#include "postgres.h"
#include "access/hash.h"
#include "access/hash_xlog.h"
#include "access/xloginsert.h"
#include "miscadmin.h"
#include "port/pg_bitutils.h"
#include "storage/predicate.h"
#include "storage/smgr.h"
#include "utils/rel.h"

Include dependency graph for hashpage.c:

Functions
static bool	_hash_alloc_buckets (Relation rel, BlockNumber firstblock, uint32 nblocks)

static void	_hash_splitbucket (Relation rel, Buffer metabuf, Bucket obucket, Bucket nbucket, Buffer obuf, Buffer nbuf, HTAB *htab, uint32 maxbucket, uint32 highmask, uint32 lowmask)

static void	log_split_page (Relation rel, Buffer buf)

Buffer	_hash_getbuf (Relation rel, BlockNumber blkno, int access, int flags)

Buffer	_hash_getbuf_with_condlock_cleanup (Relation rel, BlockNumber blkno, int flags)

Buffer	_hash_getinitbuf (Relation rel, BlockNumber blkno)

void	_hash_initbuf (Buffer buf, uint32 max_bucket, uint32 num_bucket, uint32 flag, bool initpage)

Buffer	_hash_getnewbuf (Relation rel, BlockNumber blkno, ForkNumber forkNum)

Buffer	_hash_getbuf_with_strategy (Relation rel, BlockNumber blkno, int access, int flags, BufferAccessStrategy bstrategy)

void	_hash_relbuf (Relation rel, Buffer buf)

void	_hash_dropbuf (Relation rel, Buffer buf)

void	_hash_dropscanbuf (Relation rel, HashScanOpaque so)

uint32	_hash_init (Relation rel, double num_tuples, ForkNumber forkNum)

void	_hash_init_metabuffer (Buffer buf, double num_tuples, RegProcedure procid, uint16 ffactor, bool initpage)

void	_hash_pageinit (Page page, Size size)

void	_hash_expandtable (Relation rel, Buffer metabuf)

void	_hash_finish_split (Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket, uint32 maxbucket, uint32 highmask, uint32 lowmask)

HashMetaPage	_hash_getcachedmetap (Relation rel, Buffer *metabuf, bool force_refresh)

Buffer	_hash_getbucketbuf_from_hashkey (Relation rel, uint32 hashkey, int access, HashMetaPage *cachedmetap)

Function Documentation

◆ _hash_alloc_buckets()

static bool _hash_alloc_buckets	(	Relation	rel,
		BlockNumber	firstblock,
		uint32	nblocks
	)

static

Definition at line 992 of file hashpage.c.

 {
     BlockNumber lastblock;
     PGIOAlignedBlock zerobuf;
     Page        page;
     HashPageOpaque ovflopaque;
  
     lastblock = firstblock + nblocks - 1;
  
     /*
      * Check for overflow in block number calculation; if so, we cannot extend
      * the index anymore.
      */
     if (lastblock < firstblock || lastblock == InvalidBlockNumber)
         return false;
  
     page = (Page) zerobuf.data;
  
     /*
      * Initialize the page.  Just zeroing the page won't work; see
      * _hash_freeovflpage for similar usage.  We take care to make the special
      * space valid for the benefit of tools such as pageinspect.
      */
     _hash_pageinit(page, BLCKSZ);
  
     ovflopaque = HashPageGetOpaque(page);
  
     ovflopaque->hasho_prevblkno = InvalidBlockNumber;
     ovflopaque->hasho_nextblkno = InvalidBlockNumber;
     ovflopaque->hasho_bucket = InvalidBucket;
     ovflopaque->hasho_flag = LH_UNUSED_PAGE;
     ovflopaque->hasho_page_id = HASHO_PAGE_ID;
  
     if (RelationNeedsWAL(rel))
         log_newpage(&rel->rd_locator,
                     MAIN_FORKNUM,
                     lastblock,
                     zerobuf.data,
                     true);
  
     PageSetChecksumInplace(page, lastblock);
     smgrextend(RelationGetSmgr(rel), MAIN_FORKNUM, lastblock, zerobuf.data,
                false);
  
     return true;
 }

References _hash_pageinit(), PGIOAlignedBlock::data, HashPageOpaqueData::hasho_bucket, HashPageOpaqueData::hasho_flag, HashPageOpaqueData::hasho_nextblkno, HashPageOpaqueData::hasho_page_id, HASHO_PAGE_ID, HashPageOpaqueData::hasho_prevblkno, HashPageGetOpaque, InvalidBlockNumber, InvalidBucket, LH_UNUSED_PAGE, log_newpage(), MAIN_FORKNUM, PageSetChecksumInplace(), RelationData::rd_locator, RelationGetSmgr(), RelationNeedsWAL, and smgrextend().

