SQLite

/*
** 2004 April 6
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
** $Id: btree.c,v 1.162 2004/06/09 17:37:23 drh Exp $
**
** This file implements a external (disk-based) database using BTrees.
** For a detailed discussion of BTrees, refer to
**
**     Donald E. Knuth, THE ART OF COMPUTER PROGRAMMING, Volume 3:
**     "Sorting And Searching", pages 473-480. Addison-Wesley
**     Publishing Company, Reading, Massachusetts.

**
** The pParent field points back to the parent page.  This allows us to
** walk up the BTree from any leaf to the root.  Care must be taken to
** unref() the parent page pointer when this page is no longer referenced.
** The pageDestructor() routine handles that chore.
*/
struct MemPage {
  u8 isInit;           /* True if previously initialized. MUST BE FIRST! */
  u8 idxShift;         /* True if Cell indices have changed */
  u8 nOverflow;        /* Number of overflow cell bodies in aCell[] */
  u8 intKey;           /* True if intkey flag is set */
  u8 leaf;             /* True if leaf flag is set */
  u8 zeroData;         /* True if table stores keys only */
  u8 leafData;         /* True if tables stores data on leaves only */
  u8 hasData;          /* True if this page stores data */

  if( pPage->pParent ){
    MemPage *pParent = pPage->pParent;
    pPage->pParent = 0;
    releasePage(pParent);
  }
  pPage->isInit = 0;
}

/*
** During a rollback, when the pager reloads information into the cache
** so that the cache is restored to its original state at the start of
** the transaction, for each page restored this routine is called.
**
** This routine needs to reset the extra data section at the end of the
** page to agree with the restored data.
*/
static void pageReinit(void *pData, int pageSize){
  MemPage *pPage = (MemPage*)&((char*)pData)[pageSize];
  if( pPage->isInit ){
    pPage->isInit = 0;
    initPage(pPage, pPage->pParent);
  }
}

/*
** Open a new database.
**
** Actually, this routine just sets up the internal data structures
** for accessing the database.  We do not open the database file 
** until the first page is loaded.

  if( rc!=SQLITE_OK ){
    if( pBt->pPager ) sqlite3pager_close(pBt->pPager);
    sqliteFree(pBt);
    *ppBtree = 0;
    return rc;
  }
  sqlite3pager_set_destructor(pBt->pPager, pageDestructor);
  sqlite3pager_set_reiniter(pBt->pPager, pageReinit);
  pBt->pCursor = 0;
  pBt->pPage1 = 0;
  pBt->readOnly = sqlite3pager_isreadonly(pBt->pPager);
  pBt->pageSize = SQLITE_PAGE_SIZE;  /* FIX ME - read from header */
  pBt->usableSize = pBt->pageSize;
  pBt->maxEmbedFrac = 64;            /* FIX ME - read from header */
  pBt->minEmbedFrac = 32;            /* FIX ME - read from header */

int sqlite3OsClose(OsFile*);
int sqlite3OsRead(OsFile*, void*, int amt);
int sqlite3OsWrite(OsFile*, const void*, int amt);
int sqlite3OsSeek(OsFile*, off_t offset);
int sqlite3OsSync(OsFile*);
int sqlite3OsTruncate(OsFile*, off_t size);
int sqlite3OsFileSize(OsFile*, off_t *pSize);



int sqlite3OsRandomSeed(char*);
int sqlite3OsSleep(int ms);
int sqlite3OsCurrentTime(double*);
void sqlite3OsEnterMutex(void);
void sqlite3OsLeaveMutex(void);
char *sqlite3OsFullPathname(const char*);
int sqlite3OsLock(OsFile*, int);
int sqlite3OsUnlock(OsFile*, int);
int sqlite3OsCheckReservedLock(OsFile *id);

#endif /* _SQLITE_OS_H_ */

#define TIMER_END         elapse=hwtime()-g_start
#define SEEK(X)           last_page=(X)
#define TRACE1(X)         if( sqlite3_os_trace ) sqlite3DebugPrintf(X)
#define TRACE2(X,Y)       if( sqlite3_os_trace ) sqlite3DebugPrintf(X,Y)
#define TRACE3(X,Y,Z)     if( sqlite3_os_trace ) sqlite3DebugPrintf(X,Y,Z)
#define TRACE4(X,Y,Z,A)   if( sqlite3_os_trace ) sqlite3DebugPrintf(X,Y,Z,A)
#define TRACE5(X,Y,Z,A,B) if( sqlite3_os_trace ) sqlite3DebugPrintf(X,Y,Z,A,B)
#define TRACE6(X,Y,Z,A,B,C) if(sqlite3_os_trace) sqlite3DebugPrintf(X,Y,Z,A,B,C)
#else
#define TIMER_START
#define TIMER_END
#define SEEK(X)
#define TRACE1(X)
#define TRACE2(X,Y)
#define TRACE3(X,Y,Z)
#define TRACE4(X,Y,Z,A)
#define TRACE5(X,Y,Z,A,B)
#define TRACE6(X,Y,Z,A,B,C)
#endif


/*
** If we compile with the SQLITE_TEST macro set, then the following block
** of code will give us the ability to simulate a disk I/O error.  This
** is used for testing the I/O recovery logic.

int sqlite3OsOpenReadWrite(
  const char *zFilename,
  OsFile *id,
  int *pReadonly
){
  int rc;
  id->dirfd = -1;
  id->h = open(zFilename, O_RDWR|O_CREAT|O_LARGEFILE|O_BINARY, 0644);
  if( id->h<0 ){
    id->h = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY);
    if( id->h<0 ){
      return SQLITE_CANTOPEN; 
    }
    *pReadonly = 1;
  }else{
    *pReadonly = 0;
  }
  sqlite3OsEnterMutex();
  rc = findLockInfo(id->h, &id->pLock, &id->pOpen);
  sqlite3OsLeaveMutex();
  if( rc ){
    close(id->h);
    return SQLITE_NOMEM;
  }
  id->locktype = 0;
  TRACE3("OPEN    %-3d %s\n", id->h, zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
}


/*
** Attempt to open a new file for exclusive access by this process.

*/
int sqlite3OsOpenExclusive(const char *zFilename, OsFile *id, int delFlag){
  int rc;
  if( access(zFilename, 0)==0 ){
    return SQLITE_CANTOPEN;
  }
  id->dirfd = -1;
  id->h = open(zFilename,
                O_RDWR|O_CREAT|O_EXCL|O_NOFOLLOW|O_LARGEFILE|O_BINARY, 0600);
  if( id->h<0 ){
    return SQLITE_CANTOPEN;
  }
  sqlite3OsEnterMutex();
  rc = findLockInfo(id->h, &id->pLock, &id->pOpen);
  sqlite3OsLeaveMutex();
  if( rc ){
    close(id->h);
    unlink(zFilename);
    return SQLITE_NOMEM;
  }
  id->locktype = 0;
  if( delFlag ){
    unlink(zFilename);
  }
  TRACE3("OPEN-EX %-3d %s\n", id->h, zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
}

/*
** Attempt to open a new file for read-only access.
**
** On success, write the file handle into *id and return SQLITE_OK.
**
** On failure, return SQLITE_CANTOPEN.
*/
int sqlite3OsOpenReadOnly(const char *zFilename, OsFile *id){
  int rc;
  id->dirfd = -1;
  id->h = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY);
  if( id->h<0 ){
    return SQLITE_CANTOPEN;
  }
  sqlite3OsEnterMutex();
  rc = findLockInfo(id->h, &id->pLock, &id->pOpen);
  sqlite3OsLeaveMutex();
  if( rc ){
    close(id->h);
    return SQLITE_NOMEM;
  }
  id->locktype = 0;
  TRACE3("OPEN-RO %-3d %s\n", id->h, zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
}

