pager.c

sqlite-source-2_8_16.zip 轻量级嵌入式数据库源码

** or 0 if the page is not in cache.
**
** See also sqlitepager_get().  The difference between this routine
** and sqlitepager_get() 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.
*/
void *sqlitepager_lookup(Pager *pPager, Pgno pgno){
  PgHdr *pPg;

  assert( pPager!=0 );
  assert( pgno!=0 );
  if( pPager->errMask & ~(PAGER_ERR_FULL) ){
    return 0;
  }
  /* if( pPager->nRef==0 ){
  **  return 0;
  ** }
  */
  pPg = pager_lookup(pPager, pgno);
  if( pPg==0 ) return 0;
  page_ref(pPg);
  return PGHDR_TO_DATA(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 sqlitepager_unref(void *pData){
  PgHdr *pPg;

  /* Decrement the reference count for this page
  */
  pPg = DATA_TO_PGHDR(pData);
  assert( pPg->nRef>0 );
  pPg->nRef--;
  REFINFO(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 ){
    Pager *pPager;
    pPager = pPg->pPager;
    pPg->pNextFree = 0;
    pPg->pPrevFree = pPager->pLast;
    pPager->pLast = pPg;
    if( pPg->pPrevFree ){
      pPg->pPrevFree->pNextFree = pPg;
    }else{
      pPager->pFirst = pPg;
    }
    if( pPg->needSync==0 && pPager->pFirstSynced==0 ){
      pPager->pFirstSynced = pPg;
    }
    if( pPager->xDestructor ){
      pPager->xDestructor(pData);
    }
  
    /* When all pages reach the freelist, drop the read lock from
    ** the database file.
    */
    pPager->nRef--;
    assert( pPager->nRef>=0 );
    if( pPager->nRef==0 ){
      pager_reset(pPager);
    }
  }
  return SQLITE_OK;
}

/*
** Create a journal file for pPager.  There should already be a write
** 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==SQLITE_WRITELOCK );
  assert( pPager->journalOpen==0 );
  assert( pPager->useJournal );
  sqlitepager_pagecount(pPager);
  pPager->aInJournal = sqliteMalloc( pPager->dbSize/8 + 1 );
  if( pPager->aInJournal==0 ){
    sqliteOsReadLock(&pPager->fd);
    pPager->state = SQLITE_READLOCK;
    return SQLITE_NOMEM;
  }
  rc = sqliteOsOpenExclusive(pPager->zJournal, &pPager->jfd,pPager->tempFile);
  if( rc!=SQLITE_OK ){
    sqliteFree(pPager->aInJournal);
    pPager->aInJournal = 0;
    sqliteOsReadLock(&pPager->fd);
    pPager->state = SQLITE_READLOCK;
    return SQLITE_CANTOPEN;
  }
  sqliteOsOpenDirectory(pPager->zDirectory, &pPager->jfd);
  pPager->journalOpen = 1;
  pPager->journalStarted = 0;
  pPager->needSync = 0;
  pPager->alwaysRollback = 0;
  pPager->nRec = 0;
  if( pPager->errMask!=0 ){
    rc = pager_errcode(pPager);
    return rc;
  }
  pPager->origDbSize = pPager->dbSize;
  if( journal_format==JOURNAL_FORMAT_3 ){
    rc = sqliteOsWrite(&pPager->jfd, aJournalMagic3, sizeof(aJournalMagic3));
    if( rc==SQLITE_OK ){
      rc = write32bits(&pPager->jfd, pPager->noSync ? 0xffffffff : 0);
    }
    if( rc==SQLITE_OK ){
      sqliteRandomness(sizeof(pPager->cksumInit), &pPager->cksumInit);
      rc = write32bits(&pPager->jfd, pPager->cksumInit);
    }
  }else if( journal_format==JOURNAL_FORMAT_2 ){
    rc = sqliteOsWrite(&pPager->jfd, aJournalMagic2, sizeof(aJournalMagic2));
  }else{
    assert( journal_format==JOURNAL_FORMAT_1 );
    rc = sqliteOsWrite(&pPager->jfd, aJournalMagic1, sizeof(aJournalMagic1));
  }
  if( rc==SQLITE_OK ){
    rc = write32bits(&pPager->jfd, pPager->dbSize);
  }
  if( pPager->ckptAutoopen && rc==SQLITE_OK ){
    rc = sqlitepager_ckpt_begin(pPager);
  }
  if( rc!=SQLITE_OK ){
    rc = pager_unwritelock(pPager);
    if( rc==SQLITE_OK ){
      rc = SQLITE_FULL;
    }
  }
  return rc;  
}

/*
** Acquire a write-lock on the database.  The lock is removed when
** the any of the following happen:
**
**   *  sqlitepager_commit() is called.
**   *  sqlitepager_rollback() is called.
**   *  sqlitepager_close() is called.
**   *  sqlitepager_unref() is called to on every outstanding page.
**
** The 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.
**
** 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 write-locked, this routine is a no-op.
*/
int sqlitepager_begin(void *pData){
  PgHdr *pPg = DATA_TO_PGHDR(pData);
  Pager *pPager = pPg->pPager;
  int rc = SQLITE_OK;
  assert( pPg->nRef>0 );
  assert( pPager->state!=SQLITE_UNLOCK );
  if( pPager->state==SQLITE_READLOCK ){
    assert( pPager->aInJournal==0 );
    rc = sqliteOsWriteLock(&pPager->fd);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    pPager->state = SQLITE_WRITELOCK;
    pPager->dirtyFile = 0;
    TRACE1("TRANSACTION\n");
    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 write lock on the database.  If the write
** 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 sqlitepager_commit() or sqlitepager_rollback() to
** reset.
*/
int sqlitepager_write(void *pData){
  PgHdr *pPg = DATA_TO_PGHDR(pData);
  Pager *pPager = pPg->pPager;
  int rc = SQLITE_OK;

