each tuple is unappealing
* too.
*/
hitem = (HashItem) PageGetItem(opage, PageGetItemId(opage, ooffnum));
itup = &(hitem->hash_itup);
datum = index_getattr(itup, 1, itupdesc, &null);
Assert(!null);
bucket = _hash_hashkey2bucket(_hash_datum2hashkey(rel, datum),
maxbucket, highmask, lowmask);
if (bucket == nbucket)
{
/*
* 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(hitem->hash_itup)
+ (sizeof(HashItemData) - sizeof(IndexTupleData));
itemsz = MAXALIGN(itemsz);
if (PageGetFreeSpace(npage) < itemsz)
{
/* write out nbuf and drop lock, but keep pin */
_hash_chgbufaccess(rel, nbuf, HASH_WRITE, HASH_NOLOCK);
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf);
npage = BufferGetPage(nbuf);
_hash_checkpage(rel, npage, LH_OVERFLOW_PAGE);
/* we don't need nopaque within the loop */
}
noffnum = OffsetNumberNext(PageGetMaxOffsetNumber(npage));
if (PageAddItem(npage, (Item) hitem, itemsz, noffnum, LP_USED)
== InvalidOffsetNumber)
elog(ERROR, "failed to add index item to \"%s\"",
RelationGetRelationName(rel));
/*
* now delete the tuple from the old bucket. after this
* section of code, 'ooffnum' will actually point to the
* ItemId to which we would point if we had advanced it before
* the deletion (PageIndexTupleDelete repacks the ItemId
* array). this also means that 'omaxoffnum' is exactly one
* less than it used to be, so we really can just decrement it
* instead of calling PageGetMaxOffsetNumber.
*/
PageIndexTupleDelete(opage, ooffnum);
omaxoffnum = OffsetNumberPrev(omaxoffnum);
}
else
{
/*
* the tuple stays on this page. we didn't move anything, so
* we didn't delete anything and therefore we don't have to
* change 'omaxoffnum'.
*/
Assert(bucket == obucket);
ooffnum = OffsetNumberNext(ooffnum);
}
}
/*
* We're at the end of the old bucket chain, so we're done partitioning
* the tuples. Before quitting, call _hash_squeezebucket to ensure the
* tuples remaining in the old bucket (including the overflow pages) are
* packed as tightly as possible. The new bucket is already tight.
*/
_hash_wrtbuf(rel, obuf);
_hash_wrtbuf(rel, nbuf);
_hash_squeezebucket(rel, obucket, start_oblkno);
}
/*
* replay move of page contents for squeeze operation of hash index
*/
static void
hash_xlog_move_page_contents(XLogReaderState *record)
{
XLogRecPtr lsn = record->EndRecPtr;
xl_hash_move_page_contents *xldata = (xl_hash_move_page_contents *) XLogRecGetData(record);
Buffer bucketbuf = InvalidBuffer;
Buffer writebuf = InvalidBuffer;
Buffer deletebuf = InvalidBuffer;
XLogRedoAction action;
/*
* Ensure we have a cleanup lock on primary bucket page before we start
* with the actual replay operation. This is to ensure that neither a
* scan can start nor a scan can be already-in-progress during the replay
* of this operation. If we allow scans during this operation, then they
* can miss some records or show the same record multiple times.
*/
if (xldata->is_prim_bucket_same_wrt)
action = XLogReadBufferForRedoExtended(record, 1, RBM_NORMAL, true, &writebuf);
else
{
/*
* we don't care for return value as the purpose of reading bucketbuf
* is to ensure a cleanup lock on primary bucket page.
*/
(void) XLogReadBufferForRedoExtended(record, 0, RBM_NORMAL, true, &bucketbuf);
action = XLogReadBufferForRedo(record, 1, &writebuf);
}
/* replay the record for adding entries in overflow buffer */
if (action == BLK_NEEDS_REDO)
{
Page writepage;
char *begin;
char *data;
Size datalen;
uint16 ninserted = 0;
data = begin = XLogRecGetBlockData(record, 1, &datalen);
writepage = (Page) BufferGetPage(writebuf);
if (xldata->ntups > 0)
{
OffsetNumber *towrite = (OffsetNumber *) data;
data += sizeof(OffsetNumber) * xldata->ntups;
while (data - begin < datalen)
{
IndexTuple itup = (IndexTuple) data;
Size itemsz;
OffsetNumber l;
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
data += itemsz;
l = PageAddItem(writepage, (Item) itup, itemsz, towrite[ninserted], false, false);
if (l == InvalidOffsetNumber)
elog(ERROR, "hash_xlog_move_page_contents: failed to add item to hash index page, size %d bytes",
(int) itemsz);
ninserted++;
}
}
/*
* number of tuples inserted must be same as requested in REDO record.
*/
Assert(ninserted == xldata->ntups);
PageSetLSN(writepage, lsn);
MarkBufferDirty(writebuf);
}
/* replay the record for deleting entries from overflow buffer */
if (XLogReadBufferForRedo(record, 2, &deletebuf) == BLK_NEEDS_REDO)
{
Page page;
char *ptr;
Size len;
ptr = XLogRecGetBlockData(record, 2, &len);
page = (Page) BufferGetPage(deletebuf);
if (len > 0)
{
OffsetNumber *unused;
OffsetNumber *unend;
unused = (OffsetNumber *) ptr;
unend = (OffsetNumber *) ((char *) ptr + len);
if ((unend - unused) > 0)
PageIndexMultiDelete(page, unused, unend - unused);
}
PageSetLSN(page, lsn);
MarkBufferDirty(deletebuf);
}
/*
* Replay is complete, now we can release the buffers. We release locks at
* end of replay operation to ensure that we hold lock on primary bucket
* page till end of operation. We can optimize by releasing the lock on
* write buffer as soon as the operation for same is complete, if it is
* not same as primary bucket page, but that doesn't seem to be worth
* complicating the code.
