pager.c

来自「SQLite 2.8.6 源代码,用来在Linux/Unix/Windows上编」· C语言 代码 · 共 2,008 行 · 第 1/5 页

    sqliteOsTempFileName(zFile);
    rc = sqliteOsOpenExclusive(zFile, fd, 1);
  }while( cnt>0 && rc!=SQLITE_OK );
  return rc;
}

/*
** Create a new page cache and put a pointer to the page cache in *ppPager.
** The file to be cached need not exist.  The file is not locked until
** the first call to sqlitepager_get() and is only held open until the
** last page is released using sqlitepager_unref().
**
** If zFilename is NULL then a randomly-named temporary file is created
** and used as the file to be cached.  The file will be deleted
** automatically when it is closed.
*/
int sqlitepager_open(
  Pager **ppPager,         /* Return the Pager structure here */
  const char *zFilename,   /* Name of the database file to open */
  int mxPage,              /* Max number of in-memory cache pages */
  int nExtra,              /* Extra bytes append to each in-memory page */
  int useJournal           /* TRUE to use a rollback journal on this file */
){
  Pager *pPager;
  char *zFullPathname;
  int nameLen;
  OsFile fd;
  int rc, i;
  int tempFile;
  int readOnly = 0;
  char zTemp[SQLITE_TEMPNAME_SIZE];

  *ppPager = 0;
  if( sqlite_malloc_failed ){
    return SQLITE_NOMEM;
  }
  if( zFilename ){
    zFullPathname = sqliteOsFullPathname(zFilename);
    rc = sqliteOsOpenReadWrite(zFullPathname, &fd, &readOnly);
    tempFile = 0;
  }else{
    rc = sqlitepager_opentemp(zTemp, &fd);
    zFilename = zTemp;
    zFullPathname = sqliteOsFullPathname(zFilename);
    tempFile = 1;
  }
  if( sqlite_malloc_failed ){
    return SQLITE_NOMEM;
  }
  if( rc!=SQLITE_OK ){
    sqliteFree(zFullPathname);
    return SQLITE_CANTOPEN;
  }
  nameLen = strlen(zFullPathname);
  pPager = sqliteMalloc( sizeof(*pPager) + nameLen*3 + 30 );
  if( pPager==0 ){
    sqliteOsClose(&fd);
    sqliteFree(zFullPathname);
    return SQLITE_NOMEM;
  }
  SET_PAGER(pPager);
  pPager->zFilename = (char*)&pPager[1];
  pPager->zDirectory = &pPager->zFilename[nameLen+1];
  pPager->zJournal = &pPager->zDirectory[nameLen+1];
  strcpy(pPager->zFilename, zFullPathname);
  strcpy(pPager->zDirectory, zFullPathname);
  for(i=nameLen; i>0 && pPager->zDirectory[i-1]!='/'; i--){}
  if( i>0 ) pPager->zDirectory[i-1] = 0;
  strcpy(pPager->zJournal, zFullPathname);
  sqliteFree(zFullPathname);
  strcpy(&pPager->zJournal[nameLen], "-journal");
  pPager->fd = fd;
  pPager->journalOpen = 0;
  pPager->useJournal = useJournal;
  pPager->ckptOpen = 0;
  pPager->ckptInUse = 0;
  pPager->nRef = 0;
  pPager->dbSize = -1;
  pPager->ckptSize = 0;
  pPager->ckptJSize = 0;
  pPager->nPage = 0;
  pPager->mxPage = mxPage>5 ? mxPage : 10;
  pPager->state = SQLITE_UNLOCK;
  pPager->errMask = 0;
  pPager->tempFile = tempFile;
  pPager->readOnly = readOnly;
  pPager->needSync = 0;
  pPager->noSync = pPager->tempFile || !useJournal;
  pPager->pFirst = 0;
  pPager->pFirstSynced = 0;
  pPager->pLast = 0;
  pPager->nExtra = nExtra;
  memset(pPager->aHash, 0, sizeof(pPager->aHash));
  *ppPager = pPager;
  return SQLITE_OK;
}

/*
** Set the destructor for this pager.  If not NULL, the destructor is called
** when the reference count on each page reaches zero.  The destructor can
** be used to clean up information in the extra segment appended to each page.
**
** The destructor is not called as a result sqlitepager_close().  
** Destructors are only called by sqlitepager_unref().
*/
void sqlitepager_set_destructor(Pager *pPager, void (*xDesc)(void*)){
  pPager->xDestructor = xDesc;
}

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

/*
** Forward declaration
*/
static int syncAllPages(Pager*);

