os_unix.c

一个小型的嵌入式数据库

    if( errno==EISDIR ){
      return SQLITE_CANTOPEN;
    }
#endif
    id->h = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY);
    if( id->h<0 ){
      return SQLITE_CANTOPEN; 
    }
    *pReadonly = 1;
  }else{
    *pReadonly = 0;
  }
  sqlite3OsEnterMutex();
  rc = findLockInfo(id->h, &id->pLock, &id->pOpen);
  sqlite3OsLeaveMutex();
  if( rc ){
    close(id->h);
    return SQLITE_NOMEM;
  }
  id->locktype = 0;
  id->isOpen = 1;
  TRACE3("OPEN    %-3d %s\n", id->h, zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
}


/*
** Attempt to open a new file for exclusive access by this process.
** The file will be opened for both reading and writing.  To avoid
** a potential security problem, we do not allow the file to have
** previously existed.  Nor do we allow the file to be a symbolic
** link.
**
** If delFlag is true, then make arrangements to automatically delete
** the file when it is closed.
**
** On success, write the file handle into *id and return SQLITE_OK.
**
** On failure, return SQLITE_CANTOPEN.
*/
int sqlite3OsOpenExclusive(const char *zFilename, OsFile *id, int delFlag){
  int rc;
  assert( !id->isOpen );
  if( access(zFilename, 0)==0 ){
    return SQLITE_CANTOPEN;
  }
  id->dirfd = -1;
  id->h = open(zFilename,
                O_RDWR|O_CREAT|O_EXCL|O_NOFOLLOW|O_LARGEFILE|O_BINARY, 0600);
  if( id->h<0 ){
    return SQLITE_CANTOPEN;
  }
  sqlite3OsEnterMutex();
  rc = findLockInfo(id->h, &id->pLock, &id->pOpen);
  sqlite3OsLeaveMutex();
  if( rc ){
    close(id->h);
    unlink(zFilename);
    return SQLITE_NOMEM;
  }
  id->locktype = 0;
  id->isOpen = 1;
  if( delFlag ){
    unlink(zFilename);
  }
  TRACE3("OPEN-EX %-3d %s\n", id->h, zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
}

/*
** Attempt to open a new file for read-only access.
**
** On success, write the file handle into *id and return SQLITE_OK.
**
** On failure, return SQLITE_CANTOPEN.
*/
int sqlite3OsOpenReadOnly(const char *zFilename, OsFile *id){
  int rc;
  assert( !id->isOpen );
  id->dirfd = -1;
  id->h = open(zFilename, O_RDONLY|O_LARGEFILE|O_BINARY);
  if( id->h<0 ){
    return SQLITE_CANTOPEN;
  }
  sqlite3OsEnterMutex();
  rc = findLockInfo(id->h, &id->pLock, &id->pOpen);
  sqlite3OsLeaveMutex();
  if( rc ){
    close(id->h);
    return SQLITE_NOMEM;
  }
  id->locktype = 0;
  id->isOpen = 1;
  TRACE3("OPEN-RO %-3d %s\n", id->h, zFilename);
  OpenCounter(+1);
  return SQLITE_OK;
}

/*
** Attempt to open a file descriptor for the directory that contains a
** file.  This file descriptor can be used to fsync() the directory
** in order to make sure the creation of a new file is actually written
** to disk.
**
** This routine is only meaningful for Unix.  It is a no-op under
** windows since windows does not support hard links.
**
** On success, a handle for a previously open file is at *id is
** updated with the new directory file descriptor and SQLITE_OK is
** returned.
**
** On failure, the function returns SQLITE_CANTOPEN and leaves
** *id unchanged.
*/
int sqlite3OsOpenDirectory(
  const char *zDirname,
  OsFile *id
){
  if( !id->isOpen ){
    /* Do not open the directory if the corresponding file is not already
    ** open. */
    return SQLITE_CANTOPEN;
  }
  assert( id->dirfd<0 );
  id->dirfd = open(zDirname, O_RDONLY|O_BINARY, 0644);
  if( id->dirfd<0 ){
    return SQLITE_CANTOPEN; 
  }
  TRACE3("OPENDIR %-3d %s\n", id->dirfd, zDirname);
  return SQLITE_OK;
}

/*
** If the following global variable points to a string which is the
** name of a directory, then that directory will be used to store
** temporary files.
*/
char *sqlite3_temp_directory = 0;

