pager.c

这是一个开源的数据库系统,值得学习啊, 里面用了SQL语句,与微软的SQL SERVIER,差不了多少

  pagerEnter(pPager);
  if( aHash==0 ){
    /* Failure to rehash is not an error.  It is only a performance hit. */
    return;
  }
  sqlite3_free(pPager->aHash);
  pPager->nHash = N;
  pPager->aHash = aHash;
  for(pPg=pPager->pAll; pPg; pPg=pPg->pNextAll){
    int h;
    if( pPg->pgno==0 ){
      assert( pPg->pNextHash==0 && pPg->pPrevHash==0 );
      continue;
    }
    h = pPg->pgno & (N-1);
    pPg->pNextHash = aHash[h];
    if( aHash[h] ){
      aHash[h]->pPrevHash = pPg;
    }
    aHash[h] = pPg;
    pPg->pPrevHash = 0;
  }
}

/*
** Read a 32-bit integer from the given file descriptor.  Store the integer
** that is read in *pRes.  Return SQLITE_OK if everything worked, or an
** error code is something goes wrong.
**
** All values are stored on disk as big-endian.
*/
static int read32bits(sqlite3_file *fd, i64 offset, u32 *pRes){
  unsigned char ac[4];
  int rc = sqlite3OsRead(fd, ac, sizeof(ac), offset);
  if( rc==SQLITE_OK ){
    *pRes = sqlite3Get4byte(ac);
  }
  return rc;
}

/*
** Write a 32-bit integer into a string buffer in big-endian byte order.
*/
#define put32bits(A,B)  sqlite3Put4byte((u8*)A,B)

/*
** Write a 32-bit integer into the given file descriptor.  Return SQLITE_OK
** on success or an error code is something goes wrong.
*/
static int write32bits(sqlite3_file *fd, i64 offset, u32 val){
  char ac[4];
  put32bits(ac, val);
  return sqlite3OsWrite(fd, ac, 4, offset);
}

/*
** If file pFd is open, call sqlite3OsUnlock() on it.
*/
static int osUnlock(sqlite3_file *pFd, int eLock){
  if( !pFd->pMethods ){
    return SQLITE_OK;
  }
  return sqlite3OsUnlock(pFd, eLock);
}

/*
** This function determines whether or not the atomic-write optimization
** can be used with this pager. The optimization can be used if:
**
**  (a) the value returned by OsDeviceCharacteristics() indicates that
**      a database page may be written atomically, and
**  (b) the value returned by OsSectorSize() is less than or equal
**      to the page size.
**
** If the optimization cannot be used, 0 is returned. If it can be used,
** then the value returned is the size of the journal file when it
** contains rollback data for exactly one page.
*/
#ifdef SQLITE_ENABLE_ATOMIC_WRITE
static int jrnlBufferSize(Pager *pPager){
  int dc;           /* Device characteristics */
  int nSector;      /* Sector size */
  int nPage;        /* Page size */
  sqlite3_file *fd = pPager->fd;

  if( fd->pMethods ){
    dc = sqlite3OsDeviceCharacteristics(fd);
    nSector = sqlite3OsSectorSize(fd);
    nPage = pPager->pageSize;
  }

  assert(SQLITE_IOCAP_ATOMIC512==(512>>8));
  assert(SQLITE_IOCAP_ATOMIC64K==(65536>>8));

  if( !fd->pMethods || (dc&(SQLITE_IOCAP_ATOMIC|(nPage>>8))&&nSector<=nPage) ){
    return JOURNAL_HDR_SZ(pPager) + JOURNAL_PG_SZ(pPager);
  }
  return 0;
}
#endif

/*
** This function should be called when an error occurs within the pager
** code. The first argument is a pointer to the pager structure, the
** second the error-code about to be returned by a pager API function. 
** The value returned is a copy of the second argument to this function. 
**
** If the second argument is SQLITE_IOERR, SQLITE_CORRUPT, or SQLITE_FULL
** the error becomes persistent. All subsequent API calls on this Pager
** will immediately return the same error code.
*/
static int pager_error(Pager *pPager, int rc){
  int rc2 = rc & 0xff;
  assert(
       pPager->errCode==SQLITE_FULL ||
       pPager->errCode==SQLITE_OK ||
       (pPager->errCode & 0xff)==SQLITE_IOERR
  );
  if(
    rc2==SQLITE_FULL ||
    rc2==SQLITE_IOERR ||
    rc2==SQLITE_CORRUPT
  ){
    pPager->errCode = rc;
  }
  return rc;
}

