数据库系统

开发平台：

C/C++

pager.c：源码内容

assert( pPager->state>=PAGER_SHARED || pPager->dbSize<0 || MEMDB );
if( pPager->state>=locktype ){
rc = SQLITE_OK;
}else{
if( pPager->pBusyHandler ) pPager->pBusyHandler->nBusy = 0;
do {
rc = sqlite3OsLock(pPager->fd, locktype);
}while( rc==SQLITE_BUSY && sqlite3InvokeBusyHandler(pPager->pBusyHandler) );
if( rc==SQLITE_OK ){
pPager->state = locktype;
IOTRACE(("LOCK %p %dn", pPager, locktype))
}
}
return rc;
}
/*
** Truncate the file to the number of pages specified.
*/
int sqlite3PagerTruncate(Pager *pPager, Pgno nPage){
int rc;
assert( pPager->state>=PAGER_SHARED || MEMDB );
sqlite3PagerPagecount(pPager);
if( pPager->errCode ){
rc = pPager->errCode;
return rc;
}
if( nPage>=(unsigned)pPager->dbSize ){
return SQLITE_OK;
}
if( MEMDB ){
pPager->dbSize = nPage;
pager_truncate_cache(pPager);
return SQLITE_OK;
}
pagerEnter(pPager);
rc = syncJournal(pPager);
pagerLeave(pPager);
if( rc!=SQLITE_OK ){
return rc;
}
/* Get an exclusive lock on the database before truncating. */
pagerEnter(pPager);
rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);
pagerLeave(pPager);
if( rc!=SQLITE_OK ){
return rc;
}
rc = pager_truncate(pPager, nPage);
return rc;
}
/*
** Shutdown the page cache. Free all memory and close all files.
**
** If a transaction was in progress when this routine is called, that
** transaction is rolled back. All outstanding pages are invalidated
** and their memory is freed. Any attempt to use a page associated
** with this page cache after this function returns will likely
** result in a coredump.
**
** This function always succeeds. If a transaction is active an attempt
** is made to roll it back. If an error occurs during the rollback
** a hot journal may be left in the filesystem but no error is returned
** to the caller.
*/
int sqlite3PagerClose(Pager *pPager){
#ifdef SQLITE_ENABLE_MEMORY_MANAGEMENT
if( !MEMDB ){
sqlite3_mutex *mutex = sqlite3_mutex_alloc(SQLITE_MUTEX_STATIC_MEM2);
sqlite3_mutex_enter(mutex);
if( pPager->pPrev ){
pPager->pPrev->pNext = pPager->pNext;
}else{
sqlite3PagerList = pPager->pNext;
}
if( pPager->pNext ){
pPager->pNext->pPrev = pPager->pPrev;
}
sqlite3_mutex_leave(mutex);
}
#endif
disable_simulated_io_errors();
pPager->errCode = 0;
pPager->exclusiveMode = 0;
pager_reset(pPager);
pagerUnlockAndRollback(pPager);
enable_simulated_io_errors();
PAGERTRACE2("CLOSE %dn", PAGERID(pPager));
IOTRACE(("CLOSE %pn", pPager))
assert( pPager->errCode || (pPager->journalOpen==0 && pPager->stmtOpen==0) );
if( pPager->journalOpen ){
sqlite3OsClose(pPager->jfd);
}
sqlite3BitvecDestroy(pPager->pInJournal);
if( pPager->stmtOpen ){
sqlite3OsClose(pPager->stfd);
}
sqlite3OsClose(pPager->fd);
/* Temp files are automatically deleted by the OS
** if( pPager->tempFile ){
** sqlite3OsDelete(pPager->zFilename);
** }
*/
sqlite3_free(pPager->aHash);
sqlite3_free(pPager->pTmpSpace);
sqlite3_free(pPager);
return SQLITE_OK;
}
#if !defined(NDEBUG) || defined(SQLITE_TEST)
/*
** Return the page number for the given page data.
*/
Pgno sqlite3PagerPagenumber(DbPage *p){
return p->pgno;
}
#endif
/*
** The page_ref() function increments the reference count for a page.
** If the page is currently on the freelist (the reference count is zero) then
** remove it from the freelist.
**
** For non-test systems, page_ref() is a macro that calls _page_ref()
** online of the reference count is zero. For test systems, page_ref()
** is a real function so that we can set breakpoints and trace it.
*/
static void _page_ref(PgHdr *pPg){
if( pPg->nRef==0 ){
/* The page is currently on the freelist. Remove it. */
lruListRemove(pPg);
pPg->pPager->nRef++;
}
pPg->nRef++;
}
#ifdef SQLITE_DEBUG
static void page_ref(PgHdr *pPg){
if( pPg->nRef==0 ){
_page_ref(pPg);
}else{
pPg->nRef++;
}
}
#else
# define page_ref(P) ((P)->nRef==0?_page_ref(P):(void)(P)->nRef++)
#endif
/*
** Increment the reference count for a page. The input pointer is
** a reference to the page data.
*/
int sqlite3PagerRef(DbPage *pPg){
pagerEnter(pPg->pPager);
page_ref(pPg);
pagerLeave(pPg->pPager);
return SQLITE_OK;
}
/*
** Sync the journal. In other words, make sure all the pages that have
** been written to the journal have actually reached the surface of the
** disk. It is not safe to modify the original database file until after
** the journal has been synced. If the original database is modified before
** the journal is synced and a power failure occurs, the unsynced journal
** data would be lost and we would be unable to completely rollback the
** database changes. Database corruption would occur.
**
** This routine also updates the nRec field in the header of the journal.
** (See comments on the pager_playback() routine for additional information.)
** If the sync mode is FULL, two syncs will occur. First the whole journal
** is synced, then the nRec field is updated, then a second sync occurs.
**
** For temporary databases, we do not care if we are able to rollback
** after a power failure, so no sync occurs.
**
** If the IOCAP_SEQUENTIAL flag is set for the persistent media on which
** the database is stored, then OsSync() is never called on the journal
** file. In this case all that is required is to update the nRec field in
** the journal header.
**
** This routine clears the needSync field of every page current held in
** memory.
*/
static int syncJournal(Pager *pPager){
PgHdr *pPg;
int rc = SQLITE_OK;
/* Sync the journal before modifying the main database
** (assuming there is a journal and it needs to be synced.)
*/
if( pPager->needSync ){
if( !pPager->tempFile ){
int iDc = sqlite3OsDeviceCharacteristics(pPager->fd);
assert( pPager->journalOpen );
/* assert( !pPager->noSync ); // noSync might be set if synchronous
** was turned off after the transaction was started. Ticket #615 */
#ifndef NDEBUG
{
/* Make sure the pPager->nRec counter we are keeping agrees
** with the nRec computed from the size of the journal file.
*/
i64 jSz;
rc = sqlite3OsFileSize(pPager->jfd, &jSz);
if( rc!=0 ) return rc;
assert( pPager->journalOff==jSz );
}
#endif
if( 0==(iDc&SQLITE_IOCAP_SAFE_APPEND) ){
/* Write the nRec value into the journal file header. If in
** full-synchronous mode, sync the journal first. This ensures that
** all data has really hit the disk before nRec is updated to mark
** it as a candidate for rollback.
**
** This is not required if the persistent media supports the
** SAFE_APPEND property. Because in this case it is not possible
** for garbage data to be appended to the file, the nRec field
** is populated with 0xFFFFFFFF when the journal header is written
** and never needs to be updated.
*/
i64 jrnlOff;
if( pPager->fullSync && 0==(iDc&SQLITE_IOCAP_SEQUENTIAL) ){
PAGERTRACE2("SYNC journal of %dn", PAGERID(pPager));
IOTRACE(("JSYNC %pn", pPager))
rc = sqlite3OsSync(pPager->jfd, pPager->sync_flags);
if( rc!=0 ) return rc;
}
jrnlOff = pPager->journalHdr + sizeof(aJournalMagic);
IOTRACE(("JHDR %p %lld %dn", pPager, jrnlOff, 4));
rc = write32bits(pPager->jfd, jrnlOff, pPager->nRec);
if( rc ) return rc;
}
if( 0==(iDc&SQLITE_IOCAP_SEQUENTIAL) ){
PAGERTRACE2("SYNC journal of %dn", PAGERID(pPager));
IOTRACE(("JSYNC %pn", pPager))
rc = sqlite3OsSync(pPager->jfd, pPager->sync_flags|
(pPager->sync_flags==SQLITE_SYNC_FULL?SQLITE_SYNC_DATAONLY:0)
);
if( rc!=0 ) return rc;
}
pPager->journalStarted = 1;
}
pPager->needSync = 0;
/* Erase the needSync flag from every page.
*/
for(pPg=pPager->pAll; pPg; pPg=pPg->pNextAll){
pPg->needSync = 0;
}
lruListSetFirstSynced(pPager);
}
#ifndef NDEBUG
/* If the Pager.needSync flag is clear then the PgHdr.needSync
** flag must also be clear for all pages. Verify that this
** invariant is true.
*/
else{
for(pPg=pPager->pAll; pPg; pPg=pPg->pNextAll){
assert( pPg->needSync==0 );
}
assert( pPager->lru.pFirstSynced==pPager->lru.pFirst );
}
#endif
return rc;
}
/*
** Merge two lists of pages connected by pDirty and in pgno order.
** Do not both fixing the pPrevDirty pointers.
*/
static PgHdr *merge_pagelist(PgHdr *pA, PgHdr *pB){
PgHdr result, *pTail;
pTail = &result;
while( pA && pB ){
if( pA->pgno<pB->pgno ){
pTail->pDirty = pA;
pTail = pA;
pA = pA->pDirty;
}else{
pTail->pDirty = pB;
pTail = pB;
pB = pB->pDirty;
}
}
if( pA ){
pTail->pDirty = pA;
}else if( pB ){
pTail->pDirty = pB;
}else{
pTail->pDirty = 0;
}
return result.pDirty;
}
/*
** Sort the list of pages in accending order by pgno. Pages are
** connected by pDirty pointers. The pPrevDirty pointers are
** corrupted by this sort.
*/
#define N_SORT_BUCKET_ALLOC 25
#define N_SORT_BUCKET 25
#ifdef SQLITE_TEST
int sqlite3_pager_n_sort_bucket = 0;
#undef N_SORT_BUCKET
#define N_SORT_BUCKET
(sqlite3_pager_n_sort_bucket?sqlite3_pager_n_sort_bucket:N_SORT_BUCKET_ALLOC)
#endif
static PgHdr *sort_pagelist(PgHdr *pIn){
PgHdr *a[N_SORT_BUCKET_ALLOC], *p;
int i;
memset(a, 0, sizeof(a));
while( pIn ){
p = pIn;
pIn = p->pDirty;
p->pDirty = 0;
for(i=0; i<N_SORT_BUCKET-1; i++){
if( a[i]==0 ){
a[i] = p;
break;
}else{
p = merge_pagelist(a[i], p);
a[i] = 0;
}
}
if( i==N_SORT_BUCKET-1 ){
/* Coverage: To get here, there need to be 2^(N_SORT_BUCKET)