  /* Check for errors
  */
  if( pPager->errMask ){ 
    return pager_errcode(pPager);
  }
  if( pPager->readOnly ){
    return SQLITE_PERM;
  }

  /* 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->inCkpt || pPager->ckptInUse==0) ){
    pPager->dirtyFile = 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!=SQLITE_UNLOCK );
  rc = sqlitepager_begin(pData);
  if( rc!=SQLITE_OK ){
    return rc;
  }
  assert( pPager->state==SQLITE_WRITELOCK );
  if( !pPager->journalOpen && pPager->useJournal ){
    rc = pager_open_journal(pPager);
    if( rc!=SQLITE_OK ) return rc;
  }
  assert( pPager->journalOpen || !pPager->useJournal );
  pPager->dirtyFile = 1;

  /* The transaction journal now exists and we have a write 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 ){
    if( (int)pPg->pgno <= pPager->origDbSize ){
      int szPg;
      u32 saved;
      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 = sqliteOsWrite(&pPager->jfd, &((char*)pData)[-4], szPg);
      TRACE3("JOURNAL %d %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 ){
        sqlitepager_rollback(pPager);
        pPager->errMask |= PAGER_ERR_FULL;
        return rc;
      }
      pPager->nRec++;
      assert( pPager->aInJournal!=0 );
      pPager->aInJournal[pPg->pgno/8] |= 1<<(pPg->pgno&7);
      pPg->needSync = !pPager->noSync;
      pPg->inJournal = 1;
      if( pPager->ckptInUse ){
        pPager->aInCkpt[pPg->pgno/8] |= 1<<(pPg->pgno&7);
        page_add_to_ckpt_list(pPg);
      }
    }else{
      pPg->needSync = !pPager->journalStarted && !pPager->noSync;
      TRACE3("APPEND %d %d\n", pPg->pgno, pPg->needSync);
    }
    if( pPg->needSync ){
      pPager->needSync = 1;
    }
  }

  /* If the checkpoint journal is open and the page is not in it,
  ** then write the current page to the checkpoint journal.  Note that
  ** the checkpoint journal always uses the simplier format 2 that lacks
  ** checksums.  The header is also omitted from the checkpoint journal.
  */
  if( pPager->ckptInUse && !pPg->inCkpt && (int)pPg->pgno<=pPager->ckptSize ){
    assert( pPg->inJournal || (int)pPg->pgno>pPager->origDbSize );
    store32bits(pPg->pgno, pPg, -4);
    CODEC(pPager, pData, pPg->pgno, 7);
    rc = sqliteOsWrite(&pPager->cpfd, &((char*)pData)[-4], SQLITE_PAGE_SIZE+4);
    TRACE2("CKPT-JOURNAL %d\n", pPg->pgno);
    CODEC(pPager, pData, pPg->pgno, 0);
    if( rc!=SQLITE_OK ){
      sqlitepager_rollback(pPager);
      pPager->errMask |= PAGER_ERR_FULL;
      return rc;
    }
    pPager->ckptNRec++;
    assert( pPager->aInCkpt!=0 );
    pPager->aInCkpt[pPg->pgno/8] |= 1<<(pPg->pgno&7);
    page_add_to_ckpt_list(pPg);
  }

  /* Update the database size and return.
  */
  if( pPager->dbSize<(int)pPg->pgno ){
    pPager->dbSize = pPg->pgno;
  }
  return rc;
}

/*
** Return TRUE if the page given in the argument was previously passed
** to sqlitepager_write().  In other words, return TRUE if it is ok
** to change the content of the page.
*/
int sqlitepager_iswriteable(void *pData){
  PgHdr *pPg = DATA_TO_PGHDR(pData);
  return pPg->dirty;
}

/*
** Replace the content of a single page with the information in the third
** argument.
*/
int sqlitepager_overwrite(Pager *pPager, Pgno pgno, void *pData){
  void *pPage;
  int rc;

  rc = sqlitepager_get(pPager, pgno, &pPage);
  if( rc==SQLITE_OK ){
    rc = sqlitepager_write(pPage);
    if( rc==SQLITE_OK ){
      memcpy(pPage, pData, SQLITE_PAGE_SIZE);
    }
    sqlitepager_unref(pPage);
  }
  return rc;
}

/*
** A call to this routine tells the pager that it is not necessary to
** write the information on page "pgno" 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
** sqlitepager_dont_rollback() 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 sqlitepager_dont_rollback() 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 dont_rollback() routine is called.  But because the page contains
** critical data, we still need to be sure it gets rolled back in spite
** of the dont_rollback() call.
*/
void sqlitepager_dont_write(Pager *pPager, Pgno pgno){
  PgHdr *pPg;

  pPg = pager_lookup(pPager, pgno);
  pPg->alwaysRollback = 1;
  if( pPg && pPg->dirty ){
    if( pPage