mirrors/postgres
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

nbtree: Allocate new pages in separate function.

Browse Source 
Split nbtree's _bt_getbuf function is two: code that read locks or write
locks existing pages remains in _bt_getbuf, while code that deals with
allocating new pages is moved to a new, dedicated function called
_bt_allocbuf.  This simplifies most _bt_getbuf callers, since it is no
longer necessary for them to pass a heaprel argument.  Many of the
changes to nbtree from commit 61b313e4 can be reverted.  This minimizes
the divergence between HEAD/PostgreSQL 16 and earlier release branches.

_bt_allocbuf replaces the previous nbtree idiom of passing P_NEW to
_bt_getbuf.  There are only 3 affected call sites, all of which continue
to pass a heaprel for recovery conflict purposes.  Note that nbtree's
use of P_NEW was superficial; nbtree never actually relied on the P_NEW
code paths in bufmgr.c, so this change is strictly mechanical.

GiST already took the same approach; it has a dedicated function for
allocating new pages called gistNewBuffer().  That factor allowed commit
61b313e4 to make much more targeted changes to GiST.

Author: Peter Geoghegan <pg@bowt.ie>
Reviewed-By: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/CAH2-Wz=8Z9qY58bjm_7TAHgtW6RzZ5Ke62q5emdCEy9BAzwhmg@mail.gmail.com
This commit is contained in:
Peter Geoghegan 2023-06-10 14:08:25 -07:00
parent fe879ae3a8
commit d088ba5a5a
12 changed files with 271 additions and 257 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
contrib/amcheck/verify_nbtree.c
View File

@ -183,7 +183,6 @@ static inline bool invariant_l_nontarget_offset(BtreeCheckState *state,
OffsetNumber upperbound); OffsetNumber upperbound);
static Page palloc_btree_page(BtreeCheckState *state, BlockNumber blocknum); static Page palloc_btree_page(BtreeCheckState *state, BlockNumber blocknum);
static inline BTScanInsert bt_mkscankey_pivotsearch(Relation rel, static inline BTScanInsert bt_mkscankey_pivotsearch(Relation rel,
Relation heaprel,
IndexTuple itup); IndexTuple itup);
static ItemId PageGetItemIdCareful(BtreeCheckState *state, BlockNumber block, static ItemId PageGetItemIdCareful(BtreeCheckState *state, BlockNumber block,
Page page, OffsetNumber offset); Page page, OffsetNumber offset);
@ -332,7 +331,7 @@ bt_index_check_internal(Oid indrelid, bool parentcheck, bool heapallindexed,
RelationGetRelationName(indrel)))); RelationGetRelationName(indrel))));
/* Extract metadata from metapage, and sanitize it in passing */ /* Extract metadata from metapage, and sanitize it in passing */
_bt_metaversion(indrel, heaprel, &heapkeyspace, &allequalimage); _bt_metaversion(indrel, &heapkeyspace, &allequalimage);
if (allequalimage && !heapkeyspace) if (allequalimage && !heapkeyspace)
ereport(ERROR, ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED), (errcode(ERRCODE_INDEX_CORRUPTED),
@ -1259,7 +1258,7 @@ bt_target_page_check(BtreeCheckState *state)
} }
/* Build insertion scankey for current page offset */ /* Build insertion scankey for current page offset */
skey = bt_mkscankey_pivotsearch(state->rel, state->heaprel, itup); skey = bt_mkscankey_pivotsearch(state->rel, itup);
/* /*
* Make sure tuple size does not exceed the relevant BTREE_VERSION * Make sure tuple size does not exceed the relevant BTREE_VERSION
@ -1769,7 +1768,7 @@ bt_right_page_check_scankey(BtreeCheckState *state)
* memory remaining allocated. * memory remaining allocated.
*/ */
firstitup = (IndexTuple) PageGetItem(rightpage, rightitem); firstitup = (IndexTuple) PageGetItem(rightpage, rightitem);
return bt_mkscankey_pivotsearch(state->rel, state->heaprel, firstitup); return bt_mkscankey_pivotsearch(state->rel, firstitup);
} }
/* /*
@ -2682,7 +2681,7 @@ bt_rootdescend(BtreeCheckState *state, IndexTuple itup)
Buffer lbuf; Buffer lbuf;
bool exists; bool exists;
key = _bt_mkscankey(state->rel, state->heaprel, itup); key = _bt_mkscankey(state->rel, itup);
Assert(key->heapkeyspace && key->scantid != NULL); Assert(key->heapkeyspace && key->scantid != NULL);
/* /*
@ -2695,7 +2694,7 @@ bt_rootdescend(BtreeCheckState *state, IndexTuple itup)
*/ */
Assert(state->readonly && state->rootdescend); Assert(state->readonly && state->rootdescend);
exists = false; exists = false;
stack = _bt_search(state->rel, state->heaprel, key, &lbuf, BT_READ, NULL); stack = _bt_search(state->rel, NULL, key, &lbuf, BT_READ, NULL);
if (BufferIsValid(lbuf)) if (BufferIsValid(lbuf))
{ {
@ -3134,11 +3133,11 @@ palloc_btree_page(BtreeCheckState *state, BlockNumber blocknum)
* the scankey is greater. * the scankey is greater.
*/ */
static inline BTScanInsert static inline BTScanInsert
bt_mkscankey_pivotsearch(Relation rel, Relation heaprel, IndexTuple itup) bt_mkscankey_pivotsearch(Relation rel, IndexTuple itup)
{ {
BTScanInsert skey; BTScanInsert skey;
skey = _bt_mkscankey(rel, heaprel, itup); skey = _bt_mkscankey(rel, itup);
skey->pivotsearch = true; skey->pivotsearch = true;
return skey; return skey;

9
src/backend/access/heap/heapam_handler.c
View File

@ -731,14 +731,9 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
*multi_cutoff); *multi_cutoff);
/* /* Set up sorting if wanted */
* Set up sorting if wanted. NewHeap is being passed to
* tuplesort_begin_cluster(), it could have been OldHeap too. It does not
* really matter, as the goal is to have a heap relation being passed to
* _bt_log_reuse_page() (which should not be called from this code path).
*/
if (use_sort) if (use_sort)
tuplesort = tuplesort_begin_cluster(oldTupDesc, OldIndex, NewHeap, tuplesort = tuplesort_begin_cluster(oldTupDesc, OldIndex,
maintenance_work_mem, maintenance_work_mem,
NULL, TUPLESORT_NONE); NULL, TUPLESORT_NONE);
else else