/*
** Attempt to open a file descriptor for the directory that contains a
** file.  This file descriptor can be used to fsync() the directory

** On failure, the function returns SQLITE_CANTOPEN and leaves
** *id unchanged.
*/
int sqlite3OsOpenDirectory(
  const char *zDirname,
  OsFile *id
){
  if( id->h<0 ){
    /* Do not open the directory if the corresponding file is not already
    ** open. */
    return SQLITE_CANTOPEN;
  }
  assert( id->dirfd<0 );
  id->dirfd = open(zDirname, O_RDONLY|O_BINARY, 0644);
  if( id->dirfd<0 ){

  return SQLITE_OK; 
}

/*
** Close a file.
*/
int sqlite3OsClose(OsFile *id){
  sqlite3OsUnlock(id, NO_LOCK);
  if( id->dirfd>=0 ) close(id->dirfd);
  id->dirfd = -1;
  sqlite3OsEnterMutex();
  if( id->pOpen->nLock ){
    /* If there are outstanding locks, do not actually close the file just
    ** yet because that would clear those locks.  Instead, add the file
    ** descriptor to pOpen->aPending.  It will be automatically closed when
    ** the last lock is cleared.
    */
    int *aNew;
    struct openCnt *pOpen = id->pOpen;
    pOpen->nPending++;
    aNew = sqliteRealloc( pOpen->aPending, pOpen->nPending*sizeof(int) );
    if( aNew==0 ){
      /* If a malloc fails, just leak the file descriptor */
    }else{
      pOpen->aPending = aNew;
      pOpen->aPending[pOpen->nPending-1] = id->h;
    }
  }else{
    /* There are no outstanding locks so we can close the file immediately */
    close(id->h);
  }
  releaseLockInfo(id->pLock);
  releaseOpenCnt(id->pOpen);
  sqlite3OsLeaveMutex();
  TRACE2("CLOSE   %-3d\n", id->h);
  OpenCounter(-1);
  return SQLITE_OK;
}

/*
** Read data from a file into a buffer.  Return SQLITE_OK if all
** bytes were read successfully and SQLITE_IOERR if anything goes
** wrong.
*/
int sqlite3OsRead(OsFile *id, void *pBuf, int amt){
  int got;
  SimulateIOError(SQLITE_IOERR);
  TIMER_START;
  got = read(id->h, pBuf, amt);
  TIMER_END;
  TRACE4("READ    %-3d %7d %d\n", id->h, last_page, elapse);
  SEEK(0);
  /* if( got<0 ) got = 0; */
  if( got==amt ){
    return SQLITE_OK;
  }else{
    return SQLITE_IOERR;
  }
}

/*
** Write data from a buffer into a file.  Return SQLITE_OK on success
** or some other error code on failure.
*/
int sqlite3OsWrite(OsFile *id, const void *pBuf, int amt){
  int wrote = 0;
  SimulateIOError(SQLITE_IOERR);
  TIMER_START;
  while( amt>0 && (wrote = write(id->h, pBuf, amt))>0 ){
    amt -= wrote;
    pBuf = &((char*)pBuf)[wrote];
  }
  TIMER_END;
  TRACE4("WRITE   %-3d %7d %d\n", id->h, last_page, elapse);
  SEEK(0);
  if( amt>0 ){
    return SQLITE_FULL;
  }
  return SQLITE_OK;
}

/*
** Move the read/write pointer in a file.
*/
int sqlite3OsSeek(OsFile *id, off_t offset){
  SEEK(offset/1024 + 1);
  lseek(id->h, offset, SEEK_SET);
  return SQLITE_OK;
}

/*
** Make sure all writes to a particular file are committed to disk.
**
** Under Unix, also make sure that the directory entry for the file
** has been created by fsync-ing the directory that contains the file.
** If we do not do this and we encounter a power failure, the directory
** entry for the journal might not exist after we reboot.  The next
** SQLite to access the file will not know that the journal exists (because
** the directory entry for the journal was never created) and the transaction
** will not roll back - possibly leading to database corruption.
*/
int sqlite3OsSync(OsFile *id){
  SimulateIOError(SQLITE_IOERR);
  TRACE2("SYNC    %-3d\n", id->h);
  if( fsync(id->h) ){
    return SQLITE_IOERR;
  }else{
    if( id->dirfd>=0 ){
      TRACE2("DIRSYNC %-3d\n", id->dirfd);
      fsync(id->dirfd);
      close(id->dirfd);  /* Only need to sync once, so close the directory */
      id->dirfd = -1;    /* when we are done. */
    }
    return SQLITE_OK;
  }
}

/*
** Truncate an open file to a specified size
*/
int sqlite3OsTruncate(OsFile *id, off_t nByte){
  SimulateIOError(SQLITE_IOERR);
  return ftruncate(id->h, nByte)==0 ? SQLITE_OK : SQLITE_IOERR;
}

/*
** Determine the current size of a file in bytes
*/
int sqlite3OsFileSize(OsFile *id, off_t *pSize){
  struct stat buf;
  SimulateIOError(SQLITE_IOERR);
  if( fstat(id->h, &buf)!=0 ){
    return SQLITE_IOERR;
  }
  *pSize = buf.st_size;
  return SQLITE_OK;
}

/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, return
** non-zero.  If the file is unlocked or holds only SHARED locks, then
** return zero.
*/
int sqlite3OsCheckReservedLock(OsFile *id){
  int r = 0;

  sqlite3OsEnterMutex(); /* Needed because id->pLock is shared across threads */

  /* Check if a thread in this process holds such a lock */
  if( id->pLock->locktype>SHARED_LOCK ){
    r = 1;
  }

  /* Otherwise see if some other process holds it.
  */
  if( !r ){
    struct flock lock;
    lock.l_whence = SEEK_SET;
    lock.l_start = RESERVED_BYTE;
    lock.l_len = 1;
    lock.l_type = F_WRLCK;
    fcntl(id->h, F_GETLK, &lock);
    if( lock.l_type!=F_UNLCK ){
      r = 1;
    }
  }
  
  sqlite3OsLeaveMutex();
  TRACE3("TEST WR-LOCK %d %d\n", id->h, r);

  return r;
}

/*
** Lock the file with the lock specified by parameter locktype - one
** of the following:

**
**    UNLOCKED -> SHARED
**    SHARED -> RESERVED
**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()


** routine to lower a locking level.
*/
int sqlite3OsLock(OsFile *id, int locktype){
  int rc = SQLITE_OK;
  struct lockInfo *pLock = id->pLock;
  struct flock lock;
  int s;

  TRACE6("LOCK %d %d was %d(%d,%d)\n",
          id->h, locktype, id->locktype, pLock->locktype, pLock->cnt);

  /* If there is already a lock of this type or more restrictive on the
  ** OsFile, do nothing. Don't use the end_lock: exit path, as
  ** sqlite3OsEnterMutex() hasn't been called yet.
  */
  if( id->locktype>=locktype ){
    return SQLITE_OK;

    assert( pLock->locktype==0 );
  
    /* Temporarily grab a PENDING lock.  This prevents new SHARED locks from
    ** being formed if a PENDING lock is already held.
    */
    lock.l_type = F_RDLCK;
    lock.l_start = PENDING_BYTE;
    s = fcntl(id->h, F_SETLK, &lock);
    if( s ){
      rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY;
      goto end_lock;
    }

    /* Now get the read-lock */
    lock.l_start = SHARED_FIRST;
    lock.l_len = SHARED_SIZE;
    s = fcntl(id->h, F_SETLK, &lock);

    /* Drop the temporary PENDING lock */
    lock.l_start = PENDING_BYTE;
    lock.l_len = 1L;
    lock.l_type = F_UNLCK;
    fcntl(id->h, F_SETLK, &lock);
    if( s ){
      rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY;
    }else{
      id->locktype = SHARED_LOCK;
      id->pOpen->nLock++;
      pLock->cnt = 1;
    }

      case EXCLUSIVE_LOCK:
        lock.l_start = SHARED_FIRST;
        lock.l_len = SHARED_SIZE;
        break;
      default:
        assert(0);
    }
    s = fcntl(id->h, F_SETLK, &lock);
    if( s ){
      rc = (errno==EINVAL) ? SQLITE_NOLFS : SQLITE_BUSY;
    }
  }
  
  if( rc==SQLITE_OK ){
    id->locktype = locktype;
    pLock->locktype = locktype;
  }

end_lock:
  sqlite3OsLeaveMutex();
  TRACE4("LOCK %d %d %s\n", id->h, locktype, rc==SQLITE_OK ? "ok" : "failed");
  return rc;
}

/*
** Lower the locking level on file descriptor id to locktype.  locktype
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
**
** It is not possible for this routine to fail.