** elements in the input list. This is possible, but impractical.
** Testing this line is the point of global variable
** sqlite3_pager_n_sort_bucket.
*/
a[i] = merge_pagelist(a[i], p);
}
}
p = a[0];
for(i=1; i<N_SORT_BUCKET; i++){
p = merge_pagelist(p, a[i]);
}
return p;
}
/*
** Given a list of pages (connected by the PgHdr.pDirty pointer) write
** every one of those pages out to the database file and mark them all
** as clean.
*/
static int pager_write_pagelist(PgHdr *pList){
Pager *pPager;
PgHdr *p;
int rc;
if( pList==0 ) return SQLITE_OK;
pPager = pList->pPager;
/* At this point there may be either a RESERVED or EXCLUSIVE lock on the
** database file. If there is already an EXCLUSIVE lock, the following
** calls to sqlite3OsLock() are no-ops.
**
** Moving the lock from RESERVED to EXCLUSIVE actually involves going
** through an intermediate state PENDING. A PENDING lock prevents new
** readers from attaching to the database but is unsufficient for us to
** write. The idea of a PENDING lock is to prevent new readers from
** coming in while we wait for existing readers to clear.
**
** While the pager is in the RESERVED state, the original database file
** is unchanged and we can rollback without having to playback the
** journal into the original database file. Once we transition to
** EXCLUSIVE, it means the database file has been changed and any rollback
** will require a journal playback.
*/
rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);
if( rc!=SQLITE_OK ){
return rc;
}
pList = sort_pagelist(pList);
for(p=pList; p; p=p->pDirty){
assert( p->dirty );
p->dirty = 0;
}
while( pList ){
/* If the file has not yet been opened, open it now. */
if( !pPager->fd->pMethods ){
assert(pPager->tempFile);
rc = sqlite3PagerOpentemp(pPager->pVfs, pPager->fd, pPager->zFilename,
pPager->vfsFlags);
if( rc ) return rc;
}
/* If there are dirty pages in the page cache with page numbers greater
** than Pager.dbSize, this means sqlite3PagerTruncate() was called to
** make the file smaller (presumably by auto-vacuum code). Do not write
** any such pages to the file.
*/
if( pList->pgno<=pPager->dbSize ){
i64 offset = (pList->pgno-1)*(i64)pPager->pageSize;
char *pData = CODEC2(pPager, PGHDR_TO_DATA(pList), pList->pgno, 6);
PAGERTRACE4("STORE %d page %d hash(%08x)n",
PAGERID(pPager), pList->pgno, pager_pagehash(pList));
IOTRACE(("PGOUT %p %dn", pPager, pList->pgno));
rc = sqlite3OsWrite(pPager->fd, pData, pPager->pageSize, offset);
PAGER_INCR(sqlite3_pager_writedb_count);
PAGER_INCR(pPager->nWrite);
if( pList->pgno==1 ){
memcpy(&pPager->dbFileVers, &pData[24], sizeof(pPager->dbFileVers));
}
}
#ifndef NDEBUG
else{
PAGERTRACE3("NOSTORE %d page %dn", PAGERID(pPager), pList->pgno);
}
#endif
if( rc ) return rc;
#ifdef SQLITE_CHECK_PAGES
pList->pageHash = pager_pagehash(pList);
#endif
pList = pList->pDirty;
}
return SQLITE_OK;
}
/*
** Collect every dirty page into a dirty list and
** return a pointer to the head of that list. All pages are
** collected even if they are still in use.
*/
static PgHdr *pager_get_all_dirty_pages(Pager *pPager){
#ifndef NDEBUG
/* Verify the sanity of the dirty list when we are running
** in debugging mode. This is expensive, so do not
** do this on a normal build. */
int n1 = 0;
int n2 = 0;
PgHdr *p;
for(p=pPager->pAll; p; p=p->pNextAll){ if( p->dirty ) n1++; }
for(p=pPager->pDirty; p; p=p->pDirty){ n2++; }
assert( n1==n2 );
#endif
return pPager->pDirty;
}
/*
** Return 1 if there is a hot journal on the given pager.
** A hot journal is one that needs to be played back.
**
** If the current size of the database file is 0 but a journal file
** exists, that is probably an old journal left over from a prior
** database with the same name. Just delete the journal.
**
** Return negative if unable to determine the status of the journal.
*/
static int hasHotJournal(Pager *pPager){
sqlite3_vfs *pVfs = pPager->pVfs;
int rc;
if( !pPager->useJournal ) return 0;
if( !pPager->fd->pMethods ) return 0;
rc = sqlite3OsAccess(pVfs, pPager->zJournal, SQLITE_ACCESS_EXISTS);
if( rc<=0 ){
return rc;
}
if( sqlite3OsCheckReservedLock(pPager->fd) ){
return 0;
}
if( sqlite3PagerPagecount(pPager)==0 ){
sqlite3OsDelete(pVfs, pPager->zJournal, 0);
return 0;
}else{
return 1;
}
}
/*
** Try to find a page in the cache that can be recycled.
**
** This routine may return SQLITE_IOERR, SQLITE_FULL or SQLITE_OK. It
** does not set the pPager->errCode variable.
*/
static int pager_recycle(Pager *pPager, PgHdr **ppPg){
PgHdr *pPg;
*ppPg = 0;
/* It is illegal to call this function unless the pager object
** pointed to by pPager has at least one free page (page with nRef==0).
*/
assert(!MEMDB);
assert(pPager->lru.pFirst);
/* Find a page to recycle. Try to locate a page that does not
** require us to do an fsync() on the journal.
*/
pPg = pPager->lru.pFirstSynced;
/* If we could not find a page that does not require an fsync()
** on the journal file then fsync the journal file. This is a
** very slow operation, so we work hard to avoid it. But sometimes
** it can't be helped.
*/
if( pPg==0 && pPager->lru.pFirst){
int iDc = sqlite3OsDeviceCharacteristics(pPager->fd);
int rc = syncJournal(pPager);
if( rc!=0 ){
return rc;
}
if( pPager->fullSync && 0==(iDc&SQLITE_IOCAP_SAFE_APPEND) ){
/* If in full-sync mode, write a new journal header into the
** journal file. This is done to avoid ever modifying a journal
** header that is involved in the rollback of pages that have
** already been written to the database (in case the header is
** trashed when the nRec field is updated).
*/
pPager->nRec = 0;
assert( pPager->journalOff > 0 );
assert( pPager->doNotSync==0 );
rc = writeJournalHdr(pPager);
if( rc!=0 ){
return rc;
}
}
pPg = pPager->lru.pFirst;
}
assert( pPg->nRef==0 );
/* Write the page to the database file if it is dirty.
*/
if( pPg->dirty ){
int rc;
assert( pPg->needSync==0 );
makeClean(pPg);
pPg->dirty = 1;
pPg->pDirty = 0;
rc = pager_write_pagelist( pPg );
pPg->dirty = 0;
if( rc!=SQLITE_OK ){
return rc;
}
}
assert( pPg->dirty==0 );
/* If the page we are recycling is marked as alwaysRollback, then
** set the global alwaysRollback flag, thus disabling the
** sqlite3PagerDontRollback() optimization for the rest of this transaction.
** It is necessary to do this because the page marked alwaysRollback
** might be reloaded at a later time but at that point we won't remember
** that is was marked alwaysRollback. This means that all pages must
** be marked as alwaysRollback from here on out.
*/
if( pPg->alwaysRollback ){
IOTRACE(("ALWAYS_ROLLBACK %pn", pPager))
pPager->alwaysRollback = 1;
}
/* Unlink the old page from the free list and the hash table
*/
unlinkPage(pPg);
assert( pPg->pgno==0 );
*ppPg = pPg;
return SQLITE_OK;
}
#ifdef SQLITE_ENABLE_MEMORY_MANAGEMENT
/*
** This function is called to free superfluous dynamically allocated memory
** held by the pager system. Memory in use by any SQLite pager allocated
** by the current thread may be sqlite3_free()ed.
**
** nReq is the number of bytes of memory required. Once this much has
** been released, the function returns. The return value is the total number
** of bytes of memory released.
*/
int sqlite3PagerReleaseMemory(int nReq){
int nReleased = 0; /* Bytes of memory released so far */
sqlite3_mutex *mutex; /* The MEM2 mutex */
Pager *pPager; /* For looping over pagers */
BusyHandler *savedBusy; /* Saved copy of the busy handler */
int rc = SQLITE_OK;
/* Acquire the memory-management mutex
*/
mutex = sqlite3_mutex_alloc(SQLITE_MUTEX_STATIC_MEM2);
sqlite3_mutex_enter(mutex);
/* Signal all database connections that memory management wants
** to have access to the pagers.
*/
for(pPager=sqlite3PagerList; pPager; pPager=pPager->pNext){
pPager->iInUseMM = 1;
}
while( rc==SQLITE_OK && (nReq<0 || nReleased<nReq) ){
PgHdr *pPg;
PgHdr *pRecycled;
/* Try to find a page to recycle that does not require a sync(). If
** this is not possible, find one that does require a sync().
*/
sqlite3_mutex_enter(sqlite3_mutex_alloc(SQLITE_MUTEX_STATIC_LRU));
pPg = sqlite3LruPageList.pFirstSynced;
while( pPg && (pPg->needSync || pPg->pPager->iInUseDB) ){
pPg = pPg->gfree.pNext;
}
if( !pPg ){
pPg = sqlite3LruPageList.pFirst;
while( pPg && pPg->pPager->iInUseDB ){
pPg = pPg->gfree.pNext;
}
}
sqlite3_mutex_leave(sqlite3_mutex_alloc(SQLITE_MUTEX_STATIC_LRU));
/* If pPg==0, then the block above has failed to find a page to
** recycle. In this case return early - no further memory will
** be released.
*/
if( !pPg ) break;
pPager = pPg->pPager;
assert(!pPg->needSync || pPg==pPager->lru.pFirst);
assert(pPg->needSync || pPg==pPager->lru.pFirstSynced);
savedBusy = pPager->pBusyHandler;
pPager->pBusyHandler = 0;
rc = pager_recycle(pPager, &pRecycled);
pPager->pBusyHandler = savedBusy;
assert(pRecycled==pPg || rc!=SQLITE_OK);
if( rc==SQLITE_OK ){
/* We've found a page to free. At this point the page has been
** removed from the page hash-table, free-list and synced-list
** (pFirstSynced). It is still in the all pages (pAll) list.
** Remove it from this list before freeing.
**
** Todo: Check the Pager.pStmt list to make sure this is Ok. It
** probably is though.
*/
PgHdr *pTmp;
assert( pPg );
if( pPg==pPager->pAll ){
pPager->pAll = pPg->pNextAll;
}else{
for( pTmp=pPager->pAll; pTmp->pNextAll!=pPg; pTmp=pTmp->pNextAll ){}
pTmp->pNextAll = pPg->pNextAll;
}
nReleased += (
sizeof(*pPg) + pPager->pageSize
+ sizeof(u32) + pPager->nExtra
+ MEMDB*sizeof(PgHistory)
);
IOTRACE(("PGFREE %p %d *n", pPager, pPg->pgno));
PAGER_INCR(sqlite3_pager_pgfree_count);
sqlite3_free(pPg->pData);
sqlite3_free(pPg);
pPager->nPage--;
