summaryrefslogtreecommitdiff
path: root/src/btree.c
diff options
context:
space:
mode:
Diffstat (limited to 'src/btree.c')
-rw-r--r--src/btree.c501
1 files changed, 335 insertions, 166 deletions
diff --git a/src/btree.c b/src/btree.c
index d34d6ee..3ca6058 100644
--- a/src/btree.c
+++ b/src/btree.c
@@ -43,6 +43,25 @@ int sqlite3BtreeTrace=1; /* True to enable tracing */
*/
#define get2byteNotZero(X) (((((int)get2byte(X))-1)&0xffff)+1)
+/*
+** Values passed as the 5th argument to allocateBtreePage()
+*/
+#define BTALLOC_ANY 0 /* Allocate any page */
+#define BTALLOC_EXACT 1 /* Allocate exact page if possible */
+#define BTALLOC_LE 2 /* Allocate any page <= the parameter */
+
+/*
+** Macro IfNotOmitAV(x) returns (x) if SQLITE_OMIT_AUTOVACUUM is not
+** defined, or 0 if it is. For example:
+**
+** bIncrVacuum = IfNotOmitAV(pBtShared->incrVacuum);
+*/
+#ifndef SQLITE_OMIT_AUTOVACUUM
+#define IfNotOmitAV(expr) (expr)
+#else
+#define IfNotOmitAV(expr) 0
+#endif
+
#ifndef SQLITE_OMIT_SHARED_CACHE
/*
** A list of BtShared objects that are eligible for participation
@@ -557,6 +576,19 @@ static void btreeClearHasContent(BtShared *pBt){
}
/*
+** Release all of the apPage[] pages for a cursor.
+*/
+static void btreeReleaseAllCursorPages(BtCursor *pCur){
+ int i;
+ for(i=0; i<=pCur->iPage; i++){
+ releasePage(pCur->apPage[i]);
+ pCur->apPage[i] = 0;
+ }
+ pCur->iPage = -1;
+}
+
+
+/*
** Save the current cursor position in the variables BtCursor.nKey
** and BtCursor.pKey. The cursor's state is set to CURSOR_REQUIRESEEK.
**
@@ -595,12 +627,7 @@ static int saveCursorPosition(BtCursor *pCur){
assert( !pCur->apPage[0]->intKey || !pCur->pKey );
if( rc==SQLITE_OK ){
- int i;
- for(i=0; i<=pCur->iPage; i++){
- releasePage(pCur->apPage[i]);
- pCur->apPage[i] = 0;
- }
- pCur->iPage = -1;
+ btreeReleaseAllCursorPages(pCur);
pCur->eState = CURSOR_REQUIRESEEK;
}
@@ -618,11 +645,15 @@ static int saveAllCursors(BtShared *pBt, Pgno iRoot, BtCursor *pExcept){
assert( sqlite3_mutex_held(pBt->mutex) );
assert( pExcept==0 || pExcept->pBt==pBt );
for(p=pBt->pCursor; p; p=p->pNext){
- if( p!=pExcept && (0==iRoot || p->pgnoRoot==iRoot) &&
- p->eState==CURSOR_VALID ){
- int rc = saveCursorPosition(p);
- if( SQLITE_OK!=rc ){
- return rc;
+ if( p!=pExcept && (0==iRoot || p->pgnoRoot==iRoot) ){
+ if( p->eState==CURSOR_VALID ){
+ int rc = saveCursorPosition(p);
+ if( SQLITE_OK!=rc ){
+ return rc;
+ }
+ }else{
+ testcase( p->iPage>0 );
+ btreeReleaseAllCursorPages(p);
}
}
}
@@ -1550,13 +1581,17 @@ static int btreeGetPage(
BtShared *pBt, /* The btree */
Pgno pgno, /* Number of the page to fetch */
MemPage **ppPage, /* Return the page in this parameter */
- int noContent /* Do not load page content if true */
+ int noContent, /* Do not load page content if true */
+ int bReadonly /* True if a read-only (mmap) page is ok */
){
int rc;
DbPage *pDbPage;
+ int flags = (noContent ? PAGER_ACQUIRE_NOCONTENT : 0)
+ | (bReadonly ? PAGER_ACQUIRE_READONLY : 0);
+ assert( noContent==0 || bReadonly==0 );
assert( sqlite3_mutex_held(pBt->mutex) );
- rc = sqlite3PagerAcquire(pBt->pPager, pgno, (DbPage**)&pDbPage, noContent);
+ rc = sqlite3PagerAcquire(pBt->pPager, pgno, (DbPage**)&pDbPage, flags);
if( rc ) return rc;
*ppPage = btreePageFromDbPage(pDbPage, pgno, pBt);
return SQLITE_OK;
@@ -1599,9 +1634,10 @@ u32 sqlite3BtreeLastPage(Btree *p){
** may remain unchanged, or it may be set to an invalid value.
*/
static int getAndInitPage(
- BtShared *pBt, /* The database file */
- Pgno pgno, /* Number of the page to get */
- MemPage **ppPage /* Write the page pointer here */
+ BtShared *pBt, /* The database file */
+ Pgno pgno, /* Number of the page to get */
+ MemPage **ppPage, /* Write the page pointer here */
+ int bReadonly /* True if a read-only (mmap) page is ok */
){
int rc;
assert( sqlite3_mutex_held(pBt->mutex) );
@@ -1609,7 +1645,7 @@ static int getAndInitPage(
if( pgno>btreePagecount(pBt) ){
rc = SQLITE_CORRUPT_BKPT;
}else{
- rc = btreeGetPage(pBt, pgno, ppPage, 0);
+ rc = btreeGetPage(pBt, pgno, ppPage, 0, bReadonly);
if( rc==SQLITE_OK ){
rc = btreeInitPage(*ppPage);
if( rc!=SQLITE_OK ){
@@ -1840,6 +1876,7 @@ int sqlite3BtreeOpen(
rc = sqlite3PagerOpen(pVfs, &pBt->pPager, zFilename,
EXTRA_SIZE, flags, vfsFlags, pageReinit);
if( rc==SQLITE_OK ){
+ sqlite3PagerSetMmapLimit(pBt->pPager, db->szMmap);
rc = sqlite3PagerReadFileheader(pBt->pPager,sizeof(zDbHeader),zDbHeader);
}
if( rc!=SQLITE_OK ){
@@ -2107,6 +2144,19 @@ int sqlite3BtreeSetCacheSize(Btree *p, int mxPage){
}
/*
+** Change the limit on the amount of the database file that may be
+** memory mapped.