*/
int sqlite3OsUnlock(OsFile *id, int locktype){
  struct lockInfo *pLock;
  struct flock lock;

  TRACE6("UNLOCK %d %d was %d(%d,%d)\n",
          id->h, locktype, id->locktype, id->pLock->locktype, id->pLock->cnt);

  assert( locktype<=SHARED_LOCK );
  if( id->locktype<=locktype ){
    return SQLITE_OK;
  }
  sqlite3OsEnterMutex();
  pLock = id->pLock;
  assert( pLock->cnt!=0 );
  if( id->locktype>SHARED_LOCK ){


    assert( pLock->locktype==id->locktype );

    lock.l_type = F_UNLCK;
    lock.l_whence = SEEK_SET;
    lock.l_start = PENDING_BYTE;
    lock.l_len = 2L;  assert( PENDING_BYTE+1==RESERVED_BYTE );
    fcntl(id->h, F_SETLK, &lock);





    pLock->locktype = SHARED_LOCK;
  }
  if( locktype==NO_LOCK ){
    struct openCnt *pOpen;

    /* Decrement the shared lock counter.  Release the lock using an
    ** OS call only when all threads in this same process have released
    ** the lock.
    */
    pLock->cnt--;
    if( pLock->cnt==0 ){
      lock.l_type = F_UNLCK;
      lock.l_whence = SEEK_SET;
      lock.l_start = lock.l_len = 0L;
      fcntl(id->h, F_SETLK, &lock);
      pLock->locktype = NO_LOCK;
    }

    /* Decrement the count of locks against this same file.  When the
    ** count reaches zero, close any other file descriptors whose close
    ** was deferred because of outstanding locks.
    */
    pOpen = id->pOpen;
    pOpen->nLock--;
    assert( pOpen->nLock>=0 );
    if( pOpen->nLock==0 && pOpen->nPending>0 ){
      int i;
      for(i=0; i<pOpen->nPending; i++){
        close(pOpen->aPending[i]);
      }
      sqliteFree(pOpen->aPending);
      pOpen->nPending = 0;
      pOpen->aPending = 0;
    }
  }
  sqlite3OsLeaveMutex();
  id->locktype = locktype;
  return SQLITE_OK;
}

/*
** Get information to seed the random number generator.  The seed
** is written into the buffer zBuf[256].  The calling function must
** supply a sufficiently large buffer.
*/

** OsFile.locktype takes one of the values SHARED_LOCK, RESERVED_LOCK,
** PENDING_LOCK or EXCLUSIVE_LOCK.
*/
typedef struct OsFile OsFile;
struct OsFile {
  struct openCnt *pOpen;    /* Info about all open fd's on this inode */
  struct lockInfo *pLock;   /* Info about locks on this inode */
  int h;                    /* The file descriptor */
  int locktype;             /* The type of lock held on this fd */
  int dirfd;                /* File descriptor for the directory */
};

/*
** Maximum number of characters in a temporary file name
*/

}

/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, return
** non-zero, otherwise zero.
*/
int sqlite3OsCheckReservedLock(OsFile *id){
  int rc;
  if( id->locktype>=RESERVED_LOCK ){
    rc = 1;
    TRACE3("TEST WR-LOCK %d %d (local)\n", id->h, rc);
  }else{
    rc = LockFile(id->h, RESERVED_BYTE, 0, 1, 0);
    if( rc ){
      UnlockFile(id->h, RESERVED_BYTE, 0, 1, 0);
    }
    rc = !rc;
    TRACE3("TEST WR-LOCK %d %d (remote)\n", id->h, rc);
  }
  return rc;
}

/*
** Lower the locking level on file descriptor id to locktype.  locktype
** must be either NO_LOCK or SHARED_LOCK.
**
** If the locking level of the file descriptor is already at or below
** the requested locking level, this routine is a no-op.
**
** It is not possible for this routine to fail.

*/
int sqlite3OsUnlock(OsFile *id, int locktype){
  int rc, type;
  assert( locktype<=SHARED_LOCK );
  TRACE4("UNLOCK %d to %d was %d(%d)\n", id->h, locktype,
          id->locktype, id->sharedLockByte);
  type = id->locktype;
  if( type>=EXCLUSIVE_LOCK ){
    UnlockFile(id->h, SHARED_FIRST, 0, SHARED_SIZE, 0);
  }
  if( type>=RESERVED_LOCK ){
    UnlockFile(id->h, RESERVED_BYTE, 0, 1, 0);
  }
  if( locktype==NO_LOCK && type>=SHARED_LOCK && type<EXCLUSIVE_LOCK ){
    unlockReadLock(id);
  }
  if( type>=PENDING_LOCK ){
    UnlockFile(id->h, PENDING_BYTE, 0, 1, 0);
  }
  id->locktype = locktype;
  return SQLITE_OK;
}

/*
** Get information to seed the random number generator.  The seed
** is written into the buffer zBuf[256].  The calling function must
** supply a sufficiently large buffer.

** The pager is used to access a database disk file.  It implements
** atomic commit and rollback through the use of a journal file that
** is separate from the database file.  The pager also implements file
** locking to prevent two processes from writing the same database
** file simultaneously, or one process from reading the database while
** another is writing.
**
** @(#) $Id: pager.c,v 1.115 2004/06/09 17:37:28 drh Exp $
*/
#include "os.h"         /* Must be first to enable large file support */
#include "sqliteInt.h"
#include "pager.h"
#include <assert.h>
#include <string.h>

#endif


/*
** The page cache as a whole is always in one of the following
** states:
**
**   PAGER_UNLOCK        The page cache is not currently reading or 
**                       writing the database file.  There is no
**                       data held in memory.  This is the initial
**                       state.
**
**   PAGER_SHARED        The page cache is reading the database.
**                       Writing is not permitted.  There can be
**                       multiple readers accessing the same database
**                       file at the same time.
**
**   PAGER_RESERVED      Writing is permitted to the page cache only.
**                       The original database file has not been modified.
**                       Other processes may still be reading the on-disk
**                       database file.
**
**   PAGER_EXCLUSIVE     The page cache is writing the database.
**                       Access is exclusive.  No other processes or
**                       threads can be reading or writing while one
**                       process is writing.
**
** The page cache comes up in PAGER_UNLOCK.  The first time a
** sqlite_page_get() occurs, the state transitions to PAGER_SHARED.
** After all pages have been released using sqlite_page_unref(),
** the state transitions back to PAGER_UNLOCK.  The first time
** that sqlite_page_write() is called, the state transitions to
** PAGER_RESERVED.  (Note that sqlite_page_write() can only be
** called on an outstanding page which means that the pager must
** be in PAGER_SHARED before it transitions to PAGER_RESERVED.)
** The sqlite_page_rollback() and sqlite_page_commit() functions 
** transition the state from PAGER_RESERVED to PAGER_EXCLUSIVE to
** PAGER_SHARED.
*/
#define PAGER_UNLOCK      0
#define PAGER_SHARED      1
#define PAGER_RESERVED    2
#define PAGER_EXCLUSIVE   3