/*
** If SQLITE_CHECK_PAGES is defined then we do some sanity checking
** on the cache using a hash function.  This is used for testing
** and debugging only.
*/
#ifdef SQLITE_CHECK_PAGES
/*
** Return a 32-bit hash of the page data for pPage.
*/
static u32 pager_datahash(int nByte, unsigned char *pData){
  u32 hash = 0;
  int i;
  for(i=0; i<nByte; i++){
    hash = (hash*1039) + pData[i];
  }
  return hash;
}
static u32 pager_pagehash(PgHdr *pPage){
  return pager_datahash(pPage->pPager->pageSize, 
                        (unsigned char *)PGHDR_TO_DATA(pPage));
}

/*
** The CHECK_PAGE macro takes a PgHdr* as an argument. If SQLITE_CHECK_PAGES
** is defined, and NDEBUG is not defined, an assert() statement checks
** that the page is either dirty or still matches the calculated page-hash.
*/
#define CHECK_PAGE(x) checkPage(x)
static void checkPage(PgHdr *pPg){
  Pager *pPager = pPg->pPager;
  assert( !pPg->pageHash || pPager->errCode || MEMDB || pPg->dirty || 
      pPg->pageHash==pager_pagehash(pPg) );
}

#else
#define pager_datahash(X,Y)  0
#define pager_pagehash(X)  0
#define CHECK_PAGE(x)
#endif

/*
** When this is called the journal file for pager pPager must be open.
** The master journal file name is read from the end of the file and 
** written into memory supplied by the caller. 
**
** zMaster must point to a buffer of at least nMaster bytes allocated by
** the caller. This should be sqlite3_vfs.mxPathname+1 (to ensure there is
** enough space to write the master journal name). If the master journal
** name in the journal is longer than nMaster bytes (including a
** nul-terminator), then this is handled as if no master journal name
** were present in the journal.
**
** If no master journal file name is present zMaster[0] is set to 0 and
** SQLITE_OK returned.
*/
static int readMasterJournal(sqlite3_file *pJrnl, char *zMaster, int nMaster){
  int rc;
  u32 len;
  i64 szJ;
  u32 cksum;
  int i;
  unsigned char aMagic[8]; /* A buffer to hold the magic header */

  zMaster[0] = '\0';

  rc = sqlite3OsFileSize(pJrnl, &szJ);
  if( rc!=SQLITE_OK || szJ<16 ) return rc;

  rc = read32bits(pJrnl, szJ-16, &len);
  if( rc!=SQLITE_OK ) return rc;

  if( len>=nMaster ){
    return SQLITE_OK;
  }

  rc = read32bits(pJrnl, szJ-12, &cksum);
  if( rc!=SQLITE_OK ) return rc;

  rc = sqlite3OsRead(pJrnl, aMagic, 8, szJ-8);
  if( rc!=SQLITE_OK || memcmp(aMagic, aJournalMagic, 8) ) return rc;

  rc = sqlite3OsRead(pJrnl, zMaster, len, szJ-16-len);
  if( rc!=SQLITE_OK ){
    return rc;
  }
  zMaster[len] = '\0';

  /* See if the checksum matches the master journal name */
  for(i=0; i<len; i++){
    cksum -= zMaster[i];
   }
  if( cksum ){
    /* If the checksum doesn't add up, then one or more of the disk sectors
    ** containing the master journal filename is corrupted. This means
    ** definitely roll back, so just return SQLITE_OK and report a (nul)
    ** master-journal filename.
    */
    zMaster[0] = '\0';
  }
   
  return SQLITE_OK;
}

/*
** Seek the journal file descriptor to the next sector boundary where a
** journal header may be read or written. Pager.journalOff is updated with
** the new seek offset.
**
** i.e for a sector size of 512:
**
** Input Offset              Output Offset
** ---------------------------------------
** 0                         0
** 512                       512
** 100                       512
** 2000                      2048
** 
*/
static void seekJournalHdr(Pager *pPager){
  i64 offset = 0;
  i64 c = pPager->journalOff;
  if( c ){
    offset = ((c-1)/JOURNAL_HDR_SZ(pPager) + 1) * JOURNAL_HDR_SZ(pPager);
  }
  assert( offset%JOURNAL_HDR_SZ(pPager)==0 );
  assert( offset>=c );
  assert( (offset-c)<JOURNAL_HDR_SZ(pPager) );
  pPager->journalOff = offset;
}