*/
if (BufferIsValid(deletebuf))
UnlockReleaseBuffer(deletebuf);
if (BufferIsValid(writebuf))
UnlockReleaseBuffer(writebuf);
if (BufferIsValid(bucketbuf))
UnlockReleaseBuffer(bucketbuf);
}
/*
* replay squeeze page operation of hash index
*/
static void
hash_xlog_squeeze_page(XLogReaderState *record)
{
XLogRecPtr lsn = record->EndRecPtr;
xl_hash_squeeze_page *xldata = (xl_hash_squeeze_page *) XLogRecGetData(record);
Buffer bucketbuf = InvalidBuffer;
Buffer writebuf;
Buffer ovflbuf;
Buffer prevbuf = InvalidBuffer;
Buffer mapbuf;
XLogRedoAction action;
/*
* Ensure we have a cleanup lock on primary bucket page before we start
* with the actual replay operation. This is to ensure that neither a
* scan can start nor a scan can be already-in-progress during the replay
* of this operation. If we allow scans during this operation, then they
* can miss some records or show the same record multiple times.
*/
if (xldata->is_prim_bucket_same_wrt)
action = XLogReadBufferForRedoExtended(record, 1, RBM_NORMAL, true, &writebuf);
else
{
/*
* we don't care for return value as the purpose of reading bucketbuf
* is to ensure a cleanup lock on primary bucket page.
*/
(void) XLogReadBufferForRedoExtended(record, 0, RBM_NORMAL, true, &bucketbuf);
action = XLogReadBufferForRedo(record, 1, &writebuf);
}
/* replay the record for adding entries in overflow buffer */
if (action == BLK_NEEDS_REDO)
{
Page writepage;
char *begin;
char *data;
Size datalen;
uint16 ninserted = 0;
data = begin = XLogRecGetBlockData(record, 1, &datalen);
writepage = (Page) BufferGetPage(writebuf);
if (xldata->ntups > 0)
{
OffsetNumber *towrite = (OffsetNumber *) data;
data += sizeof(OffsetNumber) * xldata->ntups;
while (data - begin < datalen)
{
IndexTuple itup = (IndexTuple) data;
Size itemsz;
OffsetNumber l;
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
data += itemsz;
l = PageAddItem(writepage, (Item) itup, itemsz, towrite[ninserted], false, false);
if (l == InvalidOffsetNumber)
elog(ERROR, "hash_xlog_squeeze_page: failed to add item to hash index page, size %d bytes",
(int) itemsz);
ninserted++;
}
}
/*
* number of tuples inserted must be same as requested in REDO record.
*/
Assert(ninserted == xldata->ntups);
/*
* if the page on which are adding tuples is a page previous to freed
* overflow page, then update its nextblno.
*/
if (xldata->is_prev_bucket_same_wrt)
{
HashPageOpaque writeopaque = (HashPageOpaque) PageGetSpecialPointer(writepage);
writeopaque->hasho_nextblkno = xldata->nextblkno;
}
PageSetLSN(writepage, lsn);
MarkBufferDirty(writebuf);
}
/* replay the record for initializing overflow buffer */
if (XLogReadBufferForRedo(record, 2, &ovflbuf) == BLK_NEEDS_REDO)
{
Page ovflpage;
ovflpage = BufferGetPage(ovflbuf);
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
PageSetLSN(ovflpage, lsn);
MarkBufferDirty(ovflbuf);
}
if (BufferIsValid(ovflbuf))
UnlockReleaseBuffer(ovflbuf);
/* replay the record for page previous to the freed overflow page */
if (!xldata->is_prev_bucket_same_wrt &&
XLogReadBufferForRedo(record, 3, &prevbuf) == BLK_NEEDS_REDO)
{
Page prevpage = BufferGetPage(prevbuf);
HashPageOpaque prevopaque = (HashPageOpaque) PageGetSpecialPointer(prevpage);
prevopaque->hasho_nextblkno = xldata->nextblkno;
PageSetLSN(prevpage, lsn);
MarkBufferDirty(prevbuf);
}
if (BufferIsValid(prevbuf))
UnlockReleaseBuffer(prevbuf);
/* replay the record for page next to the freed overflow page */
if (XLogRecHasBlockRef(record, 4))
{
Buffer nextbuf;
if (XLogReadBufferForRedo(record, 4, &nextbuf) == BLK_NEEDS_REDO)
{
Page nextpage = BufferGetPage(nextbuf);
HashPageOpaque nextopaque = (HashPageOpaque) PageGetSpecialPointer(nextpage);
nextopaque->hasho_prevblkno = xldata->prevblkno;
PageSetLSN(nextpage, lsn);
MarkBufferDirty(nextbuf);
}
if (BufferIsValid(nextbuf))
UnlockReleaseBuffer(nextbuf);
}
if (BufferIsValid(writebuf))
UnlockReleaseBuffer(writebuf);
if (BufferIsValid(bucketbuf))
UnlockReleaseBuffer(bucketbuf);
/*
* Note: in normal operation, we'd update the bitmap and meta page while
* still holding lock on the primary bucket page and overflow pages. But
* during replay it's not necessary to hold those locks, since no other
* index updates can be happening concurrently.