/*
** Each in-memory image of a page begins with the following header.
** This header is only visible to this pager module.  The client
** code that calls pager sees only the data that follows the header.
**

  int stmtSize;               /* Size of database (in pages) at stmt_begin() */
  off_t stmtJSize;            /* Size of journal at stmt_begin() */
  int nRec;                   /* Number of pages written to the journal */
  u32 cksumInit;              /* Quasi-random value added to every checksum */
  int stmtNRec;               /* Number of records in stmt subjournal */
  int nExtra;                 /* Add this many bytes to each in-memory page */
  void (*xDestructor)(void*,int); /* Call this routine when freeing pages */
  void (*xReiniter)(void*,int);   /* Call this routine when reloading pages */
  int pageSize;               /* Number of bytes in a page */
  int nPage;                  /* Total number of in-memory pages */
  int nRef;                   /* Number of in-memory pages with PgHdr.nRef>0 */
  int mxPage;                 /* Maximum number of pages to hold in cache */
  int nHit, nMiss, nOvfl;     /* Cache hits, missing, and LRU overflows */
  void (*xCodec)(void*,void*,Pgno,int); /* Routine for en/decoding data */
  void *pCodecArg;            /* First argument to xCodec() */
  u8 journalOpen;             /* True if journal file descriptors is valid */
  u8 journalStarted;          /* True if header of journal is synced */
  u8 useJournal;              /* Use a rollback journal on this file */
  u8 stmtOpen;                /* True if the statement subjournal is open */
  u8 stmtInUse;               /* True we are in a statement subtransaction */
  u8 stmtAutoopen;            /* Open stmt journal when main journal is opened*/
  u8 noSync;                  /* Do not sync the journal if true */
  u8 fullSync;                /* Do extra syncs of the journal for robustness */
  u8 state;                   /* PAGER_UNLOCK, _SHARED, _RESERVED, etc. */
  u8 errMask;                 /* One of several kinds of errors */
  u8 tempFile;                /* zFilename is a temporary file */
  u8 readOnly;                /* True for a read-only database */
  u8 needSync;                /* True if an fsync() is needed on the journal */
  u8 dirtyCache;              /* True if cached pages have changed */
  u8 alwaysRollback;          /* Disable dont_rollback() for all pages */
  u8 memDb;                   /* True to inhibit all file I/O */
  u8 *aInJournal;             /* One bit for each page in the database file */
  u8 *aInStmt;                /* One bit for each page in the database */
  PgHdr *pFirst, *pLast;      /* List of free pages */
  PgHdr *pFirstSynced;        /* First free page with PgHdr.needSync==0 */
  PgHdr *pAll;                /* List of all pages */

  }
  pPager->pFirst = 0;
  pPager->pFirstSynced = 0;
  pPager->pLast = 0;
  pPager->pAll = 0;
  memset(pPager->aHash, 0, sizeof(pPager->aHash));
  pPager->nPage = 0;
  if( pPager->state>=PAGER_RESERVED ){
    sqlite3pager_rollback(pPager);
  }
  sqlite3OsUnlock(&pPager->fd, NO_LOCK);
  pPager->state = PAGER_UNLOCK;
  pPager->dbSize = -1;
  pPager->nRef = 0;
  assert( pPager->journalOpen==0 );
}

/*
** When this routine is called, the pager has the journal file open and
** a RESERVED or EXCLUSIVE lock on the database.  This routine releases
** the database lock and acquires a SHARED lock in its place.  The journal
** file is deleted and closed.
**
** TODO: Consider keeping the journal file open for temporary databases.
** This might give a performance improvement on windows where opening
** a file is an expensive operation.
*/
static int pager_unwritelock(Pager *pPager){

  PgHdr *pPg;
  if( pPager->state<PAGER_RESERVED ){
    return SQLITE_OK;
  }
  sqlite3pager_stmt_commit(pPager);
  if( pPager->stmtOpen ){
    sqlite3OsClose(&pPager->stfd);
    pPager->stmtOpen = 0;
  }
  if( pPager->journalOpen ){
    sqlite3OsClose(&pPager->jfd);
    pPager->journalOpen = 0;
    sqlite3OsDelete(pPager->zJournal);
    sqliteFree( pPager->aInJournal );
    pPager->aInJournal = 0;
    for(pPg=pPager->pAll; pPg; pPg=pPg->pNextAll){
      pPg->inJournal = 0;
      pPg->dirty = 0;
      pPg->needSync = 0;
    }
  }else{
    assert( pPager->dirtyCache==0 || pPager->useJournal==0 );
  }
  sqlite3OsUnlock(&pPager->fd, SHARED_LOCK);







  pPager->state = PAGER_SHARED;

  return SQLITE_OK;
}

/*
** Compute and return a checksum for the page of data.
**
** This is not a real checksum.  It is really just the sum of the 
** random initial value and the page number.  We considered do a checksum

    }
  }

  /* Playback the page.  Update the in-memory copy of the page
  ** at the same time, if there is one.
  */
  pPg = pager_lookup(pPager, pgRec.pgno);
  TRACE2("PLAYBACK page %d\n", pgRec.pgno);
  sqlite3OsSeek(&pPager->fd, (pgRec.pgno-1)*(off_t)SQLITE_PAGE_SIZE);
  rc = sqlite3OsWrite(&pPager->fd, pgRec.aData, SQLITE_PAGE_SIZE);
  if( pPg ){
    /* No page should ever be rolled back that is in use, except for page
    ** 1 which is held in use in order to keep the lock on the database
    ** active.
    */

    sqliteFree(zMasterJournal);
  }  
  if( master_open ){
    sqlite3OsClose(&master);
  }
  return rc;
}

/*
** Make every page in the cache agree with what is on disk.  In other words,
** reread the disk to reset the state of the cache.
**
** This routine is called after a rollback in which some of the dirty cache
** pages had never been written out to disk.  We need to roll back the
** cache content and the easiest way to do that is to reread the old content
** back from the disk.
*/
static int pager_reload_cache(Pager *pPager){
  PgHdr *pPg;
  int rc = SQLITE_OK;
  for(pPg=pPager->pAll; pPg; pPg=pPg->pNextAll){
    char zBuf[SQLITE_PAGE_SIZE];
    if( !pPg->dirty ) continue;
    if( (int)pPg->pgno <= pPager->origDbSize ){
      sqlite3OsSeek(&pPager->fd, SQLITE_PAGE_SIZE*(off_t)(pPg->pgno-1));
      rc = sqlite3OsRead(&pPager->fd, zBuf, SQLITE_PAGE_SIZE);
      TRACE2("REFETCH page %d\n", pPg->pgno);
      CODEC(pPager, zBuf, pPg->pgno, 2);
      if( rc ) break;
    }else{
      memset(zBuf, 0, SQLITE_PAGE_SIZE);
    }
    if( pPg->nRef==0 || memcmp(zBuf, PGHDR_TO_DATA(pPg), SQLITE_PAGE_SIZE) ){
      memcpy(PGHDR_TO_DATA(pPg), zBuf, SQLITE_PAGE_SIZE);
      if( pPager->xReiniter ){
        pPager->xReiniter(PGHDR_TO_DATA(pPg), pPager->pageSize);
      }else{
        memset(PGHDR_TO_EXTRA(pPg), 0, pPager->nExtra);
      }
    }
    pPg->needSync = 0;
    pPg->dirty = 0;
  }
  return rc;
}


/*
** Playback the journal and thus restore the database file to
** the state it was in before we started making changes.  
**
** The journal file format is as follows: 
**