}else{
/* An error occured whilst writing to the database file or
** journal in pager_recycle(). The error is not returned to the
** caller of this function. Instead, set the Pager.errCode variable.
** The error will be returned to the user (or users, in the case
** of a shared pager cache) of the pager for which the error occured.
*/
assert(
(rc&0xff)==SQLITE_IOERR ||
rc==SQLITE_FULL ||
rc==SQLITE_BUSY
);
assert( pPager->state>=PAGER_RESERVED );
pager_error(pPager, rc);
}
}
/* Clear the memory management flags and release the mutex
*/
for(pPager=sqlite3PagerList; pPager; pPager=pPager->pNext){
pPager->iInUseMM = 0;
}
sqlite3_mutex_leave(mutex);
/* Return the number of bytes released
*/
return nReleased;
}
#endif /* SQLITE_ENABLE_MEMORY_MANAGEMENT */
/*
** Read the content of page pPg out of the database file.
*/
static int readDbPage(Pager *pPager, PgHdr *pPg, Pgno pgno){
int rc;
i64 offset;
assert( MEMDB==0 );
assert(pPager->fd->pMethods||pPager->tempFile);
if( !pPager->fd->pMethods ){
return SQLITE_IOERR_SHORT_READ;
}
offset = (pgno-1)*(i64)pPager->pageSize;
rc = sqlite3OsRead(pPager->fd, PGHDR_TO_DATA(pPg), pPager->pageSize, offset);
PAGER_INCR(sqlite3_pager_readdb_count);
PAGER_INCR(pPager->nRead);
IOTRACE(("PGIN %p %dn", pPager, pgno));
if( pgno==1 ){
memcpy(&pPager->dbFileVers, &((u8*)PGHDR_TO_DATA(pPg))[24],
sizeof(pPager->dbFileVers));
}
CODEC1(pPager, PGHDR_TO_DATA(pPg), pPg->pgno, 3);
PAGERTRACE4("FETCH %d page %d hash(%08x)n",
PAGERID(pPager), pPg->pgno, pager_pagehash(pPg));
return rc;
}
/*
** This function is called to obtain the shared lock required before
** data may be read from the pager cache. If the shared lock has already
** been obtained, this function is a no-op.
**
** Immediately after obtaining the shared lock (if required), this function
** checks for a hot-journal file. If one is found, an emergency rollback
** is performed immediately.
*/
static int pagerSharedLock(Pager *pPager){
int rc = SQLITE_OK;
int isHot = 0;
/* If this database is opened for exclusive access, has no outstanding
** page references and is in an error-state, now is the chance to clear
** the error. Discard the contents of the pager-cache and treat any
** open journal file as a hot-journal.
*/
if( !MEMDB && pPager->exclusiveMode && pPager->nRef==0 && pPager->errCode ){
if( pPager->journalOpen ){
isHot = 1;
}
pager_reset(pPager);
pPager->errCode = SQLITE_OK;
}
/* If the pager is still in an error state, do not proceed. The error
** state will be cleared at some point in the future when all page
** references are dropped and the cache can be discarded.
*/
if( pPager->errCode && pPager->errCode!=SQLITE_FULL ){
return pPager->errCode;
}
if( pPager->state==PAGER_UNLOCK || isHot ){
sqlite3_vfs *pVfs = pPager->pVfs;
if( !MEMDB ){
assert( pPager->nRef==0 );
if( !pPager->noReadlock ){
rc = pager_wait_on_lock(pPager, SHARED_LOCK);
if( rc!=SQLITE_OK ){
return pager_error(pPager, rc);
}
assert( pPager->state>=SHARED_LOCK );
}
/* If a journal file exists, and there is no RESERVED lock on the
** database file, then it either needs to be played back or deleted.
*/
rc = hasHotJournal(pPager);
if( rc<0 ){
return SQLITE_IOERR_NOMEM;
}
if( rc==1 || isHot ){
/* Get an EXCLUSIVE lock on the database file. At this point it is
** important that a RESERVED lock is not obtained on the way to the
** EXCLUSIVE lock. If it were, another process might open the
** database file, detect the RESERVED lock, and conclude that the
** database is safe to read while this process is still rolling it
** back.
**
** Because the intermediate RESERVED lock is not requested, the
** second process will get to this point in the code and fail to
** obtain its own EXCLUSIVE lock on the database file.
*/
if( pPager->state<EXCLUSIVE_LOCK ){
rc = sqlite3OsLock(pPager->fd, EXCLUSIVE_LOCK);
if( rc!=SQLITE_OK ){
pager_unlock(pPager);
return pager_error(pPager, rc);
}
pPager->state = PAGER_EXCLUSIVE;
}
/* Open the journal for reading only. Return SQLITE_BUSY if
** we are unable to open the journal file.
**
** The journal file does not need to be locked itself. The
** journal file is never open unless the main database file holds
** a write lock, so there is never any chance of two or more
** processes opening the journal at the same time.
**
** Open the journal for read/write access. This is because in
** exclusive-access mode the file descriptor will be kept open and
** possibly used for a transaction later on. On some systems, the
** OsTruncate() call used in exclusive-access mode also requires
** a read/write file handle.
*/
if( !isHot ){
int res = sqlite3OsAccess(pVfs,pPager->zJournal,SQLITE_ACCESS_EXISTS);
if( res==1 ){
int fout = 0;
int f = SQLITE_OPEN_READWRITE|SQLITE_OPEN_MAIN_JOURNAL;
assert( !pPager->tempFile );
rc = sqlite3OsOpen(pVfs, pPager->zJournal, pPager->jfd, f, &fout);
assert( rc!=SQLITE_OK || pPager->jfd->pMethods );
if( fout&SQLITE_OPEN_READONLY ){
rc = SQLITE_BUSY;
sqlite3OsClose(pPager->jfd);
}
}else{
rc = (res==0?SQLITE_BUSY:SQLITE_IOERR_NOMEM);
}
}
if( rc!=SQLITE_OK ){
pager_unlock(pPager);
switch( rc ){
case SQLITE_NOMEM:
case SQLITE_IOERR_UNLOCK:
case SQLITE_IOERR_NOMEM:
return rc;
default:
return SQLITE_BUSY;
}
}
pPager->journalOpen = 1;
pPager->journalStarted = 0;
pPager->journalOff = 0;
pPager->setMaster = 0;
pPager->journalHdr = 0;
/* Playback and delete the journal. Drop the database write
** lock and reacquire the read lock.
*/
rc = pager_playback(pPager, 1);
if( rc!=SQLITE_OK ){
return pager_error(pPager, rc);
}
assert(pPager->state==PAGER_SHARED ||
(pPager->exclusiveMode && pPager->state>PAGER_SHARED)
);
}
if( pPager->pAll ){
/* The shared-lock has just been acquired on the database file
** and there are already pages in the cache (from a previous
** read or write transaction). Check to see if the database
** has been modified. If the database has changed, flush the
** cache.
**
** Database changes is detected by looking at 15 bytes beginning
** at offset 24 into the file. The first 4 of these 16 bytes are
** a 32-bit counter that is incremented with each change. The
** other bytes change randomly with each file change when
** a codec is in use.
**
** There is a vanishingly small chance that a change will not be
** detected. The chance of an undetected change is so small that
** it can be neglected.
*/
char dbFileVers[sizeof(pPager->dbFileVers)];
sqlite3PagerPagecount(pPager);
if( pPager->errCode ){
return pPager->errCode;
}
if( pPager->dbSize>0 ){
IOTRACE(("CKVERS %p %dn", pPager, sizeof(dbFileVers)));
rc = sqlite3OsRead(pPager->fd, &dbFileVers, sizeof(dbFileVers), 24);
if( rc!=SQLITE_OK ){
return rc;
}
}else{
memset(dbFileVers, 0, sizeof(dbFileVers));
}
if( memcmp(pPager->dbFileVers, dbFileVers, sizeof(dbFileVers))!=0 ){
pager_reset(pPager);
}
}
}
assert( pPager->exclusiveMode || pPager->state<=PAGER_SHARED );
if( pPager->state==PAGER_UNLOCK ){
pPager->state = PAGER_SHARED;
}
}
return rc;
}
/*
** Allocate a PgHdr object. Either create a new one or reuse
** an existing one that is not otherwise in use.
**
** A new PgHdr structure is created if any of the following are
** true:
**
** (1) We have not exceeded our maximum allocated cache size
** as set by the "PRAGMA cache_size" command.
**
** (2) There are no unused PgHdr objects available at this time.
**
** (3) This is an in-memory database.
**
** (4) There are no PgHdr objects that do not require a journal
** file sync and a sync of the journal file is currently
** prohibited.
**
** Otherwise, reuse an existing PgHdr. In other words, reuse an
** existing PgHdr if all of the following are true:
**
** (1) We have reached or exceeded the maximum cache size
** allowed by "PRAGMA cache_size".
**
** (2) There is a PgHdr available with PgHdr->nRef==0
**
** (3) We are not in an in-memory database
**
** (4) Either there is an available PgHdr that does not need
** to be synced to disk or else disk syncing is currently
** allowed.
*/
static int pagerAllocatePage(Pager *pPager, PgHdr **ppPg){
int rc = SQLITE_OK;
PgHdr *pPg;
int nByteHdr;
/* Create a new PgHdr if any of the four conditions defined
** above are met: */
if( pPager->nPage<pPager->mxPage
|| pPager->lru.pFirst==0
|| MEMDB
|| (pPager->lru.pFirstSynced==0 && pPager->doNotSync)
){
void *pData;
if( pPager->nPage>=pPager->nHash ){
pager_resize_hash_table(pPager,
pPager->nHash<256 ? 256 : pPager->nHash*2);
if( pPager->nHash==0 ){
rc = SQLITE_NOMEM;
goto pager_allocate_out;
}
}
pagerLeave(pPager);
nByteHdr = sizeof(*pPg) + sizeof(u32) + pPager->nExtra
+ MEMDB*sizeof(PgHistory);
pPg = sqlite3_malloc( nByteHdr );
if( pPg ){
pData = sqlite3_malloc( pPager->pageSize );
if( pData==0 ){
sqlite3_free(pPg);
pPg = 0;
}
}
pagerEnter(pPager);
if( pPg==0 ){
rc = SQLITE_NOMEM;
goto pager_allocate_out;
}
memset(pPg, 0, nByteHdr);
pPg->pData = pData;
pPg->pPager = pPager;
pPg->pNextAll = pPager->pAll;
pPager->pAll = pPg;
pPager->nPage++;
}else{
/* Recycle an existing page with a zero ref-count. */
rc = pager_recycle(pPager, &pPg);
if( rc==SQLITE_BUSY ){
rc = SQLITE_IOERR_BLOCKED;
}
if( rc!=SQLITE_OK ){
goto pager_allocate_out;
}
assert( pPager->state>=SHARED_LOCK );
assert(pPg);
}
*ppPg = pPg;
pager_allocate_out:
return rc;