Referenced by _hash_expandtable().

◆ _hash_dropbuf()

void _hash_dropbuf	(	Relation	rel,
		Buffer	buf
	)

Definition at line 277 of file hashpage.c.

 {
     ReleaseBuffer(buf);
 }

References buf, and ReleaseBuffer().

Referenced by _hash_doinsert(), _hash_dropscanbuf(), _hash_expandtable(), _hash_finish_split(), _hash_first(), _hash_getbucketbuf_from_hashkey(), _hash_next(), _hash_readprev(), and hashbulkdelete().

◆ _hash_dropscanbuf()

void _hash_dropscanbuf	(	Relation	rel,
		HashScanOpaque	so
	)

Definition at line 289 of file hashpage.c.

 {
     /* release pin we hold on primary bucket page */
     if (BufferIsValid(so->hashso_bucket_buf) &&
         so->hashso_bucket_buf != so->currPos.buf)
         _hash_dropbuf(rel, so->hashso_bucket_buf);
     so->hashso_bucket_buf = InvalidBuffer;
  
     /* release pin we hold on primary bucket page  of bucket being split */
     if (BufferIsValid(so->hashso_split_bucket_buf) &&
         so->hashso_split_bucket_buf != so->currPos.buf)
         _hash_dropbuf(rel, so->hashso_split_bucket_buf);
     so->hashso_split_bucket_buf = InvalidBuffer;
  
     /* release any pin we still hold */
     if (BufferIsValid(so->currPos.buf))
         _hash_dropbuf(rel, so->currPos.buf);
     so->currPos.buf = InvalidBuffer;
  
     /* reset split scan */
     so->hashso_buc_populated = false;
     so->hashso_buc_split = false;
 }

References _hash_dropbuf(), HashScanPosData::buf, BufferIsValid(), HashScanOpaqueData::currPos, HashScanOpaqueData::hashso_buc_populated, HashScanOpaqueData::hashso_buc_split, HashScanOpaqueData::hashso_bucket_buf, HashScanOpaqueData::hashso_split_bucket_buf, and InvalidBuffer.

Referenced by _hash_next(), hashendscan(), and hashrescan().

◆ _hash_expandtable()

void _hash_expandtable	(	Relation	rel,
		Buffer	metabuf
	)

Definition at line 614 of file hashpage.c.

 {
     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 to prevent failure of pg_ceil_log2_32() 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 = HashPageGetOpaque(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.
      *
      * XXX It doesn't make sense to call _hash_getnewbuf first, zeroing the
      * buffer, and then only afterwards check whether we have a cleanup lock.
      * However, since no scan can be accessing the buffer yet, any concurrent
      * accesses will just be from processes like the bgwriter or checkpointer
      * which don't care about its contents, so it doesn't really matter.
      */
     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 = HashPageGetOpaque(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 = HashPageGetOpaque(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);
 }

References _hash_alloc_buckets(), _hash_checkpage(), _hash_dropbuf(), _hash_finish_split(), _hash_get_totalbuckets(), _hash_getbuf_with_condlock_cleanup(), _hash_getnewbuf(), _hash_relbuf(), _hash_spareindex(), _hash_splitbucket(), Assert, BUCKET_TO_BLKNO, BUFFER_LOCK_EXCLUSIVE, BUFFER_LOCK_UNLOCK, BufferGetPage(), END_CRIT_SECTION, xl_hash_split_allocate_page::flags, H_BUCKET_BEING_SPLIT, H_NEEDS_SPLIT_CLEANUP, hashbucketcleanup(), HashMetaPageData::hashm_ffactor, HashMetaPageData::hashm_highmask, HashMetaPageData::hashm_lowmask, HashMetaPageData::hashm_maxbucket, HashMetaPageData::hashm_ntuples, HashMetaPageData::hashm_ovflpoint, HashMetaPageData::hashm_spares, HashPageOpaqueData::hasho_bucket, HashPageOpaqueData::hasho_flag, HashPageOpaqueData::hasho_nextblkno, HashPageOpaqueData::hasho_page_id, HASHO_PAGE_ID, HashPageOpaqueData::hasho_prevblkno, HashPageGetMeta, HashPageGetOpaque, InvalidBlockNumber, IsBufferCleanupOK(), LH_BUCKET_BEING_POPULATED, LH_BUCKET_BEING_SPLIT, LH_BUCKET_PAGE, LH_META_PAGE, LockBuffer(), MAIN_FORKNUM, MarkBufferDirty(), xl_hash_split_allocate_page::new_bucket, xl_hash_split_allocate_page::new_bucket_flag, xl_hash_split_allocate_page::old_bucket_flag, PageSetLSN(), REGBUF_STANDARD, REGBUF_WILL_INIT, RelationNeedsWAL, SizeOfHashSplitAllocPage, START_CRIT_SECTION, XLH_SPLIT_META_UPDATE_MASKS, XLH_SPLIT_META_UPDATE_SPLITPOINT, XLOG_HASH_SPLIT_ALLOCATE_PAGE, XLogBeginInsert(), XLogInsert(), XLogRegisterBufData(), XLogRegisterBuffer(), and XLogRegisterData().