48
src/backend/access/nbtree/nbtinsert.c
View File

@ -59,7 +59,7 @@ static Buffer _bt_split(Relation rel, Relation heaprel, BTScanInsert itup_key,
IndexTuple nposting, uint16 postingoff); IndexTuple nposting, uint16 postingoff);
static void _bt_insert_parent(Relation rel, Relation heaprel, Buffer buf, static void _bt_insert_parent(Relation rel, Relation heaprel, Buffer buf,
Buffer rbuf, BTStack stack, bool isroot, bool isonly); Buffer rbuf, BTStack stack, bool isroot, bool isonly);
static Buffer _bt_newroot(Relation rel, Relation heaprel, Buffer lbuf, Buffer rbuf); static Buffer _bt_newlevel(Relation rel, Relation heaprel, Buffer lbuf, Buffer rbuf);
static inline bool _bt_pgaddtup(Page page, Size itemsize, IndexTuple itup, static inline bool _bt_pgaddtup(Page page, Size itemsize, IndexTuple itup,
OffsetNumber itup_off, bool newfirstdataitem); OffsetNumber itup_off, bool newfirstdataitem);
static void _bt_delete_or_dedup_one_page(Relation rel, Relation heapRel, static void _bt_delete_or_dedup_one_page(Relation rel, Relation heapRel,
@ -110,7 +110,7 @@ _bt_doinsert(Relation rel, IndexTuple itup,
bool checkingunique = (checkUnique != UNIQUE_CHECK_NO); bool checkingunique = (checkUnique != UNIQUE_CHECK_NO);
/* we need an insertion scan key to do our search, so build one */ /* we need an insertion scan key to do our search, so build one */
itup_key = _bt_mkscankey(rel, heapRel, itup); itup_key = _bt_mkscankey(rel, itup);
if (checkingunique) if (checkingunique)
{ {
@ -1024,13 +1024,15 @@ _bt_findinsertloc(Relation rel,
* indexes. * indexes.
*/ */
static void static void
_bt_stepright(Relation rel, Relation heaprel, BTInsertState insertstate, BTStack stack) _bt_stepright(Relation rel, Relation heaprel, BTInsertState insertstate,
BTStack stack)
{ {
Page page; Page page;
BTPageOpaque opaque; BTPageOpaque opaque;
Buffer rbuf; Buffer rbuf;
BlockNumber rblkno; BlockNumber rblkno;
Assert(heaprel != NULL);
page = BufferGetPage(insertstate->buf); page = BufferGetPage(insertstate->buf);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
@ -1145,7 +1147,7 @@ _bt_insertonpg(Relation rel,
/* /*
* Every internal page should have exactly one negative infinity item at * Every internal page should have exactly one negative infinity item at
* all times. Only _bt_split() and _bt_newroot() should add items that * all times. Only _bt_split() and _bt_newlevel() should add items that
* become negative infinity items through truncation, since they're the * become negative infinity items through truncation, since they're the
* only routines that allocate new internal pages. * only routines that allocate new internal pages.
*/ */
@ -1250,14 +1252,14 @@ _bt_insertonpg(Relation rel,
* only one on its tree level, but was not the root, it may have been * only one on its tree level, but was not the root, it may have been
* the "fast root". We need to ensure that the fast root link points * the "fast root". We need to ensure that the fast root link points
* at or above the current page. We can safely acquire a lock on the * at or above the current page. We can safely acquire a lock on the
* metapage here --- see comments for _bt_newroot(). * metapage here --- see comments for _bt_newlevel().
*/ */
if (unlikely(split_only_page)) if (unlikely(split_only_page))
{ {
Assert(!isleaf); Assert(!isleaf);
Assert(BufferIsValid(cbuf)); Assert(BufferIsValid(cbuf));
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_WRITE); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_WRITE);
metapg = BufferGetPage(metabuf); metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg); metad = BTPageGetMeta(metapg);
@ -1421,7 +1423,7 @@ _bt_insertonpg(Relation rel,
* call _bt_getrootheight while holding a buffer lock. * call _bt_getrootheight while holding a buffer lock.
*/ */
if (BlockNumberIsValid(blockcache) && if (BlockNumberIsValid(blockcache) &&
_bt_getrootheight(rel, heaprel) >= BTREE_FASTPATH_MIN_LEVEL) _bt_getrootheight(rel) >= BTREE_FASTPATH_MIN_LEVEL)
RelationSetTargetBlock(rel, blockcache); RelationSetTargetBlock(rel, blockcache);
} }
@ -1715,7 +1717,7 @@ _bt_split(Relation rel, Relation heaprel, BTScanInsert itup_key, Buffer buf,
* way because it avoids an unnecessary PANIC when either origpage or its * way because it avoids an unnecessary PANIC when either origpage or its
* existing sibling page are corrupt. * existing sibling page are corrupt.
*/ */
rbuf = _bt_getbuf(rel, heaprel, P_NEW, BT_WRITE); rbuf = _bt_allocbuf(rel, heaprel);
rightpage = BufferGetPage(rbuf); rightpage = BufferGetPage(rbuf);
rightpagenumber = BufferGetBlockNumber(rbuf); rightpagenumber = BufferGetBlockNumber(rbuf);
/* rightpage was initialized by _bt_getbuf */ /* rightpage was initialized by _bt_getbuf */
@ -1888,7 +1890,7 @@ _bt_split(Relation rel, Relation heaprel, BTScanInsert itup_key, Buffer buf,
*/ */
if (!isrightmost) if (!isrightmost)
{ {
sbuf = _bt_getbuf(rel, heaprel, oopaque->btpo_next, BT_WRITE); sbuf = _bt_getbuf(rel, oopaque->btpo_next, BT_WRITE);
spage = BufferGetPage(sbuf); spage = BufferGetPage(sbuf);
sopaque = BTPageGetOpaque(spage); sopaque = BTPageGetOpaque(spage);
if (sopaque->btpo_prev != origpagenumber) if (sopaque->btpo_prev != origpagenumber)
@ -2102,6 +2104,8 @@ _bt_insert_parent(Relation rel,
bool isroot, bool isroot,
bool isonly) bool isonly)
{ {
Assert(heaprel != NULL);
/* /*
* Here we have to do something Lehman and Yao don't talk about: deal with * Here we have to do something Lehman and Yao don't talk about: deal with
* a root split and construction of a new root. If our stack is empty * a root split and construction of a new root. If our stack is empty
@ -2121,8 +2125,8 @@ _bt_insert_parent(Relation rel,
Assert(stack == NULL); Assert(stack == NULL);
Assert(isonly); Assert(isonly);
/* create a new root node and update the metapage */ /* create a new root node one level up and update the metapage */
rootbuf = _bt_newroot(rel, heaprel, buf, rbuf); rootbuf = _bt_newlevel(rel, heaprel, buf, rbuf);
/* release the split buffers */ /* release the split buffers */
_bt_relbuf(rel, rootbuf); _bt_relbuf(rel, rootbuf);
_bt_relbuf(rel, rbuf); _bt_relbuf(rel, rbuf);
@ -2161,8 +2165,7 @@ _bt_insert_parent(Relation rel,
BlockNumberIsValid(RelationGetTargetBlock(rel)))); BlockNumberIsValid(RelationGetTargetBlock(rel))));
/* Find the leftmost page at the next level up */ /* Find the leftmost page at the next level up */
pbuf = _bt_get_endpoint(rel, heaprel, opaque->btpo_level + 1, false, pbuf = _bt_get_endpoint(rel, opaque->btpo_level + 1, false, NULL);
NULL);
/* Set up a phony stack entry pointing there */ /* Set up a phony stack entry pointing there */
stack = &fakestack; stack = &fakestack;
stack->bts_blkno = BufferGetBlockNumber(pbuf); stack->bts_blkno = BufferGetBlockNumber(pbuf);
@ -2230,6 +2233,9 @@ _bt_insert_parent(Relation rel,
* *
* On entry, 'lbuf' must be locked in write-mode. On exit, it is unlocked * On entry, 'lbuf' must be locked in write-mode. On exit, it is unlocked
* and unpinned. * and unpinned.
*
* Caller must provide a valid heaprel, since finishing a page split requires
* allocating a new page if and when the parent page splits in turn.
*/ */
void void
_bt_finish_split(Relation rel, Relation heaprel, Buffer lbuf, BTStack stack) _bt_finish_split(Relation rel, Relation heaprel, Buffer lbuf, BTStack stack)
@ -2243,9 +2249,10 @@ _bt_finish_split(Relation rel, Relation heaprel, Buffer lbuf, BTStack stack)
bool wasonly; bool wasonly;
Assert(P_INCOMPLETE_SPLIT(lpageop)); Assert(P_INCOMPLETE_SPLIT(lpageop));
Assert(heaprel != NULL);
/* Lock right sibling, the one missing the downlink */ /* Lock right sibling, the one missing the downlink */
rbuf = _bt_getbuf(rel, heaprel, lpageop->btpo_next, BT_WRITE); rbuf = _bt_getbuf(rel, lpageop->btpo_next, BT_WRITE);
rpage = BufferGetPage(rbuf); rpage = BufferGetPage(rbuf);
rpageop = BTPageGetOpaque(rpage); rpageop = BTPageGetOpaque(rpage);
@ -2257,7 +2264,7 @@ _bt_finish_split(Relation rel, Relation heaprel, Buffer lbuf, BTStack stack)
BTMetaPageData *metad; BTMetaPageData *metad;
/* acquire lock on the metapage */ /* acquire lock on the metapage */
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_WRITE); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_WRITE);
metapg = BufferGetPage(metabuf); metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg); metad = BTPageGetMeta(metapg);
@ -2323,10 +2330,11 @@ _bt_getstackbuf(Relation rel, Relation heaprel, BTStack stack, BlockNumber child
Page page; Page page;
BTPageOpaque opaque; BTPageOpaque opaque;
buf = _bt_getbuf(rel, heaprel, blkno, BT_WRITE); buf = _bt_getbuf(rel, blkno, BT_WRITE);
page = BufferGetPage(buf); page = BufferGetPage(buf);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
Assert(heaprel != NULL);
if (P_INCOMPLETE_SPLIT(opaque)) if (P_INCOMPLETE_SPLIT(opaque))
{ {
_bt_finish_split(rel, heaprel, buf, stack->bts_parent); _bt_finish_split(rel, heaprel, buf, stack->bts_parent);
@ -2415,7 +2423,7 @@ _bt_getstackbuf(Relation rel, Relation heaprel, BTStack stack, BlockNumber child
} }
/* /*
* _bt_newroot() -- Create a new root page for the index. * _bt_newlevel() -- Create a new level above root page.
* *
* We've just split the old root page and need to create a new one. * We've just split the old root page and need to create a new one.
* In order to do this, we add a new root page to the file, then lock * In order to do this, we add a new root page to the file, then lock
@ -2433,7 +2441,7 @@ _bt_getstackbuf(Relation rel, Relation heaprel, BTStack stack, BlockNumber child
* lbuf, rbuf & rootbuf. * lbuf, rbuf & rootbuf.
*/ */
static Buffer static Buffer
_bt_newroot(Relation rel, Relation heaprel, Buffer lbuf, Buffer rbuf) _bt_newlevel(Relation rel, Relation heaprel, Buffer lbuf, Buffer rbuf)
{ {
Buffer rootbuf; Buffer rootbuf;
Page lpage, Page lpage,
@ -2459,12 +2467,12 @@ _bt_newroot(Relation rel, Relation heaprel, Buffer lbuf, Buffer rbuf)
lopaque = BTPageGetOpaque(lpage); lopaque = BTPageGetOpaque(lpage);
/* get a new root page */ /* get a new root page */
rootbuf = _bt_getbuf(rel, heaprel, P_NEW, BT_WRITE); rootbuf = _bt_allocbuf(rel, heaprel);
rootpage = BufferGetPage(rootbuf); rootpage = BufferGetPage(rootbuf);
rootblknum = BufferGetBlockNumber(rootbuf); rootblknum = BufferGetBlockNumber(rootbuf);
/* acquire lock on the metapage */ /* acquire lock on the metapage */
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_WRITE); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_WRITE);
metapg = BufferGetPage(metabuf); metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg); metad = BTPageGetMeta(metapg);