}
/*
** Make sure we have the content for a page. If the page was
** previously acquired with noContent==1, then the content was
** just initialized to zeros instead of being read from disk.
** But now we need the real data off of disk. So make sure we
** have it. Read it in if we do not have it already.
*/
static int pager_get_content(PgHdr *pPg){
if( pPg->needRead ){
int rc = readDbPage(pPg->pPager, pPg, pPg->pgno);
if( rc==SQLITE_OK ){
pPg->needRead = 0;
}else{
return rc;
}
}
return SQLITE_OK;
}
/*
** Acquire a page.
**
** A read lock on the disk file is obtained when the first page is acquired.
** This read lock is dropped when the last page is released.
**
** This routine works for any page number greater than 0. If the database
** file is smaller than the requested page, then no actual disk
** read occurs and the memory image of the page is initialized to
** all zeros. The extra data appended to a page is always initialized
** to zeros the first time a page is loaded into memory.
**
** The acquisition might fail for several reasons. In all cases,
** an appropriate error code is returned and *ppPage is set to NULL.
**
** See also sqlite3PagerLookup(). Both this routine and Lookup() attempt
** to find a page in the in-memory cache first. If the page is not already
** in memory, this routine goes to disk to read it in whereas Lookup()
** just returns 0. This routine acquires a read-lock the first time it
** has to go to disk, and could also playback an old journal if necessary.
** Since Lookup() never goes to disk, it never has to deal with locks
** or journal files.
**
** If noContent is false, the page contents are actually read from disk.
** If noContent is true, it means that we do not care about the contents
** of the page at this time, so do not do a disk read. Just fill in the
** page content with zeros. But mark the fact that we have not read the
** content by setting the PgHdr.needRead flag. Later on, if
** sqlite3PagerWrite() is called on this page or if this routine is
** called again with noContent==0, that means that the content is needed
** and the disk read should occur at that point.
*/
static int pagerAcquire(
Pager *pPager, /* The pager open on the database file */
Pgno pgno, /* Page number to fetch */
DbPage **ppPage, /* Write a pointer to the page here */
int noContent /* Do not bother reading content from disk if true */
){
PgHdr *pPg;
int rc;
assert( pPager->state==PAGER_UNLOCK || pPager->nRef>0 || pgno==1 );
/* The maximum page number is 2^31. Return SQLITE_CORRUPT if a page
** number greater than this, or zero, is requested.
*/
if( pgno>PAGER_MAX_PGNO || pgno==0 || pgno==PAGER_MJ_PGNO(pPager) ){
return SQLITE_CORRUPT_BKPT;
}
/* Make sure we have not hit any critical errors.
*/
assert( pPager!=0 );
*ppPage = 0;
/* If this is the first page accessed, then get a SHARED lock
** on the database file. pagerSharedLock() is a no-op if
** a database lock is already held.
*/
rc = pagerSharedLock(pPager);
if( rc!=SQLITE_OK ){
return rc;
}
assert( pPager->state!=PAGER_UNLOCK );
pPg = pager_lookup(pPager, pgno);
if( pPg==0 ){
/* The requested page is not in the page cache. */
int nMax;
int h;
PAGER_INCR(pPager->nMiss);
rc = pagerAllocatePage(pPager, &pPg);
if( rc!=SQLITE_OK ){
return rc;
}
pPg->pgno = pgno;
assert( !MEMDB || pgno>pPager->stmtSize );
pPg->inJournal = sqlite3BitvecTest(pPager->pInJournal, pgno);
pPg->needSync = 0;
makeClean(pPg);
pPg->nRef = 1;
pPager->nRef++;
if( pPager->nExtra>0 ){
memset(PGHDR_TO_EXTRA(pPg, pPager), 0, pPager->nExtra);
}
nMax = sqlite3PagerPagecount(pPager);
if( pPager->errCode ){
rc = pPager->errCode;
sqlite3PagerUnref(pPg);
return rc;
}
/* Populate the page with data, either by reading from the database
** file, or by setting the entire page to zero.
*/
if( nMax<(int)pgno || MEMDB || (noContent && !pPager->alwaysRollback) ){
if( pgno>pPager->mxPgno ){
sqlite3PagerUnref(pPg);
return SQLITE_FULL;
}
memset(PGHDR_TO_DATA(pPg), 0, pPager->pageSize);
pPg->needRead = noContent && !pPager->alwaysRollback;
IOTRACE(("ZERO %p %dn", pPager, pgno));
}else{
rc = readDbPage(pPager, pPg, pgno);
if( rc!=SQLITE_OK && rc!=SQLITE_IOERR_SHORT_READ ){
pPg->pgno = 0;
sqlite3PagerUnref(pPg);
return rc;
}
pPg->needRead = 0;
}
/* Link the page into the page hash table */
h = pgno & (pPager->nHash-1);
assert( pgno!=0 );
pPg->pNextHash = pPager->aHash[h];
pPager->aHash[h] = pPg;
if( pPg->pNextHash ){
assert( pPg->pNextHash->pPrevHash==0 );
pPg->pNextHash->pPrevHash = pPg;
}
#ifdef SQLITE_CHECK_PAGES
pPg->pageHash = pager_pagehash(pPg);
#endif
}else{
/* The requested page is in the page cache. */
assert(pPager->nRef>0 || pgno==1);
PAGER_INCR(pPager->nHit);
if( !noContent ){
rc = pager_get_content(pPg);
if( rc ){
return rc;
}
}
page_ref(pPg);
}
*ppPage = pPg;
return SQLITE_OK;
}
int sqlite3PagerAcquire(
Pager *pPager, /* The pager open on the database file */
Pgno pgno, /* Page number to fetch */
DbPage **ppPage, /* Write a pointer to the page here */
int noContent /* Do not bother reading content from disk if true */
){
int rc;
pagerEnter(pPager);
rc = pagerAcquire(pPager, pgno, ppPage, noContent);
pagerLeave(pPager);
return rc;
}
/*
** Acquire a page if it is already in the in-memory cache. Do
** not read the page from disk. Return a pointer to the page,
** or 0 if the page is not in cache.
**
** See also sqlite3PagerGet(). The difference between this routine
** and sqlite3PagerGet() is that _get() will go to the disk and read
** in the page if the page is not already in cache. This routine
** returns NULL if the page is not in cache or if a disk I/O error
** has ever happened.
*/
DbPage *sqlite3PagerLookup(Pager *pPager, Pgno pgno){
PgHdr *pPg = 0;
assert( pPager!=0 );
assert( pgno!=0 );
pagerEnter(pPager);
if( pPager->state==PAGER_UNLOCK ){
assert( !pPager->pAll || pPager->exclusiveMode );
}else if( pPager->errCode && pPager->errCode!=SQLITE_FULL ){
/* Do nothing */
}else if( (pPg = pager_lookup(pPager, pgno))!=0 ){
page_ref(pPg);
}
pagerLeave(pPager);
return pPg;
}
/*
** Release a page.
**
** If the number of references to the page drop to zero, then the
** page is added to the LRU list. When all references to all pages
** are released, a rollback occurs and the lock on the database is
** removed.
*/
int sqlite3PagerUnref(DbPage *pPg){
Pager *pPager = pPg->pPager;
/* Decrement the reference count for this page
*/
assert( pPg->nRef>0 );
pagerEnter(pPg->pPager);
pPg->nRef--;
CHECK_PAGE(pPg);
/* When the number of references to a page reach 0, call the
** destructor and add the page to the freelist.
*/
if( pPg->nRef==0 ){
lruListAdd(pPg);
if( pPager->xDestructor ){
pPager->xDestructor(pPg, pPager->pageSize);
}
/* When all pages reach the freelist, drop the read lock from
** the database file.
*/
pPager->nRef--;
assert( pPager->nRef>=0 );
if( pPager->nRef==0 && (!pPager->exclusiveMode || pPager->journalOff>0) ){
pagerUnlockAndRollback(pPager);
}
}
pagerLeave(pPager);
return SQLITE_OK;
}
/*
** Create a journal file for pPager. There should already be a RESERVED
** or EXCLUSIVE lock on the database file when this routine is called.
**
** Return SQLITE_OK if everything. Return an error code and release the
** write lock if anything goes wrong.
*/
static int pager_open_journal(Pager *pPager){
sqlite3_vfs *pVfs = pPager->pVfs;
int flags = (SQLITE_OPEN_READWRITE|SQLITE_OPEN_EXCLUSIVE|SQLITE_OPEN_CREATE);
int rc;
assert( !MEMDB );
assert( pPager->state>=PAGER_RESERVED );
assert( pPager->journalOpen==0 );
assert( pPager->useJournal );
assert( pPager->pInJournal==0 );
sqlite3PagerPagecount(pPager);
pagerLeave(pPager);
pPager->pInJournal = sqlite3BitvecCreate(pPager->dbSize);
pagerEnter(pPager);
if( pPager->pInJournal==0 ){
rc = SQLITE_NOMEM;
goto failed_to_open_journal;
}
if( pPager->tempFile ){
flags |= (SQLITE_OPEN_DELETEONCLOSE|SQLITE_OPEN_TEMP_JOURNAL);
}else{
flags |= (SQLITE_OPEN_MAIN_JOURNAL);
}
#ifdef SQLITE_ENABLE_ATOMIC_WRITE
rc = sqlite3JournalOpen(
pVfs, pPager->zJournal, pPager->jfd, flags, jrnlBufferSize(pPager)
);
#else
rc = sqlite3OsOpen(pVfs, pPager->zJournal, pPager->jfd, flags, 0);
#endif
assert( rc!=SQLITE_OK || pPager->jfd->pMethods );
pPager->journalOff = 0;
pPager->setMaster = 0;
pPager->journalHdr = 0;
if( rc!=SQLITE_OK ){
if( rc==SQLITE_NOMEM ){
sqlite3OsDelete(pVfs, pPager->zJournal, 0);
}
goto failed_to_open_journal;
}
pPager->journalOpen = 1;
pPager->journalStarted = 0;
pPager->needSync = 0;
pPager->alwaysRollback = 0;
pPager->nRec = 0;
if( pPager->errCode ){
rc = pPager->errCode;
goto failed_to_open_journal;
}
pPager->origDbSize = pPager->dbSize;
rc = writeJournalHdr(pPager);
if( pPager->stmtAutoopen && rc==SQLITE_OK ){
rc = sqlite3PagerStmtBegin(pPager);
}
if( rc!=SQLITE_OK && rc!=SQLITE_NOMEM && rc!=SQLITE_IOERR_NOMEM ){
rc = pager_end_transaction(pPager);
if( rc==SQLITE_OK ){
rc = SQLITE_FULL;
}
}
return rc;
failed_to_open_journal:
sqlite3BitvecDestroy(pPager->pInJournal);
pPager->pInJournal = 0;
return rc;