Referenced by _hash_doinsert().

◆ _hash_finish_split()

void _hash_finish_split	(	Relation	rel,
		Buffer	metabuf,
		Buffer	obuf,
		Bucket	obucket,
		uint32	maxbucket,
		uint32	highmask,
		uint32	lowmask
	)

Definition at line 1356 of file hashpage.c.

 {
     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 */
     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 = HashPageGetOpaque(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 = HashPageGetOpaque(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);
 }

References _hash_dropbuf(), _hash_get_newblock_from_oldbucket(), _hash_getbuf(), _hash_relbuf(), _hash_splitbucket(), Assert, BlockNumberIsValid(), BUFFER_LOCK_UNLOCK, BufferGetPage(), ConditionalLockBufferForCleanup(), CurrentMemoryContext, HASHCTL::entrysize, FirstOffsetNumber, HASH_BLOBS, HASH_CONTEXT, hash_create(), hash_destroy(), HASH_ELEM, HASH_ENTER, HASH_READ, hash_search(), HashPageOpaqueData::hasho_bucket, HashPageOpaqueData::hasho_nextblkno, HashPageGetOpaque, HASHCTL::hcxt, InvalidBuffer, HASHCTL::keysize, LH_BUCKET_PAGE, LH_OVERFLOW_PAGE, LockBuffer(), OffsetNumberNext, PageGetItem(), PageGetItemId(), PageGetMaxOffsetNumber(), and IndexTupleData::t_tid.

Referenced by _hash_doinsert(), and _hash_expandtable().

◆ _hash_getbucketbuf_from_hashkey()

Buffer _hash_getbucketbuf_from_hashkey	(	Relation	rel,
		uint32	hashkey,
		int	access,
		HashMetaPage *	cachedmetap
	)

Definition at line 1559 of file hashpage.c.

 {
     HashMetaPage metap;
     Buffer      buf;
     Buffer      metabuf = InvalidBuffer;
     Page        page;
     Bucket      bucket;
     BlockNumber blkno;
     HashPageOpaque opaque;
  
     /* We read from target bucket buffer, hence locking is must. */
     Assert(access == HASH_READ || access == HASH_WRITE);
  
     metap = _hash_getcachedmetap(rel, &metabuf, false);
     Assert(metap != NULL);
  
     /*
      * 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);
  
         /* Fetch the primary bucket page for the bucket */
         buf = _hash_getbuf(rel, blkno, access, LH_BUCKET_PAGE);
         page = BufferGetPage(buf);
         opaque = HashPageGetOpaque(page);
         Assert(opaque->hasho_bucket == bucket);
         Assert(opaque->hasho_prevblkno != InvalidBlockNumber);
  
         /*
          * If this bucket hasn't been split, we're done.
          */
         if (opaque->hasho_prevblkno <= metap->hashm_maxbucket)
             break;
  
         /* Drop lock on this buffer, update cached metapage, and retry. */
         _hash_relbuf(rel, buf);
         metap = _hash_getcachedmetap(rel, &metabuf, true);
         Assert(metap != NULL);
     }
  
     if (BufferIsValid(metabuf))
         _hash_dropbuf(rel, metabuf);
  
     if (cachedmetap)
         *cachedmetap = metap;
  
     return buf;
 }

References _hash_dropbuf(), _hash_getbuf(), _hash_getcachedmetap(), _hash_hashkey2bucket(), _hash_relbuf(), Assert, BUCKET_TO_BLKNO, buf, BufferGetPage(), BufferIsValid(), HASH_READ, HASH_WRITE, HashMetaPageData::hashm_highmask, HashMetaPageData::hashm_lowmask, HashMetaPageData::hashm_maxbucket, HashPageOpaqueData::hasho_bucket, HashPageOpaqueData::hasho_prevblkno, HashPageGetOpaque, InvalidBlockNumber, InvalidBuffer, and LH_BUCKET_PAGE.