*/
/* replay the record for bitmap page */
if (XLogReadBufferForRedo(record, 5, &mapbuf) == BLK_NEEDS_REDO)
{
Page mappage = (Page) BufferGetPage(mapbuf);
uint32 *freep = NULL;
char *data;
uint32 *bitmap_page_bit;
Size datalen;
freep = HashPageGetBitmap(mappage);
data = XLogRecGetBlockData(record, 5, &datalen);
bitmap_page_bit = (uint32 *) data;
CLRBIT(freep, *bitmap_page_bit);
PageSetLSN(mappage, lsn);
MarkBufferDirty(mapbuf);
}
if (BufferIsValid(mapbuf))
UnlockReleaseBuffer(mapbuf);
/* replay the record for meta page */
if (XLogRecHasBlockRef(record, 6))
{
Buffer metabuf;
if (XLogReadBufferForRedo(record, 6, &metabuf) == BLK_NEEDS_REDO)
{
HashMetaPage metap;
Page page;
char *data;
uint32 *firstfree_ovflpage;
Size datalen;
data = XLogRecGetBlockData(record, 6, &datalen);
firstfree_ovflpage = (uint32 *) data;
page = BufferGetPage(metabuf);
metap = HashPageGetMeta(page);
metap->hashm_firstfree = *firstfree_ovflpage;
PageSetLSN(page, lsn);
MarkBufferDirty(metabuf);
}
if (BufferIsValid(metabuf))
UnlockReleaseBuffer(metabuf);
}
}
/*
* _hash_squeezebucket(rel, bucket)
*
* Try to squeeze the tuples onto pages occurring earlier in the
* bucket chain in an attempt to free overflow pages. When we start
* the "squeezing", the page from which we start taking tuples (the
* "read" page) is the last bucket in the bucket chain and the page
* onto which we start squeezing tuples (the "write" page) is the
* first page in the bucket chain. The read page works backward and
* the write page works forward; the procedure terminates when the
* read page and write page are the same page.
*
* At completion of this procedure, it is guaranteed that all pages in
* the bucket are nonempty, unless the bucket is totally empty (in
* which case all overflow pages will be freed). The original implementation
* required that to be true on entry as well, but it's a lot easier for
* callers to leave empty overflow pages and let this guy clean it up.
*
* Caller must acquire cleanup lock on the primary page of the target
* bucket to exclude any scans that are in progress, which could easily
* be confused into returning the same tuple more than once or some tuples
* not at all by the rearrangement we are performing here. To prevent
* any concurrent scan to cross the squeeze scan we use lock chaining
* similar to hasbucketcleanup. Refer comments atop hashbucketcleanup.
*
* We need to retain a pin on the primary bucket to ensure that no concurrent
* split can start.
*
* Since this function is invoked in VACUUM, we provide an access strategy
* parameter that controls fetches of the bucket pages.
*/
void
_hash_squeezebucket(Relation rel,
Bucket bucket,
BlockNumber bucket_blkno,
Buffer bucket_buf,
BufferAccessStrategy bstrategy)
{
BlockNumber wblkno;
BlockNumber rblkno;
Buffer wbuf;
Buffer rbuf;
Page wpage;
Page rpage;
HashPageOpaque wopaque;
HashPageOpaque ropaque;
/*
* start squeezing into the primary bucket page.
*/
wblkno = bucket_blkno;
wbuf = bucket_buf;
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
/*
* if there aren't any overflow pages, there's nothing to squeeze. caller
* is responsible for releasing the pin on primary bucket page.
*/
if (!BlockNumberIsValid(wopaque->hasho_nextblkno))
{
LockBuffer(wbuf, BUFFER_LOCK_UNLOCK);
return;
}
/*
* Find the last page in the bucket chain by starting at the base bucket
* page and working forward. Note: we assume that a hash bucket chain is
* usually smaller than the buffer ring being used by VACUUM, else using
* the access strategy here would be counterproductive.
*/
rbuf = InvalidBuffer;
ropaque = wopaque;
do
{
rblkno = ropaque->hasho_nextblkno;
if (rbuf != InvalidBuffer)
_hash_relbuf(rel, rbuf);
rbuf = _hash_getbuf_with_strategy(rel,
rblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
} while (BlockNumberIsValid(ropaque->hasho_nextblkno));
/*
* squeeze the tuples.
*/
for (;;)
{
OffsetNumber roffnum;
OffsetNumber maxroffnum;
OffsetNumber deletable[MaxOffsetNumber];
IndexTuple itups[MaxIndexTuplesPerPage];
Size tups_size[MaxIndexTuplesPerPage];
OffsetNumber itup_offsets[MaxIndexTuplesPerPage];
uint16 ndeletable = 0;
uint16 nitups = 0;
Size all_tups_size = 0;
int i;
bool retain_pin = false;
readpage:
/* Scan each tuple in "read" page */
maxroffnum = PageGetMaxOffsetNumber(rpage);
for (roffnum = FirstOffsetNumber;
roffnum <= maxroffnum;
roffnum = OffsetNumberNext(roffnum))
{
IndexTuple itup;
Size itemsz;
/* skip dead tuples */
if (ItemIdIsDead(PageGetItemId(rpage, roffnum)))
continue;
itup = (IndexTuple) PageGetItem(rpage,
PageGetItemId(rpage, roffnum));
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
/*
* Walk up the bucket chain, looking for a page big enough for
* this item and all other accumulated items. Exit if we reach
* the read page.
*/
while (PageGetFreeSpaceForMultipleTuples(wpage, nitups + 1) < (all_tups_size + itemsz))
{
Buffer next_wbuf = InvalidBuffer;
bool tups_moved = false;
Assert(!PageIsEmpty(wpage));
if (wblkno == bucket_blkno)
retain_pin = true;
wblkno = wopaque->hasho_nextblkno;
Assert(BlockNumberIsValid(wblkno));
/* don't need to move to next page if we reached the read page */
if (wblkno != rblkno)
next_wbuf = _hash_getbuf_with_strategy(rel,
wblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
if (nitups > 0)
{
Assert(nitups == ndeletable);
/*
* This operation needs to log multiple tuples, prepare
* WAL for that.
*/
if (RelationNeedsWAL(rel))
XLogEnsureRecordSpace(0, 3 + nitups);
START_CRIT_SECTION();
/*
* we have to insert tuples on the "write" page, being
* careful to preserve hashkey ordering. (If we insert
* many tuples into the same "write" page it would be
* worth qsort'ing them).