  }

  /* Pages that have been written to the journal but never synced
  ** where not restored by the loop above.  We have to restore those
  ** pages by reading them back from the original database.
  */
  if( rc==SQLITE_OK ){








    pager_reload_cache(pPager);











  }

end_playback:
  if( zMaster ){
    /* If there was a master journal and this routine will return true,
    ** see if it is possible to delete the master journal. If errors 
    ** occur during this process, ignore them.

  pPager->nRef = 0;
  pPager->dbSize = memDb-1;
  pPager->pageSize = SQLITE_PAGE_SIZE;
  pPager->stmtSize = 0;
  pPager->stmtJSize = 0;
  pPager->nPage = 0;
  pPager->mxPage = mxPage>5 ? mxPage : 10;
  pPager->state = PAGER_UNLOCK;
  pPager->errMask = 0;
  pPager->tempFile = tempFile;
  pPager->memDb = memDb;
  pPager->readOnly = readOnly;
  pPager->needSync = 0;
  pPager->noSync = pPager->tempFile || !useJournal;
  pPager->pFirst = 0;

**
** The destructor is not called as a result sqlite3pager_close().  
** Destructors are only called by sqlite3pager_unref().
*/
void sqlite3pager_set_destructor(Pager *pPager, void (*xDesc)(void*,int)){
  pPager->xDestructor = xDesc;
}

/*
** Set the reinitializer for this pager.  If not NULL, the reinitializer
** is called when the content of a page in cache is restored to its original
** value as a result of a rollback.  The callback gives higher-level code
** an opportunity to restore the EXTRA section to agree with the restored
** page data.
*/
void sqlite3pager_set_reiniter(Pager *pPager, void (*xReinit)(void*,int)){
  pPager->xReiniter = xReinit;
}

/*
** Return the total number of pages in the disk file associated with
** pPager.
*/
int sqlite3pager_pagecount(Pager *pPager){
  off_t n;
  assert( pPager!=0 );
  if( pPager->dbSize>=0 ){
    return pPager->dbSize;
  }
  if( sqlite3OsFileSize(&pPager->fd, &n)!=SQLITE_OK ){
    pPager->errMask |= PAGER_ERR_DISK;
    return 0;
  }
  n /= SQLITE_PAGE_SIZE;
  if( pPager->state!=PAGER_UNLOCK ){
    pPager->dbSize = n;
  }
  return n;
}

/*
** Forward declaration

** 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.
*/
int sqlite3pager_close(Pager *pPager){
  PgHdr *pPg, *pNext;
  switch( pPager->state ){
    case PAGER_RESERVED:
    case PAGER_EXCLUSIVE: {
      sqlite3pager_rollback(pPager);
      if( !pPager->memDb ){
        sqlite3OsUnlock(&pPager->fd, NO_LOCK);
      }
      assert( pPager->journalOpen==0 );
      break;
    }
    case PAGER_SHARED: {
      if( !pPager->memDb ){
        sqlite3OsUnlock(&pPager->fd, NO_LOCK);
      }
      break;
    }
    default: {
      /* Do nothing */
      break;
    }

        assert( pPager->nRec*pgSz+hdrSz==jSz );
      }
#endif
      if( journal_format>=3 ){
        /* Write the nRec value into the journal file header */
        off_t szJ;
        if( pPager->fullSync ){
          TRACE2("SYNC journal of %d\n", pPager->fd.h);
          rc = sqlite3OsSync(&pPager->jfd);
          if( rc!=0 ) return rc;
        }
        sqlite3OsSeek(&pPager->jfd, sizeof(aJournalMagic1));
        rc = write32bits(&pPager->jfd, pPager->nRec);
        if( rc ) return rc;

        /* Write the name of the master journal file if one is specified */
        if( zMaster ){
          assert( strlen(zMaster)<pPager->nMaster );
          rc = sqlite3OsSeek(&pPager->jfd, sizeof(aJournalMagic3) + 3*4);
          if( rc ) return rc;
          rc = sqlite3OsWrite(&pPager->jfd, zMaster, strlen(zMaster)+1);
          if( rc ) return rc;
        }

        szJ = JOURNAL_HDR_SZ(pPager, journal_format) +
                 pPager->nRec*JOURNAL_PG_SZ(journal_format);
        sqlite3OsSeek(&pPager->jfd, szJ);
      }
      TRACE2("SYNC journal of %d\n", pPager->fd.h);
      rc = sqlite3OsSync(&pPager->jfd);
      if( rc!=0 ) return rc;
      pPager->journalStarted = 1;
    }
    pPager->needSync = 0;

    /* Erase the needSync flag from every page.

  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.
  */
  do {
    rc = sqlite3OsLock(&pPager->fd, EXCLUSIVE_LOCK);
  }while( rc==SQLITE_BUSY && 
      pPager->pBusyHandler && 
      pPager->pBusyHandler->xFunc && 
      pPager->pBusyHandler->xFunc(pPager->pBusyHandler->pArg, "", busy++)
  );
  if( rc!=SQLITE_OK ){
    return rc;
  }
  pPager->state = PAGER_EXCLUSIVE;

  while( pList ){
    assert( pList->dirty );
    sqlite3OsSeek(&pPager->fd, (pList->pgno-1)*(off_t)SQLITE_PAGE_SIZE);
    CODEC(pPager, PGHDR_TO_DATA(pList), pList->pgno, 6);
    TRACE2("STORE page %d\n", pList->pgno);
    rc = sqlite3OsWrite(&pPager->fd, PGHDR_TO_DATA(pList), SQLITE_PAGE_SIZE);
    CODEC(pPager, PGHDR_TO_DATA(pList), pList->pgno, 0);
    if( rc ) return rc;
    pList->dirty = 0;
    pList = pList->pDirty;
  }
  return SQLITE_OK;

        pPager->pBusyHandler && 
        pPager->pBusyHandler->xFunc && 
        pPager->pBusyHandler->xFunc(pPager->pBusyHandler->pArg, "", busy++)
    );
    if( rc!=SQLITE_OK ){
      return rc;
    }
    pPager->state = PAGER_SHARED;

    /* 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.
    */
    if( pPager->useJournal && 
        sqlite3OsFileExists(pPager->zJournal) &&
        !sqlite3OsCheckReservedLock(&pPager->fd) 
    ){
       int rc;

       /* Get an EXCLUSIVE lock on the database file. */
       rc = sqlite3OsLock(&pPager->fd, EXCLUSIVE_LOCK);
       if( rc!=SQLITE_OK ){
         sqlite3OsUnlock(&pPager->fd, NO_LOCK);

         pPager->state = PAGER_UNLOCK;

         return 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.
       */
       rc = sqlite3OsOpenReadOnly(pPager->zJournal, &pPager->jfd);
       if( rc!=SQLITE_OK ){
         sqlite3OsUnlock(&pPager->fd, NO_LOCK);
         pPager->state = PAGER_UNLOCK;
         return SQLITE_BUSY;
       }
       pPager->journalOpen = 1;
       pPager->journalStarted = 0;