Referenced by _hash_doinsert(), and _hash_first().

◆ _hash_getbuf()

Buffer _hash_getbuf	(	Relation	rel,
		BlockNumber	blkno,
		int	access,
		int	flags
	)

Definition at line 70 of file hashpage.c.

 {
     Buffer      buf;
  
     if (blkno == P_NEW)
         elog(ERROR, "hash AM does not use P_NEW");
  
     buf = ReadBuffer(rel, blkno);
  
     if (access != HASH_NOLOCK)
         LockBuffer(buf, access);
  
     /* ref count and lock type are correct */
  
     _hash_checkpage(rel, buf, flags);
  
     return buf;
 }

References _hash_checkpage(), buf, elog, ERROR, HASH_NOLOCK, LockBuffer(), P_NEW, and ReadBuffer().

Referenced by _hash_addovflpage(), _hash_doinsert(), _hash_finish_split(), _hash_first(), _hash_freeovflpage(), _hash_get_newblock_from_oldbucket(), _hash_get_oldblock_from_newbucket(), _hash_getbucketbuf_from_hashkey(), _hash_getcachedmetap(), _hash_kill_items(), _hash_next(), _hash_readnext(), _hash_readprev(), _hash_splitbucket(), hash_bitmap_info(), hashbulkdelete(), and pgstathashindex().

◆ _hash_getbuf_with_condlock_cleanup()

Buffer _hash_getbuf_with_condlock_cleanup	(	Relation	rel,
		BlockNumber	blkno,
		int	flags
	)

Definition at line 96 of file hashpage.c.

 {
     Buffer      buf;
  
     if (blkno == P_NEW)
         elog(ERROR, "hash AM does not use P_NEW");
  
     buf = ReadBuffer(rel, blkno);
  
     if (!ConditionalLockBufferForCleanup(buf))
     {
         ReleaseBuffer(buf);
         return InvalidBuffer;
     }
  
     /* ref count and lock type are correct */
  
     _hash_checkpage(rel, buf, flags);
  
     return buf;
 }

References _hash_checkpage(), buf, ConditionalLockBufferForCleanup(), elog, ERROR, InvalidBuffer, P_NEW, ReadBuffer(), and ReleaseBuffer().

Referenced by _hash_expandtable().

◆ _hash_getbuf_with_strategy()

Buffer _hash_getbuf_with_strategy	(	Relation	rel,
		BlockNumber	blkno,
		int	access,
		int	flags,
		BufferAccessStrategy	bstrategy
	)

Definition at line 239 of file hashpage.c.

 {
     Buffer      buf;
  
     if (blkno == P_NEW)
         elog(ERROR, "hash AM does not use P_NEW");
  
     buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL, bstrategy);
  
     if (access != HASH_NOLOCK)
         LockBuffer(buf, access);
  
     /* ref count and lock type are correct */
  
     _hash_checkpage(rel, buf, flags);
  
     return buf;
 }

References _hash_checkpage(), buf, elog, ERROR, HASH_NOLOCK, LockBuffer(), MAIN_FORKNUM, P_NEW, RBM_NORMAL, and ReadBufferExtended().

Referenced by _hash_freeovflpage(), _hash_squeezebucket(), hashbucketcleanup(), and pgstat_hash_page().

◆ _hash_getcachedmetap()

HashMetaPage _hash_getcachedmetap	(	Relation	rel,
		Buffer *	metabuf,
		bool	force_refresh
	)

Definition at line 1501 of file hashpage.c.