*/
_hash_pgaddmultitup(rel, wbuf, itups, itup_offsets, nitups);
MarkBufferDirty(wbuf);
/* Delete tuples we already moved off read page */
PageIndexMultiDelete(rpage, deletable, ndeletable);
MarkBufferDirty(rbuf);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
XLogRecPtr recptr;
xl_hash_move_page_contents xlrec;
xlrec.ntups = nitups;
xlrec.is_prim_bucket_same_wrt = (wbuf == bucket_buf) ? true : false;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHashMovePageContents);
/*
* bucket buffer needs to be registered to ensure that
* we can acquire a cleanup lock on it during replay.
*/
if (!xlrec.is_prim_bucket_same_wrt)
XLogRegisterBuffer(0, bucket_buf, REGBUF_STANDARD | REGBUF_NO_IMAGE);
XLogRegisterBuffer(1, wbuf, REGBUF_STANDARD);
XLogRegisterBufData(1, (char *) itup_offsets,
nitups * sizeof(OffsetNumber));
for (i = 0; i < nitups; i++)
XLogRegisterBufData(1, (char *) itups[i], tups_size[i]);
XLogRegisterBuffer(2, rbuf, REGBUF_STANDARD);
XLogRegisterBufData(2, (char *) deletable,
ndeletable * sizeof(OffsetNumber));
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_MOVE_PAGE_CONTENTS);
PageSetLSN(BufferGetPage(wbuf), recptr);
PageSetLSN(BufferGetPage(rbuf), recptr);
}
END_CRIT_SECTION();
tups_moved = true;
}
/*
* release the lock on previous page after acquiring the lock
* on next page
*/
if (retain_pin)
LockBuffer(wbuf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, wbuf);
/* nothing more to do if we reached the read page */
if (rblkno == wblkno)
{
_hash_relbuf(rel, rbuf);
return;
}
wbuf = next_wbuf;
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
Assert(wopaque->hasho_bucket == bucket);
retain_pin = false;
/* be tidy */
for (i = 0; i < nitups; i++)
pfree(itups[i]);
nitups = 0;
all_tups_size = 0;
ndeletable = 0;
/*
* after moving the tuples, rpage would have been compacted,
* so we need to rescan it.
*/
if (tups_moved)
goto readpage;
}
/* remember tuple for deletion from "read" page */
deletable[ndeletable++] = roffnum;
/*
* we need a copy of index tuples as they can be freed as part of
* overflow page, however we need them to write a WAL record in
* _hash_freeovflpage.
*/
itups[nitups] = CopyIndexTuple(itup);
tups_size[nitups++] = itemsz;
all_tups_size += itemsz;
}
/*
* If we reach here, there are no live tuples on the "read" page ---
* it was empty when we got to it, or we moved them all. So we can
* just free the page without bothering with deleting tuples
* individually. Then advance to the previous "read" page.
*
* Tricky point here: if our read and write pages are adjacent in the
* bucket chain, our write lock on wbuf will conflict with
* _hash_freeovflpage's attempt to update the sibling links of the
* removed page. In that case, we don't need to lock it again.
*/
rblkno = ropaque->hasho_prevblkno;
Assert(BlockNumberIsValid(rblkno));
/* free this overflow page (releases rbuf) */
_hash_freeovflpage(rel, bucket_buf, rbuf, wbuf, itups, itup_offsets,
tups_size, nitups, bstrategy);
/* be tidy */
for (i = 0; i < nitups; i++)
pfree(itups[i]);
/* are we freeing the page adjacent to wbuf? */
if (rblkno == wblkno)
{
/* retain the pin on primary bucket page till end of bucket scan */
if (wblkno == bucket_blkno)
LockBuffer(wbuf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, wbuf);
return;
}
rbuf = _hash_getbuf_with_strategy(rel,
rblkno,
HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
}
/* NOTREACHED */
}
/*
* _hash_splitbucket -- split 'obucket' into 'obucket' and 'nbucket'
*
* 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, and compress out any free space in the old
* bucket.
*
* The caller must hold exclusive 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.)
*/
static void
_hash_splitbucket(Relation rel,
Buffer metabuf,
Bucket obucket,
Bucket nbucket,
BlockNumber start_oblkno,
BlockNumber start_nblkno,
uint32 maxbucket,
uint32 highmask,
uint32 lowmask)
{
BlockNumber oblkno;
BlockNumber nblkno;
Buffer obuf;
Buffer nbuf;
Page opage;
Page npage;
HashPageOpaque oopaque;
HashPageOpaque nopaque;
/*
* It should be okay to simultaneously write-lock pages from each bucket,
* since no one else can be trying to acquire buffer lock on pages of
* either bucket.
*/
oblkno = start_oblkno;
obuf = _hash_getbuf(rel, oblkno, HASH_WRITE, LH_BUCKET_PAGE);
opage = BufferGetPage(obuf);
oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
nblkno = start_nblkno;
nbuf = _hash_getnewbuf(rel, nblkno, MAIN_FORKNUM);
npage = BufferGetPage(nbuf);
/* initialize the new bucket's primary page */
nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
nopaque->hasho_prevblkno = InvalidBlockNumber;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = nbucket;
nopaque->hasho_flag = LH_BUCKET_PAGE;
nopaque->hasho_page_id = HASHO_PAGE_ID;
/*
* 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 (;;)
{
OffsetNumber ooffnum;
OffsetNumber omaxoffnum;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
/* Scan each tuple in old page */
omaxoffnum = PageGetMaxOffsetNumber(opage);
for (ooffnum = FirstOffsetNumber;
ooffnum <= omaxoffnum;
ooffnum = OffsetNumberNext(ooffnum))
{
IndexTuple itup;
Size itemsz;
Bucket bucket;
/*
* 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));
bucket = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup),
maxbucket, highmask, lowmask);
if (bucket == nbucket)
{
/*
* 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(*itup);
itemsz = MAXALIGN(itemsz);
if (PageGetFreeSpace(npage) < itemsz)
{
/* write out nbuf and drop lock, but keep pin */
_hash_chgbufaccess(rel, nbuf, HASH_WRITE, HASH_NOLOCK);
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf);
npage = BufferGetPage(nbuf);
/* we don't need nblkno or nopaque within the loop */
}
/*
* Insert tuple on new page, using _hash_pgaddtup to ensure
* correct ordering by hashkey. This is a tad inefficient
* since we may have to shuffle itempointers repeatedly.