+*/
+int sqlite3BtreeSetMmapLimit(Btree *p, sqlite3_int64 szMmap){
+ BtShared *pBt = p->pBt;
+ assert( sqlite3_mutex_held(p->db->mutex) );
+ sqlite3BtreeEnter(p);
+ sqlite3PagerSetMmapLimit(pBt->pPager, szMmap);
+ sqlite3BtreeLeave(p);
+ return SQLITE_OK;
+}
+
+/*
** Change the way data is synced to disk in order to increase or decrease
** how well the database resists damage due to OS crashes and power
** failures. Level 1 is the same as asynchronous (no syncs() occur and
@@ -2200,6 +2250,24 @@ int sqlite3BtreeGetPageSize(Btree *p){
return p->pBt->pageSize;
}
+#if defined(SQLITE_HAS_CODEC) || defined(SQLITE_DEBUG)
+/*
+** This function is similar to sqlite3BtreeGetReserve(), except that it
+** may only be called if it is guaranteed that the b-tree mutex is already
+** held.
+**
+** This is useful in one special case in the backup API code where it is
+** known that the shared b-tree mutex is held, but the mutex on the
+** database handle that owns *p is not. In this case if sqlite3BtreeEnter()
+** were to be called, it might collide with some other operation on the
+** database handle that owns *p, causing undefined behavior.
+*/
+int sqlite3BtreeGetReserveNoMutex(Btree *p){
+ assert( sqlite3_mutex_held(p->pBt->mutex) );
+ return p->pBt->pageSize - p->pBt->usableSize;
+}
+#endif /* SQLITE_HAS_CODEC || SQLITE_DEBUG */
+
#if !defined(SQLITE_OMIT_PAGER_PRAGMAS) || !defined(SQLITE_OMIT_VACUUM)
/*
** Return the number of bytes of space at the end of every page that
@@ -2313,7 +2381,7 @@ static int lockBtree(BtShared *pBt){
assert( pBt->pPage1==0 );
rc = sqlite3PagerSharedLock(pBt->pPager);
if( rc!=SQLITE_OK ) return rc;
- rc = btreeGetPage(pBt, 1, &pPage1, 0);
+ rc = btreeGetPage(pBt, 1, &pPage1, 0, 0);
if( rc!=SQLITE_OK ) return rc;
/* Do some checking to help insure the file we opened really is
@@ -2449,6 +2517,29 @@ page1_init_failed:
return rc;
}
+#ifndef NDEBUG
+/*
+** Return the number of cursors open on pBt. This is for use
+** in assert() expressions, so it is only compiled if NDEBUG is not
+** defined.
+**
+** Only write cursors are counted if wrOnly is true. If wrOnly is
+** false then all cursors are counted.
+**
+** For the purposes of this routine, a cursor is any cursor that
+** is capable of reading or writing to the databse. Cursors that
+** have been tripped into the CURSOR_FAULT state are not counted.
+*/
+static int countValidCursors(BtShared *pBt, int wrOnly){
+ BtCursor *pCur;
+ int r = 0;
+ for(pCur=pBt->pCursor; pCur; pCur=pCur->pNext){
+ if( (wrOnly==0 || pCur->wrFlag) && pCur->eState!=CURSOR_FAULT ) r++;
+ }
+ return r;
+}
+#endif
+
/*
** If there are no outstanding cursors and we are not in the middle
** of a transaction but there is a read lock on the database, then
@@ -2459,7 +2550,7 @@ page1_init_failed:
*/
static void unlockBtreeIfUnused(BtShared *pBt){
assert( sqlite3_mutex_held(pBt->mutex) );
- assert( pBt->pCursor==0 || pBt->inTransaction>TRANS_NONE );
+ assert( countValidCursors(pBt,0)==0 || pBt->inTransaction>TRANS_NONE );
if( pBt->inTransaction==TRANS_NONE && pBt->pPage1!=0 ){
assert( pBt->pPage1->aData );
assert( sqlite3PagerRefcount(pBt->pPager)==1 );
@@ -2514,6 +2605,20 @@ static int newDatabase(BtShared *pBt){
}
/*
+** Initialize the first page of the database file (creating a database
+** consisting of a single page and no schema objects). Return SQLITE_OK
+** if successful, or an SQLite error code otherwise.
+*/
+int sqlite3BtreeNewDb(Btree *p){
+ int rc;
+ sqlite3BtreeEnter(p);
+ p->pBt->nPage = 0;
+ rc = newDatabase(p->pBt);
+ sqlite3BtreeLeave(p);
+ return rc;
+}
+
+/*
** Attempt to start a new transaction. A write-transaction
** is started if the second argument is nonzero, otherwise a read-
** transaction. If the second argument is 2 or more and exclusive
@@ -2563,6 +2668,7 @@ int sqlite3BtreeBeginTrans(Btree *p, int wrflag){
if( p->inTrans==TRANS_WRITE || (p->inTrans==TRANS_READ && !wrflag) ){
goto trans_begun;
}
+ assert( IfNotOmitAV(pBt->bDoTruncate)==0 );
/* Write transactions are not possible on a read-only database */
if( (pBt->btsFlags & BTS_READ_ONLY)!=0 && wrflag ){
@@ -2857,7 +2963,7 @@ static int relocatePage(
** iPtrPage.
*/
if( eType!=PTRMAP_ROOTPAGE ){
- rc = btreeGetPage(pBt, iPtrPage, &pPtrPage, 0);
+ rc = btreeGetPage(pBt, iPtrPage, &pPtrPage, 0, 0);
if( rc!=SQLITE_OK ){
return rc;
}
@@ -2879,24 +2985,23 @@ static int relocatePage(
static int allocateBtreePage(BtShared *, MemPage **, Pgno *, Pgno, u8);
/*
-** Perform a single step of an incremental-vacuum. If successful,
-** return SQLITE_OK. If there is no work to do (and therefore no
-** point in calling this function again), return SQLITE_DONE.