 {
     Page        page;
  
     Assert(metabuf);
     if (force_refresh || rel->rd_amcache == NULL)
     {
         char       *cache = NULL;
  
         /*
          * It's important that we don't set rd_amcache to an invalid value.
          * Either MemoryContextAlloc or _hash_getbuf could fail, so don't
          * install a pointer to the newly-allocated storage in the actual
          * relcache entry until both have succeeded.
          */
         if (rel->rd_amcache == NULL)
             cache = MemoryContextAlloc(rel->rd_indexcxt,
                                        sizeof(HashMetaPageData));
  
         /* Read the metapage. */
         if (BufferIsValid(*metabuf))
             LockBuffer(*metabuf, BUFFER_LOCK_SHARE);
         else
             *metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_READ,
                                     LH_META_PAGE);
         page = BufferGetPage(*metabuf);
  
         /* Populate the cache. */
         if (rel->rd_amcache == NULL)
             rel->rd_amcache = cache;
         memcpy(rel->rd_amcache, HashPageGetMeta(page),
                sizeof(HashMetaPageData));
  
         /* Release metapage lock, but keep the pin. */
         LockBuffer(*metabuf, BUFFER_LOCK_UNLOCK);
     }
  
     return (HashMetaPage) rel->rd_amcache;
 }

References _hash_getbuf(), Assert, BUFFER_LOCK_SHARE, BUFFER_LOCK_UNLOCK, BufferGetPage(), BufferIsValid(), HASH_METAPAGE, HASH_READ, HashPageGetMeta, LH_META_PAGE, LockBuffer(), MemoryContextAlloc(), RelationData::rd_amcache, and RelationData::rd_indexcxt.

Referenced by _hash_getbucketbuf_from_hashkey(), and hashbulkdelete().

◆ _hash_getinitbuf()

Buffer _hash_getinitbuf	(	Relation	rel,
		BlockNumber	blkno
	)

Definition at line 135 of file hashpage.c.

 {
     Buffer      buf;
  
     if (blkno == P_NEW)
         elog(ERROR, "hash AM does not use P_NEW");
  
     buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_ZERO_AND_LOCK,
                              NULL);
  
     /* ref count and lock type are correct */
  
     /* initialize the page */
     _hash_pageinit(BufferGetPage(buf), BufferGetPageSize(buf));
  
     return buf;
 }

References _hash_pageinit(), buf, BufferGetPage(), BufferGetPageSize(), elog, ERROR, MAIN_FORKNUM, P_NEW, RBM_ZERO_AND_LOCK, and ReadBufferExtended().

Referenced by _hash_addovflpage().

◆ _hash_getnewbuf()

Buffer _hash_getnewbuf	(	Relation	rel,
		BlockNumber	blkno,
		ForkNumber	forkNum
	)

Definition at line 198 of file hashpage.c.

 {
     BlockNumber nblocks = RelationGetNumberOfBlocksInFork(rel, forkNum);
     Buffer      buf;
  
     if (blkno == P_NEW)
         elog(ERROR, "hash AM does not use P_NEW");
     if (blkno > nblocks)
         elog(ERROR, "access to noncontiguous page in hash index \"%s\"",
              RelationGetRelationName(rel));
  
     /* smgr insists we explicitly extend the relation */
     if (blkno == nblocks)
     {
         buf = ExtendBufferedRel(BMR_REL(rel), forkNum, NULL,
                                 EB_LOCK_FIRST | EB_SKIP_EXTENSION_LOCK);
         if (BufferGetBlockNumber(buf) != blkno)
             elog(ERROR, "unexpected hash relation size: %u, should be %u",
                  BufferGetBlockNumber(buf), blkno);
     }
     else
     {
         buf = ReadBufferExtended(rel, forkNum, blkno, RBM_ZERO_AND_LOCK,
                                  NULL);
     }
  
     /* ref count and lock type are correct */
  
     /* initialize the page */
     _hash_pageinit(BufferGetPage(buf), BufferGetPageSize(buf));
  
     return buf;
 }

References _hash_pageinit(), BMR_REL, buf, BufferGetBlockNumber(), BufferGetPage(), BufferGetPageSize(), EB_LOCK_FIRST, EB_SKIP_EXTENSION_LOCK, elog, ERROR, ExtendBufferedRel(), P_NEW, RBM_ZERO_AND_LOCK, ReadBufferExtended(), RelationGetNumberOfBlocksInFork(), and RelationGetRelationName.

Referenced by _hash_addovflpage(), _hash_expandtable(), and _hash_init().

◆ _hash_init()

uint32 _hash_init	(	Relation	rel,
		double	num_tuples,
		ForkNumber	forkNum
	)

Definition at line 327 of file hashpage.c.