* Possible future improvement: accumulate all the items for
* the new page and qsort them before insertion.
*/
(void) _hash_pgaddtup(rel, nbuf, itemsz, itup);
/*
* Mark tuple for deletion from old page.
*/
deletable[ndeletable++] = ooffnum;
}
else
{
/*
* the tuple stays on this page, so nothing to do.
*/
Assert(bucket == obucket);
}
}
oblkno = oopaque->hasho_nextblkno;
/*
* Done scanning this old page. If we moved any tuples, delete them
* from the old page.
*/
if (ndeletable > 0)
{
PageIndexMultiDelete(opage, deletable, ndeletable);
_hash_wrtbuf(rel, obuf);
}
else
_hash_relbuf(rel, obuf);
/* Exit loop if no more overflow pages in old bucket */
if (!BlockNumberIsValid(oblkno))
break;
/* Else, advance to next old page */
obuf = _hash_getbuf(rel, oblkno, HASH_WRITE, 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. Before quitting, call _hash_squeezebucket to ensure the
* tuples remaining in the old bucket (including the overflow pages) are
* packed as tightly as possible. The new bucket is already tight.
*/
_hash_wrtbuf(rel, nbuf);
_hash_squeezebucket(rel, obucket, start_oblkno, NULL);
}
/*
* _hash_squeezebucket(rel, bucket)
*
* Try to squeeze the tuples onto pages occurring earlier in the
* bucket chain in an attempt to free overflow pages. When we start
* the "squeezing", the page from which we start taking tuples (the
* "read" page) is the last bucket in the bucket chain and the page
* onto which we start squeezing tuples (the "write" page) is the
* first page in the bucket chain. The read page works backward and
* the write page works forward; the procedure terminates when the
* read page and write page are the same page.
*
* At completion of this procedure, it is guaranteed that all pages in
* the bucket are nonempty, unless the bucket is totally empty (in
* which case all overflow pages will be freed). The original implementation
* required that to be true on entry as well, but it's a lot easier for
* callers to leave empty overflow pages and let this guy clean it up.
*
* Caller must hold exclusive lock on the target bucket. This allows
* us to safely lock multiple pages in the bucket.
*/
void
_hash_squeezebucket(Relation rel,
Bucket bucket,
BlockNumber bucket_blkno)
{
Buffer wbuf;
Buffer rbuf = 0;
BlockNumber wblkno;
BlockNumber rblkno;
Page wpage;
Page rpage;
HashPageOpaque wopaque;
HashPageOpaque ropaque;
OffsetNumber woffnum;
OffsetNumber roffnum;
IndexTuple itup;
Size itemsz;
/*
* start squeezing into the base bucket page.
*/
wblkno = bucket_blkno;
wbuf = _hash_getbuf(rel, wblkno, HASH_WRITE);
_hash_checkpage(rel, wbuf, LH_BUCKET_PAGE);
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
/*
* if there aren't any overflow pages, there's nothing to squeeze.
*/
if (!BlockNumberIsValid(wopaque->hasho_nextblkno))
{
_hash_relbuf(rel, wbuf);
return;
}
/*
* find the last page in the bucket chain by starting at the base bucket
* page and working forward.
*/
ropaque = wopaque;
do
{
rblkno = ropaque->hasho_nextblkno;
if (ropaque != wopaque)
_hash_relbuf(rel, rbuf);
rbuf = _hash_getbuf(rel, rblkno, HASH_WRITE);
_hash_checkpage(rel, rbuf, LH_OVERFLOW_PAGE);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
} while (BlockNumberIsValid(ropaque->hasho_nextblkno));
/*
* squeeze the tuples.
*/
roffnum = FirstOffsetNumber;
for (;;)
{
/* this test is needed in case page is empty on entry */
if (roffnum <= PageGetMaxOffsetNumber(rpage))
{
itup = (IndexTuple) PageGetItem(rpage,
PageGetItemId(rpage, roffnum));
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz);
/*
* Walk up the bucket chain, looking for a page big enough for
* this item. Exit if we reach the read page.
*/
while (PageGetFreeSpace(wpage) < itemsz)
{
Assert(!PageIsEmpty(wpage));
wblkno = wopaque->hasho_nextblkno;
Assert(BlockNumberIsValid(wblkno));
_hash_wrtbuf(rel, wbuf);
if (rblkno == wblkno)
{
/* wbuf is already released */
_hash_wrtbuf(rel, rbuf);
return;
}
wbuf = _hash_getbuf(rel, wblkno, HASH_WRITE);
_hash_checkpage(rel, wbuf, LH_OVERFLOW_PAGE);
wpage = BufferGetPage(wbuf);
wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
Assert(wopaque->hasho_bucket == bucket);
}
/*
* we have found room so insert on the "write" page.
*/
woffnum = OffsetNumberNext(PageGetMaxOffsetNumber(wpage));
if (PageAddItem(wpage, (Item) itup, itemsz, woffnum, LP_USED)
== InvalidOffsetNumber)
elog(ERROR, "failed to add index item to \"%s\"",
RelationGetRelationName(rel));
/*
* delete the tuple from the "read" page. PageIndexTupleDelete
* repacks the ItemId array, so 'roffnum' will be "advanced" to
* the "next" ItemId.
*/
PageIndexTupleDelete(rpage, roffnum);
}
/*
* if the "read" page is now empty because of the deletion (or because
* it was empty when we got to it), free it.
*
* Tricky point here: if our read and write pages are adjacent in the
* bucket chain, our write lock on wbuf will conflict with
* _hash_freeovflpage's attempt to update the sibling links of the
* removed page. However, in that case we are done anyway, so we can
* simply drop the write lock before calling _hash_freeovflpage.