264
src/backend/access/nbtree/nbtpage.c
View File

@ -38,10 +38,9 @@
#include "utils/snapmgr.h" #include "utils/snapmgr.h"
static BTMetaPageData *_bt_getmeta(Relation rel, Buffer metabuf); static BTMetaPageData *_bt_getmeta(Relation rel, Buffer metabuf);
static void _bt_log_reuse_page(Relation rel, Relation heaprel, BlockNumber blkno, static void _bt_delitems_delete(Relation rel, Buffer buf,
FullTransactionId safexid);
static void _bt_delitems_delete(Relation rel, Relation heaprel, Buffer buf,
TransactionId snapshotConflictHorizon, TransactionId snapshotConflictHorizon,
bool isCatalogRel,
OffsetNumber *deletable, int ndeletable, OffsetNumber *deletable, int ndeletable,
BTVacuumPosting *updatable, int nupdatable); BTVacuumPosting *updatable, int nupdatable);
static char *_bt_delitems_update(BTVacuumPosting *updatable, int nupdatable, static char *_bt_delitems_update(BTVacuumPosting *updatable, int nupdatable,
@ -177,7 +176,7 @@ _bt_getmeta(Relation rel, Buffer metabuf)
* index tuples needed to be deleted. * index tuples needed to be deleted.
*/ */
bool bool
_bt_vacuum_needs_cleanup(Relation rel, Relation heaprel) _bt_vacuum_needs_cleanup(Relation rel)
{ {
Buffer metabuf; Buffer metabuf;
Page metapg; Page metapg;
@ -190,7 +189,7 @@ _bt_vacuum_needs_cleanup(Relation rel, Relation heaprel)
* *
* Note that we deliberately avoid using cached version of metapage here. * Note that we deliberately avoid using cached version of metapage here.
*/ */
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_READ); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metapg = BufferGetPage(metabuf); metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg); metad = BTPageGetMeta(metapg);
btm_version = metad->btm_version; btm_version = metad->btm_version;
@ -230,7 +229,7 @@ _bt_vacuum_needs_cleanup(Relation rel, Relation heaprel)
* finalized. * finalized.
*/ */
void void
_bt_set_cleanup_info(Relation rel, Relation heaprel, BlockNumber num_delpages) _bt_set_cleanup_info(Relation rel, BlockNumber num_delpages)
{ {
Buffer metabuf; Buffer metabuf;
Page metapg; Page metapg;
@ -254,7 +253,7 @@ _bt_set_cleanup_info(Relation rel, Relation heaprel, BlockNumber num_delpages)
* no longer used as of PostgreSQL 14. We set it to -1.0 on rewrite, just * no longer used as of PostgreSQL 14. We set it to -1.0 on rewrite, just
* to be consistent. * to be consistent.
*/ */
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_READ); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metapg = BufferGetPage(metabuf); metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg); metad = BTPageGetMeta(metapg);
@ -326,6 +325,9 @@ _bt_set_cleanup_info(Relation rel, Relation heaprel, BlockNumber num_delpages)
* NOTE that the returned root page will have only a read lock set * NOTE that the returned root page will have only a read lock set
* on it even if access = BT_WRITE! * on it even if access = BT_WRITE!
* *
* If access = BT_WRITE, heaprel must be set; otherwise caller can just
* pass NULL. See _bt_allocbuf for an explanation.
*
* The returned page is not necessarily the true root --- it could be * The returned page is not necessarily the true root --- it could be
* a "fast root" (a page that is alone in its level due to deletions). * a "fast root" (a page that is alone in its level due to deletions).
* Also, if the root page is split while we are "in flight" to it, * Also, if the root page is split while we are "in flight" to it,
@ -349,6 +351,8 @@ _bt_getroot(Relation rel, Relation heaprel, int access)
uint32 rootlevel; uint32 rootlevel;
BTMetaPageData *metad; BTMetaPageData *metad;
Assert(access == BT_READ || heaprel != NULL);
/* /*
* Try to use previously-cached metapage data to find the root. This * Try to use previously-cached metapage data to find the root. This
* normally saves one buffer access per index search, which is a very * normally saves one buffer access per index search, which is a very
@ -369,7 +373,7 @@ _bt_getroot(Relation rel, Relation heaprel, int access)
Assert(rootblkno != P_NONE); Assert(rootblkno != P_NONE);
rootlevel = metad->btm_fastlevel; rootlevel = metad->btm_fastlevel;
rootbuf = _bt_getbuf(rel, heaprel, rootblkno, BT_READ); rootbuf = _bt_getbuf(rel, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf); rootpage = BufferGetPage(rootbuf);
rootopaque = BTPageGetOpaque(rootpage); rootopaque = BTPageGetOpaque(rootpage);
@ -395,7 +399,7 @@ _bt_getroot(Relation rel, Relation heaprel, int access)
rel->rd_amcache = NULL; rel->rd_amcache = NULL;
} }
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_READ); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metad = _bt_getmeta(rel, metabuf); metad = _bt_getmeta(rel, metabuf);
/* if no root page initialized yet, do it */ /* if no root page initialized yet, do it */
@ -436,7 +440,7 @@ _bt_getroot(Relation rel, Relation heaprel, int access)
* the new root page. Since this is the first page in the tree, it's * the new root page. Since this is the first page in the tree, it's
* a leaf as well as the root. * a leaf as well as the root.
*/ */
rootbuf = _bt_getbuf(rel, heaprel, P_NEW, BT_WRITE); rootbuf = _bt_allocbuf(rel, heaprel);
rootblkno = BufferGetBlockNumber(rootbuf); rootblkno = BufferGetBlockNumber(rootbuf);
rootpage = BufferGetPage(rootbuf); rootpage = BufferGetPage(rootbuf);
rootopaque = BTPageGetOpaque(rootpage); rootopaque = BTPageGetOpaque(rootpage);
@ -573,7 +577,7 @@ _bt_getroot(Relation rel, Relation heaprel, int access)
* moving to the root --- that'd deadlock against any concurrent root split.) * moving to the root --- that'd deadlock against any concurrent root split.)
*/ */
Buffer Buffer
_bt_gettrueroot(Relation rel, Relation heaprel) _bt_gettrueroot(Relation rel)
{ {
Buffer metabuf; Buffer metabuf;
Page metapg; Page metapg;
@ -595,7 +599,7 @@ _bt_gettrueroot(Relation rel, Relation heaprel)
pfree(rel->rd_amcache); pfree(rel->rd_amcache);
rel->rd_amcache = NULL; rel->rd_amcache = NULL;
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_READ); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metapg = BufferGetPage(metabuf); metapg = BufferGetPage(metabuf);
metaopaque = BTPageGetOpaque(metapg); metaopaque = BTPageGetOpaque(metapg);
metad = BTPageGetMeta(metapg); metad = BTPageGetMeta(metapg);
@ -668,7 +672,7 @@ _bt_gettrueroot(Relation rel, Relation heaprel)
* about updating previously cached data. * about updating previously cached data.
*/ */
int int
_bt_getrootheight(Relation rel, Relation heaprel) _bt_getrootheight(Relation rel)
{ {
BTMetaPageData *metad; BTMetaPageData *metad;
@ -676,7 +680,7 @@ _bt_getrootheight(Relation rel, Relation heaprel)
{ {
Buffer metabuf; Buffer metabuf;
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_READ); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metad = _bt_getmeta(rel, metabuf); metad = _bt_getmeta(rel, metabuf);
/* /*
@ -732,7 +736,7 @@ _bt_getrootheight(Relation rel, Relation heaprel)
* pg_upgrade'd from Postgres 12. * pg_upgrade'd from Postgres 12.
*/ */
void void
_bt_metaversion(Relation rel, Relation heaprel, bool *heapkeyspace, bool *allequalimage) _bt_metaversion(Relation rel, bool *heapkeyspace, bool *allequalimage)
{ {
BTMetaPageData *metad; BTMetaPageData *metad;
@ -740,7 +744,7 @@ _bt_metaversion(Relation rel, Relation heaprel, bool *heapkeyspace, bool *allequ
{ {
Buffer metabuf; Buffer metabuf;
metabuf = _bt_getbuf(rel, heaprel, BTREE_METAPAGE, BT_READ); metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metad = _bt_getmeta(rel, metabuf); metad = _bt_getmeta(rel, metabuf);
/* /*
@ -821,37 +825,7 @@ _bt_checkpage(Relation rel, Buffer buf)
} }
/* /*
* Log the reuse of a page from the FSM. * _bt_getbuf() -- Get an existing block in a buffer, for read or write.
*/
static void
_bt_log_reuse_page(Relation rel, Relation heaprel, BlockNumber blkno,