       /* Playback and delete the journal.  Drop the database write
       ** lock and reacquire the read lock.
       */
       rc = pager_playback(pPager, 0);
       if( rc!=SQLITE_OK ){
         return rc;
       }
    }
    pPg = 0;
  }else{
    /* Search for page in cache */
    pPg = pager_lookup(pPager, pgno);
    if( pPager->memDb && pPager->state==PAGER_UNLOCK ){
      pPager->state = PAGER_SHARED;
    }
  }
  if( pPg==0 ){
    /* The requested page is not in the page cache. */
    int h;
    pPager->nMiss++;
    if( pPager->nPage<pPager->mxPage || pPager->pFirst==0 || pPager->memDb ){

    if( pPager->dbSize<(int)pgno ){
      memset(PGHDR_TO_DATA(pPg), 0, SQLITE_PAGE_SIZE);
    }else{
      int rc;
      assert( pPager->memDb==0 );
      sqlite3OsSeek(&pPager->fd, (pgno-1)*(off_t)SQLITE_PAGE_SIZE);
      rc = sqlite3OsRead(&pPager->fd, PGHDR_TO_DATA(pPg), SQLITE_PAGE_SIZE);
      TRACE2("FETCH page %d\n", pPg->pgno);
      CODEC(pPager, PGHDR_TO_DATA(pPg), pPg->pgno, 3);
      if( rc!=SQLITE_OK ){
        off_t fileSize;
        if( sqlite3OsFileSize(&pPager->fd,&fileSize)!=SQLITE_OK
               || fileSize>=pgno*SQLITE_PAGE_SIZE ){
          sqlite3pager_unref(PGHDR_TO_DATA(pPg));
          return rc;

      pager_reset(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){
  int rc;
  assert( pPager->state>=PAGER_RESERVED );
  assert( pPager->journalOpen==0 );
  assert( pPager->useJournal );
  sqlite3pager_pagecount(pPager);
  pPager->aInJournal = sqliteMalloc( pPager->dbSize/8 + 1 );
  if( pPager->aInJournal==0 ){
    sqlite3OsUnlock(&pPager->fd, SHARED_LOCK);
    pPager->state = PAGER_SHARED;
    return SQLITE_NOMEM;
  }
  rc = sqlite3OsOpenExclusive(pPager->zJournal, &pPager->jfd,pPager->tempFile);
  if( rc!=SQLITE_OK ){
    sqliteFree(pPager->aInJournal);
    pPager->aInJournal = 0;
    sqlite3OsUnlock(&pPager->fd, SHARED_LOCK);
    pPager->state = PAGER_SHARED;
    return SQLITE_CANTOPEN;
  }
  sqlite3OsOpenDirectory(pPager->zDirectory, &pPager->jfd);
  pPager->journalOpen = 1;
  pPager->journalStarted = 0;
  pPager->needSync = 0;
  pPager->alwaysRollback = 0;

** 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.
*/
int sqlite3pager_begin(void *pData, int nMaster){
  PgHdr *pPg = DATA_TO_PGHDR(pData);
  Pager *pPager = pPg->pPager;
  int rc = SQLITE_OK;
  assert( pPg->nRef>0 );
  assert( nMaster>=0 );
  assert( pPager->state!=PAGER_UNLOCK );
  if( pPager->state==PAGER_SHARED ){
    assert( pPager->aInJournal==0 );
    if( pPager->memDb ){
      pPager->state = PAGER_EXCLUSIVE;
      pPager->origDbSize = pPager->dbSize;
    }else{
      int busy = 1;
      do {





        rc = sqlite3OsLock(&pPager->fd, RESERVED_LOCK);
      }while( rc==SQLITE_BUSY && 
          pPager->pBusyHandler && 
          pPager->pBusyHandler->xFunc && 
          pPager->pBusyHandler->xFunc(pPager->pBusyHandler->pArg, "", busy++)
      );
      if( rc!=SQLITE_OK ){
        return rc;
      }
      pPager->nMaster = nMaster;
      pPager->state = PAGER_RESERVED;
      pPager->dirtyCache = 0;
      TRACE3("TRANSACTION %d nMaster=%d\n", pPager->fd.h, nMaster);
      if( pPager->useJournal && !pPager->tempFile ){
        rc = pager_open_journal(pPager);
      }
    }
  }
  return rc;
}

/*
** 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

  }

  /* Mark the page as dirty.  If the page has already been written
  ** to the journal then we can return right away.
  */
  pPg->dirty = 1;
  if( pPg->inJournal && (pPg->inStmt || pPager->stmtInUse==0) ){
    pPager->dirtyCache = 1;
    return SQLITE_OK;
  }

  /* 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 = sqlite3pager_begin(pData, 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 || pPager->memDb) ){
    if( (int)pPg->pgno <= pPager->origDbSize ){
      int szPg;
      u32 saved;
      if( pPager->memDb ){
        PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
        TRACE2("JOURNAL page %d\n", pPg->pgno);
        assert( pHist->pOrig==0 );
        pHist->pOrig = sqliteMallocRaw( pPager->pageSize );
        if( pHist->pOrig ){
          memcpy(pHist->pOrig, PGHDR_TO_DATA(pPg), pPager->pageSize);
        }
        pPg->inJournal = 1;
      }else{
        if( journal_format>=JOURNAL_FORMAT_3 ){
          u32 cksum = pager_cksum(pPager, pPg->pgno, pData);
          saved = *(u32*)PGHDR_TO_EXTRA(pPg);
          store32bits(cksum, pPg, SQLITE_PAGE_SIZE);
          szPg = SQLITE_PAGE_SIZE+8;
        }else{
          szPg = SQLITE_PAGE_SIZE+4;
        }
        store32bits(pPg->pgno, pPg, -4);
        CODEC(pPager, pData, pPg->pgno, 7);
        rc = sqlite3OsWrite(&pPager->jfd, &((char*)pData)[-4], szPg);
        TRACE3("JOURNAL page %d needSync=%d\n", pPg->pgno, pPg->needSync);
        CODEC(pPager, pData, pPg->pgno, 0);
        if( journal_format>=JOURNAL_FORMAT_3 ){
          *(u32*)PGHDR_TO_EXTRA(pPg) = saved;
        }
        if( rc!=SQLITE_OK ){
          sqlite3pager_rollback(pPager);
          pPager->errMask |= PAGER_ERR_FULL;

        if( pPager->stmtInUse ){
          pPager->aInStmt[pPg->pgno/8] |= 1<<(pPg->pgno&7);
          page_add_to_stmt_list(pPg);
        }
      }
    }else{
      pPg->needSync = !pPager->journalStarted && !pPager->noSync;
      TRACE3("APPEND page %d needSync=%d\n", pPg->pgno, pPg->needSync);
    }
    if( pPg->needSync ){
      pPager->needSync = 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 always uses the simplier format 2 that lacks
  ** checksums.  The header is also omitted from the statement journal.
  */
  if( pPager->stmtInUse && !pPg->inStmt && (int)pPg->pgno<=pPager->stmtSize ){
    assert( pPg->inJournal || (int)pPg->pgno>pPager->origDbSize );
    if( pPager->memDb ){
      PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
      assert( pHist->pStmt==0 );
      pHist->pStmt = sqliteMallocRaw( pPager->pageSize );
      if( pHist->pStmt ){
        memcpy(pHist->pStmt, PGHDR_TO_DATA(pPg), pPager->pageSize);
      }
      TRACE2("STMT-JOURNAL page %d\n", pPg->pgno);
    }else{
      store32bits(pPg->pgno, pPg, -4);
      CODEC(pPager, pData, pPg->pgno, 7);
      rc = sqlite3OsWrite(&pPager->stfd, ((char*)pData)-4, SQLITE_PAGE_SIZE+4);
      TRACE2("STMT-JOURNAL page %d\n", pPg->pgno);
      CODEC(pPager, pData, pPg->pgno, 0);
      if( rc!=SQLITE_OK ){
        sqlite3pager_rollback(pPager);
        pPager->errMask |= PAGER_ERR_FULL;
        return rc;
      }
      pPager->stmtNRec++;

      ** 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{
      TRACE3("DONT_WRITE page %d of %d\n", pgno, pPager->fd.h);
      pPg->dirty = 0;
    }
  }
}