+** Perform a single step of an incremental-vacuum. If successful, return
+** SQLITE_OK. If there is no work to do (and therefore no point in
+** calling this function again), return SQLITE_DONE. Or, if an error
+** occurs, return some other error code.
**
-** More specificly, this function attempts to re-organize the
-** database so that the last page of the file currently in use
-** is no longer in use.
+** More specificly, this function attempts to re-organize the database so
+** that the last page of the file currently in use is no longer in use.
**
-** If the nFin parameter is non-zero, this function assumes
-** that the caller will keep calling incrVacuumStep() until
-** it returns SQLITE_DONE or an error, and that nFin is the
-** number of pages the database file will contain after this
-** process is complete. If nFin is zero, it is assumed that
-** incrVacuumStep() will be called a finite amount of times
-** which may or may not empty the freelist. A full autovacuum
-** has nFin>0. A "PRAGMA incremental_vacuum" has nFin==0.
+** Parameter nFin is the number of pages that this database would contain
+** were this function called until it returns SQLITE_DONE.
+**
+** If the bCommit parameter is non-zero, this function assumes that the
+** caller will keep calling incrVacuumStep() until it returns SQLITE_DONE
+** or an error. bCommit is passed true for an auto-vacuum-on-commmit
+** operation, or false for an incremental vacuum.
*/
-static int incrVacuumStep(BtShared *pBt, Pgno nFin, Pgno iLastPg){
+static int incrVacuumStep(BtShared *pBt, Pgno nFin, Pgno iLastPg, int bCommit){
Pgno nFreeList; /* Number of pages still on the free-list */
int rc;
@@ -2921,15 +3026,15 @@ static int incrVacuumStep(BtShared *pBt, Pgno nFin, Pgno iLastPg){
}
if( eType==PTRMAP_FREEPAGE ){
- if( nFin==0 ){
+ if( bCommit==0 ){
/* Remove the page from the files free-list. This is not required
- ** if nFin is non-zero. In that case, the free-list will be
+ ** if bCommit is non-zero. In that case, the free-list will be
** truncated to zero after this function returns, so it doesn't
** matter if it still contains some garbage entries.
*/
Pgno iFreePg;
MemPage *pFreePg;
- rc = allocateBtreePage(pBt, &pFreePg, &iFreePg, iLastPg, 1);
+ rc = allocateBtreePage(pBt, &pFreePg, &iFreePg, iLastPg, BTALLOC_EXACT);
if( rc!=SQLITE_OK ){
return rc;
}
@@ -2939,34 +3044,37 @@ static int incrVacuumStep(BtShared *pBt, Pgno nFin, Pgno iLastPg){
} else {
Pgno iFreePg; /* Index of free page to move pLastPg to */
MemPage *pLastPg;
+ u8 eMode = BTALLOC_ANY; /* Mode parameter for allocateBtreePage() */
+ Pgno iNear = 0; /* nearby parameter for allocateBtreePage() */
- rc = btreeGetPage(pBt, iLastPg, &pLastPg, 0);
+ rc = btreeGetPage(pBt, iLastPg, &pLastPg, 0, 0);
if( rc!=SQLITE_OK ){
return rc;
}
- /* If nFin is zero, this loop runs exactly once and page pLastPg
+ /* If bCommit is zero, this loop runs exactly once and page pLastPg
** is swapped with the first free page pulled off the free list.
**
- ** On the other hand, if nFin is greater than zero, then keep
+ ** On the other hand, if bCommit is greater than zero, then keep
** looping until a free-page located within the first nFin pages
** of the file is found.
*/
+ if( bCommit==0 ){
+ eMode = BTALLOC_LE;
+ iNear = nFin;
+ }
do {
MemPage *pFreePg;
- rc = allocateBtreePage(pBt, &pFreePg, &iFreePg, 0, 0);
+ rc = allocateBtreePage(pBt, &pFreePg, &iFreePg, iNear, eMode);
if( rc!=SQLITE_OK ){
releasePage(pLastPg);
return rc;
}
releasePage(pFreePg);
- }while( nFin!=0 && iFreePg>nFin );
+ }while( bCommit && iFreePg>nFin );
assert( iFreePg<iLastPg );
- rc = sqlite3PagerWrite(pLastPg->pDbPage);
- if( rc==SQLITE_OK ){
- rc = relocatePage(pBt, pLastPg, eType, iPtrPage, iFreePg, nFin!=0);
- }
+ rc = relocatePage(pBt, pLastPg, eType, iPtrPage, iFreePg, bCommit);
releasePage(pLastPg);
if( rc!=SQLITE_OK ){
return rc;
@@ -2974,30 +3082,40 @@ static int incrVacuumStep(BtShared *pBt, Pgno nFin, Pgno iLastPg){
}
}
- if( nFin==0 ){
- iLastPg--;
- while( iLastPg==PENDING_BYTE_PAGE(pBt)||PTRMAP_ISPAGE(pBt, iLastPg) ){
- if( PTRMAP_ISPAGE(pBt, iLastPg) ){
- MemPage *pPg;
- rc = btreeGetPage(pBt, iLastPg, &pPg, 0);
- if( rc!=SQLITE_OK ){
- return rc;
- }
- rc = sqlite3PagerWrite(pPg->pDbPage);
- releasePage(pPg);
- if( rc!=SQLITE_OK ){
- return rc;
- }
- }
+ if( bCommit==0 ){
+ do {
iLastPg--;
- }
- sqlite3PagerTruncateImage(pBt->pPager, iLastPg);
+ }while( iLastPg==PENDING_BYTE_PAGE(pBt) || PTRMAP_ISPAGE(pBt, iLastPg) );
+ pBt->bDoTruncate = 1;
pBt->nPage = iLastPg;
}
return SQLITE_OK;
}
/*
+** The database opened by the first argument is an auto-vacuum database
+** nOrig pages in size containing nFree free pages. Return the expected
+** size of the database in pages following an auto-vacuum operation.