/*
** The journal file must be open when this routine is called. A journal
** header (JOURNAL_HDR_SZ bytes) is written into the journal file at the
** current location.
**
** The format for the journal header is as follows:
** - 8 bytes: Magic identifying journal format.
** - 4 bytes: Number of records in journal, or -1 no-sync mode is on.
** - 4 bytes: Random number used for page hash.
** - 4 bytes: Initial database page count.
** - 4 bytes: Sector size used by the process that wrote this journal.
** 
** Followed by (JOURNAL_HDR_SZ - 24) bytes of unused space.
*/
static int writeJournalHdr(Pager *pPager){
  char zHeader[sizeof(aJournalMagic)+16];
  int rc;

  if( pPager->stmtHdrOff==0 ){
    pPager->stmtHdrOff = pPager->journalOff;
  }

  seekJournalHdr(pPager);
  pPager->journalHdr = pPager->journalOff;

  memcpy(zHeader, aJournalMagic, sizeof(aJournalMagic));

  /* 
  ** Write the nRec Field - the number of page records that follow this
  ** journal header. Normally, zero is written to this value at this time.
  ** After the records are added to the journal (and the journal synced, 
  ** if in full-sync mode), the zero is overwritten with the true number
  ** of records (see syncJournal()).
  **
  ** A faster alternative is to write 0xFFFFFFFF to the nRec field. When
  ** reading the journal this value tells SQLite to assume that the
  ** rest of the journal file contains valid page records. This assumption
  ** is dangerous, as if a failure occured whilst writing to the journal
  ** file it may contain some garbage data. There are two scenarios
  ** where this risk can be ignored:
  **
  **   * When the pager is in no-sync mode. Corruption can follow a
  **     power failure in this case anyway.
  **
  **   * When the SQLITE_IOCAP_SAFE_APPEND flag is set. This guarantees
  **     that garbage data is never appended to the journal file.
  */
  assert(pPager->fd->pMethods||pPager->noSync);
  if( (pPager->noSync) 
   || (sqlite3OsDeviceCharacteristics(pPager->fd)&SQLITE_IOCAP_SAFE_APPEND) 
  ){
    put32bits(&zHeader[sizeof(aJournalMagic)], 0xffffffff);
  }else{
    put32bits(&zHeader[sizeof(aJournalMagic)], 0);
  }

  /* The random check-hash initialiser */ 
  sqlite3Randomness(sizeof(pPager->cksumInit), &pPager->cksumInit);
  put32bits(&zHeader[sizeof(aJournalMagic)+4], pPager->cksumInit);
  /* The initial database size */
  put32bits(&zHeader[sizeof(aJournalMagic)+8], pPager->dbSize);
  /* The assumed sector size for this process */
  put32bits(&zHeader[sizeof(aJournalMagic)+12], pPager->sectorSize);
  IOTRACE(("JHDR %p %lld %d\n", pPager, pPager->journalHdr, sizeof(zHeader)))
  rc = sqlite3OsWrite(pPager->jfd, zHeader, sizeof(zHeader),pPager->journalOff);
  pPager->journalOff += JOURNAL_HDR_SZ(pPager);

  /* The journal header has been written successfully. Seek the journal
  ** file descriptor to the end of the journal header sector.
  */
  if( rc==SQLITE_OK ){
    IOTRACE(("JTAIL %p %lld\n", pPager, pPager->journalOff-1))
    rc = sqlite3OsWrite(pPager->jfd, "\000", 1, pPager->journalOff-1);
  }
  return rc;
}

/*
** The journal file must be open when this is called. A journal header file
** (JOURNAL_HDR_SZ bytes) is read from the current location in the journal
** file. See comments above function writeJournalHdr() for a description of
** the journal header format.
**
** If the header is read successfully, *nRec is set to the number of
** page records following this header and *dbSize is set to the size of the
** database before the transaction began, in pages. Also, pPager->cksumInit
** is set to the value read from the journal header. SQLITE_OK is returned
** in this case.
**
** If the journal header file appears to be corrupted, SQLITE_DONE is
** returned and *nRec and *dbSize are not set.  If JOURNAL_HDR_SZ bytes
** cannot be read from the journal file an error code is returned.
*/
static int readJournalHdr(