/*
** Truncate the file to the number of pages specified.
*/
int sqlitepager_truncate(Pager *pPager, Pgno nPage){
  int rc;
  if( pPager->dbSize<0 ){
    sqlitepager_pagecount(pPager);
  }
  if( pPager->errMask!=0 ){
    rc = pager_errcode(pPager);
    return rc;
  }
  if( nPage>=(unsigned)pPager->dbSize ){
    return SQLITE_OK;
  }
  syncAllPages(pPager);
  rc = sqliteOsTruncate(&pPager->fd, SQLITE_PAGE_SIZE*(off_t)nPage);
  if( rc==SQLITE_OK ){
    pPager->dbSize = nPage;
  }
  return rc;
}

/*
** Shutdown the page cache.  Free all memory and close all files.
**
** If a transaction was in progress when this routine is called, that
** transaction is rolled back.  All outstanding pages are invalidated
** and their memory is freed.  Any attempt to use a page associated
** with this page cache after this function returns will likely
** result in a coredump.
*/
int sqlitepager_close(Pager *pPager){
  PgHdr *pPg, *pNext;
  switch( pPager->state ){
    case SQLITE_WRITELOCK: {
      sqlitepager_rollback(pPager);
      sqliteOsUnlock(&pPager->fd);
      assert( pPager->journalOpen==0 );
      break;
    }
    case SQLITE_READLOCK: {
      sqliteOsUnlock(&pPager->fd);
      break;
    }
    default: {
      /* Do nothing */
      break;
    }
  }
  for(pPg=pPager->pAll; pPg; pPg=pNext){
    pNext = pPg->pNextAll;
    sqliteFree(pPg);
  }
  sqliteOsClose(&pPager->fd);
  assert( pPager->journalOpen==0 );
  /* Temp files are automatically deleted by the OS
  ** if( pPager->tempFile ){
  **   sqliteOsDelete(pPager->zFilename);
  ** }
  */
  CLR_PAGER(pPager);
  if( pPager->zFilename!=(char*)&pPager[1] ){
    assert( 0 );  /* Cannot happen */
    sqliteFree(pPager->zFilename);
    sqliteFree(pPager->zJournal);
    sqliteFree(pPager->zDirectory);
  }
  sqliteFree(pPager);
  return SQLITE_OK;
}

/*
** Return the page number for the given page data.
*/
Pgno sqlitepager_pagenumber(void *pData){
  PgHdr *p = DATA_TO_PGHDR(pData);
  return p->pgno;
}

/*
** Increment the reference count for a page.  If the page is
** currently on the freelist (the reference count is zero) then
** remove it from the freelist.
*/
#define page_ref(P)   ((P)->nRef==0?_page_ref(P):(void)(P)->nRef++)
static void _page_ref(PgHdr *pPg){
  if( pPg->nRef==0 ){
    /* The page is currently on the freelist.  Remove it. */
    if( pPg==pPg->pPager->pFirstSynced ){
      PgHdr *p = pPg->pNextFree;
      while( p && p->needSync ){ p = p->pNextFree; }
      pPg->pPager->pFirstSynced = p;
    }
    if( pPg->pPrevFree ){
      pPg->pPrevFree->pNextFree = pPg->pNextFree;
    }else{
      pPg->pPager->pFirst = pPg->pNextFree;
    }
    if( pPg->pNextFree ){
      pPg->pNextFree->pPrevFree = pPg->pPrevFree;
    }else{
      pPg->pPager->pLast = pPg->pPrevFree;
    }
    pPg->pPager->nRef++;
  }
  pPg->nRef++;
  REFINFO(pPg);
}

/*
** Increment the reference count for a page.  The input pointer is
** a reference to the page data.
*/
int sqlitepager_ref(void *pData){
  PgHdr *pPg = DATA_TO_PGHDR(pData);
  page_ref(pPg);
  return SQLITE_OK;
}

/*
** Sync the journal and then write all free dirty pages to the database
** file.
**
** Writing all free dirty pages to the database after the sync is a
** non-obvious optimization.  fsync() is an expensive operation so we
** want to minimize the number ot times it is called. After an fsync() call,
** we are free to write dirty pages back to the database.  It is best
** to go ahead and write as many dirty pages as possible to minimize 
** the risk of having to do another fsync() later on.  Writing dirty
** free pages in this way was observed to make database operations go
** up to 10 times faster.
**
** If we are writing to temporary database, there is no need to preserve
** the integrity of the journal file, so we can save time and skip the
** fsync().
*/
static int syncAllPages(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 );
#ifndef NDEBUG
      {
        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 ){
        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);
    rc = sqliteOsWrite(&pPager->fd, PGHDR_TO_DATA(pList), SQLITE_PAGE_SIZE);
    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;