+*/
+static Pgno finalDbSize(BtShared *pBt, Pgno nOrig, Pgno nFree){
+ int nEntry; /* Number of entries on one ptrmap page */
+ Pgno nPtrmap; /* Number of PtrMap pages to be freed */
+ Pgno nFin; /* Return value */
+
+ nEntry = pBt->usableSize/5;
+ nPtrmap = (nFree-nOrig+PTRMAP_PAGENO(pBt, nOrig)+nEntry)/nEntry;
+ nFin = nOrig - nFree - nPtrmap;
+ if( nOrig>PENDING_BYTE_PAGE(pBt) && nFin<PENDING_BYTE_PAGE(pBt) ){
+ nFin--;
+ }
+ while( PTRMAP_ISPAGE(pBt, nFin) || nFin==PENDING_BYTE_PAGE(pBt) ){
+ nFin--;
+ }
+
+ return nFin;
+}
+
+/*
** A write-transaction must be opened before calling this function.
** It performs a single unit of work towards an incremental vacuum.
**
@@ -3014,11 +3132,24 @@ int sqlite3BtreeIncrVacuum(Btree *p){
if( !pBt->autoVacuum ){
rc = SQLITE_DONE;
}else{
- invalidateAllOverflowCache(pBt);
- rc = incrVacuumStep(pBt, 0, btreePagecount(pBt));
- if( rc==SQLITE_OK ){
- rc = sqlite3PagerWrite(pBt->pPage1->pDbPage);
- put4byte(&pBt->pPage1->aData[28], pBt->nPage);
+ Pgno nOrig = btreePagecount(pBt);
+ Pgno nFree = get4byte(&pBt->pPage1->aData[36]);
+ Pgno nFin = finalDbSize(pBt, nOrig, nFree);
+
+ if( nOrig<nFin ){
+ rc = SQLITE_CORRUPT_BKPT;
+ }else if( nFree>0 ){
+ rc = saveAllCursors(pBt, 0, 0);
+ if( rc==SQLITE_OK ){
+ invalidateAllOverflowCache(pBt);
+ rc = incrVacuumStep(pBt, nFin, nOrig, 0);
+ }
+ if( rc==SQLITE_OK ){
+ rc = sqlite3PagerWrite(pBt->pPage1->pDbPage);
+ put4byte(&pBt->pPage1->aData[28], pBt->nPage);
+ }
+ }else{
+ rc = SQLITE_DONE;
}
}
sqlite3BtreeLeave(p);
@@ -3045,9 +3176,7 @@ static int autoVacuumCommit(BtShared *pBt){
if( !pBt->incrVacuum ){
Pgno nFin; /* Number of pages in database after autovacuuming */
Pgno nFree; /* Number of pages on the freelist initially */
- Pgno nPtrmap; /* Number of PtrMap pages to be freed */
Pgno iFree; /* The next page to be freed */
- int nEntry; /* Number of entries on one ptrmap page */
Pgno nOrig; /* Database size before freeing */
nOrig = btreePagecount(pBt);
@@ -3060,26 +3189,20 @@ static int autoVacuumCommit(BtShared *pBt){
}
nFree = get4byte(&pBt->pPage1->aData[36]);
- nEntry = pBt->usableSize/5;
- nPtrmap = (nFree-nOrig+PTRMAP_PAGENO(pBt, nOrig)+nEntry)/nEntry;
- nFin = nOrig - nFree - nPtrmap;
- if( nOrig>PENDING_BYTE_PAGE(pBt) && nFin<PENDING_BYTE_PAGE(pBt) ){
- nFin--;
- }
- while( PTRMAP_ISPAGE(pBt, nFin) || nFin==PENDING_BYTE_PAGE(pBt) ){
- nFin--;
- }
+ nFin = finalDbSize(pBt, nOrig, nFree);
if( nFin>nOrig ) return SQLITE_CORRUPT_BKPT;
-
+ if( nFin<nOrig ){
+ rc = saveAllCursors(pBt, 0, 0);
+ }
for(iFree=nOrig; iFree>nFin && rc==SQLITE_OK; iFree--){
- rc = incrVacuumStep(pBt, nFin, iFree);
+ rc = incrVacuumStep(pBt, nFin, iFree, 1);
}
if( (rc==SQLITE_DONE || rc==SQLITE_OK) && nFree>0 ){
rc = sqlite3PagerWrite(pBt->pPage1->pDbPage);
put4byte(&pBt->pPage1->aData[32], 0);
put4byte(&pBt->pPage1->aData[36], 0);
put4byte(&pBt->pPage1->aData[28], nFin);
- sqlite3PagerTruncateImage(pBt->pPager, nFin);
+ pBt->bDoTruncate = 1;
pBt->nPage = nFin;
}
if( rc!=SQLITE_OK ){
@@ -3087,7 +3210,7 @@ static int autoVacuumCommit(BtShared *pBt){
}
}
- assert( nRef==sqlite3PagerRefcount(pPager) );
+ assert( nRef>=sqlite3PagerRefcount(pPager) );
return rc;
}
@@ -3134,6 +3257,9 @@ int sqlite3BtreeCommitPhaseOne(Btree *p, const char *zMaster){
return rc;
}
}
+ if( pBt->bDoTruncate ){
+ sqlite3PagerTruncateImage(pBt->pPager, pBt->nPage);
+ }
#endif
rc = sqlite3PagerCommitPhaseOne(pBt->pPager, zMaster, 0);
sqlite3BtreeLeave(p);
@@ -3149,7 +3275,9 @@ static void btreeEndTransaction(Btree *p){
BtShared *pBt = p->pBt;
assert( sqlite3BtreeHoldsMutex(p) );
- btreeClearHasContent(pBt);
+#ifndef SQLITE_OMIT_AUTOVACUUM
+ pBt->bDoTruncate = 0;
+#endif
if( p->inTrans>TRANS_NONE && p->db->activeVdbeCnt>1 ){
/* If there are other active statements that belong to this database
** handle, downgrade to a read-only transaction. The other statements
@@ -3224,6 +3352,7 @@ int sqlite3BtreeCommitPhaseTwo(Btree *p, int bCleanup){
return rc;
}
pBt->inTransaction = TRANS_READ;
+ btreeClearHasContent(pBt);
}
btreeEndTransaction(p);
@@ -3245,27 +3374,6 @@ int sqlite3BtreeCommit(Btree *p){
return rc;
}
-#ifndef NDEBUG
-/*
-** Return the number of write-cursors open on this handle. This is for use
-** in assert() expressions, so it is only compiled if NDEBUG is not
-** defined.