}
/*
** Acquire a write-lock on the database. The lock is removed when
** the any of the following happen:
**
** * sqlite3PagerCommitPhaseTwo() is called.
** * sqlite3PagerRollback() is called.
** * sqlite3PagerClose() is called.
** * sqlite3PagerUnref() is called to on every outstanding page.
**
** The first parameter to this routine is a pointer to any open page of the
** database file. Nothing changes about the page - it is used merely to
** acquire a pointer to the Pager structure and as proof that there is
** already a read-lock on the database.
**
** The second parameter indicates how much space in bytes to reserve for a
** master journal file-name at the start of the journal when it is created.
**
** A journal file is opened if this is not a temporary file. For temporary
** files, the opening of the journal file is deferred until there is an
** actual need to write to the journal.
**
** If the database is already reserved for writing, this routine is a no-op.
**
** If exFlag is true, go ahead and get an EXCLUSIVE lock on the file
** immediately instead of waiting until we try to flush the cache. The
** exFlag is ignored if a transaction is already active.
*/
int sqlite3PagerBegin(DbPage *pPg, int exFlag){
Pager *pPager = pPg->pPager;
int rc = SQLITE_OK;
pagerEnter(pPager);
assert( pPg->nRef>0 );
assert( pPager->state!=PAGER_UNLOCK );
if( pPager->state==PAGER_SHARED ){
assert( pPager->pInJournal==0 );
if( MEMDB ){
pPager->state = PAGER_EXCLUSIVE;
pPager->origDbSize = pPager->dbSize;
}else{
rc = sqlite3OsLock(pPager->fd, RESERVED_LOCK);
if( rc==SQLITE_OK ){
pPager->state = PAGER_RESERVED;
if( exFlag ){
rc = pager_wait_on_lock(pPager, EXCLUSIVE_LOCK);
}
}
if( rc!=SQLITE_OK ){
pagerLeave(pPager);
return rc;
}
pPager->dirtyCache = 0;
PAGERTRACE2("TRANSACTION %dn", PAGERID(pPager));
if( pPager->useJournal && !pPager->tempFile ){
rc = pager_open_journal(pPager);
}
}
}else if( pPager->journalOpen && pPager->journalOff==0 ){
/* This happens when the pager was in exclusive-access mode last
** time a (read or write) transaction was successfully concluded
** by this connection. Instead of deleting the journal file it was
** kept open and truncated to 0 bytes.
*/
assert( pPager->nRec==0 );
assert( pPager->origDbSize==0 );
assert( pPager->pInJournal==0 );
sqlite3PagerPagecount(pPager);
pagerLeave(pPager);
pPager->pInJournal = sqlite3BitvecCreate( pPager->dbSize );
pagerEnter(pPager);
if( !pPager->pInJournal ){
rc = SQLITE_NOMEM;
}else{
pPager->origDbSize = pPager->dbSize;
rc = writeJournalHdr(pPager);
}
}
assert( !pPager->journalOpen || pPager->journalOff>0 || rc!=SQLITE_OK );
pagerLeave(pPager);
return rc;
}
/*
** Make a page dirty. Set its dirty flag and add it to the dirty
** page list.
*/
static void makeDirty(PgHdr *pPg){
if( pPg->dirty==0 ){
Pager *pPager = pPg->pPager;
pPg->dirty = 1;
pPg->pDirty = pPager->pDirty;
if( pPager->pDirty ){
pPager->pDirty->pPrevDirty = pPg;
}
pPg->pPrevDirty = 0;
pPager->pDirty = pPg;
}
}
/*
** Make a page clean. Clear its dirty bit and remove it from the
** dirty page list.
*/
static void makeClean(PgHdr *pPg){
if( pPg->dirty ){
pPg->dirty = 0;
if( pPg->pDirty ){
assert( pPg->pDirty->pPrevDirty==pPg );
pPg->pDirty->pPrevDirty = pPg->pPrevDirty;
}
if( pPg->pPrevDirty ){
assert( pPg->pPrevDirty->pDirty==pPg );
pPg->pPrevDirty->pDirty = pPg->pDirty;
}else{
assert( pPg->pPager->pDirty==pPg );
pPg->pPager->pDirty = pPg->pDirty;
}
}
}
/*
** Mark a data page as writeable. The page is written into the journal
** if it is not there already. This routine must be called before making
** changes to a page.
**
** The first time this routine is called, the pager creates a new
** journal and acquires a RESERVED lock on the database. If the RESERVED
** lock could not be acquired, this routine returns SQLITE_BUSY. The
** calling routine must check for that return value and be careful not to
** change any page data until this routine returns SQLITE_OK.
**
** If the journal file could not be written because the disk is full,
** then this routine returns SQLITE_FULL and does an immediate rollback.
** All subsequent write attempts also return SQLITE_FULL until there
** is a call to sqlite3PagerCommit() or sqlite3PagerRollback() to
** reset.
*/
static int pager_write(PgHdr *pPg){
void *pData = PGHDR_TO_DATA(pPg);
Pager *pPager = pPg->pPager;
int rc = SQLITE_OK;
/* Check for errors
*/
if( pPager->errCode ){
return pPager->errCode;
}
if( pPager->readOnly ){
return SQLITE_PERM;
}
assert( !pPager->setMaster );
CHECK_PAGE(pPg);
/* If this page was previously acquired with noContent==1, that means
** we didn't really read in the content of the page. This can happen
** (for example) when the page is being moved to the freelist. But
** now we are (perhaps) moving the page off of the freelist for
** reuse and we need to know its original content so that content
** can be stored in the rollback journal. So do the read at this
** time.
*/
rc = pager_get_content(pPg);
if( rc ){
return rc;
}
/* Mark the page as dirty. If the page has already been written
** to the journal then we can return right away.
*/
makeDirty(pPg);
if( pPg->inJournal && (pageInStatement(pPg) || pPager->stmtInUse==0) ){
pPager->dirtyCache = 1;
}else{
/* If we get this far, it means that the page needs to be
** written to the transaction journal or the ckeckpoint journal
** or both.
**
** First check to see that the transaction journal exists and
** create it if it does not.
*/
assert( pPager->state!=PAGER_UNLOCK );
rc = sqlite3PagerBegin(pPg, 0);
if( rc!=SQLITE_OK ){
return rc;
}
assert( pPager->state>=PAGER_RESERVED );
if( !pPager->journalOpen && pPager->useJournal ){
rc = pager_open_journal(pPager);
if( rc!=SQLITE_OK ) return rc;
}
assert( pPager->journalOpen || !pPager->useJournal );
pPager->dirtyCache = 1;
/* The transaction journal now exists and we have a RESERVED or an
** EXCLUSIVE lock on the main database file. Write the current page to
** the transaction journal if it is not there already.
*/
if( !pPg->inJournal && (pPager->useJournal || MEMDB) ){
if( (int)pPg->pgno <= pPager->origDbSize ){
if( MEMDB ){
PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
PAGERTRACE3("JOURNAL %d page %dn", PAGERID(pPager), pPg->pgno);
assert( pHist->pOrig==0 );
pHist->pOrig = sqlite3_malloc( pPager->pageSize );
if( !pHist->pOrig ){
return SQLITE_NOMEM;
}
memcpy(pHist->pOrig, PGHDR_TO_DATA(pPg), pPager->pageSize);
}else{
u32 cksum;
char *pData2;
/* We should never write to the journal file the page that
** contains the database locks. The following assert verifies
** that we do not. */
assert( pPg->pgno!=PAGER_MJ_PGNO(pPager) );
pData2 = CODEC2(pPager, pData, pPg->pgno, 7);
cksum = pager_cksum(pPager, (u8*)pData2);
rc = write32bits(pPager->jfd, pPager->journalOff, pPg->pgno);
if( rc==SQLITE_OK ){
rc = sqlite3OsWrite(pPager->jfd, pData2, pPager->pageSize,
pPager->journalOff + 4);
pPager->journalOff += pPager->pageSize+4;
}
if( rc==SQLITE_OK ){
rc = write32bits(pPager->jfd, pPager->journalOff, cksum);
pPager->journalOff += 4;
}
IOTRACE(("JOUT %p %d %lld %dn", pPager, pPg->pgno,
pPager->journalOff, pPager->pageSize));
PAGER_INCR(sqlite3_pager_writej_count);
PAGERTRACE5("JOURNAL %d page %d needSync=%d hash(%08x)n",
PAGERID(pPager), pPg->pgno, pPg->needSync, pager_pagehash(pPg));
/* An error has occured writing to the journal file. The
** transaction will be rolled back by the layer above.
*/
if( rc!=SQLITE_OK ){
return rc;
}
pPager->nRec++;
assert( pPager->pInJournal!=0 );
sqlite3BitvecSet(pPager->pInJournal, pPg->pgno);
pPg->needSync = !pPager->noSync;
if( pPager->stmtInUse ){
sqlite3BitvecSet(pPager->pInStmt, pPg->pgno);
}
}
}else{
pPg->needSync = !pPager->journalStarted && !pPager->noSync;
PAGERTRACE4("APPEND %d page %d needSync=%dn",
PAGERID(pPager), pPg->pgno, pPg->needSync);
}
if( pPg->needSync ){
pPager->needSync = 1;
}
pPg->inJournal = 1;
}
/* If the statement journal is open and the page is not in it,
** then write the current page to the statement journal. Note that
** the statement journal format differs from the standard journal format
** in that it omits the checksums and the header.
*/
if( pPager->stmtInUse
&& !pageInStatement(pPg)
&& (int)pPg->pgno<=pPager->stmtSize
){
assert( pPg->inJournal || (int)pPg->pgno>pPager->origDbSize );
if( MEMDB ){
PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
assert( pHist->pStmt==0 );
pHist->pStmt = sqlite3_malloc( pPager->pageSize );
if( pHist->pStmt ){
memcpy(pHist->pStmt, PGHDR_TO_DATA(pPg), pPager->pageSize);
}
PAGERTRACE3("STMT-JOURNAL %d page %dn", PAGERID(pPager), pPg->pgno);
page_add_to_stmt_list(pPg);
}else{
i64 offset = pPager->stmtNRec*(4+pPager->pageSize);
char *pData2 = CODEC2(pPager, pData, pPg->pgno, 7);
rc = write32bits(pPager->stfd, offset, pPg->pgno);
if( rc==SQLITE_OK ){
rc = sqlite3OsWrite(pPager->stfd, pData2, pPager->pageSize, offset+4);
}
PAGERTRACE3("STMT-JOURNAL %d page %dn", PAGERID(pPager), pPg->pgno);
if( rc!=SQLITE_OK ){
return rc;
}
pPager->stmtNRec++;
assert( pPager->pInStmt!=0 );
sqlite3BitvecSet(pPager->pInStmt, pPg->pgno);
}
}
}
/* Update the database size and return.