/*
** Create a temporary file name in zBuf.  zBuf must be big enough to
** hold at least SQLITE_TEMPNAME_SIZE characters.
*/
int sqlite3OsTempFileName(char *zBuf){
  static const char *azDirs[] = {
     0,
     "/var/tmp",
     "/usr/tmp",
     "/tmp",
     ".",
  };
  static const unsigned char zChars[] =
    "abcdefghijklmnopqrstuvwxyz"
    "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
    "0123456789";
  int i, j;
  struct stat buf;
  const char *zDir = ".";
  azDirs[0] = sqlite3_temp_directory;
  for(i=0; i<sizeof(azDirs)/sizeof(azDirs[0]); i++){
    if( azDirs[i]==0 ) continue;
    if( stat(azDirs[i], &buf) ) continue;
    if( !S_ISDIR(buf.st_mode) ) continue;
    if( access(azDirs[i], 07) ) continue;
    zDir = azDirs[i];
    break;
  }
  do{
    sprintf(zBuf, "%s/"TEMP_FILE_PREFIX, zDir);
    j = strlen(zBuf);
    sqlite3Randomness(15, &zBuf[j]);
    for(i=0; i<15; i++, j++){
      zBuf[j] = (char)zChars[ ((unsigned char)zBuf[j])%(sizeof(zChars)-1) ];
    }
    zBuf[j] = 0;
  }while( access(zBuf,0)==0 );
  return SQLITE_OK; 
}

#ifndef SQLITE_OMIT_PAGER_PRAGMAS
/*
** Check that a given pathname is a directory and is writable 
**
*/
int sqlite3OsIsDirWritable(char *zBuf){
  struct stat buf;
  if( zBuf==0 ) return 0;
  if( zBuf[0]==0 ) return 0;
  if( stat(zBuf, &buf) ) return 0;
  if( !S_ISDIR(buf.st_mode) ) return 0;
  if( access(zBuf, 07) ) return 0;
  return 1;
}
#endif /* SQLITE_OMIT_PAGER_PRAGMAS */

/*
** Read data from a file into a buffer.  Return SQLITE_OK if all
** bytes were read successfully and SQLITE_IOERR if anything goes
** wrong.
*/
int sqlite3OsRead(OsFile *id, void *pBuf, int amt){
  int got;
  assert( id->isOpen );
  SimulateIOError(SQLITE_IOERR);
  TIMER_START;
  got = read(id->h, pBuf, amt);
  TIMER_END;
  TRACE4("READ    %-3d %7d %d\n", id->h, last_page, TIMER_ELAPSED);
  SEEK(0);
  /* if( got<0 ) got = 0; */
  if( got==amt ){
    return SQLITE_OK;
  }else{
    return SQLITE_IOERR;
  }
}

/*
** Write data from a buffer into a file.  Return SQLITE_OK on success
** or some other error code on failure.
*/
int sqlite3OsWrite(OsFile *id, const void *pBuf, int amt){
  int wrote = 0;
  assert( id->isOpen );
  assert( amt>0 );
  SimulateIOError(SQLITE_IOERR);
  SimulateDiskfullError;
  TIMER_START;
  while( amt>0 && (wrote = write(id->h, pBuf, amt))>0 ){
    amt -= wrote;
    pBuf = &((char*)pBuf)[wrote];
  }
  TIMER_END;
  TRACE4("WRITE   %-3d %7d %d\n", id->h, last_page, TIMER_ELAPSED);
  SEEK(0);
  if( amt>0 ){
    return SQLITE_FULL;
  }
  return SQLITE_OK;
}

/*
** Move the read/write pointer in a file.
*/
int sqlite3OsSeek(OsFile *id, i64 offset){
  assert( id->isOpen );
  SEEK(offset/1024 + 1);
  lseek(id->h, offset, SEEK_SET);
  return SQLITE_OK;
}

/*
** The fsync() system call does not work as advertised on many
** unix systems.  The following procedure is an attempt to make
** it work better.
**
** The SQLITE_NO_SYNC macro disables all fsync()s.  This is useful
** for testing when we want to run through the test suite quickly.
** You are strongly advised *not* to deploy with SQLITE_NO_SYNC
** enabled, however, since with SQLITE_NO_SYNC enabled, an OS crash
** or power failure will likely corrupt the database file.
*/
static int full_fsync(int fd){
#ifdef SQLITE_NO_SYNC
  return SQLITE_OK;
#else
  int rc;
#ifdef F_FULLFSYNC
  rc = fcntl(fd, F_FULLFSYNC, 0);
  if( rc ) rc = fsync(fd);
#else
  rc = fsync(fd);
#endif
  return rc;
#endif
}

/*
** Make sure all writes to a particular file are committed to disk.
**
** Under Unix, also make sure that the directory entry for the file
** has been created by fsync-ing the directory that contains the file.
** If we do not do this and we encounter a power failure, the directory
** entry for the journal might not exist after we reboot.  The next
** SQLite to access the file will not know that the journal exists (because
** the directory entry for the journal was never created) and the transaction
** will not roll back - possibly leading to database corruption.
*/
int sqlite3OsSync(OsFile *id){
  assert( id->isOpen );
  SimulateIOError(SQLITE_IOERR);
  TRACE2("SYNC    %-3d\n", id->h);
  if( full_fsync(id->h) ){
    return SQLITE_IOERR;
  }
  if( id->dirfd>=0 ){
    TRACE2("DIRSYNC %-3d\n", id->dirfd);
    full_fsync(id->dirfd);
    close(id->dirfd);  /* Only need to sync once, so close the directory */
    id->dirfd = -1;    /* when we are done. */
  }
  return SQLITE_OK;
}