/*
** 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.
*/
void sqlite3pager_dont_rollback(void *pData){
  PgHdr *pPg = DATA_TO_PGHDR(pData);
  Pager *pPager = pPg->pPager;

  if( pPager->state!=PAGER_EXCLUSIVE || pPager->journalOpen==0 ) return;
  if( pPg->alwaysRollback || pPager->alwaysRollback || pPager->memDb ) return;
  if( !pPg->inJournal && (int)pPg->pgno <= pPager->origDbSize ){
    assert( pPager->aInJournal!=0 );
    pPager->aInJournal[pPg->pgno/8] |= 1<<(pPg->pgno&7);
    pPg->inJournal = 1;
    if( pPager->stmtInUse ){
      pPager->aInStmt[pPg->pgno/8] |= 1<<(pPg->pgno&7);
      page_add_to_stmt_list(pPg);
    }
    TRACE3("DONT_ROLLBACK page %d of %d\n", pPg->pgno, pPager->fd.h);
  }
  if( pPager->stmtInUse && !pPg->inStmt && (int)pPg->pgno<=pPager->stmtSize ){
    assert( pPg->inJournal || (int)pPg->pgno>pPager->origDbSize );
    assert( pPager->aInStmt!=0 );
    pPager->aInStmt[pPg->pgno/8] |= 1<<(pPg->pgno&7);
    page_add_to_stmt_list(pPg);
  }

    }
    return rc;
  }
  if( pPager->errMask!=0 ){
    rc = pager_errcode(pPager);
    return rc;
  }
  if( pPager->state<PAGER_RESERVED ){
    return SQLITE_ERROR;
  }
  TRACE2("COMMIT %d\n", pPager->fd.h);
  if( pPager->memDb ){
    pPg = pager_get_all_dirty_pages(pPager);
    while( pPg ){
      clearHistory(PGHDR_TO_HIST(pPg, pPager));
      pPg->dirty = 0;
      pPg->inJournal = 0;
      pPg->inStmt = 0;
      pPg->pPrevStmt = pPg->pNextStmt = 0;
      pPg = pPg->pDirty;
    }
    pPager->pStmt = 0;
    pPager->state = PAGER_SHARED;
    return SQLITE_OK;
  }

  if( pPager->dirtyCache==0 ){
    /* Exit early (without doing the time-consuming sqlite3OsSync() calls)
    ** if there have been no changes to the database file. */
    assert( pPager->needSync==0 );
    rc = pager_unwritelock(pPager);
    pPager->dbSize = -1;
    return rc;
  }
  assert( pPager->journalOpen );
#if 0
  rc = syncJournal(pPager, 0);
  if( rc!=SQLITE_OK ){
    goto commit_abort;
  }
  pPg = pager_get_all_dirty_pages(pPager);
  if( pPg ){
    rc = pager_write_pagelist(pPg);
    if( rc || (!pPager->noSync && sqlite3OsSync(&pPager->fd)!=SQLITE_OK) ){
      goto commit_abort;
    }
  }
#endif
  rc = sqlite3pager_sync(pPager, 0);
  if( rc!=SQLITE_OK ){
    goto commit_abort;
  }
  rc = pager_unwritelock(pPager);
  pPager->dbSize = -1;
  return rc;

  /* Jump here if anything goes wrong during the commit process.
  */
commit_abort:
  rc = sqlite3pager_rollback(pPager);
  if( rc==SQLITE_OK ){
    rc = SQLITE_FULL;
  }
  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 (SQLITE_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 sqlite3pager_rollback(Pager *pPager){
  int rc;
  TRACE2("ROLLBACK %d\n", pPager->fd.h);
  if( pPager->memDb ){
    PgHdr *p;
    for(p=pPager->pAll; p; p=p->pNextAll){
      PgHistory *pHist;
      if( !p->dirty ) continue;
      pHist = PGHDR_TO_HIST(p, pPager);
      if( pHist->pOrig ){

      p->inStmt = 0;
      p->pPrevStmt = p->pNextStmt = 0;
    }
    pPager->pStmt = 0;
    pPager->dbSize = pPager->origDbSize;
    memoryTruncate(pPager);
    pPager->stmtInUse = 0;
    pPager->state = PAGER_SHARED;
    return SQLITE_OK;
  }

  if( !pPager->dirtyCache || !pPager->journalOpen ){
    rc = pager_unwritelock(pPager);
    pPager->dbSize = -1;
    return rc;
  }

  if( pPager->errMask!=0 && pPager->errMask!=PAGER_ERR_FULL ){
    if( pPager->state>=PAGER_EXCLUSIVE ){
      pager_playback(pPager, 1);
    }
    return pager_errcode(pPager);
  }
  if( pPager->state==PAGER_RESERVED ){
    int rc2;
    rc = pager_reload_cache(pPager);
    rc2 = pager_unwritelock(pPager);
    if( rc==SQLITE_OK ){
      rc = rc2;
    }
  }else{
    rc = pager_playback(pPager, 1);
  }
  if( rc!=SQLITE_OK ){
    rc = SQLITE_CORRUPT;
    pPager->errMask |= PAGER_ERR_CORRUPT;
  }
  pPager->dbSize = -1;
  return rc;
}

** open.  A new statement journal is created that can be used to rollback
** changes of a single SQL command within a larger transaction.
*/
int sqlite3pager_stmt_begin(Pager *pPager){
  int rc;
  char zTemp[SQLITE_TEMPNAME_SIZE];
  assert( !pPager->stmtInUse );
  TRACE2("STMT-BEGIN %d\n", pPager->fd.h);
  if( pPager->memDb ){
    pPager->stmtInUse = 1;
    pPager->stmtSize = pPager->dbSize;
    return SQLITE_OK;
  }
  if( !pPager->journalOpen ){
    pPager->stmtAutoopen = 1;

/*
** Commit a statement.
*/
int sqlite3pager_stmt_commit(Pager *pPager){
  if( pPager->stmtInUse ){
    PgHdr *pPg, *pNext;
    TRACE2("STMT-COMMIT %d\n", pPager->fd.h);
    if( !pPager->memDb ){
      sqlite3OsSeek(&pPager->stfd, 0);
      /* sqlite3OsTruncate(&pPager->stfd, 0); */
      sqliteFree( pPager->aInStmt );
      pPager->aInStmt = 0;
    }
    for(pPg=pPager->pStmt; pPg; pPg=pNext){

/*
** Rollback a statement.
*/
int sqlite3pager_stmt_rollback(Pager *pPager){
  int rc;
  if( pPager->stmtInUse ){
    TRACE2("STMT-ROLLBACK %d\n", pPager->fd.h);
    if( pPager->memDb ){
      PgHdr *pPg;
      for(pPg=pPager->pStmt; pPg; pPg=pPg->pNextStmt){
        PgHistory *pHist = PGHDR_TO_HIST(pPg, pPager);
        if( pHist->pStmt ){
          memcpy(PGHDR_TO_DATA(pPg), pHist->pStmt, pPager->pageSize);
          sqliteFree(pHist->pStmt);

*/
int sqlite3pager_sync(Pager *pPager, const char *zMaster){
  int rc = SQLITE_OK;

  /* If this is an in-memory db, or no pages have been written to, this
  ** function is a no-op.
  */
  if( !pPager->memDb && pPager->dirtyCache ){
    PgHdr *pPg;
    assert( pPager->journalOpen );

    /* Sync the journal file */
    rc = syncJournal(pPager, zMaster);
    if( rc!=SQLITE_OK ) goto sync_exit;

**    May you share freely, never taking more than you give.
**
*************************************************************************
** This header file defines the interface that the sqlite page cache
** subsystem.  The page cache subsystem reads and writes a file a page
** at a time and provides a journal for rollback.
**
** @(#) $Id: pager.h,v 1.33 2004/06/09 17:37:28 drh Exp $
*/

/*
** The size of a page.
**
** You can change this value to another (reasonable) value you want.
** It need not be a power of two, though the interface to the disk