*/
assert( pPager->state>=PAGER_SHARED );
if( pPager->dbSize<(int)pPg->pgno ){
pPager->dbSize = pPg->pgno;
if( !MEMDB && pPager->dbSize==PENDING_BYTE/pPager->pageSize ){
pPager->dbSize++;
}
}
return rc;
}
/*
** This function is used to mark a data-page as writable. It uses
** pager_write() to open a journal file (if it is not already open)
** and write the page *pData to the journal.
**
** The difference between this function and pager_write() is that this
** function also deals with the special case where 2 or more pages
** fit on a single disk sector. In this case all co-resident pages
** must have been written to the journal file before returning.
*/
int sqlite3PagerWrite(DbPage *pDbPage){
int rc = SQLITE_OK;
PgHdr *pPg = pDbPage;
Pager *pPager = pPg->pPager;
Pgno nPagePerSector = (pPager->sectorSize/pPager->pageSize);
pagerEnter(pPager);
if( !MEMDB && nPagePerSector>1 ){
Pgno nPageCount; /* Total number of pages in database file */
Pgno pg1; /* First page of the sector pPg is located on. */
int nPage; /* Number of pages starting at pg1 to journal */
int ii;
int needSync = 0;
/* Set the doNotSync flag to 1. This is because we cannot allow a journal
** header to be written between the pages journaled by this function.
*/
assert( pPager->doNotSync==0 );
pPager->doNotSync = 1;
/* This trick assumes that both the page-size and sector-size are
** an integer power of 2. It sets variable pg1 to the identifier
** of the first page of the sector pPg is located on.
*/
pg1 = ((pPg->pgno-1) & ~(nPagePerSector-1)) + 1;
nPageCount = sqlite3PagerPagecount(pPager);
if( pPg->pgno>nPageCount ){
nPage = (pPg->pgno - pg1)+1;
}else if( (pg1+nPagePerSector-1)>nPageCount ){
nPage = nPageCount+1-pg1;
}else{
nPage = nPagePerSector;
}
assert(nPage>0);
assert(pg1<=pPg->pgno);
assert((pg1+nPage)>pPg->pgno);
for(ii=0; ii<nPage && rc==SQLITE_OK; ii++){
Pgno pg = pg1+ii;
PgHdr *pPage;
if( pg==pPg->pgno || !sqlite3BitvecTest(pPager->pInJournal, pg) ){
if( pg!=PAGER_MJ_PGNO(pPager) ){
rc = sqlite3PagerGet(pPager, pg, &pPage);
if( rc==SQLITE_OK ){
rc = pager_write(pPage);
if( pPage->needSync ){
needSync = 1;
}
sqlite3PagerUnref(pPage);
}
}
}else if( (pPage = pager_lookup(pPager, pg))!=0 ){
if( pPage->needSync ){
needSync = 1;
}
}
}
/* If the PgHdr.needSync flag is set for any of the nPage pages
** starting at pg1, then it needs to be set for all of them. Because
** writing to any of these nPage pages may damage the others, the
** journal file must contain sync()ed copies of all of them
** before any of them can be written out to the database file.
*/
if( needSync ){
for(ii=0; ii<nPage && needSync; ii++){
PgHdr *pPage = pager_lookup(pPager, pg1+ii);
if( pPage ) pPage->needSync = 1;
}
assert(pPager->needSync);
}
assert( pPager->doNotSync==1 );
pPager->doNotSync = 0;
}else{
rc = pager_write(pDbPage);
}
pagerLeave(pPager);
return rc;
}
/*
** Return TRUE if the page given in the argument was previously passed
** to sqlite3PagerWrite(). In other words, return TRUE if it is ok
** to change the content of the page.
*/
#ifndef NDEBUG
int sqlite3PagerIswriteable(DbPage *pPg){
return pPg->dirty;
}
#endif
#ifndef SQLITE_OMIT_VACUUM
/*
** Replace the content of a single page with the information in the third
** argument.
*/
int sqlite3PagerOverwrite(Pager *pPager, Pgno pgno, void *pData){
PgHdr *pPg;
int rc;
pagerEnter(pPager);
rc = sqlite3PagerGet(pPager, pgno, &pPg);
if( rc==SQLITE_OK ){
rc = sqlite3PagerWrite(pPg);
if( rc==SQLITE_OK ){
memcpy(sqlite3PagerGetData(pPg), pData, pPager->pageSize);
}
sqlite3PagerUnref(pPg);
}
pagerLeave(pPager);
return rc;
}
#endif
/*
** A call to this routine tells the pager that it is not necessary to
** write the information on page pPg back to the disk, even though
** that page might be marked as dirty.
**
** The overlying software layer calls this routine when all of the data
** on the given page is unused. The pager marks the page as clean so
** that it does not get written to disk.
**
** Tests show that this optimization, together with the
** sqlite3PagerDontRollback() below, more than double the speed
** of large INSERT operations and quadruple the speed of large DELETEs.
**
** When this routine is called, set the alwaysRollback flag to true.
** Subsequent calls to sqlite3PagerDontRollback() for the same page
** will thereafter be ignored. This is necessary to avoid a problem
** where a page with data is added to the freelist during one part of
** a transaction then removed from the freelist during a later part
** of the same transaction and reused for some other purpose. When it
** is first added to the freelist, this routine is called. When reused,
** the sqlite3PagerDontRollback() routine is called. But because the
** page contains critical data, we still need to be sure it gets
** rolled back in spite of the sqlite3PagerDontRollback() call.
*/
void sqlite3PagerDontWrite(DbPage *pDbPage){
PgHdr *pPg = pDbPage;
Pager *pPager = pPg->pPager;
if( MEMDB ) return;
pagerEnter(pPager);
pPg->alwaysRollback = 1;
if( pPg->dirty && !pPager->stmtInUse ){
assert( pPager->state>=PAGER_SHARED );
if( pPager->dbSize==(int)pPg->pgno && pPager->origDbSize<pPager->dbSize ){
/* If this pages is the last page in the file and the file has grown
** during the current transaction, then do NOT mark the page as clean.
** When the database file grows, we must make sure that the last page
** gets written at least once so that the disk file will be the correct
** size. If you do not write this page and the size of the file
** on the disk ends up being too small, that can lead to database
** corruption during the next transaction.
*/
}else{
PAGERTRACE3("DONT_WRITE page %d of %dn", pPg->pgno, PAGERID(pPager));
IOTRACE(("CLEAN %p %dn", pPager, pPg->pgno))
makeClean(pPg);
#ifdef SQLITE_CHECK_PAGES
pPg->pageHash = pager_pagehash(pPg);
#endif
}
}
pagerLeave(pPager);
}
/*
** A call to this routine tells the pager that if a rollback occurs,
** it is not necessary to restore the data on the given page. This
** means that the pager does not have to record the given page in the
** rollback journal.
**
** If we have not yet actually read the content of this page (if
** the PgHdr.needRead flag is set) then this routine acts as a promise
** that we will never need to read the page content in the future.
** so the needRead flag can be cleared at this point.
*/
void sqlite3PagerDontRollback(DbPage *pPg){
Pager *pPager = pPg->pPager;
pagerEnter(pPager);
assert( pPager->state>=PAGER_RESERVED );
/* If the journal file is not open, or DontWrite() has been called on
** this page (DontWrite() sets the alwaysRollback flag), then this
** function is a no-op.
*/
if( pPager->journalOpen==0 || pPg->alwaysRollback || pPager->alwaysRollback ){
pagerLeave(pPager);
return;
}
assert( !MEMDB ); /* For a memdb, pPager->journalOpen is always 0 */
#ifdef SQLITE_SECURE_DELETE
if( pPg->inJournal || (int)pPg->pgno > pPager->origDbSize ){
return;
}
#endif
/* If SECURE_DELETE is disabled, then there is no way that this
** routine can be called on a page for which sqlite3PagerDontWrite()
** has not been previously called during the same transaction.
** And if DontWrite() has previously been called, the following
** conditions must be met.
*/
assert( !pPg->inJournal && (int)pPg->pgno <= pPager->origDbSize );
assert( pPager->pInJournal!=0 );
sqlite3BitvecSet(pPager->pInJournal, pPg->pgno);
pPg->inJournal = 1;
pPg->needRead = 0;
if( pPager->stmtInUse ){
assert( pPager->stmtSize >= pPager->origDbSize );
sqlite3BitvecSet(pPager->pInStmt, pPg->pgno);
}
PAGERTRACE3("DONT_ROLLBACK page %d of %dn", pPg->pgno, PAGERID(pPager));
IOTRACE(("GARBAGE %p %dn", pPager, pPg->pgno))
pagerLeave(pPager);
}
/*
** This routine is called to increment the database file change-counter,
** stored at byte 24 of the pager file.
*/
static int pager_incr_changecounter(Pager *pPager, int isDirect){
PgHdr *pPgHdr;
u32 change_counter;
int rc = SQLITE_OK;
if( !pPager->changeCountDone ){
/* Open page 1 of the file for writing. */
rc = sqlite3PagerGet(pPager, 1, &pPgHdr);
if( rc!=SQLITE_OK ) return rc;
if( !isDirect ){
rc = sqlite3PagerWrite(pPgHdr);
if( rc!=SQLITE_OK ){
sqlite3PagerUnref(pPgHdr);
return rc;
}
}
/* Increment the value just read and write it back to byte 24. */
change_counter = sqlite3Get4byte((u8*)pPager->dbFileVers);
change_counter++;
put32bits(((char*)PGHDR_TO_DATA(pPgHdr))+24, change_counter);
if( isDirect && pPager->fd->pMethods ){
const void *zBuf = PGHDR_TO_DATA(pPgHdr);
rc = sqlite3OsWrite(pPager->fd, zBuf, pPager->pageSize, 0);
}
/* Release the page reference. */
sqlite3PagerUnref(pPgHdr);
pPager->changeCountDone = 1;
}
return rc;
}
/*
** Sync the pager file to disk.
*/
int sqlite3PagerSync(Pager *pPager){
int rc;
pagerEnter(pPager);
rc = sqlite3OsSync(pPager->fd, pPager->sync_flags);
pagerLeave(pPager);
return rc;
}
/*
** Sync the database file for the pager pPager. zMaster points to the name
** of a master journal file that should be written into the individual
** journal file. zMaster may be NULL, which is interpreted as no master
** journal (a single database transaction).
**
** This routine ensures that the journal is synced, all dirty pages written
** to the database file and the database file synced. The only thing that
** remains to commit the transaction is to delete the journal file (or
** master journal file if specified).
**
** Note that if zMaster==NULL, this does not overwrite a previous value
** passed to an sqlite3PagerCommitPhaseOne() call.
**
** If parameter nTrunc is non-zero, then the pager file is truncated to
** nTrunc pages (this is used by auto-vacuum databases).
**
** If the final parameter - noSync - is true, then the database file itself
** is not synced. The caller must call sqlite3PagerSync() directly to
** sync the database file before calling CommitPhaseTwo() to delete the
** journal file in this case.
*/
int sqlite3PagerCommitPhaseOne(
Pager *pPager,
const char *zMaster,
Pgno nTrunc,
int noSync
){
int rc = SQLITE_OK;
PAGERTRACE4("DATABASE SYNC: File=%s zMaster=%s nTrunc=%dn",
pPager->zFilename, zMaster, nTrunc);
pagerEnter(pPager);
