pager.c

sqlite-source-2_8_16.zip 轻量级嵌入式数据库源码

** 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 sync occurs.
**
** 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 ){
      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.
        */
        off_t hdrSz, pgSz, jSz;
        hdrSz = JOURNAL_HDR_SZ(journal_format);
        pgSz = JOURNAL_PG_SZ(journal_format);
        rc = sqliteOsFileSize(&pPager->jfd, &jSz);
        if( rc!=0 ) return rc;
        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 ){
          TRACE1("SYNC\n");
          rc = sqliteOsSync(&pPager->jfd);
          if( rc!=0 ) return rc;
        }
        sqliteOsSeek(&pPager->jfd, sizeof(aJournalMagic1));
        rc = write32bits(&pPager->jfd, pPager->nRec);
        if( rc ) return rc;
        szJ = JOURNAL_HDR_SZ(journal_format) +
                 pPager->nRec*JOURNAL_PG_SZ(journal_format);
        sqliteOsSeek(&pPager->jfd, szJ);
      }
      TRACE1("SYNC\n");
      rc = sqliteOsSync(&pPager->jfd);
      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;
    }
    pPager->pFirstSynced = pPager->pFirst;
  }

#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->pFirstSynced==pPager->pFirst );
  }
#endif

  return rc;
}

/*
** 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;
  int rc;

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

/*
** 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){
  PgHdr *p, *pList;
  pList = 0;
  for(p=pPager->pAll; p; p=p->pNextAll){
    if( p->dirty ){
      p->pDirty = pList;
      pList = p;
    }
  }
  return pList;
}

/*
** 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.
**
** A _get 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 sqlitepager_lookup().  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.
*/
int sqlitepager_get(Pager *pPager, Pgno pgno, void **ppPage){
  PgHdr *pPg;
  int rc;

  /* Make sure we have not hit any critical errors.
  */ 
  assert( pPager!=0 );
  assert( pgno!=0 );
  *ppPage = 0;
  if( pPager->errMask & ~(PAGER_ERR_FULL) ){
    return pager_errcode(pPager);
  }

  /* If this is the first page accessed, then get a read lock
  ** on the database file.
  */
  if( pPager->nRef==0 ){
    rc = sqliteOsReadLock(&pPager->fd);
    if( rc!=SQLITE_OK ){
      return rc;
    }
    pPager->state = SQLITE_READLOCK;

    /* If a journal file exists, try to play it back.
    */
    if( pPager->useJournal && sqliteOsFileExists(pPager->zJournal) ){
       int rc;

       /* Get a write lock on the database
       */
       rc = sqliteOsWriteLock(&pPager->fd);
       if( rc!=SQLITE_OK ){
         if( sqliteOsUnlock(&pPager->fd)!=SQLITE_OK ){
           /* This should never happen! */
           rc = SQLITE_INTERNAL;
         }
         return rc;
       }
       pPager->state = SQLITE_WRITELOCK;

       /* 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 = sqliteOsOpenReadOnly(pPager->zJournal, &pPager->jfd);
       if( rc!=SQLITE_OK ){
         rc = sqliteOsUnlock(&pPager->fd);
         assert( rc==SQLITE_OK );
         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( pPg==0 ){
    /* The requested page is not in the page cache. */
    int h;
    pPager->nMiss++;
    if( pPager->nPage<pPager->mxPage || pPager->pFirst==0 ){
      /* Create a new page */
      pPg = sqliteMallocRaw( sizeof(*pPg) + SQLITE_PAGE_SIZE 
                              + sizeof(u32) + pPager->nExtra );
      if( pPg==0 ){
        pager_unwritelock(pPager);
        pPager->errMask |= PAGER_ERR_MEM;
        return SQLITE_NOMEM;
      }
      memset(pPg, 0, sizeof(*pPg));
      pPg->pPager = pPager;
      pPg->pNextAll = pPager->pAll;
      if( pPager->pAll ){
        pPager->pAll->pPrevAll = pPg;
      }
      pPg->pPrevAll = 0;
      pPager->pAll = pPg;
      pPager->nPage++;
    }else{
      /* Find a page to recycle.  Try to locate a page that does not
      ** require us to do an fsync() on the journal.
      */
      pPg = pPager->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 ){
        int rc = syncJournal(pPager);
        if( rc!=0 ){
          sqlitepager_rollback(pPager);
          return SQLITE_IOERR;
        }
        pPg = pPager->pFirst;
      }
      assert( pPg->nRef==0 );