** See source code comments for a detailed description of the following
** routines:
*/
int sqlite3pager_open(Pager **ppPager, const char *zFilename,
                     int nPage, int nExtra, int useJournal,
                     void *pBusyHandler);
void sqlite3pager_set_destructor(Pager*, void(*)(void*,int));
void sqlite3pager_set_reiniter(Pager*, void(*)(void*,int));
void sqlite3pager_set_cachesize(Pager*, int);
int sqlite3pager_close(Pager *pPager);
int sqlite3pager_get(Pager *pPager, Pgno pgno, void **ppPage);
void *sqlite3pager_lookup(Pager *pPager, Pgno pgno);
int sqlite3pager_ref(void*);
int sqlite3pager_unref(void*);
Pgno sqlite3pager_pagenumber(void*);

**    May you share freely, never taking more than you give.
**
*************************************************************************
** Code for testing the printf() interface to SQLite.  This code
** is not included in the SQLite library.  It is used for automated
** testing of the SQLite library.
**
** $Id: test1.c,v 1.74 2004/06/09 17:37:28 drh Exp $
*/
#include "sqliteInt.h"
#include "tcl.h"
#include "os.h"
#include <stdlib.h>
#include <string.h>

       Tcl_GetString(objv[0]), " filehandle", 0);
    return TCL_ERROR;
  }

  if( getFilePointer(interp, Tcl_GetString(objv[1]), &pFile) ){
    return TCL_ERROR;
  }
  rc = sqlite3OsUnlock(pFile, NO_LOCK);
  if( rc!=SQLITE_OK ){
    Tcl_SetResult(interp, (char *)errorName(rc), TCL_STATIC);
    return TCL_ERROR;
  }
  return TCL_OK;
}

    }
  }

  /* The simple case - no more than one database file (not counting the TEMP
  ** database) has a transaction active.   There is no need for the
  ** master-journal.
  */
  if( nTrans<=100 ){  /**** FIX ME ****/
    for(i=0; rc==SQLITE_OK && i<db->nDb; i++){ 
      Btree *pBt = db->aDb[i].pBt;
      if( pBt ){
        rc = sqlite3BtreeSync(pBt, 0);
      }
    }

  else{
    char *zMaster = 0;   /* File-name for the master journal */
    char const *zMainFile = sqlite3BtreeGetFilename(db->aDb[0].pBt);
    OsFile master;

    /* Select a master journal file name */
    do {
      u32 random;

      sqliteFree(zMaster);

      sqlite3Randomness(sizeof(random), &random);
      zMaster = sqlite3MPrintf("%s-mj%08X", zMainFile, random);
      if( !zMaster ){
        return SQLITE_NOMEM;
      }
    }while( sqlite3OsFileExists(zMaster) );

    /* Open the master journal. */
    rc = sqlite3OsOpenExclusive(zMaster, &master, 0);

#    May you share freely, never taking more than you give.
#
#***********************************************************************
# This file implements regression tests for SQLite library.  The
# focus of this script is testing the ATTACH and DETACH commands
# and related functionality.
#
# $Id: attach2.test,v 1.14 2004/06/09 17:37:29 drh Exp $
#


set testdir [file dirname $argv0]
source $testdir/tester.tcl


# Ticket #354
#

  set rc
} {0}

db close
for {set i 2} {$i<=15} {incr i} {
  catch {db$i close}
}

# A procedure to verify the status of locks on a database.
#
proc lock_status {testnum db expected_result} {
  do_test attach2-$testnum [subst {
    execsql {PRAGMA lock_status} $db
  }] $expected_result
}
set sqlite_os_trace 0

# Tests attach2-4.* test that read-locks work correctly with attached
# databases.
do_test attach2-4.1 {
  sqlite db test.db
  sqlite db2 test.db
  execsql {ATTACH 'test2.db' as file2}
  execsql {ATTACH 'test2.db' as file2} db2
} {}

lock_status 4.1.1 db {main unlocked temp unlocked file2 unlocked}
lock_status 4.1.2 db2 {main unlocked temp unlocked file2 unlocked}

do_test attach2-4.2 {
  # Handle 'db' read-locks test.db
  execsql {BEGIN}
  execsql {SELECT * FROM t1}
  # Lock status:
  #    db  - shared(main)
  #    db2 -
} {}

lock_status 4.2.1 db {main shared temp shared file2 unlocked}
lock_status 4.2.2 db2 {main unlocked temp unlocked file2 unlocked}

do_test attach2-4.3 {
  # The read lock held by db does not prevent db2 from reading test.db
  execsql {SELECT * FROM t1} db2
} {}

lock_status 4.3.1 db {main shared temp shared file2 unlocked}
lock_status 4.3.2 db2 {main unlocked temp unlocked file2 unlocked}

do_test attach2-4.4 {
  # db is holding a read lock on test.db, so we should not be able
  # to commit a write to test.db from db2
  catchsql {
    INSERT INTO t1 VALUES(1, 2)
  } db2 
} {1 {database is locked}}

lock_status 4.4.1 db {main shared temp shared file2 unlocked}
lock_status 4.4.2 db2 {main unlocked temp unlocked file2 unlocked}

do_test attach2-4.5 {
  # Handle 'db2' reserves file2.
  execsql {BEGIN} db2
  execsql {INSERT INTO file2.t1 VALUES(1, 2)} db2
  # Lock status:
  #    db  - shared(main)
  #    db2 - reserved(file2)
} {}

lock_status 4.5.1 db {main shared temp shared file2 unlocked}
lock_status 4.5.2 db2 {main unlocked temp reserved file2 reserved}

do_test attach2-4.6.1 {
  # Reads are allowed against a reserved database.
  catchsql {
    SELECT * FROM file2.t1;
  }
  # Lock status:
  #    db  - shared(main), shared(file2)
  #    db2 - reserved(file2)
} {0 {}}

lock_status 4.6.1.1 db {main shared temp shared file2 shared}
lock_status 4.6.1.2 db2 {main unlocked temp reserved file2 reserved}

do_test attach2-4.6.2 {
  # Writes against a reserved database are not allowed.
  catchsql {
    UPDATE file2.t1 SET a=0;
  }
} {1 {database is locked}}

lock_status 4.6.2.1 db {main shared temp reserved file2 shared}
lock_status 4.6.2.2 db2 {main unlocked temp reserved file2 reserved}

do_test attach2-4.7 {
  # Ensure handle 'db' retains the lock on the main file after
  # failing to obtain a write-lock on file2.
  catchsql {
    INSERT INTO t1 VALUES(1, 2)
  } db2 
} {0 {}}

lock_status 4.7.1 db {main shared temp reserved file2 shared}
lock_status 4.7.2 db2 {main reserved temp reserved file2 reserved}

do_test attach2-4.8 {


  # We should still be able to read test.db from db2
  execsql {SELECT * FROM t1} db2
} {1 2}

lock_status 4.8.1 db {main shared temp reserved file2 shared}
lock_status 4.8.2 db2 {main reserved temp reserved file2 reserved}

do_test attach2-4.9 {
  # Try to upgrade the handle 'db' lock.
  catchsql {
    INSERT INTO t1 VALUES(1, 2)
  }

} {1 {database is locked}}

lock_status 4.9.1 db {main shared temp reserved file2 shared}
lock_status 4.9.2 db2 {main reserved temp reserved file2 reserved}

do_test attach2-4.10 {
  # Release the locks held by handle 'db2'
  execsql {COMMIT} db2
} {}

lock_status 4.10.1 db {main shared temp reserved file2 shared}
lock_status 4.10.2 db2 {main unlocked temp unlocked file2 unlocked}

do_test attach2-4.11 {
  execsql {SELECT * FROM file2.t1}
} {1 2}
do_test attach2-4.12 {
  execsql {INSERT INTO t1 VALUES(1, 2)}
} {}
do_test attach2-4.13 {