/*
** Sync the directory zDirname. This is a no-op on operating systems other
** than UNIX.
*/
int sqlite3OsSyncDirectory(const char *zDirname){
  int fd;
  int r;
  SimulateIOError(SQLITE_IOERR);
  fd = open(zDirname, O_RDONLY|O_BINARY, 0644);
  TRACE3("DIRSYNC %-3d (%s)\n", fd, zDirname);
  if( fd<0 ){
    return SQLITE_CANTOPEN; 
  }
  r = fsync(fd);
  close(fd);
  return ((r==0)?SQLITE_OK:SQLITE_IOERR);
}

/*
** Truncate an open file to a specified size
*/
int sqlite3OsTruncate(OsFile *id, i64 nByte){
  assert( id->isOpen );
  SimulateIOError(SQLITE_IOERR);
  return ftruncate(id->h, nByte)==0 ? SQLITE_OK : SQLITE_IOERR;
}

/*
** Determine the current size of a file in bytes
*/
int sqlite3OsFileSize(OsFile *id, i64 *pSize){
  struct stat buf;
  assert( id->isOpen );
  SimulateIOError(SQLITE_IOERR);
  if( fstat(id->h, &buf)!=0 ){
    return SQLITE_IOERR;
  }
  *pSize = buf.st_size;
  return SQLITE_OK;
}

/*
** This routine checks if there is a RESERVED lock held on the specified
** file by this or any other process. If such a lock is held, return
** non-zero.  If the file is unlocked or holds only SHARED locks, then
** return zero.
*/
int sqlite3OsCheckReservedLock(OsFile *id){
  int r = 0;

  assert( id->isOpen );
  sqlite3OsEnterMutex(); /* Needed because id->pLock is shared across threads */

  /* Check if a thread in this process holds such a lock */
  if( id->pLock->locktype>SHARED_LOCK ){
    r = 1;
  }

  /* Otherwise see if some other process holds it.
  */
  if( !r ){
    struct flock lock;
    lock.l_whence = SEEK_SET;
    lock.l_start = RESERVED_BYTE;
    lock.l_len = 1;
    lock.l_type = F_WRLCK;
    fcntl(id->h, F_GETLK, &lock);
    if( lock.l_type!=F_UNLCK ){
      r = 1;
    }
  }
  
  sqlite3OsLeaveMutex();
  TRACE3("TEST WR-LOCK %d %d\n", id->h, r);

  return r;
}

#ifdef SQLITE_DEBUG
/*
** Helper function for printing out trace information from debugging
** binaries. This returns the string represetation of the supplied
** integer lock-type.
*/
static const char * locktypeName(int locktype){
  switch( locktype ){
  case NO_LOCK: return "NONE";
  case SHARED_LOCK: return "SHARED";
  case RESERVED_LOCK: return "RESERVED";
  case PENDING_LOCK: return "PENDING";
  case EXCLUSIVE_LOCK: return "EXCLUSIVE";
  }
  return "ERROR";
}
#endif

/*
** Lock the file with the lock specified by parameter locktype - one
** of the following:
**
**     (1) SHARED_LOCK
**     (2) RESERVED_LOCK
**     (3) PENDING_LOCK
**     (4) EXCLUSIVE_LOCK
**
** Sometimes when requesting one lock state, additional lock states
** are inserted in between.  The locking might fail on one of the later
** transitions leaving the lock state different from what it started but
** still short of its goal.  The following chart shows the allowed
** transitions and the inserted intermediate states:
**
**    UNLOCKED -> SHARED
**    SHARED -> RESERVED
**    SHARED -> (PENDING) -> EXCLUSIVE
**    RESERVED -> (PENDING) -> EXCLUSIVE
**    PENDING -> EXCLUSIVE
**
** This routine will only increase a lock.  Use the sqlite3OsUnlock()
** routine to lower a locking level.
*/
int sqlite3OsLock(OsFile *id, int locktype){
  /* The following describes the implementation of the various locks and
  ** lock transitions in terms of the POSIX advisory shared and exclusive
  ** lock primitives (called read-locks and write-locks below, to avoid
  ** confusion with SQLite lock names). The algorithms are complicated
  ** slightly in order to be compatible with windows systems simultaneously
  ** accessing the same database file, in case that is ever required.
  **
  ** Symbols defined in os.h indentify the 'pending byte' and the 'reserved
  ** byte', each single bytes at well known offsets, and the 'shared byte
  ** range', a range of 510 bytes at a well known offset.