** 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 1.278 2007/01/02 18:41:55 drh Exp $
*/
#ifndef SQLITE_OMIT_DISKIO
#include "sqliteInt.h"
#include "os.h"
#include "pager.h"
#include <assert.h>
#include <string.h>
................................................................................
  if( pPager->aHash==0 ) return 0;
  p = pPager->aHash[pgno & (pPager->nHash-1)];
  while( p && p->pgno!=pgno ){
    p = p->pNextHash;
  }
  return p;
}

















/*
** Unlock the database and clear the in-memory cache.  This routine
** sets the state of the pager back to what it was when it was first
** opened.  Any outstanding pages are invalidated and subsequent attempts
** to access those pages will likely result in a coredump.
*/
................................................................................
  pPager->nHash = 0;
  sqliteFree(pPager->aHash);
  pPager->nPage = 0;
  pPager->aHash = 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
................................................................................
  }
  
  /* Truncate the database back to its original size.
  */
  if( pPager->state>=PAGER_EXCLUSIVE ){
    rc = pager_truncate(pPager, pPager->stmtSize);
  }

  pPager->dbSize = pPager->stmtSize;

  /* Figure out how many records are in the statement journal.
  */
  assert( pPager->stmtInUse && pPager->journalOpen );
  sqlite3OsSeek(pPager->stfd, 0);
  nRec = pPager->stmtNRec;
................................................................................
** false or until the lock succeeds.
**
** Return SQLITE_OK on success and an error code if we cannot obtain
** the lock.
*/
static int pager_wait_on_lock(Pager *pPager, int locktype){
  int rc;


  assert( PAGER_SHARED==SHARED_LOCK );
  assert( PAGER_RESERVED==RESERVED_LOCK );
  assert( PAGER_EXCLUSIVE==EXCLUSIVE_LOCK );




  if( pPager->state>=locktype ){
    rc = SQLITE_OK;
  }else{
    do {
      rc = sqlite3OsLock(pPager->fd, locktype);
    }while( rc==SQLITE_BUSY && sqlite3InvokeBusyHandler(pPager->pBusyHandler) );
    if( rc==SQLITE_OK ){
................................................................................
}

/*
** Truncate the file to the number of pages specified.
*/
int sqlite3pager_truncate(Pager *pPager, Pgno nPage){
  int rc;

  sqlite3pager_pagecount(pPager);
  if( pPager->errCode ){
    rc = pPager->errCode;
    return rc;
  }
  if( nPage>=(unsigned)pPager->dbSize ){
    return SQLITE_OK;
................................................................................
      /* We ignore any IO errors that occur during the rollback
      ** operation. So disable IO error simulation so that testing
      ** works more easily.
      */
      disable_simulated_io_errors();
      sqlite3pager_rollback(pPager);
      enable_simulated_io_errors();
      if( !MEMDB ){
        sqlite3OsUnlock(pPager->fd, NO_LOCK);
      }
      assert( pPager->errCode || pPager->journalOpen==0 );
      break;
    }
    case PAGER_SHARED: {
      if( !MEMDB ){
        sqlite3OsUnlock(pPager->fd, NO_LOCK);
      }
      break;
    }
    default: {
      /* Do nothing */
      break;
    }
  }
................................................................................
       ** 
       ** Because the intermediate RESERVED lock is not requested, the
       ** second process will get to this point in the code and fail to
       ** obtain it's own 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 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.
       */
       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;
       pPager->journalOff = 0;
       pPager->setMaster = 0;
       pPager->journalHdr = 0;
................................................................................
    /* If this was a malloc() failure, then we will not be closing the pager
    ** file. So delete any journal file we may have just created. Otherwise,
    ** the system will get confused, we have a read-lock on the file and a
    ** mysterious journal has appeared in the filesystem.
    */
    sqlite3OsDelete(pPager->zJournal);
  }else{
    sqlite3OsUnlock(pPager->fd, NO_LOCK);
    pPager->state = PAGER_UNLOCK;
  }
  return rc;
}

/*
** Acquire a write-lock on the database.  The lock is removed when
** the any of the following happen:
................................................................................
      }
      page_add_to_stmt_list(pPg);
    }
  }

  /* Update the database size and return.
  */

  if( pPager->dbSize<(int)pPg->pgno ){
    pPager->dbSize = pPg->pgno;
    if( !MEMDB && pPager->dbSize==PENDING_BYTE/pPager->pageSize ){
      pPager->dbSize++;
    }
  }
  return rc;
................................................................................

  if( MEMDB ) return;

  pPg = pager_lookup(pPager, pgno);
  assert( pPg!=0 );  /* We never call _dont_write unless the page is in mem */
  pPg->alwaysRollback = 1;
  if( pPg->dirty && !pPager->stmtInUse ){

    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
................................................................................
** 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 );

  assert( pPager->dbSize>=0 );
  TRACE2("STMT-BEGIN %d\n", PAGERID(pPager));
  if( MEMDB ){
    pPager->stmtInUse = 1;
    pPager->stmtSize = pPager->dbSize;
    return SQLITE_OK;
  }

** 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 1.279 2007/01/03 15:34:30 drh Exp $
*/
#ifndef SQLITE_OMIT_DISKIO
#include "sqliteInt.h"
#include "os.h"
#include "pager.h"
#include <assert.h>
#include <string.h>
................................................................................
  if( pPager->aHash==0 ) return 0;
  p = pPager->aHash[pgno & (pPager->nHash-1)];
  while( p && p->pgno!=pgno ){
    p = p->pNextHash;
  }
  return p;
}