 {
     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 = HashGetTargetPageUsage(rel) / 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 | REGBUF_STANDARD);
  
         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_locator,
                         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;
 }

References _hash_getnewbuf(), _hash_init_metabuffer(), _hash_initbitmapbuffer(), _hash_initbuf(), _hash_relbuf(), xl_hash_init_bitmap_page::bmsize, BUCKET_TO_BLKNO, buf, BUFFER_LOCK_EXCLUSIVE, BUFFER_LOCK_UNLOCK, BufferGetPage(), CHECK_FOR_INTERRUPTS, elog, ereport, errcode(), errmsg(), ERROR, xl_hash_init_meta_page::ffactor, HASH_MAX_BITMAPS, HASH_METAPAGE, HashGetTargetPageUsage, HashMetaPageData::hashm_bmsize, HashMetaPageData::hashm_ffactor, HashMetaPageData::hashm_mapp, HashMetaPageData::hashm_maxbucket, HashMetaPageData::hashm_nmaps, HashMetaPageData::hashm_procid, HashPageGetMeta, HASHSTANDARD_PROC, i, index_getprocid(), INIT_FORKNUM, LH_BUCKET_PAGE, LockBuffer(), log_newpage(), MarkBufferDirty(), MAXALIGN, xl_hash_init_meta_page::num_tuples, PageSetLSN(), xl_hash_init_meta_page::procid, RelationData::rd_locator, REGBUF_STANDARD, REGBUF_WILL_INIT, RelationGetNumberOfBlocksInFork(), RelationGetRelationName, RelationNeedsWAL, SizeOfHashInitBitmapPage, SizeOfHashInitMetaPage, XLOG_HASH_INIT_BITMAP_PAGE, XLOG_HASH_INIT_META_PAGE, XLogBeginInsert(), XLogInsert(), XLogRegisterBuffer(), and XLogRegisterData().

Referenced by hashbuild(), and hashbuildempty().

◆ _hash_init_metabuffer()

void _hash_init_metabuffer	(	Buffer	buf,
		double	num_tuples,
		RegProcedure	procid,
		uint16	ffactor,
		bool	initpage
	)

Definition at line 498 of file hashpage.c.

 {
     HashMetaPage metap;
     HashPageOpaque pageopaque;
     Page        page;
     double      dnumbuckets;
     uint32      num_buckets;
     uint32      spare_index;
     uint32      lshift;
  
     /*
      * Choose the number of initial bucket pages to match the fill factor
      * given the estimated number of tuples.  We round up the result to the
      * total number of buckets which has to be allocated before using its
      * hashm_spares element. However always force at least 2 bucket pages. The
      * upper limit is determined by considerations explained in
      * _hash_expandtable().
      */
     dnumbuckets = num_tuples / ffactor;
     if (dnumbuckets <= 2.0)
         num_buckets = 2;
     else if (dnumbuckets >= (double) 0x40000000)
         num_buckets = 0x40000000;
     else
         num_buckets = _hash_get_totalbuckets(_hash_spareindex(dnumbuckets));
  
     spare_index = _hash_spareindex(num_buckets);
     Assert(spare_index < HASH_MAX_SPLITPOINTS);
  
     page = BufferGetPage(buf);
     if (initpage)
         _hash_pageinit(page, BufferGetPageSize(buf));
  
     pageopaque = HashPageGetOpaque(page);
     pageopaque->hasho_prevblkno = InvalidBlockNumber;
     pageopaque->hasho_nextblkno = InvalidBlockNumber;
     pageopaque->hasho_bucket = InvalidBucket;
     pageopaque->hasho_flag = LH_META_PAGE;
     pageopaque->hasho_page_id = HASHO_PAGE_ID;
  
     metap = HashPageGetMeta(page);
  
     metap->hashm_magic = HASH_MAGIC;
     metap->hashm_version = HASH_VERSION;
     metap->hashm_ntuples = 0;
     metap->hashm_nmaps = 0;
     metap->hashm_ffactor = ffactor;
     metap->hashm_bsize = HashGetMaxBitmapSize(page);
  
     /* find largest bitmap array size that will fit in page size */
     lshift = pg_leftmost_one_pos32(metap->hashm_bsize);
     Assert(lshift > 0);
     metap->hashm_bmsize = 1 << lshift;
     metap->hashm_bmshift = lshift + BYTE_TO_BIT;
     Assert((1 << BMPG_SHIFT(metap)) == (BMPG_MASK(metap) + 1));
  
     /*
      * Label the index with its primary hash support function's OID.  This is
      * pretty useless for normal operation (in fact, hashm_procid is not used
      * anywhere), but it might be handy for forensic purposes so we keep it.
      */
     metap->hashm_procid = procid;
  
     /*
      * We initialize the index with N buckets, 0 .. N-1, occupying physical
      * blocks 1 to N.  The first freespace bitmap page is in block N+1.
      */
     metap->hashm_maxbucket = num_buckets - 1;
  