*/
if (PageIsEmpty(rpage))
{
rblkno = ropaque->hasho_prevblkno;
Assert(BlockNumberIsValid(rblkno));
/* are we freeing the page adjacent to wbuf? */
if (rblkno == wblkno)
{
/* yes, so release wbuf lock first */
_hash_wrtbuf(rel, wbuf);
/* free this overflow page (releases rbuf) */
_hash_freeovflpage(rel, rbuf);
/* done */
return;
}
/* free this overflow page, then get the previous one */
_hash_freeovflpage(rel, rbuf);
rbuf = _hash_getbuf(rel, rblkno, HASH_WRITE);
_hash_checkpage(rel, rbuf, LH_OVERFLOW_PAGE);
rpage = BufferGetPage(rbuf);
ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
Assert(ropaque->hasho_bucket == bucket);
roffnum = FirstOffsetNumber;
}
}
/* NOTREACHED */
}
/*
* _hash_doinsert() -- Handle insertion of a single index tuple.
*
* This routine is called by the public interface routines, hashbuild
* and hashinsert. By here, itup is completely filled in.
*/
void
_hash_doinsert(Relation rel, IndexTuple itup)
{
Buffer buf;
Buffer metabuf;
HashMetaPage metap;
BlockNumber blkno;
Page page;
HashPageOpaque pageopaque;
Size itemsz;
bool do_expand;
uint32 hashkey;
Bucket bucket;
/*
* Get the hash key for the item (it's stored in the index tuple itself).
*/
hashkey = _hash_get_indextuple_hashkey(itup);
/* compute item size too */
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz); /* be safe, PageAddItem will do this but we
* need to be consistent */
/*
* Acquire shared split lock so we can compute the target bucket safely
* (see README).
*/
_hash_getlock(rel, 0, HASH_SHARE);
/* Read the metapage */
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_READ, LH_META_PAGE);
metap = HashPageGetMeta(BufferGetPage(metabuf));
/*
* Check whether the item can fit on a hash page at all. (Eventually, we
* ought to try to apply TOAST methods if not.) Note that at this point,
* itemsz doesn't include the ItemId.
*
* XXX this is useless code if we are only storing hash keys.
*/
if (itemsz > HashMaxItemSize((Page) metap))
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index row size %lu exceeds hash maximum %lu",
(unsigned long) itemsz,
(unsigned long) HashMaxItemSize((Page) metap)),
errhint("Values larger than a buffer page cannot be indexed.")));
/*
* Compute the target bucket number, and convert to block number.
*/
bucket = _hash_hashkey2bucket(hashkey,
metap->hashm_maxbucket,
metap->hashm_highmask,
metap->hashm_lowmask);
blkno = BUCKET_TO_BLKNO(metap, bucket);
/* release lock on metapage, but keep pin since we'll need it again */
_hash_chgbufaccess(rel, metabuf, HASH_READ, HASH_NOLOCK);
/*
* Acquire share lock on target bucket; then we can release split lock.
*/
_hash_getlock(rel, blkno, HASH_SHARE);
_hash_droplock(rel, 0, HASH_SHARE);
/* Fetch the primary bucket page for the bucket */
buf = _hash_getbuf(rel, blkno, HASH_WRITE, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(pageopaque->hasho_bucket == bucket);
/* Do the insertion */
while (PageGetFreeSpace(page) < itemsz)
{
/*
* no space on this page; check for an overflow page
*/
BlockNumber nextblkno = pageopaque->hasho_nextblkno;
if (BlockNumberIsValid(nextblkno))
{
/*
* ovfl page exists; go get it. if it doesn't have room, we'll
* find out next pass through the loop test above.
*/
_hash_relbuf(rel, buf);
buf = _hash_getbuf(rel, nextblkno, HASH_WRITE, LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
}
else
{
/*
* we're at the end of the bucket chain and we haven't found a
* page with enough room. allocate a new overflow page.
*/
/* release our write lock without modifying buffer */
_hash_chgbufaccess(rel, buf, HASH_READ, HASH_NOLOCK);
/* chain to a new overflow page */
buf = _hash_addovflpage(rel, metabuf, buf);
page = BufferGetPage(buf);
/* should fit now, given test above */
Assert(PageGetFreeSpace(page) >= itemsz);
}
pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(pageopaque->hasho_flag == LH_OVERFLOW_PAGE);
Assert(pageopaque->hasho_bucket == bucket);
}
/* found page with enough space, so add the item here */
(void) _hash_pgaddtup(rel, buf, itemsz, itup);
/* write and release the modified page */
_hash_wrtbuf(rel, buf);
/* We can drop the bucket lock now */
_hash_droplock(rel, blkno, HASH_SHARE);
/*
* Write-lock the metapage so we can increment the tuple count. After
* incrementing it, check to see if it's time for a split.
*/
_hash_chgbufaccess(rel, metabuf, HASH_NOLOCK, HASH_WRITE);
metap->hashm_ntuples += 1;
/* Make sure this stays in sync with _hash_expandtable() */
do_expand = metap->hashm_ntuples >
(double) metap->hashm_ffactor * (metap->hashm_maxbucket + 1);
/* Write out the metapage and drop lock, but keep pin */
_hash_chgbufaccess(rel, metabuf, HASH_WRITE, HASH_NOLOCK);
/* Attempt to split if a split is needed */
if (do_expand)
_hash_expandtable(rel, metabuf);
/* Finally drop our pin on the metapage */
_hash_dropbuf(rel, metabuf);
}
/*
* _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_doinsert() -- Handle insertion of a single HashItem in the table.
*
* This routine is called by the public interface routines, hashbuild
* and hashinsert. By here, hashitem is completely filled in.
* The datum to be used as a "key" is in the hashitem.