-**
-** For the purposes of this routine, a write-cursor is any cursor that
-** is capable of writing to the databse. That means the cursor was
-** originally opened for writing and the cursor has not be disabled
-** by having its state changed to CURSOR_FAULT.
-*/
-static int countWriteCursors(BtShared *pBt){
- BtCursor *pCur;
- int r = 0;
- for(pCur=pBt->pCursor; pCur; pCur=pCur->pNext){
- if( pCur->wrFlag && pCur->eState!=CURSOR_FAULT ) r++;
- }
- return r;
-}
-#endif
-
/*
** This routine sets the state to CURSOR_FAULT and the error
** code to errCode for every cursor on BtShared that pBtree
@@ -3337,7 +3445,7 @@ int sqlite3BtreeRollback(Btree *p, int tripCode){
/* The rollback may have destroyed the pPage1->aData value. So
** call btreeGetPage() on page 1 again to make
** sure pPage1->aData is set correctly. */
- if( btreeGetPage(pBt, 1, &pPage1, 0)==SQLITE_OK ){
+ if( btreeGetPage(pBt, 1, &pPage1, 0, 0)==SQLITE_OK ){
int nPage = get4byte(28+(u8*)pPage1->aData);
testcase( nPage==0 );
if( nPage==0 ) sqlite3PagerPagecount(pBt->pPager, &nPage);
@@ -3345,8 +3453,9 @@ int sqlite3BtreeRollback(Btree *p, int tripCode){
pBt->nPage = nPage;
releasePage(pPage1);
}
- assert( countWriteCursors(pBt)==0 );
+ assert( countValidCursors(pBt, 1)==0 );
pBt->inTransaction = TRANS_READ;
+ btreeClearHasContent(pBt);
}
btreeEndTransaction(p);
@@ -3771,7 +3880,7 @@ static int getOverflowPage(
assert( next==0 || rc==SQLITE_DONE );
if( rc==SQLITE_OK ){
- rc = btreeGetPage(pBt, ovfl, &pPage, 0);
+ rc = btreeGetPage(pBt, ovfl, &pPage, 0, (ppPage==0));
assert( rc==SQLITE_OK || pPage==0 );
if( rc==SQLITE_OK ){
next = get4byte(pPage->aData);
@@ -3992,7 +4101,9 @@ static int accessPayload(
{
DbPage *pDbPage;
- rc = sqlite3PagerGet(pBt->pPager, nextPage, &pDbPage);
+ rc = sqlite3PagerAcquire(pBt->pPager, nextPage, &pDbPage,
+ (eOp==0 ? PAGER_ACQUIRE_READONLY : 0)
+ );
if( rc==SQLITE_OK ){
aPayload = sqlite3PagerGetData(pDbPage);
nextPage = get4byte(aPayload);
@@ -4171,10 +4282,11 @@ static int moveToChild(BtCursor *pCur, u32 newPgno){
assert( cursorHoldsMutex(pCur) );
assert( pCur->eState==CURSOR_VALID );
assert( pCur->iPage<BTCURSOR_MAX_DEPTH );
+ assert( pCur->iPage>=0 );
if( pCur->iPage>=(BTCURSOR_MAX_DEPTH-1) ){
return SQLITE_CORRUPT_BKPT;
}
- rc = getAndInitPage(pBt, newPgno, &pNewPage);
+ rc = getAndInitPage(pBt, newPgno, &pNewPage, (pCur->wrFlag==0));
if( rc ) return rc;
pCur->apPage[i+1] = pNewPage;
pCur->aiIdx[i+1] = 0;
@@ -4291,7 +4403,7 @@ static int moveToRoot(BtCursor *pCur){
pCur->eState = CURSOR_INVALID;
return SQLITE_OK;
}else{
- rc = getAndInitPage(pBt, pCur->pgnoRoot, &pCur->apPage[0]);
+ rc = getAndInitPage(pBt, pCur->pgnoRoot, &pCur->apPage[0], pCur->wrFlag==0);
if( rc!=SQLITE_OK ){
pCur->eState = CURSOR_INVALID;
return rc;
@@ -4821,21 +4933,23 @@ int sqlite3BtreePrevious(BtCursor *pCur, int *pRes){
** an error. *ppPage and *pPgno are undefined in the event of an error.
** Do not invoke sqlite3PagerUnref() on *ppPage if an error is returned.
**
-** If the "nearby" parameter is not 0, then a (feeble) effort is made to
+** If the "nearby" parameter is not 0, then an effort is made to
** locate a page close to the page number "nearby". This can be used in an
** attempt to keep related pages close to each other in the database file,
** which in turn can make database access faster.
**
-** If the "exact" parameter is not 0, and the page-number nearby exists
-** anywhere on the free-list, then it is guarenteed to be returned. This
-** is only used by auto-vacuum databases when allocating a new table.
+** If the eMode parameter is BTALLOC_EXACT and the nearby page exists
+** anywhere on the free-list, then it is guaranteed to be returned. If
+** eMode is BTALLOC_LT then the page returned will be less than or equal
+** to nearby if any such page exists. If eMode is BTALLOC_ANY then there
+** are no restrictions on which page is returned.