     /*
      * Set highmask as next immediate ((2 ^ x) - 1), which should be
      * sufficient to cover num_buckets.
      */
     metap->hashm_highmask = pg_nextpower2_32(num_buckets + 1) - 1;
     metap->hashm_lowmask = (metap->hashm_highmask >> 1);
  
     MemSet(metap->hashm_spares, 0, sizeof(metap->hashm_spares));
     MemSet(metap->hashm_mapp, 0, sizeof(metap->hashm_mapp));
  
     /* Set up mapping for one spare page after the initial splitpoints */
     metap->hashm_spares[spare_index] = 1;
     metap->hashm_ovflpoint = spare_index;
     metap->hashm_firstfree = 0;
  
     /*
      * Set pd_lower just past the end of the metadata.  This is essential,
      * because without doing so, metadata will be lost if xlog.c compresses
      * the page.
      */
     ((PageHeader) page)->pd_lower =
         ((char *) metap + sizeof(HashMetaPageData)) - (char *) page;
 }

Referenced by _hash_init(), and hash_xlog_init_meta_page().

◆ _hash_initbuf()

void _hash_initbuf	(	Buffer	buf,
		uint32	max_bucket,
		uint32	num_bucket,
		uint32	flag,
		bool	initpage
	)

Definition at line 157 of file hashpage.c.

 {
     HashPageOpaque pageopaque;
     Page        page;
  
     page = BufferGetPage(buf);
  
     /* initialize the page */
     if (initpage)
         _hash_pageinit(page, BufferGetPageSize(buf));
  
     pageopaque = HashPageGetOpaque(page);
  
     /*
      * Set hasho_prevblkno with current hashm_maxbucket. This value will be
      * used to validate cached HashMetaPageData. See
      * _hash_getbucketbuf_from_hashkey().
      */
     pageopaque->hasho_prevblkno = max_bucket;
     pageopaque->hasho_nextblkno = InvalidBlockNumber;
     pageopaque->hasho_bucket = num_bucket;
     pageopaque->hasho_flag = flag;
     pageopaque->hasho_page_id = HASHO_PAGE_ID;
 }

References _hash_pageinit(), buf, BufferGetPage(), BufferGetPageSize(), flag(), HashPageOpaqueData::hasho_bucket, HashPageOpaqueData::hasho_flag, HashPageOpaqueData::hasho_nextblkno, HashPageOpaqueData::hasho_page_id, HASHO_PAGE_ID, HashPageOpaqueData::hasho_prevblkno, HashPageGetOpaque, and InvalidBlockNumber.

Referenced by _hash_init(), hash_xlog_add_ovfl_page(), and hash_xlog_split_allocate_page().

◆ _hash_pageinit()

void _hash_pageinit	(	Page	page,
		Size	size
	)

Definition at line 596 of file hashpage.c.

 {
     PageInit(page, size, sizeof(HashPageOpaqueData));
 }

References PageInit(), and size.

Referenced by _hash_alloc_buckets(), _hash_freeovflpage(), _hash_getinitbuf(), _hash_getnewbuf(), _hash_init_metabuffer(), _hash_initbitmapbuffer(), _hash_initbuf(), and hash_xlog_squeeze_page().

◆ _hash_relbuf()

void _hash_relbuf	(	Relation	rel,
		Buffer	buf
	)

Definition at line 266 of file hashpage.c.

 {
     UnlockReleaseBuffer(buf);
 }

References buf, and UnlockReleaseBuffer().

Referenced by _hash_addovflpage(), _hash_doinsert(), _hash_expandtable(), _hash_finish_split(), _hash_freeovflpage(), _hash_get_newblock_from_oldbucket(), _hash_get_oldblock_from_newbucket(), _hash_getbucketbuf_from_hashkey(), _hash_init(), _hash_kill_items(), _hash_readnext(), _hash_readpage(), _hash_readprev(), _hash_splitbucket(), _hash_squeezebucket(), hash_bitmap_info(), hashbucketcleanup(), hashbulkdelete(), pgstat_hash_page(), and pgstathashindex().

◆ _hash_splitbucket()

static void _hash_splitbucket	(	Relation	rel,
		Buffer	metabuf,
		Bucket	obucket,
		Bucket	nbucket,
		Buffer	obuf,
		Buffer	nbuf,
		HTAB *	htab,
		uint32	maxbucket,
		uint32	highmask,
		uint32	lowmask
	)

static

Definition at line 1073 of file hashpage.c.