FullTransactionId safexid)
{
xl_btree_reuse_page xlrec_reuse;
/*
* Note that we don't register the buffer with the record, because this
* operation doesn't modify the page. This record only exists to provide a
* conflict point for Hot Standby.
*/
/* XLOG stuff */
xlrec_reuse.isCatalogRel = RelationIsAccessibleInLogicalDecoding(heaprel);
xlrec_reuse.locator = rel->rd_locator;
xlrec_reuse.block = blkno;
xlrec_reuse.snapshotConflictHorizon = safexid;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec_reuse, SizeOfBtreeReusePage);
XLogInsert(RM_BTREE_ID, XLOG_BTREE_REUSE_PAGE);
}
/*
* _bt_getbuf() -- Get a buffer by block number for read or write.
*
* blkno == P_NEW means to get an unallocated index page. The page
* will be initialized before returning it.
* *
* The general rule in nbtree is that it's never okay to access a * The general rule in nbtree is that it's never okay to access a
* page without holding both a buffer pin and a buffer lock on * page without holding both a buffer pin and a buffer lock on
@ -860,55 +834,69 @@ _bt_log_reuse_page(Relation rel, Relation heaprel, BlockNumber blkno,
* When this routine returns, the appropriate lock is set on the * When this routine returns, the appropriate lock is set on the
* requested buffer and its reference count has been incremented * requested buffer and its reference count has been incremented
* (ie, the buffer is "locked and pinned"). Also, we apply * (ie, the buffer is "locked and pinned"). Also, we apply
* _bt_checkpage to sanity-check the page (except in P_NEW case), * _bt_checkpage to sanity-check the page, and perform Valgrind
* and perform Valgrind client requests that help Valgrind detect * client requests that help Valgrind detect unsafe page accesses.
* unsafe page accesses.
* *
* Note: raw LockBuffer() calls are disallowed in nbtree; all * Note: raw LockBuffer() calls are disallowed in nbtree; all
* buffer lock requests need to go through wrapper functions such * buffer lock requests need to go through wrapper functions such
* as _bt_lockbuf(). * as _bt_lockbuf().
*/ */
Buffer Buffer
_bt_getbuf(Relation rel, Relation heaprel, BlockNumber blkno, int access) _bt_getbuf(Relation rel, BlockNumber blkno, int access)
{ {
Buffer buf; Buffer buf;
if (blkno != P_NEW) Assert(BlockNumberIsValid(blkno));
{
/* Read an existing block of the relation */ /* Read an existing block of the relation */
buf = ReadBuffer(rel, blkno); buf = ReadBuffer(rel, blkno);
_bt_lockbuf(rel, buf, access); _bt_lockbuf(rel, buf, access);
_bt_checkpage(rel, buf); _bt_checkpage(rel, buf);
}
else return buf;
{ }
/*
* _bt_allocbuf() -- Allocate a new block/page.
*
* Returns a write-locked buffer containing an unallocated nbtree page.
*
* Callers are required to pass a valid heaprel. We need heaprel so that we
* can handle generating a snapshotConflictHorizon that makes reusing a page
* from the FSM safe for queries that may be running on standbys.
*/
Buffer
_bt_allocbuf(Relation rel, Relation heaprel)
{
Buffer buf;
BlockNumber blkno;
Page page; Page page;
Assert(access == BT_WRITE); Assert(heaprel != NULL);
/* /*
* First see if the FSM knows of any free pages. * First see if the FSM knows of any free pages.
* *
* We can't trust the FSM's report unreservedly; we have to check that * We can't trust the FSM's report unreservedly; we have to check that the
* the page is still free. (For example, an already-free page could * page is still free. (For example, an already-free page could have been
* have been re-used between the time the last VACUUM scanned it and * re-used between the time the last VACUUM scanned it and the time the
* the time the VACUUM made its FSM updates.) * VACUUM made its FSM updates.)
* *
* In fact, it's worse than that: we can't even assume that it's safe * In fact, it's worse than that: we can't even assume that it's safe to
* to take a lock on the reported page. If somebody else has a lock * take a lock on the reported page. If somebody else has a lock on it,
* on it, or even worse our own caller does, we could deadlock. (The * or even worse our own caller does, we could deadlock. (The own-caller
* own-caller scenario is actually not improbable. Consider an index * scenario is actually not improbable. Consider an index on a serial or
* on a serial or timestamp column. Nearly all splits will be at the * timestamp column. Nearly all splits will be at the rightmost page, so
* rightmost page, so it's entirely likely that _bt_split will call us * it's entirely likely that _bt_split will call us while holding a lock
* while holding a lock on the page most recently acquired from FSM. A * on the page most recently acquired from FSM. A VACUUM running
* VACUUM running concurrently with the previous split could well have * concurrently with the previous split could well have placed that page
* placed that page back in FSM.) * back in FSM.)
* *
* To get around that, we ask for only a conditional lock on the * To get around that, we ask for only a conditional lock on the reported
* reported page. If we fail, then someone else is using the page, * page. If we fail, then someone else is using the page, and we may
* and we may reasonably assume it's not free. (If we happen to be * reasonably assume it's not free. (If we happen to be wrong, the worst
* wrong, the worst consequence is the page will be lost to use till * consequence is the page will be lost to use till the next VACUUM, which
* the next VACUUM, which is no big problem.) * is no big problem.)
*/ */
for (;;) for (;;)
{ {
@ -922,10 +910,10 @@ _bt_getbuf(Relation rel, Relation heaprel, BlockNumber blkno, int access)
/* /*
* It's possible to find an all-zeroes page in an index. For * It's possible to find an all-zeroes page in an index. For
* example, a backend might successfully extend the relation * example, a backend might successfully extend the relation one
* one page and then crash before it is able to make a WAL * page and then crash before it is able to make a WAL entry for
* entry for adding the page. If we find a zeroed page then * adding the page. If we find a zeroed page then reclaim it
* reclaim it immediately. * immediately.
*/ */
if (PageIsNew(page)) if (PageIsNew(page))
{ {
@ -937,14 +925,33 @@ _bt_getbuf(Relation rel, Relation heaprel, BlockNumber blkno, int access)
if (BTPageIsRecyclable(page, heaprel)) if (BTPageIsRecyclable(page, heaprel))
{ {
/* /*
* If we are generating WAL for Hot Standby then create a * If we are generating WAL for Hot Standby then create a WAL
* WAL record that will allow us to conflict with queries * record that will allow us to conflict with queries running
* running on standby, in case they have snapshots older * on standby, in case they have snapshots older than safexid
* than safexid value * value
*/ */
if (XLogStandbyInfoActive() && RelationNeedsWAL(rel)) if (RelationNeedsWAL(rel) && XLogStandbyInfoActive())
_bt_log_reuse_page(rel, heaprel, blkno, {
BTPageGetDeleteXid(page)); xl_btree_reuse_page xlrec_reuse;
/*
* Note that we don't register the buffer with the record,
* because this operation doesn't modify the page (that
* already happened, back when VACUUM deleted the page).
* This record only exists to provide a conflict point for
* Hot Standby. See record REDO routine comments.
*/
xlrec_reuse.locator = rel->rd_locator;
xlrec_reuse.block = blkno;
xlrec_reuse.snapshotConflictHorizon = BTPageGetDeleteXid(page);
xlrec_reuse.isCatalogRel =
RelationIsAccessibleInLogicalDecoding(heaprel);
XLogBeginInsert();
XLogRegisterData((char *) &xlrec_reuse, SizeOfBtreeReusePage);
XLogInsert(RM_BTREE_ID, XLOG_BTREE_REUSE_PAGE);
}
/* Okay to use page. Re-initialize and return it. */ /* Okay to use page. Re-initialize and return it. */
_bt_pageinit(page, BufferGetPageSize(buf)); _bt_pageinit(page, BufferGetPageSize(buf));