*/
static int allocateBtreePage(
- BtShared *pBt,
- MemPage **ppPage,
- Pgno *pPgno,
- Pgno nearby,
- u8 exact
+ BtShared *pBt, /* The btree */
+ MemPage **ppPage, /* Store pointer to the allocated page here */
+ Pgno *pPgno, /* Store the page number here */
+ Pgno nearby, /* Search for a page near this one */
+ u8 eMode /* BTALLOC_EXACT, BTALLOC_LT, or BTALLOC_ANY */
){
MemPage *pPage1;
int rc;
@@ -4846,6 +4960,7 @@ static int allocateBtreePage(
Pgno mxPage; /* Total size of the database file */
assert( sqlite3_mutex_held(pBt->mutex) );
+ assert( eMode==BTALLOC_ANY || (nearby>0 && IfNotOmitAV(pBt->autoVacuum)) );
pPage1 = pBt->pPage1;
mxPage = btreePagecount(pBt);
n = get4byte(&pPage1->aData[36]);
@@ -4858,21 +4973,24 @@ static int allocateBtreePage(
Pgno iTrunk;
u8 searchList = 0; /* If the free-list must be searched for 'nearby' */
- /* If the 'exact' parameter was true and a query of the pointer-map
+ /* If eMode==BTALLOC_EXACT and a query of the pointer-map
** shows that the page 'nearby' is somewhere on the free-list, then
** the entire-list will be searched for that page.
*/
#ifndef SQLITE_OMIT_AUTOVACUUM
- if( exact && nearby<=mxPage ){
- u8 eType;
- assert( nearby>0 );
- assert( pBt->autoVacuum );
- rc = ptrmapGet(pBt, nearby, &eType, 0);
- if( rc ) return rc;
- if( eType==PTRMAP_FREEPAGE ){
- searchList = 1;
+ if( eMode==BTALLOC_EXACT ){
+ if( nearby<=mxPage ){
+ u8 eType;
+ assert( nearby>0 );
+ assert( pBt->autoVacuum );
+ rc = ptrmapGet(pBt, nearby, &eType, 0);
+ if( rc ) return rc;
+ if( eType==PTRMAP_FREEPAGE ){
+ searchList = 1;
+ }
}
- *pPgno = nearby;
+ }else if( eMode==BTALLOC_LE ){
+ searchList = 1;
}
#endif
@@ -4885,7 +5003,8 @@ static int allocateBtreePage(
/* The code within this loop is run only once if the 'searchList' variable
** is not true. Otherwise, it runs once for each trunk-page on the
- ** free-list until the page 'nearby' is located.
+ ** free-list until the page 'nearby' is located (eMode==BTALLOC_EXACT)
+ ** or until a page less than 'nearby' is located (eMode==BTALLOC_LT)
*/
do {
pPrevTrunk = pTrunk;
@@ -4898,7 +5017,7 @@ static int allocateBtreePage(
if( iTrunk>mxPage ){
rc = SQLITE_CORRUPT_BKPT;
}else{
- rc = btreeGetPage(pBt, iTrunk, &pTrunk, 0);
+ rc = btreeGetPage(pBt, iTrunk, &pTrunk, 0, 0);
}
if( rc ){
pTrunk = 0;
@@ -4927,11 +5046,13 @@ static int allocateBtreePage(
rc = SQLITE_CORRUPT_BKPT;
goto end_allocate_page;
#ifndef SQLITE_OMIT_AUTOVACUUM
- }else if( searchList && nearby==iTrunk ){
+ }else if( searchList
+ && (nearby==iTrunk || (iTrunk<nearby && eMode==BTALLOC_LE))
+ ){
/* The list is being searched and this trunk page is the page
** to allocate, regardless of whether it has leaves.
*/
- assert( *pPgno==iTrunk );
+ *pPgno = iTrunk;
*ppPage = pTrunk;
searchList = 0;
rc = sqlite3PagerWrite(pTrunk->pDbPage);
@@ -4960,7 +5081,7 @@ static int allocateBtreePage(
goto end_allocate_page;
}
testcase( iNewTrunk==mxPage );
- rc = btreeGetPage(pBt, iNewTrunk, &pNewTrunk, 0);
+ rc = btreeGetPage(pBt, iNewTrunk, &pNewTrunk, 0, 0);
if( rc!=SQLITE_OK ){
goto end_allocate_page;
}
@@ -4994,14 +5115,24 @@ static int allocateBtreePage(
unsigned char *aData = pTrunk->aData;
if( nearby>0 ){
u32 i;
- int dist;
closest = 0;
- dist = sqlite3AbsInt32(get4byte(&aData[8]) - nearby);
- for(i=1; i<k; i++){
- int d2 = sqlite3AbsInt32(get4byte(&aData[8+i*4]) - nearby);
- if( d2<dist ){
- closest = i;
- dist = d2;
+ if( eMode==BTALLOC_LE ){
+ for(i=0; i<k; i++){
+ iPage = get4byte(&aData[8+i*4]);
+ if( iPage<=nearby ){
+ closest = i;
+ break;
+ }
+ }
+ }else{
+ int dist;
+ dist = sqlite3AbsInt32(get4byte(&aData[8]) - nearby);
+ for(i=1; i<k; i++){
+ int d2 = sqlite3AbsInt32(get4byte(&aData[8+i*4]) - nearby);
+ if( d2<dist ){
+ closest = i;
+ dist = d2;
+ }
}
}
}else{
@@ -5015,7 +5146,9 @@ static int allocateBtreePage(
goto end_allocate_page;
}
testcase( iPage==mxPage );
- if( !searchList || iPage==nearby ){
+ if( !searchList
+ || (iPage==nearby || (iPage<nearby && eMode==BTALLOC_LE))
+ ){
int noContent;
*pPgno = iPage;
TRACE(("ALLOCATE: %d was leaf %d of %d on trunk %d"
@@ -5028,7 +5161,7 @@ static int allocateBtreePage(
}
put4byte(&aData[4], k-1);
noContent = !btreeGetHasContent(pBt, *pPgno);
- rc = btreeGetPage(pBt, *pPgno, ppPage, noContent);
+ rc = btreeGetPage(pBt, *pPgno, ppPage, noContent, 0);
if( rc==SQLITE_OK ){
rc = sqlite3PagerWrite((*ppPage)->pDbPage);
if( rc!=SQLITE_OK ){
@@ -5042,8 +5175,26 @@ static int allocateBtreePage(
pPrevTrunk = 0;
}while( searchList );
}else{
- /* There are no pages on the freelist, so create a new page at the
- ** end of the file */
+ /* There are no pages on the freelist, so append a new page to the
+ ** database image.