*/
InsertIndexResult
_hash_doinsert(Relation rel, HashItem hitem)
{
Buffer buf;
Buffer metabuf;
HashMetaPage metap;
IndexTuple itup;
BlockNumber itup_blkno;
OffsetNumber itup_off;
InsertIndexResult res;
BlockNumber blkno;
Page page;
HashPageOpaque pageopaque;
Size itemsz;
bool do_expand;
uint32 hashkey;
Bucket bucket;
Datum datum;
bool isnull;
/*
* Compute the hash key for the item. We do this first so as not to
* need to hold any locks while running the hash function.
*/
itup = &(hitem->hash_itup);
if (rel->rd_rel->relnatts != 1)
elog(ERROR, "hash indexes support only one index key");
datum = index_getattr(itup, 1, RelationGetDescr(rel), &isnull);
Assert(!isnull);
hashkey = _hash_datum2hashkey(rel, datum);
/* compute item size too */
itemsz = IndexTupleDSize(hitem->hash_itup)
+ (sizeof(HashItemData) - sizeof(IndexTupleData));
itemsz = MAXALIGN(itemsz); /* be safe, PageAddItem will do this but
* we need to be consistent */
/*
* Acquire shared split lock so we can compute the target bucket
* safely (see README).
*/
_hash_getlock(rel, 0, HASH_SHARE);
/* Read the metapage */
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_READ);
metap = (HashMetaPage) BufferGetPage(metabuf);
_hash_checkpage(rel, (Page) metap, LH_META_PAGE);
/*
* Check whether the item can fit on a hash page at all. (Eventually,
* we ought to try to apply TOAST methods if not.) Note that at this
* point, itemsz doesn't include the ItemId.
*/
if (itemsz > HashMaxItemSize((Page) metap))
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index row size %lu exceeds hash maximum %lu",
(unsigned long) itemsz,
(unsigned long) HashMaxItemSize((Page) metap))));
/*
* Compute the target bucket number, and convert to block number.
*/
bucket = _hash_hashkey2bucket(hashkey,
metap->hashm_maxbucket,
metap->hashm_highmask,
metap->hashm_lowmask);
blkno = BUCKET_TO_BLKNO(metap, bucket);
/* release lock on metapage, but keep pin since we'll need it again */
_hash_chgbufaccess(rel, metabuf, HASH_READ, HASH_NOLOCK);
/*
* Acquire share lock on target bucket; then we can release split lock.
*/
_hash_getlock(rel, blkno, HASH_SHARE);
_hash_droplock(rel, 0, HASH_SHARE);
/* Fetch the primary bucket page for the bucket */
buf = _hash_getbuf(rel, blkno, HASH_WRITE);
page = BufferGetPage(buf);
_hash_checkpage(rel, page, LH_BUCKET_PAGE);
pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(pageopaque->hasho_bucket == bucket);
/* Do the insertion */
while (PageGetFreeSpace(page) < itemsz)
{
/*
* no space on this page; check for an overflow page
*/
BlockNumber nextblkno = pageopaque->hasho_nextblkno;
if (BlockNumberIsValid(nextblkno))
{
/*
* ovfl page exists; go get it. if it doesn't have room,
* we'll find out next pass through the loop test above.
*/
_hash_relbuf(rel, buf);
buf = _hash_getbuf(rel, nextblkno, HASH_WRITE);
page = BufferGetPage(buf);
}
else
{
/*
* we're at the end of the bucket chain and we haven't found a
* page with enough room. allocate a new overflow page.
*/
/* release our write lock without modifying buffer */
_hash_chgbufaccess(rel, buf, HASH_READ, HASH_NOLOCK);
/* chain to a new overflow page */
buf = _hash_addovflpage(rel, metabuf, buf);
page = BufferGetPage(buf);
/* should fit now, given test above */
Assert(PageGetFreeSpace(page) >= itemsz);
}
_hash_checkpage(rel, page, LH_OVERFLOW_PAGE);
pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(pageopaque->hasho_bucket == bucket);
}
/* found page with enough space, so add the item here */
itup_off = _hash_pgaddtup(rel, buf, itemsz, hitem);
itup_blkno = BufferGetBlockNumber(buf);
/* write and release the modified page */
_hash_wrtbuf(rel, buf);
/* We can drop the bucket lock now */
_hash_droplock(rel, blkno, HASH_SHARE);
/*
* Write-lock the metapage so we can increment the tuple count.
* After incrementing it, check to see if it's time for a split.
*/
_hash_chgbufaccess(rel, metabuf, HASH_NOLOCK, HASH_WRITE);
metap->hashm_ntuples += 1;
/* Make sure this stays in sync with _hash_expandtable() */
do_expand = metap->hashm_ntuples >
(double) metap->hashm_ffactor * (metap->hashm_maxbucket + 1);
/* Write out the metapage and drop lock, but keep pin */
_hash_chgbufaccess(rel, metabuf, HASH_WRITE, HASH_NOLOCK);
/* Attempt to split if a split is needed */
if (do_expand)
_hash_expandtable(rel, metabuf);
/* Finally drop our pin on the metapage */
_hash_dropbuf(rel, metabuf);
/* Create the return data structure */
res = (InsertIndexResult) palloc(sizeof(InsertIndexResultData));
ItemPointerSet(&(res->pointerData), itup_blkno, itup_off);
return res;
}
/*
* _hash_doinsert() -- Handle insertion of a single index tuple.
*
* This routine is called by the public interface routines, hashbuild
* and hashinsert. By here, itup is completely filled in.
*/
void
_hash_doinsert(Relation rel, IndexTuple itup)
{
Buffer buf = InvalidBuffer;
Buffer bucket_buf;
Buffer metabuf;
HashMetaPage metap;
BlockNumber blkno;
BlockNumber oldblkno;
bool retry;
Page page;
HashPageOpaque pageopaque;
Size itemsz;
bool do_expand;
uint32 hashkey;
Bucket bucket;
uint32 maxbucket;
uint32 highmask;
uint32 lowmask;
/*
* Get the hash key for the item (it's stored in the index tuple itself).