@ -962,14 +969,13 @@ _bt_getbuf(Relation rel, Relation heaprel, BlockNumber blkno, int access)
} }
/* /*
* Extend the relation by one page. Need to use RBM_ZERO_AND_LOCK or * Extend the relation by one page. Need to use RBM_ZERO_AND_LOCK or we
* we risk a race condition against btvacuumscan --- see comments * risk a race condition against btvacuumscan --- see comments therein.
* therein. This forces us to repeat the valgrind request that * This forces us to repeat the valgrind request that _bt_lockbuf()
* _bt_lockbuf() otherwise would make, as we can't use _bt_lockbuf() * otherwise would make, as we can't use _bt_lockbuf() without introducing
* without introducing a race. * a race.
*/ */
buf = ExtendBufferedRel(EB_REL(rel), MAIN_FORKNUM, NULL, buf = ExtendBufferedRel(EB_REL(rel), MAIN_FORKNUM, NULL, EB_LOCK_FIRST);
EB_LOCK_FIRST);
if (!RelationUsesLocalBuffers(rel)) if (!RelationUsesLocalBuffers(rel))
VALGRIND_MAKE_MEM_DEFINED(BufferGetPage(buf), BLCKSZ); VALGRIND_MAKE_MEM_DEFINED(BufferGetPage(buf), BLCKSZ);
@ -977,19 +983,16 @@ _bt_getbuf(Relation rel, Relation heaprel, BlockNumber blkno, int access)
page = BufferGetPage(buf); page = BufferGetPage(buf);
Assert(PageIsNew(page)); Assert(PageIsNew(page));
_bt_pageinit(page, BufferGetPageSize(buf)); _bt_pageinit(page, BufferGetPageSize(buf));
}
/* ref count and lock type are correct */
return buf; return buf;
} }
/* /*
* _bt_relandgetbuf() -- release a locked buffer and get another one. * _bt_relandgetbuf() -- release a locked buffer and get another one.
* *
* This is equivalent to _bt_relbuf followed by _bt_getbuf, with the * This is equivalent to _bt_relbuf followed by _bt_getbuf. Also, if obuf is
* exception that blkno may not be P_NEW. Also, if obuf is InvalidBuffer * InvalidBuffer then it reduces to just _bt_getbuf; allowing this case
* then it reduces to just _bt_getbuf; allowing this case simplifies some * simplifies some callers.
* callers.
* *
* The original motivation for using this was to avoid two entries to the * The original motivation for using this was to avoid two entries to the
* bufmgr when one would do. However, now it's mainly just a notational * bufmgr when one would do. However, now it's mainly just a notational
@ -1001,7 +1004,7 @@ _bt_relandgetbuf(Relation rel, Buffer obuf, BlockNumber blkno, int access)
{ {
Buffer buf; Buffer buf;
Assert(blkno != P_NEW); Assert(BlockNumberIsValid(blkno));
if (BufferIsValid(obuf)) if (BufferIsValid(obuf))
_bt_unlockbuf(rel, obuf); _bt_unlockbuf(rel, obuf);
buf = ReleaseAndReadBuffer(obuf, rel, blkno); buf = ReleaseAndReadBuffer(obuf, rel, blkno);
@ -1272,14 +1275,14 @@ _bt_delitems_vacuum(Relation rel, Buffer buf,
* (a version that lacks the TIDs that are to be deleted). * (a version that lacks the TIDs that are to be deleted).
* *
* This is nearly the same as _bt_delitems_vacuum as far as what it does to * This is nearly the same as _bt_delitems_vacuum as far as what it does to
* the page, but it needs its own snapshotConflictHorizon (caller gets this * the page, but it needs its own snapshotConflictHorizon and isCatalogRel
* from tableam). This is used by the REDO routine to generate recovery * (from the tableam). This is used by the REDO routine to generate recovery
* conflicts. The other difference is that only _bt_delitems_vacuum will * conflicts. The other difference is that only _bt_delitems_vacuum will
* clear page's VACUUM cycle ID. * clear page's VACUUM cycle ID.
*/ */
static void static void
_bt_delitems_delete(Relation rel, Relation heaprel, Buffer buf, _bt_delitems_delete(Relation rel, Buffer buf,
TransactionId snapshotConflictHorizon, TransactionId snapshotConflictHorizon, bool isCatalogRel,
OffsetNumber *deletable, int ndeletable, OffsetNumber *deletable, int ndeletable,
BTVacuumPosting *updatable, int nupdatable) BTVacuumPosting *updatable, int nupdatable)
{ {
@ -1343,10 +1346,10 @@ _bt_delitems_delete(Relation rel, Relation heaprel, Buffer buf,
XLogRecPtr recptr; XLogRecPtr recptr;
xl_btree_delete xlrec_delete; xl_btree_delete xlrec_delete;
xlrec_delete.isCatalogRel = RelationIsAccessibleInLogicalDecoding(heaprel);
xlrec_delete.snapshotConflictHorizon = snapshotConflictHorizon; xlrec_delete.snapshotConflictHorizon = snapshotConflictHorizon;
xlrec_delete.ndeleted = ndeletable; xlrec_delete.ndeleted = ndeletable;
xlrec_delete.nupdated = nupdatable; xlrec_delete.nupdated = nupdatable;
xlrec_delete.isCatalogRel = isCatalogRel;
XLogBeginInsert(); XLogBeginInsert();
XLogRegisterBuffer(0, buf, REGBUF_STANDARD); XLogRegisterBuffer(0, buf, REGBUF_STANDARD);
@ -1517,6 +1520,7 @@ _bt_delitems_delete_check(Relation rel, Buffer buf, Relation heapRel,
{ {
Page page = BufferGetPage(buf); Page page = BufferGetPage(buf);
TransactionId snapshotConflictHorizon; TransactionId snapshotConflictHorizon;
bool isCatalogRel;
OffsetNumber postingidxoffnum = InvalidOffsetNumber; OffsetNumber postingidxoffnum = InvalidOffsetNumber;
int ndeletable = 0, int ndeletable = 0,
nupdatable = 0; nupdatable = 0;
@ -1525,6 +1529,7 @@ _bt_delitems_delete_check(Relation rel, Buffer buf, Relation heapRel,
/* Use tableam interface to determine which tuples to delete first */ /* Use tableam interface to determine which tuples to delete first */
snapshotConflictHorizon = table_index_delete_tuples(heapRel, delstate); snapshotConflictHorizon = table_index_delete_tuples(heapRel, delstate);
isCatalogRel = RelationIsAccessibleInLogicalDecoding(heapRel);
/* Should not WAL-log snapshotConflictHorizon unless it's required */ /* Should not WAL-log snapshotConflictHorizon unless it's required */
if (!XLogStandbyInfoActive()) if (!XLogStandbyInfoActive())
@ -1670,8 +1675,8 @@ _bt_delitems_delete_check(Relation rel, Buffer buf, Relation heapRel,
} }
/* Physically delete tuples (or TIDs) using deletable (or updatable) */ /* Physically delete tuples (or TIDs) using deletable (or updatable) */
_bt_delitems_delete(rel, heapRel, buf, snapshotConflictHorizon, deletable, _bt_delitems_delete(rel, buf, snapshotConflictHorizon, isCatalogRel,
ndeletable, updatable, nupdatable); deletable, ndeletable, updatable, nupdatable);
/* be tidy */ /* be tidy */
for (int i = 0; i < nupdatable; i++) for (int i = 0; i < nupdatable; i++)
@ -1692,8 +1697,7 @@ _bt_delitems_delete_check(Relation rel, Buffer buf, Relation heapRel,
* same level must always be locked left to right to avoid deadlocks. * same level must always be locked left to right to avoid deadlocks.
*/ */
static bool static bool
_bt_leftsib_splitflag(Relation rel, Relation heaprel, BlockNumber leftsib, _bt_leftsib_splitflag(Relation rel, BlockNumber leftsib, BlockNumber target)
BlockNumber target)
{ {
Buffer buf; Buffer buf;
Page page; Page page;
@ -1704,7 +1708,7 @@ _bt_leftsib_splitflag(Relation rel, Relation heaprel, BlockNumber leftsib,
if (leftsib == P_NONE) if (leftsib == P_NONE)
return false; return false;
buf = _bt_getbuf(rel, heaprel, leftsib, BT_READ); buf = _bt_getbuf(rel, leftsib, BT_READ);
page = BufferGetPage(buf); page = BufferGetPage(buf);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
@ -1750,7 +1754,7 @@ _bt_leftsib_splitflag(Relation rel, Relation heaprel, BlockNumber leftsib,
* to-be-deleted subtree.) * to-be-deleted subtree.)
*/ */
static bool static bool
_bt_rightsib_halfdeadflag(Relation rel, Relation heaprel, BlockNumber leafrightsib) _bt_rightsib_halfdeadflag(Relation rel, BlockNumber leafrightsib)
{ {
Buffer buf; Buffer buf;
Page page; Page page;
@ -1759,7 +1763,7 @@ _bt_rightsib_halfdeadflag(Relation rel, Relation heaprel, BlockNumber leafrights
Assert(leafrightsib != P_NONE); Assert(leafrightsib != P_NONE);
buf = _bt_getbuf(rel, heaprel, leafrightsib, BT_READ); buf = _bt_getbuf(rel, leafrightsib, BT_READ);
page = BufferGetPage(buf); page = BufferGetPage(buf);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