/*
** Unlock the database file.
**
** Once all locks have been removed from the database file, other
** processes or threads might change the file.  So make sure all of
** our internal cache is invalidated.
*/
static void pager_unlock(Pager *pPager){
  if( !MEMDB ){
    sqlite3OsUnlock(pPager->fd, NO_LOCK);
    pPager->dbSize = -1;
  }
  pPager->state = PAGER_UNLOCK;
}


/*
** Unlock the database and clear the in-memory cache.  This routine
** sets the state of the pager back to what it was when it was first
** opened.  Any outstanding pages are invalidated and subsequent attempts
** to access those pages will likely result in a coredump.
*/
................................................................................
  pPager->nHash = 0;
  sqliteFree(pPager->aHash);
  pPager->nPage = 0;
  pPager->aHash = 0;
  if( pPager->state>=PAGER_RESERVED ){
    sqlite3pager_rollback(pPager);
  }
  pager_unlock(pPager);


  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
................................................................................
  }
  
  /* Truncate the database back to its original size.
  */
  if( pPager->state>=PAGER_EXCLUSIVE ){
    rc = pager_truncate(pPager, pPager->stmtSize);
  }
  assert( pPager->state>=PAGER_SHARED );
  pPager->dbSize = pPager->stmtSize;

  /* Figure out how many records are in the statement journal.
  */
  assert( pPager->stmtInUse && pPager->journalOpen );
  sqlite3OsSeek(pPager->stfd, 0);
  nRec = pPager->stmtNRec;
................................................................................
** false or until the lock succeeds.
**
** Return SQLITE_OK on success and an error code if we cannot obtain
** the lock.
*/
static int pager_wait_on_lock(Pager *pPager, int locktype){
  int rc;

  /* The OS lock values must be the same as the Pager lock values */
  assert( PAGER_SHARED==SHARED_LOCK );
  assert( PAGER_RESERVED==RESERVED_LOCK );
  assert( PAGER_EXCLUSIVE==EXCLUSIVE_LOCK );

  /* If the file is currently unlocked then the size must be unknown */
  assert( pPager->state>=PAGER_SHARED || pPager->dbSize<0 || MEMDB );

  if( pPager->state>=locktype ){
    rc = SQLITE_OK;
  }else{
    do {
      rc = sqlite3OsLock(pPager->fd, locktype);
    }while( rc==SQLITE_BUSY && sqlite3InvokeBusyHandler(pPager->pBusyHandler) );
    if( rc==SQLITE_OK ){
................................................................................
}

/*
** Truncate the file to the number of pages specified.
*/
int sqlite3pager_truncate(Pager *pPager, Pgno nPage){
  int rc;
  assert( pPager->state>=PAGER_SHARED || MEMDB );
  sqlite3pager_pagecount(pPager);
  if( pPager->errCode ){
    rc = pPager->errCode;
    return rc;
  }
  if( nPage>=(unsigned)pPager->dbSize ){
    return SQLITE_OK;
................................................................................
      /* We ignore any IO errors that occur during the rollback
      ** operation. So disable IO error simulation so that testing
      ** works more easily.
      */
      disable_simulated_io_errors();
      sqlite3pager_rollback(pPager);
      enable_simulated_io_errors();

      pager_unlock(pPager);

      assert( pPager->errCode || pPager->journalOpen==0 );
      break;
    }
    case PAGER_SHARED: {

      pager_unlock(pPager);

      break;
    }
    default: {
      /* Do nothing */
      break;
    }
  }
................................................................................
       ** 
       ** Because the intermediate RESERVED lock is not requested, the
       ** second process will get to this point in the code and fail to
       ** obtain it's own EXCLUSIVE lock on the database file.
       */
       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.
       */
       rc = sqlite3OsOpenReadOnly(pPager->zJournal, &pPager->jfd);
       if( rc!=SQLITE_OK ){
         pager_unlock(pPager);

         return SQLITE_BUSY;
       }
       pPager->journalOpen = 1;
       pPager->journalStarted = 0;
       pPager->journalOff = 0;
       pPager->setMaster = 0;
       pPager->journalHdr = 0;
................................................................................
    /* If this was a malloc() failure, then we will not be closing the pager
    ** file. So delete any journal file we may have just created. Otherwise,
    ** the system will get confused, we have a read-lock on the file and a
    ** mysterious journal has appeared in the filesystem.
    */
    sqlite3OsDelete(pPager->zJournal);
  }else{
    pager_unlock(pPager);

  }
  return rc;
}

/*
** Acquire a write-lock on the database.  The lock is removed when
** the any of the following happen:
................................................................................
      }
      page_add_to_stmt_list(pPg);
    }
  }

  /* 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;
................................................................................

  if( MEMDB ) return;

  pPg = pager_lookup(pPager, pgno);
  assert( pPg!=0 );  /* We never call _dont_write unless the page is in mem */
  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
................................................................................
** 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 );
  assert( pPager->state>=PAGER_SHARED );
  assert( pPager->dbSize>=0 );
  TRACE2("STMT-BEGIN %d\n", PAGERID(pPager));
  if( MEMDB ){
    pPager->stmtInUse = 1;
    pPager->stmtSize = pPager->dbSize;
    return SQLITE_OK;
  }