 {
     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 = HashPageGetOpaque(opage);
  
     bucket_nbuf = nbuf;
     npage = BufferGetPage(nbuf);
     nopaque = HashPageGetOpaque(npage);
  
     /* Copy the predicate locks from old bucket to new bucket. */
     PredicateLockPageSplit(rel,
                            BufferGetBlockNumber(bucket_obuf),
                            BufferGetBlockNumber(bucket_nbuf));
  
     /*
      * 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 = IndexTupleSize(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));
                     npage = BufferGetPage(nbuf);
                     nopaque = HashPageGetOpaque(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 = HashPageGetOpaque(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 = HashPageGetOpaque(opage);
  
     LockBuffer(bucket_nbuf, BUFFER_LOCK_EXCLUSIVE);
     npage = BufferGetPage(bucket_nbuf);
     nopaque = HashPageGetOpaque(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);
     }
 }

References _hash_addovflpage(), _hash_get_indextuple_hashkey(), _hash_getbuf(), _hash_hashkey2bucket(), _hash_pgaddmultitup(), _hash_relbuf(), Assert, BlockNumberIsValid(), BUFFER_LOCK_EXCLUSIVE, BUFFER_LOCK_UNLOCK, BufferGetBlockNumber(), BufferGetPage(), CopyIndexTuple(), END_CRIT_SECTION, FirstOffsetNumber, HASH_FIND, HASH_READ, hash_search(), hashbucketcleanup(), HashPageOpaqueData::hasho_flag, HashPageOpaqueData::hasho_nextblkno, HashPageGetOpaque, i, INDEX_MOVED_BY_SPLIT_MASK, IndexTupleSize, IsBufferCleanupOK(), ItemIdIsDead, LH_BUCKET_BEING_POPULATED, LH_BUCKET_BEING_SPLIT, LH_BUCKET_NEEDS_SPLIT_CLEANUP, LH_OVERFLOW_PAGE, LockBuffer(), log_split_page(), MarkBufferDirty(), MAXALIGN, MaxIndexTuplesPerPage, xl_hash_split_complete::new_bucket_flag, OffsetNumberNext, xl_hash_split_complete::old_bucket_flag, PageGetFreeSpaceForMultipleTuples(), PageGetItem(), PageGetItemId(), PageGetMaxOffsetNumber(), PageSetLSN(), pfree(), PredicateLockPageSplit(), REGBUF_STANDARD, RelationNeedsWAL, SizeOfHashSplitComplete, START_CRIT_SECTION, IndexTupleData::t_info, IndexTupleData::t_tid, XLOG_HASH_SPLIT_COMPLETE, XLogBeginInsert(), XLogInsert(), XLogRegisterBuffer(), and XLogRegisterData().

Referenced by _hash_expandtable(), and _hash_finish_split().

◆ log_split_page()

static void log_split_page	(	Relation	rel,
		Buffer	buf
	)

static

Definition at line 1474 of file hashpage.c.

 {
     if (RelationNeedsWAL(rel))
     {
         XLogRecPtr  recptr;
  
         XLogBeginInsert();
  
         XLogRegisterBuffer(0, buf, REGBUF_FORCE_IMAGE | REGBUF_STANDARD);
  
         recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_SPLIT_PAGE);
  
         PageSetLSN(BufferGetPage(buf), recptr);
     }
 }

References buf, BufferGetPage(), PageSetLSN(), REGBUF_FORCE_IMAGE, REGBUF_STANDARD, RelationNeedsWAL, XLOG_HASH_SPLIT_PAGE, XLogBeginInsert(), XLogInsert(), and XLogRegisterBuffer().

Referenced by _hash_splitbucket().

Functions

Function Documentation

◆ _hash_alloc_buckets()

◆ _hash_dropbuf()

◆ _hash_dropscanbuf()

◆ _hash_expandtable()

◆ _hash_finish_split()

◆ _hash_getbucketbuf_from_hashkey()

◆ _hash_getbuf()

◆ _hash_getbuf_with_condlock_cleanup()

◆ _hash_getbuf_with_strategy()

◆ _hash_getcachedmetap()

◆ _hash_getinitbuf()

◆ _hash_getnewbuf()

◆ _hash_init()

◆ _hash_init_metabuffer()

◆ _hash_initbuf()

◆ _hash_pageinit()

◆ _hash_relbuf()

◆ _hash_splitbucket()

◆ log_split_page()