@ -1948,18 +1952,18 @@ _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate)
* marked with INCOMPLETE_SPLIT flag before proceeding * marked with INCOMPLETE_SPLIT flag before proceeding
*/ */
Assert(leafblkno == scanblkno); Assert(leafblkno == scanblkno);
if (_bt_leftsib_splitflag(rel, vstate->info->heaprel, leftsib, leafblkno)) if (_bt_leftsib_splitflag(rel, leftsib, leafblkno))
{ {
ReleaseBuffer(leafbuf); ReleaseBuffer(leafbuf);
return; return;
} }
/* we need an insertion scan key for the search, so build one */ /* we need an insertion scan key for the search, so build one */
itup_key = _bt_mkscankey(rel, vstate->info->heaprel, targetkey); itup_key = _bt_mkscankey(rel, targetkey);
/* find the leftmost leaf page with matching pivot/high key */ /* find the leftmost leaf page with matching pivot/high key */
itup_key->pivotsearch = true; itup_key->pivotsearch = true;
stack = _bt_search(rel, vstate->info->heaprel, itup_key, stack = _bt_search(rel, NULL, itup_key, &sleafbuf, BT_READ,
&sleafbuf, BT_READ, NULL); NULL);
/* won't need a second lock or pin on leafbuf */ /* won't need a second lock or pin on leafbuf */
_bt_relbuf(rel, sleafbuf); _bt_relbuf(rel, sleafbuf);
@ -1990,7 +1994,8 @@ _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate)
* leafbuf page half-dead. * leafbuf page half-dead.
*/ */
Assert(P_ISLEAF(opaque) && !P_IGNORE(opaque)); Assert(P_ISLEAF(opaque) && !P_IGNORE(opaque));
if (!_bt_mark_page_halfdead(rel, vstate->info->heaprel, leafbuf, stack)) if (!_bt_mark_page_halfdead(rel, vstate->info->heaprel, leafbuf,
stack))
{ {
_bt_relbuf(rel, leafbuf); _bt_relbuf(rel, leafbuf);
return; return;
@ -2053,7 +2058,7 @@ _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate)
if (!rightsib_empty) if (!rightsib_empty)
break; break;
leafbuf = _bt_getbuf(rel, vstate->info->heaprel, rightsib, BT_WRITE); leafbuf = _bt_getbuf(rel, rightsib, BT_WRITE);
} }
} }
@ -2066,6 +2071,10 @@ _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate)
* but may include additional internal pages (at most one per level of the * but may include additional internal pages (at most one per level of the
* tree below the root). * tree below the root).
* *
* Caller must pass a valid heaprel, since it's just about possible that our
* call to _bt_lock_subtree_parent will need to allocate a new index page to
* complete a page split. Every call to _bt_allocbuf needs to pass a heaprel.
*
* Returns 'false' if leafbuf is unsafe to delete, usually because leafbuf is * Returns 'false' if leafbuf is unsafe to delete, usually because leafbuf is
* the rightmost child of its parent (and parent has more than one downlink). * the rightmost child of its parent (and parent has more than one downlink).
* Returns 'true' when the first stage of page deletion completed * Returns 'true' when the first stage of page deletion completed
@ -2094,6 +2103,7 @@ _bt_mark_page_halfdead(Relation rel, Relation heaprel, Buffer leafbuf,
Assert(!P_RIGHTMOST(opaque) && !P_ISROOT(opaque) && Assert(!P_RIGHTMOST(opaque) && !P_ISROOT(opaque) &&
P_ISLEAF(opaque) && !P_IGNORE(opaque) && P_ISLEAF(opaque) && !P_IGNORE(opaque) &&
P_FIRSTDATAKEY(opaque) > PageGetMaxOffsetNumber(page)); P_FIRSTDATAKEY(opaque) > PageGetMaxOffsetNumber(page));
Assert(heaprel != NULL);
/* /*
* Save info about the leaf page. * Save info about the leaf page.
@ -2108,7 +2118,7 @@ _bt_mark_page_halfdead(Relation rel, Relation heaprel, Buffer leafbuf,
* delete the downlink. It would fail the "right sibling of target page * delete the downlink. It would fail the "right sibling of target page
* is also the next child in parent page" cross-check below. * is also the next child in parent page" cross-check below.
*/ */
if (_bt_rightsib_halfdeadflag(rel, heaprel, leafrightsib)) if (_bt_rightsib_halfdeadflag(rel, leafrightsib))
{ {
elog(DEBUG1, "could not delete page %u because its right sibling %u is half-dead", elog(DEBUG1, "could not delete page %u because its right sibling %u is half-dead",
leafblkno, leafrightsib); leafblkno, leafrightsib);
@ -2352,7 +2362,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
Assert(target != leafblkno); Assert(target != leafblkno);
/* Fetch the block number of the target's left sibling */ /* Fetch the block number of the target's left sibling */
buf = _bt_getbuf(rel, vstate->info->heaprel, target, BT_READ); buf = _bt_getbuf(rel, target, BT_READ);
page = BufferGetPage(buf); page = BufferGetPage(buf);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
leftsib = opaque->btpo_prev; leftsib = opaque->btpo_prev;
@ -2379,7 +2389,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
_bt_lockbuf(rel, leafbuf, BT_WRITE); _bt_lockbuf(rel, leafbuf, BT_WRITE);
if (leftsib != P_NONE) if (leftsib != P_NONE)
{ {
lbuf = _bt_getbuf(rel, vstate->info->heaprel, leftsib, BT_WRITE); lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf); page = BufferGetPage(lbuf);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
while (P_ISDELETED(opaque) || opaque->btpo_next != target) while (P_ISDELETED(opaque) || opaque->btpo_next != target)
@ -2427,7 +2437,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
CHECK_FOR_INTERRUPTS(); CHECK_FOR_INTERRUPTS();
/* step right one page */ /* step right one page */
lbuf = _bt_getbuf(rel, vstate->info->heaprel, leftsib, BT_WRITE); lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf); page = BufferGetPage(lbuf);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
} }
@ -2491,7 +2501,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
* And next write-lock the (current) right sibling. * And next write-lock the (current) right sibling.
*/ */
rightsib = opaque->btpo_next; rightsib = opaque->btpo_next;
rbuf = _bt_getbuf(rel, vstate->info->heaprel, rightsib, BT_WRITE); rbuf = _bt_getbuf(rel, rightsib, BT_WRITE);
page = BufferGetPage(rbuf); page = BufferGetPage(rbuf);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
@ -2538,7 +2548,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
* of doing so are slim, and the locking considerations daunting.) * of doing so are slim, and the locking considerations daunting.)
* *
* We can safely acquire a lock on the metapage here --- see comments for * We can safely acquire a lock on the metapage here --- see comments for
* _bt_newroot(). * _bt_newlevel().
*/ */
if (leftsib == P_NONE && rightsib_is_rightmost) if (leftsib == P_NONE && rightsib_is_rightmost)
{ {
@ -2547,8 +2557,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
if (P_RIGHTMOST(opaque)) if (P_RIGHTMOST(opaque))
{ {
/* rightsib will be the only one left on the level */ /* rightsib will be the only one left on the level */
metabuf = _bt_getbuf(rel, vstate->info->heaprel, BTREE_METAPAGE, metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_WRITE);
BT_WRITE);
metapg = BufferGetPage(metabuf); metapg = BufferGetPage(metabuf);
metad = BTPageGetMeta(metapg); metad = BTPageGetMeta(metapg);
@ -2906,7 +2915,7 @@ _bt_lock_subtree_parent(Relation rel, Relation heaprel, BlockNumber child,
* *
* Note: We deliberately avoid completing incomplete splits here. * Note: We deliberately avoid completing incomplete splits here.
*/ */
if (_bt_leftsib_splitflag(rel, heaprel, leftsibparent, parent)) if (_bt_leftsib_splitflag(rel, leftsibparent, parent))
return false; return false;
/* Recurse to examine child page's grandparent page */ /* Recurse to examine child page's grandparent page */
@ -2976,6 +2985,7 @@ _bt_pendingfsm_finalize(Relation rel, BTVacState *vstate)
Relation heaprel = vstate->info->heaprel; Relation heaprel = vstate->info->heaprel;
Assert(stats->pages_newly_deleted >= vstate->npendingpages); Assert(stats->pages_newly_deleted >= vstate->npendingpages);
Assert(heaprel != NULL);
if (vstate->npendingpages == 0) if (vstate->npendingpages == 0)
{ {