/* If this is an in-memory db, or no pages have been written to, or this
** function has already been called, it is a no-op.
*/
if( pPager->state!=PAGER_SYNCED && !MEMDB && pPager->dirtyCache ){
PgHdr *pPg;
#ifdef SQLITE_ENABLE_ATOMIC_WRITE
/* The atomic-write optimization can be used if all of the
** following are true:
**
** + The file-system supports the atomic-write property for
** blocks of size page-size, and
** + This commit is not part of a multi-file transaction, and
** + Exactly one page has been modified and store in the journal file.
**
** If the optimization can be used, then the journal file will never
** be created for this transaction.
*/
int useAtomicWrite = (
!zMaster &&
pPager->journalOff==jrnlBufferSize(pPager) &&
nTrunc==0 &&
(0==pPager->pDirty || 0==pPager->pDirty->pDirty)
);
if( useAtomicWrite ){
/* Update the nRec field in the journal file. */
int offset = pPager->journalHdr + sizeof(aJournalMagic);
assert(pPager->nRec==1);
rc = write32bits(pPager->jfd, offset, pPager->nRec);
/* Update the db file change counter. The following call will modify
** the in-memory representation of page 1 to include the updated
** change counter and then write page 1 directly to the database
** file. Because of the atomic-write property of the host file-system,
** this is safe.
*/
if( rc==SQLITE_OK ){
rc = pager_incr_changecounter(pPager, 1);
}
}else{
rc = sqlite3JournalCreate(pPager->jfd);
}
if( !useAtomicWrite && rc==SQLITE_OK )
#endif
/* If a master journal file name has already been written to the
** journal file, then no sync is required. This happens when it is
** written, then the process fails to upgrade from a RESERVED to an
** EXCLUSIVE lock. The next time the process tries to commit the
** transaction the m-j name will have already been written.
*/
if( !pPager->setMaster ){
assert( pPager->journalOpen );
rc = pager_incr_changecounter(pPager, 0);
if( rc!=SQLITE_OK ) goto sync_exit;
#ifndef SQLITE_OMIT_AUTOVACUUM
if( nTrunc!=0 ){
/* If this transaction has made the database smaller, then all pages
** being discarded by the truncation must be written to the journal
** file.
*/
Pgno i;
int iSkip = PAGER_MJ_PGNO(pPager);
for( i=nTrunc+1; i<=pPager->origDbSize; i++ ){
if( !sqlite3BitvecTest(pPager->pInJournal, i) && i!=iSkip ){
rc = sqlite3PagerGet(pPager, i, &pPg);
if( rc!=SQLITE_OK ) goto sync_exit;
rc = sqlite3PagerWrite(pPg);
sqlite3PagerUnref(pPg);
if( rc!=SQLITE_OK ) goto sync_exit;
}
}
}
#endif
rc = writeMasterJournal(pPager, zMaster);
if( rc!=SQLITE_OK ) goto sync_exit;
rc = syncJournal(pPager);
}
if( rc!=SQLITE_OK ) goto sync_exit;
#ifndef SQLITE_OMIT_AUTOVACUUM
if( nTrunc!=0 ){
rc = sqlite3PagerTruncate(pPager, nTrunc);
if( rc!=SQLITE_OK ) goto sync_exit;
}
#endif
/* Write all dirty pages to the database file */
pPg = pager_get_all_dirty_pages(pPager);
rc = pager_write_pagelist(pPg);
if( rc!=SQLITE_OK ){
assert( rc!=SQLITE_IOERR_BLOCKED );
/* The error might have left the dirty list all fouled up here,
** but that does not matter because if the if the dirty list did
** get corrupted, then the transaction will roll back and
** discard the dirty list. There is an assert in
** pager_get_all_dirty_pages() that verifies that no attempt
** is made to use an invalid dirty list.
*/
goto sync_exit;
}
pPager->pDirty = 0;
/* Sync the database file. */
if( !pPager->noSync && !noSync ){
rc = sqlite3OsSync(pPager->fd, pPager->sync_flags);
}
IOTRACE(("DBSYNC %pn", pPager))
pPager->state = PAGER_SYNCED;
}else if( MEMDB && nTrunc!=0 ){
rc = sqlite3PagerTruncate(pPager, nTrunc);
}
sync_exit:
if( rc==SQLITE_IOERR_BLOCKED ){
/* pager_incr_changecounter() may attempt to obtain an exclusive
* lock to spill the cache and return IOERR_BLOCKED. But since
* there is no chance the cache is inconsistent, it is
* better to return SQLITE_BUSY.
*/
rc = SQLITE_BUSY;
}
pagerLeave(pPager);
return rc;
}
/*
** Commit all changes to the database and release the write lock.
**
** If the commit fails for any reason, a rollback attempt is made
** and an error code is returned. If the commit worked, SQLITE_OK
** is returned.
*/
int sqlite3PagerCommitPhaseTwo(Pager *pPager){
int rc;
PgHdr *pPg;
if( pPager->errCode ){
return pPager->errCode;
}
if( pPager->state<PAGER_RESERVED ){
return SQLITE_ERROR;
}
pagerEnter(pPager);
PAGERTRACE2("COMMIT %dn", PAGERID(pPager));
if( MEMDB ){
pPg = pager_get_all_dirty_pages(pPager);
while( pPg ){
PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
clearHistory(pHist);
pPg->dirty = 0;
pPg->inJournal = 0;
pHist->inStmt = 0;
pPg->needSync = 0;
pHist->pPrevStmt = pHist->pNextStmt = 0;
pPg = pPg->pDirty;
}
pPager->pDirty = 0;
#ifndef NDEBUG
for(pPg=pPager->pAll; pPg; pPg=pPg->pNextAll){
PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
assert( !pPg->alwaysRollback );
assert( !pHist->pOrig );
assert( !pHist->pStmt );
}
#endif
pPager->pStmt = 0;
pPager->state = PAGER_SHARED;
pagerLeave(pPager);
return SQLITE_OK;
}
assert( pPager->journalOpen || !pPager->dirtyCache );
assert( pPager->state==PAGER_SYNCED || !pPager->dirtyCache );
rc = pager_end_transaction(pPager);
rc = pager_error(pPager, rc);
pagerLeave(pPager);
return rc;
}
/*
** Rollback all changes. The database falls back to PAGER_SHARED mode.
** All in-memory cache pages revert to their original data contents.
** The journal is deleted.
**
** This routine cannot fail unless some other process is not following
** the correct locking protocol or unless some other
** process is writing trash into the journal file (SQLITE_CORRUPT) or
** unless a prior malloc() failed (SQLITE_NOMEM). Appropriate error
** codes are returned for all these occasions. Otherwise,
** SQLITE_OK is returned.
*/
int sqlite3PagerRollback(Pager *pPager){
int rc;
PAGERTRACE2("ROLLBACK %dn", PAGERID(pPager));
if( MEMDB ){
PgHdr *p;
for(p=pPager->pAll; p; p=p->pNextAll){
PgHistory *pHist;
assert( !p->alwaysRollback );
if( !p->dirty ){
assert( !((PgHistory *)PGHDR_TO_HIST(p, pPager))->pOrig );
assert( !((PgHistory *)PGHDR_TO_HIST(p, pPager))->pStmt );
continue;
}
pHist = PGHDR_TO_HIST(p, pPager);
if( pHist->pOrig ){
memcpy(PGHDR_TO_DATA(p), pHist->pOrig, pPager->pageSize);
PAGERTRACE3("ROLLBACK-PAGE %d of %dn", p->pgno, PAGERID(pPager));
}else{
PAGERTRACE3("PAGE %d is clean on %dn", p->pgno, PAGERID(pPager));
}
clearHistory(pHist);
p->dirty = 0;
p->inJournal = 0;
pHist->inStmt = 0;
pHist->pPrevStmt = pHist->pNextStmt = 0;
if( pPager->xReiniter ){
pPager->xReiniter(p, pPager->pageSize);
}
}
pPager->pDirty = 0;
pPager->pStmt = 0;
pPager->dbSize = pPager->origDbSize;
pager_truncate_cache(pPager);
pPager->stmtInUse = 0;
pPager->state = PAGER_SHARED;
return SQLITE_OK;
}
pagerEnter(pPager);
if( !pPager->dirtyCache || !pPager->journalOpen ){
rc = pager_end_transaction(pPager);
pagerLeave(pPager);
return rc;
}
if( pPager->errCode && pPager->errCode!=SQLITE_FULL ){
if( pPager->state>=PAGER_EXCLUSIVE ){
pager_playback(pPager, 0);
}
pagerLeave(pPager);
return pPager->errCode;
}
if( pPager->state==PAGER_RESERVED ){
int rc2;
rc = pager_playback(pPager, 0);
rc2 = pager_end_transaction(pPager);
if( rc==SQLITE_OK ){
rc = rc2;
}
}else{
rc = pager_playback(pPager, 0);
}
/* pager_reset(pPager); */
pPager->dbSize = -1;
/* If an error occurs during a ROLLBACK, we can no longer trust the pager
** cache. So call pager_error() on the way out to make any error
** persistent.
*/
rc = pager_error(pPager, rc);
pagerLeave(pPager);
return rc;
}
/*
** Return TRUE if the database file is opened read-only. Return FALSE
** if the database is (in theory) writable.
*/
int sqlite3PagerIsreadonly(Pager *pPager){
return pPager->readOnly;
}
/*
** Return the number of references to the pager.
*/
int sqlite3PagerRefcount(Pager *pPager){
return pPager->nRef;
}
#ifdef SQLITE_TEST
/*
** This routine is used for testing and analysis only.
*/
int *sqlite3PagerStats(Pager *pPager){
static int a[11];
a[0] = pPager->nRef;
a[1] = pPager->nPage;
a[2] = pPager->mxPage;
a[3] = pPager->dbSize;
a[4] = pPager->state;
a[5] = pPager->errCode;
a[6] = pPager->nHit;
a[7] = pPager->nMiss;
a[8] = 0; /* Used to be pPager->nOvfl */
a[9] = pPager->nRead;
a[10] = pPager->nWrite;
return a;