+ **
+ ** Normally, new pages allocated by this block can be requested from the
+ ** pager layer with the 'no-content' flag set. This prevents the pager
+ ** from trying to read the pages content from disk. However, if the
+ ** current transaction has already run one or more incremental-vacuum
+ ** steps, then the page we are about to allocate may contain content
+ ** that is required in the event of a rollback. In this case, do
+ ** not set the no-content flag. This causes the pager to load and journal
+ ** the current page content before overwriting it.
+ **
+ ** Note that the pager will not actually attempt to load or journal
+ ** content for any page that really does lie past the end of the database
+ ** file on disk. So the effects of disabling the no-content optimization
+ ** here are confined to those pages that lie between the end of the
+ ** database image and the end of the database file.
+ */
+ int bNoContent = (0==IfNotOmitAV(pBt->bDoTruncate));
+
rc = sqlite3PagerWrite(pBt->pPage1->pDbPage);
if( rc ) return rc;
pBt->nPage++;
@@ -5058,7 +5209,7 @@ static int allocateBtreePage(
MemPage *pPg = 0;
TRACE(("ALLOCATE: %d from end of file (pointer-map page)\n", pBt->nPage));
assert( pBt->nPage!=PENDING_BYTE_PAGE(pBt) );
- rc = btreeGetPage(pBt, pBt->nPage, &pPg, 1);
+ rc = btreeGetPage(pBt, pBt->nPage, &pPg, bNoContent, 0);
if( rc==SQLITE_OK ){
rc = sqlite3PagerWrite(pPg->pDbPage);
releasePage(pPg);
@@ -5072,7 +5223,7 @@ static int allocateBtreePage(
*pPgno = pBt->nPage;
assert( *pPgno!=PENDING_BYTE_PAGE(pBt) );
- rc = btreeGetPage(pBt, *pPgno, ppPage, 1);
+ rc = btreeGetPage(pBt, *pPgno, ppPage, bNoContent, 0);
if( rc ) return rc;
rc = sqlite3PagerWrite((*ppPage)->pDbPage);
if( rc!=SQLITE_OK ){
@@ -5140,7 +5291,7 @@ static int freePage2(BtShared *pBt, MemPage *pMemPage, Pgno iPage){
/* If the secure_delete option is enabled, then
** always fully overwrite deleted information with zeros.
*/
- if( (!pPage && ((rc = btreeGetPage(pBt, iPage, &pPage, 0))!=0) )
+ if( (!pPage && ((rc = btreeGetPage(pBt, iPage, &pPage, 0, 0))!=0) )
|| ((rc = sqlite3PagerWrite(pPage->pDbPage))!=0)
){
goto freepage_out;
@@ -5167,7 +5318,7 @@ static int freePage2(BtShared *pBt, MemPage *pMemPage, Pgno iPage){
u32 nLeaf; /* Initial number of leaf cells on trunk page */
iTrunk = get4byte(&pPage1->aData[32]);
- rc = btreeGetPage(pBt, iTrunk, &pTrunk, 0);
+ rc = btreeGetPage(pBt, iTrunk, &pTrunk, 0, 0);
if( rc!=SQLITE_OK ){
goto freepage_out;
}
@@ -5213,7 +5364,7 @@ static int freePage2(BtShared *pBt, MemPage *pMemPage, Pgno iPage){
** first trunk in the free-list is full. Either way, the page being freed
** will become the new first trunk page in the free-list.
*/
- if( pPage==0 && SQLITE_OK!=(rc = btreeGetPage(pBt, iPage, &pPage, 0)) ){
+ if( pPage==0 && SQLITE_OK!=(rc = btreeGetPage(pBt, iPage, &pPage, 0, 0)) ){
goto freepage_out;
}
rc = sqlite3PagerWrite(pPage->pDbPage);
@@ -5256,7 +5407,7 @@ static int clearCell(MemPage *pPage, unsigned char *pCell){
return SQLITE_OK; /* No overflow pages. Return without doing anything */
}
if( pCell+info.iOverflow+3 > pPage->aData+pPage->maskPage ){
- return SQLITE_CORRUPT; /* Cell extends past end of page */
+ return SQLITE_CORRUPT_BKPT; /* Cell extends past end of page */
}
ovflPgno = get4byte(&pCell[info.iOverflow]);
assert( pBt->usableSize > 4 );
@@ -5400,7 +5551,7 @@ static int fillInCell(
** If this is the first overflow page, then write a partial entry
** to the pointer-map. If we write nothing to this pointer-map slot,
** then the optimistic overflow chain processing in clearCell()
- ** may misinterpret the uninitialised values and delete the
+ ** may misinterpret the uninitialized values and delete the
** wrong pages from the database.
*/
if( pBt->autoVacuum && rc==SQLITE_OK ){
@@ -5712,7 +5863,7 @@ static int balance_quick(MemPage *pParent, MemPage *pPage, u8 *pSpace){
assert( pPage->nOverflow==1 );
/* This error condition is now caught prior to reaching this function */
- if( pPage->nCell<=0 ) return SQLITE_CORRUPT_BKPT;
+ if( pPage->nCell==0 ) return SQLITE_CORRUPT_BKPT;
/* Allocate a new page. This page will become the right-sibling of
** pPage. Make the parent page writable, so that the new divider cell
@@ -6014,7 +6165,7 @@ static int balance_nonroot(
}
pgno = get4byte(pRight);
while( 1 ){
- rc = getAndInitPage(pBt, pgno, &apOld[i]);
+ rc = getAndInitPage(pBt, pgno, &apOld[i], 0);
if( rc ){
memset(apOld, 0, (i+1)*sizeof(MemPage*));
goto balance_cleanup;
@@ -6873,7 +7024,7 @@ int sqlite3BtreeInsert(
insertCell(pPage, idx, newCell, szNew, 0, 0, &rc);
assert( rc!=SQLITE_OK || pPage->nCell>0 || pPage->nOverflow>0 );
- /* If no error has occured and pPage has an overflow cell, call balance()
+ /* If no error has occurred and pPage has an overflow cell, call balance()