4
src/backend/access/nbtree/nbtree.c
View File

@ -835,7 +835,7 @@ btvacuumcleanup(IndexVacuumInfo *info, IndexBulkDeleteResult *stats)
if (stats == NULL) if (stats == NULL)
{ {
/* Check if VACUUM operation can entirely avoid btvacuumscan() call */ /* Check if VACUUM operation can entirely avoid btvacuumscan() call */
if (!_bt_vacuum_needs_cleanup(info->index, info->heaprel)) if (!_bt_vacuum_needs_cleanup(info->index))
return NULL; return NULL;
/* /*
@ -871,7 +871,7 @@ btvacuumcleanup(IndexVacuumInfo *info, IndexBulkDeleteResult *stats)
*/ */
Assert(stats->pages_deleted >= stats->pages_free); Assert(stats->pages_deleted >= stats->pages_free);
num_delpages = stats->pages_deleted - stats->pages_free; num_delpages = stats->pages_deleted - stats->pages_free;
_bt_set_cleanup_info(info->index, info->heaprel, num_delpages); _bt_set_cleanup_info(info->index, num_delpages);
/* /*
* It's quite possible for us to be fooled by concurrent page splits into * It's quite possible for us to be fooled by concurrent page splits into

48
src/backend/access/nbtree/nbtsearch.c
View File

@ -42,8 +42,7 @@ static bool _bt_steppage(IndexScanDesc scan, ScanDirection dir);
static bool _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir); static bool _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir);
static bool _bt_parallel_readpage(IndexScanDesc scan, BlockNumber blkno, static bool _bt_parallel_readpage(IndexScanDesc scan, BlockNumber blkno,
ScanDirection dir); ScanDirection dir);
static Buffer _bt_walk_left(Relation rel, Relation heaprel, Buffer buf, static Buffer _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot);
Snapshot snapshot);
static bool _bt_endpoint(IndexScanDesc scan, ScanDirection dir); static bool _bt_endpoint(IndexScanDesc scan, ScanDirection dir);
static inline void _bt_initialize_more_data(BTScanOpaque so, ScanDirection dir); static inline void _bt_initialize_more_data(BTScanOpaque so, ScanDirection dir);
@ -92,6 +91,9 @@ _bt_drop_lock_and_maybe_pin(IndexScanDesc scan, BTScanPos sp)
* When access = BT_READ, an empty index will result in *bufP being set to * When access = BT_READ, an empty index will result in *bufP being set to
* InvalidBuffer. Also, in BT_WRITE mode, any incomplete splits encountered * InvalidBuffer. Also, in BT_WRITE mode, any incomplete splits encountered
* during the search will be finished. * during the search will be finished.
*
* heaprel must be provided by callers that pass access = BT_WRITE, since we
* might need to allocate a new root page for caller -- see _bt_allocbuf.
*/ */
BTStack BTStack
_bt_search(Relation rel, Relation heaprel, BTScanInsert key, Buffer *bufP, _bt_search(Relation rel, Relation heaprel, BTScanInsert key, Buffer *bufP,
@ -100,6 +102,10 @@ _bt_search(Relation rel, Relation heaprel, BTScanInsert key, Buffer *bufP,
BTStack stack_in = NULL; BTStack stack_in = NULL;
int page_access = BT_READ; int page_access = BT_READ;
/* heaprel must be set whenever _bt_allocbuf is reachable */
Assert(access == BT_READ || access == BT_WRITE);
Assert(access == BT_READ || heaprel != NULL);
/* Get the root page to start with */ /* Get the root page to start with */
*bufP = _bt_getroot(rel, heaprel, access); *bufP = _bt_getroot(rel, heaprel, access);
@ -222,8 +228,8 @@ _bt_search(Relation rel, Relation heaprel, BTScanInsert key, Buffer *bufP,
* *
* If forupdate is true, we will attempt to finish any incomplete splits * If forupdate is true, we will attempt to finish any incomplete splits
* that we encounter. This is required when locking a target page for an * that we encounter. This is required when locking a target page for an
* insertion, because we don't allow inserting on a page before the split * insertion, because we don't allow inserting on a page before the split is
* is completed. 'stack' is only used if forupdate is true. * completed. 'heaprel' and 'stack' are only used if forupdate is true.
* *
* On entry, we have the buffer pinned and a lock of the type specified by * On entry, we have the buffer pinned and a lock of the type specified by
* 'access'. If we move right, we release the buffer and lock and acquire * 'access'. If we move right, we release the buffer and lock and acquire
@ -247,6 +253,8 @@ _bt_moveright(Relation rel,
BTPageOpaque opaque; BTPageOpaque opaque;
int32 cmpval; int32 cmpval;
Assert(!forupdate || heaprel != NULL);
/* /*
* When nextkey = false (normal case): if the scan key that brought us to * When nextkey = false (normal case): if the scan key that brought us to
* this page is > the high key stored on the page, then the page has split * this page is > the high key stored on the page, then the page has split
@ -295,7 +303,7 @@ _bt_moveright(Relation rel,
_bt_relbuf(rel, buf); _bt_relbuf(rel, buf);
/* re-acquire the lock in the right mode, and re-check */ /* re-acquire the lock in the right mode, and re-check */
buf = _bt_getbuf(rel, heaprel, blkno, access); buf = _bt_getbuf(rel, blkno, access);
continue; continue;
} }
@ -862,7 +870,6 @@ bool
_bt_first(IndexScanDesc scan, ScanDirection dir) _bt_first(IndexScanDesc scan, ScanDirection dir)
{ {
Relation rel = scan->indexRelation; Relation rel = scan->indexRelation;
Relation heaprel = scan->heapRelation;
BTScanOpaque so = (BTScanOpaque) scan->opaque; BTScanOpaque so = (BTScanOpaque) scan->opaque;
Buffer buf; Buffer buf;
BTStack stack; BTStack stack;
@ -1355,7 +1362,7 @@ _bt_first(IndexScanDesc scan, ScanDirection dir)
} }
/* Initialize remaining insertion scan key fields */ /* Initialize remaining insertion scan key fields */
_bt_metaversion(rel, heaprel, &inskey.heapkeyspace, &inskey.allequalimage); _bt_metaversion(rel, &inskey.heapkeyspace, &inskey.allequalimage);
inskey.anynullkeys = false; /* unused */ inskey.anynullkeys = false; /* unused */
inskey.nextkey = nextkey; inskey.nextkey = nextkey;
inskey.pivotsearch = false; inskey.pivotsearch = false;
@ -1366,7 +1373,7 @@ _bt_first(IndexScanDesc scan, ScanDirection dir)
* Use the manufactured insertion scan key to descend the tree and * Use the manufactured insertion scan key to descend the tree and
* position ourselves on the target leaf page. * position ourselves on the target leaf page.
*/ */
stack = _bt_search(rel, heaprel, &inskey, &buf, BT_READ, scan->xs_snapshot); stack = _bt_search(rel, NULL, &inskey, &buf, BT_READ, scan->xs_snapshot);
/* don't need to keep the stack around... */ /* don't need to keep the stack around... */
_bt_freestack(stack); _bt_freestack(stack);
@ -2007,7 +2014,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
/* check for interrupts while we're not holding any buffer lock */ /* check for interrupts while we're not holding any buffer lock */
CHECK_FOR_INTERRUPTS(); CHECK_FOR_INTERRUPTS();
/* step right one page */ /* step right one page */
so->currPos.buf = _bt_getbuf(rel, scan->heapRelation, blkno, BT_READ); so->currPos.buf = _bt_getbuf(rel, blkno, BT_READ);
page = BufferGetPage(so->currPos.buf); page = BufferGetPage(so->currPos.buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page); TestForOldSnapshot(scan->xs_snapshot, rel, page);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
@ -2081,8 +2088,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
if (BTScanPosIsPinned(so->currPos)) if (BTScanPosIsPinned(so->currPos))
_bt_lockbuf(rel, so->currPos.buf, BT_READ); _bt_lockbuf(rel, so->currPos.buf, BT_READ);
else else
so->currPos.buf = _bt_getbuf(rel, scan->heapRelation, so->currPos.buf = _bt_getbuf(rel, so->currPos.currPage, BT_READ);
so->currPos.currPage, BT_READ);
for (;;) for (;;)
{ {
@ -2096,8 +2102,8 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
} }
/* Step to next physical page */ /* Step to next physical page */
so->currPos.buf = _bt_walk_left(rel, scan->heapRelation, so->currPos.buf = _bt_walk_left(rel, so->currPos.buf,
so->currPos.buf, scan->xs_snapshot); scan->xs_snapshot);
/* if we're physically at end of index, return failure */ /* if we're physically at end of index, return failure */
if (so->currPos.buf == InvalidBuffer) if (so->currPos.buf == InvalidBuffer)
@ -2144,8 +2150,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
BTScanPosInvalidate(so->currPos); BTScanPosInvalidate(so->currPos);
return false; return false;
} }
so->currPos.buf = _bt_getbuf(rel, scan->heapRelation, blkno, so->currPos.buf = _bt_getbuf(rel, blkno, BT_READ);
BT_READ);
} }
} }
} }
@ -2190,7 +2195,7 @@ _bt_parallel_readpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
* again if it's important. * again if it's important.
*/ */
static Buffer static Buffer
_bt_walk_left(Relation rel, Relation heaprel, Buffer buf, Snapshot snapshot) _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
{ {
Page page; Page page;
BTPageOpaque opaque; BTPageOpaque opaque;
@ -2218,7 +2223,7 @@ _bt_walk_left(Relation rel, Relation heaprel, Buffer buf, Snapshot snapshot)
_bt_relbuf(rel, buf); _bt_relbuf(rel, buf);
/* check for interrupts while we're not holding any buffer lock */ /* check for interrupts while we're not holding any buffer lock */
CHECK_FOR_INTERRUPTS(); CHECK_FOR_INTERRUPTS();
buf = _bt_getbuf(rel, heaprel, blkno, BT_READ); buf = _bt_getbuf(rel, blkno, BT_READ);
page = BufferGetPage(buf); page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page); TestForOldSnapshot(snapshot, rel, page);
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
@ -2309,7 +2314,7 @@ _bt_walk_left(Relation rel, Relation heaprel, Buffer buf, Snapshot snapshot)
* The returned buffer is pinned and read-locked. * The returned buffer is pinned and read-locked.
*/ */
Buffer Buffer
_bt_get_endpoint(Relation rel, Relation heaprel, uint32 level, bool rightmost, _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
Snapshot snapshot) Snapshot snapshot)
{ {
Buffer buf; Buffer buf;
@ -2325,9 +2330,9 @@ _bt_get_endpoint(Relation rel, Relation heaprel, uint32 level, bool rightmost,
* smarter about intermediate levels.) * smarter about intermediate levels.)
*/ */
if (level == 0) if (level == 0)
buf = _bt_getroot(rel, heaprel, BT_READ); buf = _bt_getroot(rel, NULL, BT_READ);
else else
buf = _bt_gettrueroot(rel, heaprel); buf = _bt_gettrueroot(rel);
if (!BufferIsValid(buf)) if (!BufferIsValid(buf))
return InvalidBuffer; return InvalidBuffer;
@ -2408,8 +2413,7 @@ _bt_endpoint(IndexScanDesc scan, ScanDirection dir)
* version of _bt_search(). We don't maintain a stack since we know we * version of _bt_search(). We don't maintain a stack since we know we
* won't need it. * won't need it.
*/ */
buf = _bt_get_endpoint(rel, scan->heapRelation, 0, buf = _bt_get_endpoint(rel, 0, ScanDirectionIsBackward(dir), scan->xs_snapshot);
ScanDirectionIsBackward(dir), scan->xs_snapshot);
if (!BufferIsValid(buf)) if (!BufferIsValid(buf))
{ {

2
src/backend/access/nbtree/nbtsort.c
View File

@ -566,7 +566,7 @@ _bt_leafbuild(BTSpool *btspool, BTSpool *btspool2)
wstate.heap = btspool->heap; wstate.heap = btspool->heap;
wstate.index = btspool->index; wstate.index = btspool->index;
wstate.inskey = _bt_mkscankey(wstate.index, btspool->heap, NULL); wstate.inskey = _bt_mkscankey(wstate.index, NULL);
/* _bt_mkscankey() won't set allequalimage without metapage */ /* _bt_mkscankey() won't set allequalimage without metapage */
wstate.inskey->allequalimage = _bt_allequalimage(wstate.index, true); wstate.inskey->allequalimage = _bt_allequalimage(wstate.index, true);
wstate.btws_use_wal = RelationNeedsWAL(wstate.index); wstate.btws_use_wal = RelationNeedsWAL(wstate.index);