}
#endif
/*
** Set the statement rollback point.
**
** This routine should be called with the transaction journal already
** open. A new statement journal is created that can be used to rollback
** changes of a single SQL command within a larger transaction.
*/
static int pagerStmtBegin(Pager *pPager){
int rc;
assert( !pPager->stmtInUse );
assert( pPager->state>=PAGER_SHARED );
assert( pPager->dbSize>=0 );
PAGERTRACE2("STMT-BEGIN %dn", PAGERID(pPager));
if( MEMDB ){
pPager->stmtInUse = 1;
pPager->stmtSize = pPager->dbSize;
return SQLITE_OK;
}
if( !pPager->journalOpen ){
pPager->stmtAutoopen = 1;
return SQLITE_OK;
}
assert( pPager->journalOpen );
pagerLeave(pPager);
assert( pPager->pInStmt==0 );
pPager->pInStmt = sqlite3BitvecCreate(pPager->dbSize);
pagerEnter(pPager);
if( pPager->pInStmt==0 ){
/* sqlite3OsLock(pPager->fd, SHARED_LOCK); */
return SQLITE_NOMEM;
}
#ifndef NDEBUG
rc = sqlite3OsFileSize(pPager->jfd, &pPager->stmtJSize);
if( rc ) goto stmt_begin_failed;
assert( pPager->stmtJSize == pPager->journalOff );
#endif
pPager->stmtJSize = pPager->journalOff;
pPager->stmtSize = pPager->dbSize;
pPager->stmtHdrOff = 0;
pPager->stmtCksum = pPager->cksumInit;
if( !pPager->stmtOpen ){
rc = sqlite3PagerOpentemp(pPager->pVfs, pPager->stfd, pPager->zStmtJrnl,
SQLITE_OPEN_SUBJOURNAL);
if( rc ){
goto stmt_begin_failed;
}
pPager->stmtOpen = 1;
pPager->stmtNRec = 0;
}
pPager->stmtInUse = 1;
return SQLITE_OK;
stmt_begin_failed:
if( pPager->pInStmt ){
sqlite3BitvecDestroy(pPager->pInStmt);
pPager->pInStmt = 0;
}
return rc;
}
int sqlite3PagerStmtBegin(Pager *pPager){
int rc;
pagerEnter(pPager);
rc = pagerStmtBegin(pPager);
pagerLeave(pPager);
return rc;
}
/*
** Commit a statement.
*/
int sqlite3PagerStmtCommit(Pager *pPager){
pagerEnter(pPager);
if( pPager->stmtInUse ){
PgHdr *pPg, *pNext;
PAGERTRACE2("STMT-COMMIT %dn", PAGERID(pPager));
if( !MEMDB ){
/* sqlite3OsTruncate(pPager->stfd, 0); */
sqlite3BitvecDestroy(pPager->pInStmt);
pPager->pInStmt = 0;
}else{
for(pPg=pPager->pStmt; pPg; pPg=pNext){
PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
pNext = pHist->pNextStmt;
assert( pHist->inStmt );
pHist->inStmt = 0;
pHist->pPrevStmt = pHist->pNextStmt = 0;
sqlite3_free(pHist->pStmt);
pHist->pStmt = 0;
}
}
pPager->stmtNRec = 0;
pPager->stmtInUse = 0;
pPager->pStmt = 0;
}
pPager->stmtAutoopen = 0;
pagerLeave(pPager);
return SQLITE_OK;
}
/*
** Rollback a statement.
*/
int sqlite3PagerStmtRollback(Pager *pPager){
int rc;
pagerEnter(pPager);
if( pPager->stmtInUse ){
PAGERTRACE2("STMT-ROLLBACK %dn", PAGERID(pPager));
if( MEMDB ){
PgHdr *pPg;
PgHistory *pHist;
for(pPg=pPager->pStmt; pPg; pPg=pHist->pNextStmt){
pHist = PGHDR_TO_HIST(pPg, pPager);
if( pHist->pStmt ){
memcpy(PGHDR_TO_DATA(pPg), pHist->pStmt, pPager->pageSize);
sqlite3_free(pHist->pStmt);
pHist->pStmt = 0;
}
}
pPager->dbSize = pPager->stmtSize;
pager_truncate_cache(pPager);
rc = SQLITE_OK;
}else{
rc = pager_stmt_playback(pPager);
}
sqlite3PagerStmtCommit(pPager);
}else{
rc = SQLITE_OK;
}
pPager->stmtAutoopen = 0;
pagerLeave(pPager);
return rc;
}
/*
** Return the full pathname of the database file.
*/
const char *sqlite3PagerFilename(Pager *pPager){
return pPager->zFilename;
}
/*
** Return the VFS structure for the pager.
*/
const sqlite3_vfs *sqlite3PagerVfs(Pager *pPager){
return pPager->pVfs;
}
/*
** Return the file handle for the database file associated
** with the pager. This might return NULL if the file has
** not yet been opened.
*/
sqlite3_file *sqlite3PagerFile(Pager *pPager){
return pPager->fd;
}
/*
** Return the directory of the database file.
*/
const char *sqlite3PagerDirname(Pager *pPager){
return pPager->zDirectory;
}
/*
** Return the full pathname of the journal file.
*/
const char *sqlite3PagerJournalname(Pager *pPager){
return pPager->zJournal;
}
/*
** Return true if fsync() calls are disabled for this pager. Return FALSE
** if fsync()s are executed normally.
*/
int sqlite3PagerNosync(Pager *pPager){
return pPager->noSync;
}
#ifdef SQLITE_HAS_CODEC
/*
** Set the codec for this pager
*/
void sqlite3PagerSetCodec(
Pager *pPager,
void *(*xCodec)(void*,void*,Pgno,int),
void *pCodecArg
){
pPager->xCodec = xCodec;
pPager->pCodecArg = pCodecArg;
}
#endif
#ifndef SQLITE_OMIT_AUTOVACUUM
/*
** Move the page pPg to location pgno in the file.
**
** There must be no references to the page previously located at
** pgno (which we call pPgOld) though that page is allowed to be
** in cache. If the page previous located at pgno is not already
** in the rollback journal, it is not put there by by this routine.
**
** References to the page pPg remain valid. Updating any
** meta-data associated with pPg (i.e. data stored in the nExtra bytes
** allocated along with the page) is the responsibility of the caller.
**
** A transaction must be active when this routine is called. It used to be
** required that a statement transaction was not active, but this restriction
** has been removed (CREATE INDEX needs to move a page when a statement
** transaction is active).
*/
int sqlite3PagerMovepage(Pager *pPager, DbPage *pPg, Pgno pgno){
PgHdr *pPgOld; /* The page being overwritten. */
int h;
Pgno needSyncPgno = 0;
pagerEnter(pPager);
assert( pPg->nRef>0 );
PAGERTRACE5("MOVE %d page %d (needSync=%d) moves to %dn",
PAGERID(pPager), pPg->pgno, pPg->needSync, pgno);
IOTRACE(("MOVE %p %d %dn", pPager, pPg->pgno, pgno))
pager_get_content(pPg);
if( pPg->needSync ){
needSyncPgno = pPg->pgno;
assert( pPg->inJournal || (int)pgno>pPager->origDbSize );
assert( pPg->dirty );
assert( pPager->needSync );
}
/* Unlink pPg from its hash-chain */
unlinkHashChain(pPager, pPg);
/* If the cache contains a page with page-number pgno, remove it
** from its hash chain. Also, if the PgHdr.needSync was set for
** page pgno before the 'move' operation, it needs to be retained
** for the page moved there.
*/
pPg->needSync = 0;
pPgOld = pager_lookup(pPager, pgno);
if( pPgOld ){
assert( pPgOld->nRef==0 );
unlinkHashChain(pPager, pPgOld);
makeClean(pPgOld);
pPg->needSync = pPgOld->needSync;
}else{
pPg->needSync = 0;
}
pPg->inJournal = sqlite3BitvecTest(pPager->pInJournal, pgno);
/* Change the page number for pPg and insert it into the new hash-chain. */
assert( pgno!=0 );
pPg->pgno = pgno;
h = pgno & (pPager->nHash-1);
if( pPager->aHash[h] ){
assert( pPager->aHash[h]->pPrevHash==0 );
pPager->aHash[h]->pPrevHash = pPg;
}
pPg->pNextHash = pPager->aHash[h];
pPager->aHash[h] = pPg;
pPg->pPrevHash = 0;
makeDirty(pPg);
pPager->dirtyCache = 1;
if( needSyncPgno ){
/* If needSyncPgno is non-zero, then the journal file needs to be
** sync()ed before any data is written to database file page needSyncPgno.
** Currently, no such page exists in the page-cache and the
** Pager.pInJournal bit has been set. This needs to be remedied by loading
** the page into the pager-cache and setting the PgHdr.needSync flag.
**
** If the attempt to load the page into the page-cache fails, (due
** to a malloc() or IO failure), clear the bit in the pInJournal[]
** array. Otherwise, if the page is loaded and written again in
** this transaction, it may be written to the database file before
** it is synced into the journal file. This way, it may end up in
** the journal file twice, but that is not a problem.
**
** The sqlite3PagerGet() call may cause the journal to sync. So make
** sure the Pager.needSync flag is set too.
*/
int rc;
PgHdr *pPgHdr;
assert( pPager->needSync );
rc = sqlite3PagerGet(pPager, needSyncPgno, &pPgHdr);
if( rc!=SQLITE_OK ){
if( pPager->pInJournal && (int)needSyncPgno<=pPager->origDbSize ){
sqlite3BitvecClear(pPager->pInJournal, needSyncPgno);
}
pagerLeave(pPager);
return rc;
}
pPager->needSync = 1;
pPgHdr->needSync = 1;
pPgHdr->inJournal = 1;
makeDirty(pPgHdr);
sqlite3PagerUnref(pPgHdr);
}
pagerLeave(pPager);
return SQLITE_OK;
}
#endif
/*
** Return a pointer to the data for the specified page.
*/
void *sqlite3PagerGetData(DbPage *pPg){
return PGHDR_TO_DATA(pPg);
}
/*
** Return a pointer to the Pager.nExtra bytes of "extra" space
** allocated along with the specified page.
*/
void *sqlite3PagerGetExtra(DbPage *pPg){
Pager *pPager = pPg->pPager;
return (pPager?PGHDR_TO_EXTRA(pPg, pPager):0);
}
/*
** Get/set the locking-mode for this pager. Parameter eMode must be one
** of PAGER_LOCKINGMODE_QUERY, PAGER_LOCKINGMODE_NORMAL or
** PAGER_LOCKINGMODE_EXCLUSIVE. If the parameter is not _QUERY, then
** the locking-mode is set to the value specified.
**
** The returned value is either PAGER_LOCKINGMODE_NORMAL or
** PAGER_LOCKINGMODE_EXCLUSIVE, indicating the current (possibly updated)
** locking-mode.
*/
int sqlite3PagerLockingMode(Pager *pPager, int eMode){
assert( eMode==PAGER_LOCKINGMODE_QUERY
|| eMode==PAGER_LOCKINGMODE_NORMAL
|| eMode==PAGER_LOCKINGMODE_EXCLUSIVE );
assert( PAGER_LOCKINGMODE_QUERY<0 );
assert( PAGER_LOCKINGMODE_NORMAL>=0 && PAGER_LOCKINGMODE_EXCLUSIVE>=0 );
if( eMode>=0 && !pPager->tempFile ){
pPager->exclusiveMode = eMode;
}
return (int)pPager->exclusiveMode;
}
#ifdef SQLITE_TEST
/*
** Print a listing of all referenced pages and their ref count.
*/
void sqlite3PagerRefdump(Pager *pPager){
PgHdr *pPg;
for(pPg=pPager->pAll; pPg; pPg=pPg->pNextAll){
if( pPg->nRef<=0 ) continue;
sqlite3DebugPrintf("PAGE %3d addr=%p nRef=%dn",
pPg->pgno, PGHDR_TO_DATA(pPg), pPg->nRef);
}
}
#endif
#endif /* SQLITE_OMIT_DISKIO */