** to redistribute the cells within the tree. Since balance() may move
** the cursor, zero the BtCursor.info.nSize and BtCursor.validNKey
** variables.
@@ -7087,7 +7238,7 @@ static int btreeCreateTable(Btree *p, int *piTable, int createTabFlags){
** be moved to the allocated page (unless the allocated page happens
** to reside at pgnoRoot).
*/
- rc = allocateBtreePage(pBt, &pPageMove, &pgnoMove, pgnoRoot, 1);
+ rc = allocateBtreePage(pBt, &pPageMove, &pgnoMove, pgnoRoot, BTALLOC_EXACT);
if( rc!=SQLITE_OK ){
return rc;
}
@@ -7102,10 +7253,17 @@ static int btreeCreateTable(Btree *p, int *piTable, int createTabFlags){
u8 eType = 0;
Pgno iPtrPage = 0;
+ /* Save the positions of any open cursors. This is required in
+ ** case they are holding a reference to an xFetch reference
+ ** corresponding to page pgnoRoot. */
+ rc = saveAllCursors(pBt, 0, 0);
releasePage(pPageMove);
+ if( rc!=SQLITE_OK ){
+ return rc;
+ }
/* Move the page currently at pgnoRoot to pgnoMove. */
- rc = btreeGetPage(pBt, pgnoRoot, &pRoot, 0);
+ rc = btreeGetPage(pBt, pgnoRoot, &pRoot, 0, 0);
if( rc!=SQLITE_OK ){
return rc;
}
@@ -7126,7 +7284,7 @@ static int btreeCreateTable(Btree *p, int *piTable, int createTabFlags){
if( rc!=SQLITE_OK ){
return rc;
}
- rc = btreeGetPage(pBt, pgnoRoot, &pRoot, 0);
+ rc = btreeGetPage(pBt, pgnoRoot, &pRoot, 0, 0);
if( rc!=SQLITE_OK ){
return rc;
}
@@ -7202,7 +7360,7 @@ static int clearDatabasePage(
return SQLITE_CORRUPT_BKPT;
}
- rc = getAndInitPage(pBt, pgno, &pPage);
+ rc = getAndInitPage(pBt, pgno, &pPage, 0);
if( rc ) return rc;
for(i=0; i<pPage->nCell; i++){
pCell = findCell(pPage, i);
@@ -7304,7 +7462,7 @@ static int btreeDropTable(Btree *p, Pgno iTable, int *piMoved){
return SQLITE_LOCKED_SHAREDCACHE;
}
- rc = btreeGetPage(pBt, (Pgno)iTable, &pPage, 0);
+ rc = btreeGetPage(pBt, (Pgno)iTable, &pPage, 0, 0);
if( rc ) return rc;
rc = sqlite3BtreeClearTable(p, iTable, 0);
if( rc ){
@@ -7339,7 +7497,7 @@ static int btreeDropTable(Btree *p, Pgno iTable, int *piMoved){
*/
MemPage *pMove;
releasePage(pPage);
- rc = btreeGetPage(pBt, maxRootPgno, &pMove, 0);
+ rc = btreeGetPage(pBt, maxRootPgno, &pMove, 0, 0);
if( rc!=SQLITE_OK ){
return rc;
}
@@ -7349,7 +7507,7 @@ static int btreeDropTable(Btree *p, Pgno iTable, int *piMoved){
return rc;
}
pMove = 0;
- rc = btreeGetPage(pBt, maxRootPgno, &pMove, 0);
+ rc = btreeGetPage(pBt, maxRootPgno, &pMove, 0, 0);
freePage(pMove, &rc);
releasePage(pMove);
if( rc!=SQLITE_OK ){
@@ -7761,7 +7919,7 @@ static int checkTreePage(
usableSize = pBt->usableSize;
if( iPage==0 ) return 0;
if( checkRef(pCheck, iPage, zParentContext) ) return 0;
- if( (rc = btreeGetPage(pBt, (Pgno)iPage, &pPage, 0))!=0 ){
+ if( (rc = btreeGetPage(pBt, (Pgno)iPage, &pPage, 0, 0))!=0 ){
checkAppendMsg(pCheck, zContext,
"unable to get the page. error code=%d", rc);
return 0;
@@ -7994,7 +8152,7 @@ char *sqlite3BtreeIntegrityCheck(
}
i = PENDING_BYTE_PAGE(pBt);
if( i<=sCheck.nPage ) setPageReferenced(&sCheck, i);
- sqlite3StrAccumInit(&sCheck.errMsg, zErr, sizeof(zErr), 20000);
+ sqlite3StrAccumInit(&sCheck.errMsg, zErr, sizeof(zErr), SQLITE_MAX_LENGTH);
sCheck.errMsg.useMalloc = 2;
/* Check the integrity of the freelist
@@ -8233,6 +8391,17 @@ int sqlite3BtreePutData(BtCursor *pCsr, u32 offset, u32 amt, void *z){
return SQLITE_ABORT;
}
+ /* Save the positions of all other cursors open on this table. This is
+ ** required in case any of them are holding references to an xFetch
+ ** version of the b-tree page modified by the accessPayload call below.
+ **
+ ** Note that pCsr must be open on a BTREE_INTKEY table and saveCursorPosition()
+ ** and hence saveAllCursors() cannot fail on a BTREE_INTKEY table, hence
+ ** saveAllCursors can only return SQLITE_OK.
+ */
+ VVA_ONLY(rc =) saveAllCursors(pCsr->pBt, pCsr->pgnoRoot, pCsr);
+ assert( rc==SQLITE_OK );
+
/* Check some assumptions:
** (a) the cursor is open for writing,
** (b) there is a read/write transaction open,
generated by cgit v1.2.3 (git 2.25.1) at 2024-05-08 06:27:13 +0000