7
src/backend/access/nbtree/nbtutils.c
View File

@ -87,7 +87,7 @@ static int _bt_keep_natts(Relation rel, IndexTuple lastleft,
* field themselves. * field themselves.
*/ */
BTScanInsert BTScanInsert
_bt_mkscankey(Relation rel, Relation heaprel, IndexTuple itup) _bt_mkscankey(Relation rel, IndexTuple itup)
{ {
BTScanInsert key; BTScanInsert key;
ScanKey skey; ScanKey skey;
@ -112,7 +112,7 @@ _bt_mkscankey(Relation rel, Relation heaprel, IndexTuple itup)
key = palloc(offsetof(BTScanInsertData, scankeys) + key = palloc(offsetof(BTScanInsertData, scankeys) +
sizeof(ScanKeyData) * indnkeyatts); sizeof(ScanKeyData) * indnkeyatts);
if (itup) if (itup)
_bt_metaversion(rel, heaprel, &key->heapkeyspace, &key->allequalimage); _bt_metaversion(rel, &key->heapkeyspace, &key->allequalimage);
else else
{ {
/* Utility statement callers can set these fields themselves */ /* Utility statement callers can set these fields themselves */
@ -1761,8 +1761,7 @@ _bt_killitems(IndexScanDesc scan)
droppedpin = true; droppedpin = true;
/* Attempt to re-read the buffer, getting pin and lock. */ /* Attempt to re-read the buffer, getting pin and lock. */
buf = _bt_getbuf(scan->indexRelation, scan->heapRelation, buf = _bt_getbuf(scan->indexRelation, so->currPos.currPage, BT_READ);
so->currPos.currPage, BT_READ);
page = BufferGetPage(buf); page = BufferGetPage(buf);
if (BufferGetLSNAtomic(buf) == so->currPos.lsn) if (BufferGetLSNAtomic(buf) == so->currPos.lsn)

2
src/backend/optimizer/util/plancat.c
View File

@ -462,7 +462,7 @@ get_relation_info(PlannerInfo *root, Oid relationObjectId, bool inhparent,
* For btrees, get tree height while we have the index * For btrees, get tree height while we have the index
* open * open
*/ */
info->tree_height = _bt_getrootheight(indexRelation, relation); info->tree_height = _bt_getrootheight(indexRelation);
} }
else else
{ {

5
src/backend/utils/sort/tuplesortvariants.c
View File

@ -207,7 +207,6 @@ tuplesort_begin_heap(TupleDesc tupDesc,
Tuplesortstate * Tuplesortstate *
tuplesort_begin_cluster(TupleDesc tupDesc, tuplesort_begin_cluster(TupleDesc tupDesc,
Relation indexRel, Relation indexRel,
Relation heaprel,
int workMem, int workMem,
SortCoordinate coordinate, int sortopt) SortCoordinate coordinate, int sortopt)
{ {
@ -261,7 +260,7 @@ tuplesort_begin_cluster(TupleDesc tupDesc,
arg->tupDesc = tupDesc; /* assume we need not copy tupDesc */ arg->tupDesc = tupDesc; /* assume we need not copy tupDesc */
indexScanKey = _bt_mkscankey(indexRel, heaprel, NULL); indexScanKey = _bt_mkscankey(indexRel, NULL);
if (arg->indexInfo->ii_Expressions != NULL) if (arg->indexInfo->ii_Expressions != NULL)
{ {
@ -362,7 +361,7 @@ tuplesort_begin_index_btree(Relation heapRel,
arg->enforceUnique = enforceUnique; arg->enforceUnique = enforceUnique;
arg->uniqueNullsNotDistinct = uniqueNullsNotDistinct; arg->uniqueNullsNotDistinct = uniqueNullsNotDistinct;
indexScanKey = _bt_mkscankey(indexRel, heapRel, NULL); indexScanKey = _bt_mkscankey(indexRel, NULL);
/* Prepare SortSupport data for each column */ /* Prepare SortSupport data for each column */
base->sortKeys = (SortSupport) palloc0(base->nKeys * base->sortKeys = (SortSupport) palloc0(base->nKeys *

26
src/include/access/nbtree.h
View File

@ -293,11 +293,14 @@ BTPageIsRecyclable(Page page, Relation heaprel)
BTPageOpaque opaque; BTPageOpaque opaque;
Assert(!PageIsNew(page)); Assert(!PageIsNew(page));
Assert(heaprel != NULL);
/* Recycling okay iff page is deleted and safexid is old enough */ /* Recycling okay iff page is deleted and safexid is old enough */
opaque = BTPageGetOpaque(page); opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque)) if (P_ISDELETED(opaque))
{ {
FullTransactionId safexid = BTPageGetDeleteXid(page);
/* /*
* The page was deleted, but when? If it was just deleted, a scan * The page was deleted, but when? If it was just deleted, a scan
* might have seen the downlink to it, and will read the page later. * might have seen the downlink to it, and will read the page later.
@ -308,7 +311,7 @@ BTPageIsRecyclable(Page page, Relation heaprel)
* anyone. If not, then no scan that's still in progress could have * anyone. If not, then no scan that's still in progress could have
* seen its downlink, and we can recycle it. * seen its downlink, and we can recycle it.
*/ */
return GlobalVisCheckRemovableFullXid(heaprel, BTPageGetDeleteXid(page)); return GlobalVisCheckRemovableFullXid(heaprel, safexid);
} }
return false; return false;
@ -1194,18 +1197,17 @@ extern OffsetNumber _bt_findsplitloc(Relation rel, Page origpage,
*/ */
extern void _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level, extern void _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level,
bool allequalimage); bool allequalimage);
extern bool _bt_vacuum_needs_cleanup(Relation rel, Relation heaprel); extern bool _bt_vacuum_needs_cleanup(Relation rel);
extern void _bt_set_cleanup_info(Relation rel, Relation heaprel, extern void _bt_set_cleanup_info(Relation rel, BlockNumber num_delpages);
BlockNumber num_delpages);
extern void _bt_upgrademetapage(Page page); extern void _bt_upgrademetapage(Page page);
extern Buffer _bt_getroot(Relation rel, Relation heaprel, int access); extern Buffer _bt_getroot(Relation rel, Relation heaprel, int access);
extern Buffer _bt_gettrueroot(Relation rel, Relation heaprel); extern Buffer _bt_gettrueroot(Relation rel);
extern int _bt_getrootheight(Relation rel, Relation heaprel); extern int _bt_getrootheight(Relation rel);
extern void _bt_metaversion(Relation rel, Relation heaprel, bool *heapkeyspace, extern void _bt_metaversion(Relation rel, bool *heapkeyspace,
bool *allequalimage); bool *allequalimage);
extern void _bt_checkpage(Relation rel, Buffer buf); extern void _bt_checkpage(Relation rel, Buffer buf);
extern Buffer _bt_getbuf(Relation rel, Relation heaprel, BlockNumber blkno, extern Buffer _bt_getbuf(Relation rel, BlockNumber blkno, int access);
int access); extern Buffer _bt_allocbuf(Relation rel, Relation heaprel);
extern Buffer _bt_relandgetbuf(Relation rel, Buffer obuf, extern Buffer _bt_relandgetbuf(Relation rel, Buffer obuf,
BlockNumber blkno, int access); BlockNumber blkno, int access);
extern void _bt_relbuf(Relation rel, Buffer buf); extern void _bt_relbuf(Relation rel, Buffer buf);
@ -1237,13 +1239,13 @@ extern OffsetNumber _bt_binsrch_insert(Relation rel, BTInsertState insertstate);
extern int32 _bt_compare(Relation rel, BTScanInsert key, Page page, OffsetNumber offnum); extern int32 _bt_compare(Relation rel, BTScanInsert key, Page page, OffsetNumber offnum);
extern bool _bt_first(IndexScanDesc scan, ScanDirection dir); extern bool _bt_first(IndexScanDesc scan, ScanDirection dir);
extern bool _bt_next(IndexScanDesc scan, ScanDirection dir); extern bool _bt_next(IndexScanDesc scan, ScanDirection dir);
extern Buffer _bt_get_endpoint(Relation rel, Relation heaprel, uint32 level, extern Buffer _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
bool rightmost, Snapshot snapshot); Snapshot snapshot);
/* /*
* prototypes for functions in nbtutils.c * prototypes for functions in nbtutils.c
*/ */
extern BTScanInsert _bt_mkscankey(Relation rel, Relation heaprel, IndexTuple itup); extern BTScanInsert _bt_mkscankey(Relation rel, IndexTuple itup);
extern void _bt_freestack(BTStack stack); extern void _bt_freestack(BTStack stack);
extern void _bt_preprocess_array_keys(IndexScanDesc scan); extern void _bt_preprocess_array_keys(IndexScanDesc scan);
extern void _bt_start_array_keys(IndexScanDesc scan, ScanDirection dir); extern void _bt_start_array_keys(IndexScanDesc scan, ScanDirection dir);

4
src/include/utils/tuplesort.h
View File

@ -399,9 +399,7 @@ extern Tuplesortstate *tuplesort_begin_heap(TupleDesc tupDesc,
int workMem, SortCoordinate coordinate, int workMem, SortCoordinate coordinate,
int sortopt); int sortopt);
extern Tuplesortstate *tuplesort_begin_cluster(TupleDesc tupDesc, extern Tuplesortstate *tuplesort_begin_cluster(TupleDesc tupDesc,
Relation indexRel, Relation indexRel, int workMem,
Relation heaprel,
int workMem,
SortCoordinate coordinate, SortCoordinate coordinate,
int sortopt); int sortopt);
extern Tuplesortstate *tuplesort_begin_index_btree(Relation heapRel, extern Tuplesortstate *tuplesort_begin_index_btree(Relation heapRel,