*/
hashkey = _hash_get_indextuple_hashkey(itup);
/* compute item size too */
itemsz = IndexTupleDSize(*itup);
itemsz = MAXALIGN(itemsz); /* be safe, PageAddItem will do this but we
* need to be consistent */
restart_insert:
/* Read the metapage */
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_READ, LH_META_PAGE);
metap = HashPageGetMeta(BufferGetPage(metabuf));
/*
* Check whether the item can fit on a hash page at all. (Eventually, we
* ought to try to apply TOAST methods if not.) Note that at this point,
* itemsz doesn't include the ItemId.
*
* XXX this is useless code if we are only storing hash keys.
*/
if (itemsz > HashMaxItemSize((Page) metap))
ereport(ERROR,
(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
errmsg("index row size %zu exceeds hash maximum %zu",
itemsz, HashMaxItemSize((Page) metap)),
errhint("Values larger than a buffer page cannot be indexed.")));
oldblkno = InvalidBlockNumber;
retry = false;
/*
* Loop until we get a lock on the correct target bucket.
*/
for (;;)
{
/*
* Compute the target bucket number, and convert to block number.
*/
bucket = _hash_hashkey2bucket(hashkey,
metap->hashm_maxbucket,
metap->hashm_highmask,
metap->hashm_lowmask);
blkno = BUCKET_TO_BLKNO(metap, bucket);
/*
* Copy bucket mapping info now; refer the comment in
* _hash_expandtable 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 metapage lock, but keep pin. */
_hash_chgbufaccess(rel, metabuf, HASH_READ, HASH_NOLOCK);
/*
* If the previous iteration of this loop locked the primary page of
* what is still the correct target bucket, we are done. Otherwise,
* drop any old lock before acquiring the new one.
*/
if (retry)
{
if (oldblkno == blkno)
break;
_hash_relbuf(rel, buf);
}
/* Fetch and lock the primary bucket page for the target bucket */
buf = _hash_getbuf(rel, blkno, HASH_WRITE, LH_BUCKET_PAGE);
/*
* Reacquire metapage lock and check that no bucket split has taken
* place while we were awaiting the bucket lock.
*/
_hash_chgbufaccess(rel, metabuf, HASH_NOLOCK, HASH_READ);
oldblkno = blkno;
retry = true;
}
/* remember the primary bucket buffer to release the pin on it at end. */
bucket_buf = buf;
page = BufferGetPage(buf);
pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(pageopaque->hasho_bucket == bucket);
/*
* If this bucket is in the process of being split, try to finish the
* split before inserting, because that might create room for the
* insertion to proceed without allocating an additional overflow page.
* It's only interesting to finish the split if we're trying to insert
* into the bucket from which we're removing tuples (the "old" bucket),
* not if we're trying to insert into the bucket into which tuples are
* being moved (the "new" bucket).
*/
if (H_BUCKET_BEING_SPLIT(pageopaque) && IsBufferCleanupOK(buf))
{
/* release the lock on bucket buffer, before completing the split. */
_hash_chgbufaccess(rel, buf, HASH_READ, HASH_NOLOCK);
_hash_finish_split(rel, metabuf, buf, pageopaque->hasho_bucket,
maxbucket, highmask, lowmask);
/* release the pin on old and meta buffer. retry for insert. */
_hash_dropbuf(rel, buf);
_hash_dropbuf(rel, metabuf);
goto restart_insert;
}
/* Do the insertion */
while (PageGetFreeSpace(page) < itemsz)
{
/*
* no space on this page; check for an overflow page
*/
BlockNumber nextblkno = pageopaque->hasho_nextblkno;
if (BlockNumberIsValid(nextblkno))
{
/*
* ovfl page exists; go get it. if it doesn't have room, we'll
* find out next pass through the loop test above. we always
* release both the lock and pin if this is an overflow page, but
* only the lock if this is the primary bucket page, since the pin
* on the primary bucket must be retained throughout the scan.
*/
if (buf != bucket_buf)
_hash_relbuf(rel, buf);
else
_hash_chgbufaccess(rel, buf, HASH_READ, HASH_NOLOCK);
buf = _hash_getbuf(rel, nextblkno, HASH_WRITE, LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
}
else
{
/*
* we're at the end of the bucket chain and we haven't found a
* page with enough room. allocate a new overflow page.
*/
/* release our write lock without modifying buffer */
_hash_chgbufaccess(rel, buf, HASH_READ, HASH_NOLOCK);
/* chain to a new overflow page */
buf = _hash_addovflpage(rel, metabuf, buf, (buf == bucket_buf) ? true : false);
page = BufferGetPage(buf);
/* should fit now, given test above */
Assert(PageGetFreeSpace(page) >= itemsz);
}
pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(pageopaque->hasho_flag == LH_OVERFLOW_PAGE);
Assert(pageopaque->hasho_bucket == bucket);
}
/* found page with enough space, so add the item here */
(void) _hash_pgaddtup(rel, buf, itemsz, itup);
/*
* dirty and release the modified page. if the page we modified was an
* overflow page, we also need to separately drop the pin we retained on
* the primary bucket page.
*/
MarkBufferDirty(buf);
_hash_relbuf(rel, buf);
if (buf != bucket_buf)
_hash_dropbuf(rel, bucket_buf);
/*
* Write-lock the metapage so we can increment the tuple count. After
* incrementing it, check to see if it's time for a split.
*/
_hash_chgbufaccess(rel, metabuf, HASH_NOLOCK, HASH_WRITE);
metap->hashm_ntuples += 1;
/* Make sure this stays in sync with _hash_expandtable() */
do_expand = metap->hashm_ntuples >
(double) metap->hashm_ffactor * (metap->hashm_maxbucket + 1);
/* Write out the metapage and drop lock, but keep pin */
_hash_chgbufaccess(rel, metabuf, HASH_WRITE, HASH_NOLOCK);
/* Attempt to split if a split is needed */
if (do_expand)
_hash_expandtable(rel, metabuf);
/* Finally drop our pin on the metapage */
_hash_dropbuf(rel, metabuf);
}