      /* Write the page to the database file if it is dirty.
      */
      if( pPg->dirty ){
        assert( pPg->needSync==0 );
        pPg->pDirty = 0;
        rc = pager_write_pagelist( pPg );
        if( rc!=SQLITE_OK ){
          sqlitepager_rollback(pPager);
          return SQLITE_IOERR;
        }
      }
      assert( pPg->dirty==0 );

      /* If the page we are recycling is marked as alwaysRollback, then
      ** set the global alwaysRollback flag, thus disabling the
      ** sqlite_dont_rollback() 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 ){
        pPager->alwaysRollback = 1;
      }

      /* Unlink the old page from the free list and the hash table
      */
      if( pPg==pPager->pFirstSynced ){
        PgHdr *p = pPg->pNextFree;
        while( p && p->needSync ){ p = p->pNextFree; }
        pPager->pFirstSynced = p;
      }
      if( pPg->pPrevFree ){
        pPg->pPrevFree->pNextFree = pPg->pNextFree;
      }else{
        assert( pPager->pFirst==pPg );
        pPager->pFirst = pPg->pNextFree;
      }
      if( pPg->pNextFree ){
        pPg->pNextFree->pPrevFree = pPg->pPrevFree;
      }else{
        assert( pPager->pLast==pPg );
        pPager->pLast = pPg->pPrevFree;
      }
      pPg->pNextFree = pPg->pPrevFree = 0;
      if( pPg->pNextHash ){
        pPg->pNextHash->pPrevHash = pPg->pPrevHash;
      }
      if( pPg->pPrevHash ){
        pPg->pPrevHash->pNextHash = pPg->pNextHash;
      }else{
        h = pager_hash(pPg->pgno);
        assert( pPager->aHash[h]==pPg );
        pPager->aHash[h] = pPg->pNextHash;
      }
      pPg->pNextHash = pPg->pPrevHash = 0;
      pPager->nOvfl++;
    }
    pPg->pgno = pgno;
    if( pPager->aInJournal && (int)pgno<=pPager->origDbSize ){
      sqliteCheckMemory(pPager->aInJournal, pgno/8);
      assert( pPager->journalOpen );
      pPg->inJournal = (pPager->aInJournal[pgno/8] & (1<<(pgno&7)))!=0;
      pPg->needSync = 0;
    }else{
      pPg->inJournal = 0;
      pPg->needSync = 0;
    }
    if( pPager->aInCkpt && (int)pgno<=pPager->ckptSize
             && (pPager->aInCkpt[pgno/8] & (1<<(pgno&7)))!=0 ){
      page_add_to_ckpt_list(pPg);
    }else{
      page_remove_from_ckpt_list(pPg);
    }
    pPg->dirty = 0;
    pPg->nRef = 1;
    REFINFO(pPg);
    pPager->nRef++;
    h = pager_hash(pgno);
    pPg->pNextHash = pPager->aHash[h];
    pPager->aHash[h] = pPg;
    if( pPg->pNextHash ){
      assert( pPg->pNextHash->pPrevHash==0 );
      pPg->pNextHash->pPrevHash = pPg;
    }
    if( pPager->nExtra>0 ){
      memset(PGHDR_TO_EXTRA(pPg), 0, pPager->nExtra);
    }
    if( pPager->dbSize<0 ) sqlitepager_pagecount(pPager);
    if( pPager->errMask!=0 ){
      sqlitepager_unref(PGHDR_TO_DATA(pPg));
      rc = pager_errcode(pPager);
      return rc;
    }
    if( pPager->dbSize<(int)pgno ){
      memset(PGHDR_TO_DATA(pPg), 0, SQLITE_PAGE_SIZE);
    }else{
      int rc;
      sqliteOsSeek(&pPager->fd, (pgno-1)*(off_t)SQLITE_PAGE_SIZE);
      rc = sqliteOsRead(&pPager->fd, PGHDR_TO_DATA(pPg), SQLITE_PAGE_SIZE);
      TRACE2("FETCH %d\n", pPg->pgno);
      CODEC(pPager, PGHDR_TO_DATA(pPg), pPg->pgno, 3);
      if( rc!=SQLITE_OK ){
        off_t fileSize;
        if( sqliteOsFileSize(&pPager->fd,&fileSize)!=SQLITE_OK
               || fileSize>=pgno*SQLITE_PAGE_SIZE ){
          sqlitepager_unref(PGHDR_TO_DATA(pPg));
          return rc;
        }else{
          memset(PGHDR_TO_DATA(pPg), 0, SQLITE_PAGE_SIZE);
        }
      }
    }
  }else{
    /* The requested page is in the page cache. */
    pPager->nHit++;
    page_ref(pPg);
  }
  *ppPage = PGHDR_TO_DATA(pPg);
  return SQLITE_OK;
}

/*
** 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,